Implement rudimentary support for the PowerPC SPE APU

[jhibbits]

Summary

This adds support for the Signal Processing Engine, a non-standard FPU
and vector unit on some PowerPC cores (Freescale e500, maybe others).

Currently supports:

• asm parsing and printing for almost all instructions
• Code generation following the SPE ABI at the llvm IR level (function call)
• Single- and Double-precision math at the level supported by the APU, including conversion between precisions and with integers
• Support for some vector operations (float math, some integer math)

Currently not fully compliant with the SPEPIM; anything that takes a
__ev64_opaque argument in intrinsics instead takes a v2i32.

Along with this, add the Freescale e500 scheduler, which is slightly
different from the e500mc scheduler.

[chmeee]

Tests are in the works right now. Until they're ready, this is a preliminary review, to see if I did anything unnecessary, or if I'm missing anything obvious.

[chmeee]

Add tests, add intrinsics (required for one test). Fix several bugs found via tests.

[chmeee]

Add patterns for intrinsics for codegen, and fix the intrinsic definitions.

[chmeee]

Adding some more powerpc regulars in hopes of getting review. Feel free to remove yourself if you're not interested.

[chmeee]

Correct e500 instruction scheduling.

[nemanjai]

Seems like this patch should include more tests:

• How about disassembly? (maybe test/MC/Disassembler/PowerPC/ppc64-encoding-e500.txt)
• It doesn't seem that the single CodeGen test case has enough coverage to exercise every single one of the patterns
• How about FastISEL test cases since you added FastISEL support?

lib/Target/PowerPC/PPCFastISel.cpp

558

Debugging artifact left behind?

660

The code above uses the ternary operator and is more concise and readable. Could you use it here as well?

812

It seems very strange to me that we don't need to modify this at all. The SPE comparisons seem to always set bit 1 of the respective CR field (leaving the other 3 bits undefined). As such, won't we have to always add PPC::PRED_GT rather than PPCPred?

Or am I missing something?

If my assumption is correct here and the predicate needs to change, I would actually imagine it would be better if we define a new predicate - say PPC::PRED_SPE for this. The reason is that we have transformations that will convert one predicate into another and we certainly don't want to use undefined CR bits for SPE.

1254

I think just using the ternary operator here is more readable.

2008

Nit: variables start with capitals.

2009

This on the other hand, I think looks a little messy as a ternary operator now. I think it would be more readable as an if.

2020

Similarly here - the ternary operator looks messy.

lib/Target/PowerPC/PPCFrameLowering.cpp

178

Nit: complete sentences in comments.

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2475

Is the reason for this change to ensure that we don't enter the block when the value type of the LHS is an SPE vector? If so, wouldn't it be clearer to actually check for that (i.e. that the LHS is a vector type that is not an SPE vector type on an SPE target)?

I guess what I'm getting at is whether it is possible to end up in this block if the LHS has an SPE vector type if the target happens to have both Altivec and SPE (if that's even possible).

lib/Target/PowerPC/PPCInstrInfo.td

1217

It seems like this might be a bit fragile. With this approach, we have to ensure that future FP patterns we add have this predicate so they don't break SPE.

I'm not sure how we can do this more reliably though. Is there any precedent for this on other targets?
I suppose we could use the AddedComplexity hack to ensure that when the SPE feature is available, we ensure we use the patterns for SPE. That seems more hacky, but less fragile as none of the patterns we add in the future will be selected over the SPE patterns so we shouldn't break SPE.

3619

Why does this block need to be in this file rather than PPCInstrSPE.td?

3862

This is part of the reason I don't like the HasTraditionalFPU solution. It is not clear or obvious why we don't need that pattern here.

lib/Target/PowerPC/PPCInstrSPE.td

14

I am really not reviewing the changes to this file as it would be too time-consuming to cross-reference this to the ISA. I assume there's some thorough functional testing for this enablement on a processor with SPE.

test/MC/PowerPC/ppc64-encoding-spe.s

1

These two lines have no actual change? Just an extra space? If so, please refrain from that.

[chmeee]

Thanks for the review. It is in need of more tests, I know.

• I can add to the Disassemble tests. There were not disassemble tests for the existing SPE support, so I missed that section.
• What more CodeGen tests are needed? I know what I have is not exhaustive for error cases or edge cases. Some code paths couldn't be tested with simple tests (efdabs doesn't get generated with function calls, so I have to get creative, and haven't found anything easy for it)
• I'll see what I can add for the fast isel tests.

lib/Target/PowerPC/PPCFastISel.cpp

558

Yeah. I don't even remember what I was debugging with it now.

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2475

Since the opcodes overlap it is impossible to have both SPE and Altivec on the same target.

lib/Target/PowerPC/PPCInstrInfo.td

1217

I admit this is really fragile, but I couldn't think of a better way. What's the AddedComplexity hack?

3619

To be honest, I don't recall. I had added it to PPCInstrSPE.td and ran into problems, but it magically worked in here. I'll have to try that again.

test/MC/PowerPC/ppc64-encoding-spe.s

1

I think my editor screwed this up, I didn't intentionally change it. Fixing.

[nemanjai]

In D38778#938619, @chmeee wrote:

Thanks for the review. It is in need of more tests, I know.

• I can add to the Disassemble tests. There were not disassemble tests for the existing SPE support, so I missed that section.

OK. Cool, that would be good.

• What more CodeGen tests are needed? I know what I have is not exhaustive for error cases or edge cases. Some code paths couldn't be tested with simple tests (efdabs doesn't get generated with function calls, so I have to get creative, and haven't found anything easy for it)

• There should be a test case for every instruction that can be matched with a pattern (including those that are matched to intrinsics which appear to be numerous)
• There should also be a test case for each instruction that is emitted through custom logic - things like the various setcc/selectcc patterns, etc.

• I'll see what I can add for the fast isel tests.

Yeah, pretty much anything you're emitting through FISel (IIRC you have loads, stores, compares, etc.)

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2475

OK. Fair enough.

lib/Target/PowerPC/PPCInstrInfo.td

1217

On targets that have VSX, we prefer to use VSX opcodes for scalar floating point operations. To ensure we don't select VSX patterns when we don't have VSX, we use a predicate. And to ensure we select the VSX pattern rather than an FPU pattern, we wrap the patterns in let AddedComplexity = 400 blocks. The selector will always choose the highest complexity pattern when multiple patterns produce a match.
Of course, it's a hack, but it has an obvious advantage over this approach - when someone adds a pattern, they'll add a test case to ensure that pattern is matched, they won't necessarily think to add a test case to ensure their pattern isn't matched on subtargets that have some other feature enabled.

Ideally, the selector would use the fact that the output pattern has operands with registers that aren't available on the target. Perhaps there's a way to do that - maybe @kparzysz would know.

[chmeee]

lib/Target/PowerPC/PPCInstrInfo.td

1217

There is one other problem with using the AddedComplexity hack: Not all "traditional" FPU constructs are available in the SPE, so if it doesn't match an SPE operation it would still end up falling back to the FPU constructs, which obviously wouldn't work.

[nemanjai]

lib/Target/PowerPC/PPCInstrInfo.td

1217

Sure, but if such a construct made it to ISEL with your current solution, you'd get a failure to select. I suppose a compile failure is better than SIGILL at runtime, but it is still not the desired behaviour.

I think that any operations that aren't available in SPE should not be marked legal. Furthermore, I think it would be a good idea to add an assert somewhere (perhaps the asm streamer) that would trip if you ever allocate an FPR on a subtarget that has SPE.

[kparzysz]

lib/Target/PowerPC/PPCInstrInfo.td

1217

There is no explicit association between registers and targets, so the instruction selector does not have that knowledge. I cannot think of any elegant solution to this. I suspect that the only viable approach is to always assume the most restricted target and include any extra features under predicates. This means that if you add support for a target that is more restricted than everything else so far, then that "everything" now becomes "extensions" to the new "most restricted baseline".

What could be a possibility is that the DAG operations that are not available on SPE are replaced with special SPE-specific ISD opcodes (during DAG preprocessing), so that during pattern matching they could only match the SPE-specific patterns (which would have to be provided for each such operation).

[hfinkel]

lib/Target/PowerPC/PPCCallingConv.td

187

Please update this comment for SPE.

lib/Target/PowerPC/PPCFastISel.cpp

583

Remove spaces to line up break on this line with the break on the previous line.

812

I agree with @nemanjai, I think that we need a new predicate value to handle this kind of usage of the CR bits. efscmpeq and friends seem to just set bit 1 of the CR register (and, moreover, leave the others explicitly undefined).

996

Line is too long.

1234

You should do this.

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2475

If you do it this way, you'll break the QPX check below (which is also mutually-exclusive with Altivec). Why can't you just add the SPE check to the QPX check below?

test/CodeGen/PowerPC/spe.ll

3

Please also add tests for fast-isel, spill/restore (you can use inline asm to force spilling), inline asm register constraints, full coverage for all fcmp operations. Not all of these tests actually check their output, please fix that.

315

Why are these commented out?

498

Why are these commented out?

[chmeee]

lib/Target/PowerPC/PPCFastISel.cpp

1234

Nope, the comment is wrong, as is the containing 'else' block, so deleting.

[chmeee]

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2475

Oops, sorry about that. I'm surprised existing tests didn't pick up on that.

[chmeee]

Address most comments. I think tests are pretty complete now, and all bugs
found have been fixed. Removed selection for setcc on vectors.

[nemanjai]

I have gotten to PPCInstrInfo.td and wanted to submit the comments rather than delaying the review further until I am completely done with it. I'll continue to review it and add comments over the remainder of the week.
A couple of general notes:

• In retrospect, the review process would have probably been much quicker and easier if this was split up into a bunch of smaller patches. Perhaps the target features and calling convention in the first, then the new register classes and spill/restore bits, then legalization of small groups of instructions along with the associated intrinsics/builtins/tests, etc.
• Considering we have a mutually exclusive situation between FPU and SPE and the sheer prevalence of FPU cores and developers adding code for those, it would probably be a good idea to add some checks that we don't break SPE. What I'm thinking would be a good idea is a linear traversal in SDAG post-processing that would ensure that all the MachineSDNodes in the DAG are available on SPE. I think this can simply be done by checking the operand/result register classes to make sure you don't end up with any F4RC or F8RC registers. Of course, this check would only fire on SPE subtargets.
• Is there a Clang portion to follow? To define builtins for all the intrinsics in C/C++. Or do you not plan to target SPE from C/C++?

lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h

52 ↗

(On Diff #128842)

Just a comment stating something like SPE instructions always set the GT bit for comparisons.

lib/Target/PowerPC/PPCCallingConv.td

255

Please add a comment to make it clearer here that we're splitting the CSR list into GPR and FPR since SPE targets don't use FPR's and therefore can't have them in the CSR list.

lib/Target/PowerPC/PPCFastISel.cpp

160

If you've changed all the call sites, I think it'd be good to remove the default arg.

1094

Shouldn't there be a user for this result? I think at this point, we use updateValueMap() to map the instruction to the vreg we are defining with the newly emitted instruction.

lib/Target/PowerPC/PPCFrameLowering.cpp

1752

Seems that we would have to handle SPE registers here, wouldn't we? I don't think we can rely on the ordering of SPE registers with VR registers in order to determine the lowest numbered register that needs to be spilled (and therefore how large the spill area is).

For the GPRC/SPE4RC class, we should be safe since they are classes that have the same registers in them, but SPERC/VRRC do not have the same registers.

1890

Do the SPE register spills require 16-byte alignment? It may very well be so, but I'd prefer that this be explicit in a comment.

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

2478

Why not in a single if statement with an or condition? In any case, the else is redundant.

2508

Nit: unnecessary parentheses.

3051

Not sure if anyone will object to this, but I find that even though the else/else if binds to the right if in situations like this, I prefer to have braces around the block for the outer if.

When reading this, I find it more clear with the braces when the block contains more than 2-3 lines.

3066

Do you think it would make sense to assert here that you have a type that SPE can handle? I know legalization should clean this up, but I think it's a good way to catch any missed legalization here if any bleeds through.

I probably would have done the same thing for the fallback path to PPC::SELECT_CC_VRRC there too, so I think it's no worse for you to leave this as the fallback for SPE.

lib/Target/PowerPC/PPCInstrInfo.cpp

279

Why are PPC::EVLDD and PPC::SPELWZ not here?

986

Can we actually get here? Won't the registers be in GPRCRegClass and be caught above?

lib/Target/PowerPC/PPCRegisterInfo.td

225

Maybe just a quick comment as to what SPEACC is and why it doesn't need to be encoded.

[nemanjai]

In D38778#1000445, @nemanjai wrote:

• Considering we have a mutually exclusive situation between FPU and SPE and the sheer prevalence of FPU cores and developers adding code for those, it would probably be a good idea to add some checks that we don't break SPE. What I'm thinking would be a good idea is a linear traversal in SDAG post-processing that would ensure that all the MachineSDNodes in the DAG are available on SPE. I think this can simply be done by checking the operand/result register classes to make sure you don't end up with any F4RC or F8RC registers. Of course, this check would only fire on SPE subtargets.

Of course, this won't catch late additions such as regclass copies and MachineInstr's we add in various MI-level passes and late pseudo-expansion, so perhaps the better choice would be to catch this in EmitInstruction by getting the MCInstrDesc for the MI and checking the register class there.

[chmeee]

lib/Target/PowerPC/PPCFastISel.cpp

160

Not all have been changed (line 1172 lacks it).

lib/Target/PowerPC/PPCFrameLowering.cpp

1752

You're right. It's better to split it up. Even though the code itself is identical between the two, it's better to be more explicit of the purpose of the block than to be clever to save the code, particularly considering, as you said, we shouldn't rely on ordering of the register definitions.

1890

Yes, it does require 16-byte alignment. But, as you pointed out above, it's better to split them up to be more explicit.

lib/Target/PowerPC/PPCInstrInfo.cpp

279

Oversight.

[chmeee]

Is there a Clang portion to follow? To define builtins for all the intrinsics in C/C++. Or do you not plan to target SPE from C/C++?

Yes, there is a Clang portion to follow, and without the intrinsics it's a trivial patch. I'm waiting on this to be completed and committed before I post that, in case things have to change for it. The trivial no-intrinsics part is done and works. It merges -mspe and -mabi=spe of GCC into a single -mspe, since I'm not splitting the codegen into code and ABI here.

[codeman.consulting]

I have a couple of questions (along with a nitpicky inline comment).

The e500 docs from Freescale specify e500v1 which supports basically everything added except for the efd* instructions and the various vector insts; e500v2 is required for efd* inst. / double support. They also have different scheduling models. It looks you're just implementing e500v2, should this be specified at all, or is the e500v1 not needed as a target?

Probably less important: The same manual also mentions that IEEE 754 compliance isn't full in hardware with respect to denorms; Is this handled at the OS level or does some fp exception handler wind up getting linked in for full support?

lib/Target/PowerPC/PPCFastISel.cpp

483

Personally this ternary is (and nearly already was) complicated enough I'd rework it into an if statement. It's not unreadable but takes longer to parse than necessary.

[chmeee]

In D38778#1018576, @codeman.consulting wrote:

I have a couple of questions (along with a nitpicky inline comment).

The e500 docs from Freescale specify e500v1 which supports basically everything added except for the efd* instructions and the various vector insts; e500v2 is required for efd* inst. / double support. They also have different scheduling models. It looks you're just implementing e500v2, should this be specified at all, or is the e500v1 not needed as a target?

Yes, this is intended only for e500v2, not e500v1. I can't find any current SoC that uses the e500v1 core anywhere, so didn't put much thought into splitting that part out (it could probably be done trivially, just not adding the SPERCRegisterClass in the PPCISelLowering bit). Do you know of any current SoC with e500v1? Additionally, this is to target the powerpcspe target for FreeBSD, which is modeled after the powerpcspe target for Debian, which itself targets e500v2.

Probably less important: The same manual also mentions that IEEE 754 compliance isn't full in hardware with respect to denorms; Is this handled at the OS level or does some fp exception handler wind up getting linked in for full support?

This is expected to be handled at the OS level.

[jhibbits]

I just did a search on NXP's website, and it appears the last e500v1-based chip went into production around 2001/2002, and had a Longevity-program lifetime of 10 years (MPC8560), superseded by rev 2.0 of the silicon, which used e500v2. So it appears there are no current e500v1-based SoCs in production at this time. e500v2-based SoCs appear to have another 8 years or so, as the latest one went into the longevity program in 2010, with a 15 year product lifetime.

[nemanjai]

A few more comments inline and a couple of general comments.

1. Due to the size of this patch, a thorough review is difficult and time-consuming. Furthermore, it is not feasible that anyone will review in detail whether the tests cover all instructions, all intrinsics. I think much of this support would have been upstream if it were done in small incremental patches.
2. I am still not thrilled that FP ops in general now need a predicate defined in the .td file. There are a couple of issues with doing that. The obvious one is that anyone adding patterns in the future now has to be careful to set the predicate. Of course, this isn't all that likely to happen a lot since any new ISA implementations will have to be guarded by their own predicates. So the problem only really exists if someone is adding new patterns that are to work with FPU instructions on all the CPU's that have it. The second issue with this is a bit more subtle. Namely something like this:

let Predicates = [ PredicateOne ] in {
  // Many many lines of code ...
  let Predicates = [ PredicateTwo ] in {
    // More code
  }
}

If I recall correctly, // More code actually only requires PredicateTwo to be satisfied.

3. I did not review the instruction definitions or the scheduling info since I don't know anything about SPE and there's a whole lot of code.
4. I will apply your next update to ToT and try to do some extensive testing on PPC64LE and PPC64BE in case this patch causes unforeseen failures.

lib/Target/PowerPC/PPCFastISel.cpp

813

Is it an option to just update getComparePred() to do the right thing and then not need this conditional op?

lib/Target/PowerPC/PPCISelLowering.cpp

120

I am hoping for us to clean this section up to make it more readable in the future. Namely, I would like for us to minimize the nesting based on target features. As such, is it possible for you to add things that change as a result of the subtarget having SPE into a single section. Perhaps a section in the end that will reset everything that needs to be reset with SPE.

Of course, this may be somewhat controversial, so if you or others don't agree, I suppose it can stay this way until we actually get to the clean-up effort.

346

Sorry, this is a big patch so it's hard to keep track of things. Does SPE have these conversions for the vector types? In any case, in places where you're adding legalization actions for SPE, if v2f32 and v2f32 don't need to be mentioned, perhaps just mention it in a comment.

lib/Target/PowerPC/PPCSubtarget.cpp

147 ↗

(On Diff #128842)

So -mcpu=pwr7 -mattr=-fpu will not actually disable FPU. I don't think that's a problem. I'm thinking more along the lines of whether this affects soft-float support, but I've never really looked all that closely into soft-float.

[jhibbits]

I will attempt to break this up into (slightly) smaller chunks. My current thought is:

• New instructions, and encoding tests
• e500 scheduling
• floating point and ABI
• intrinsics
• vector

Each will build upon the previous, but it should be smaller and easier to review. Does this seem reasonable?

Apologies long after the fact, I did intend to start this as only adding the instructions and floating point bits, but it grew as it sat waiting for review in the initial stages.

[nemanjai]

Is this patch ready to be abandoned now that it has been broken up into smaller patches which were committed?

[jhibbits]

Yes, yes it is

include/llvm/IR/IntrinsicsPowerPC.td

 def int_ppc_tsuspend : GCCBuiltin<"__builtin_tsuspend">,
       Intrinsic<[llvm_i32_ty], [], []>;
 
 def int_ppc_ttest : GCCBuiltin<"__builtin_ttest">,
       Intrinsic<[llvm_i64_ty], [], []>;
 
 def int_ppc_cfence : Intrinsic<[], [llvm_anyint_ty], []>;
 }
+
+//===----------------------------------------------------------------------===//
+// PowerPC SPE Intrinsic Definitions.
+
+let TargetPrefix = "ppc" in {  // All intrinsics start with "llvm.ppc.".
+
+/// PowerPC_SPE_Intrinsic - Base class for all SPE intrinsics.
+class PowerPC_SPE_Intrinsic<string GCCIntSuffix, list<LLVMType> ret_types,
+                            list<LLVMType> param_types,
+                            list<IntrinsicProperty> properties>
+  : GCCBuiltin<!strconcat("__builtin_spe_", GCCIntSuffix)>,
+    Intrinsic<ret_types, param_types, properties>;
+
+/// Single argument
+class PowerPC_SPE_Vec_Intrinsic21<string GCCIntSuffix,
+                                  LLVMType rt, LLVMType vec_type>
+      : PowerPC_SPE_Intrinsic<GCCIntSuffix, [rt], [vec_type], [IntrNoMem]>;
+
+class PowerPC_SPE_Vec_Intrinsic1<string GCCIntSuffix, LLVMType vec_type>
+      : PowerPC_SPE_Intrinsic<GCCIntSuffix, [vec_type], [vec_type], [IntrNoMem]>;
+
+/// Two argument (vectors)
+class PowerPC_SPE_Vec_Intrinsic2<string GCCIntSuffix, LLVMType vec_type>
+      : PowerPC_SPE_Intrinsic<GCCIntSuffix, [vec_type],
+                              [vec_type, vec_type], [IntrNoMem]>;
+
+/// Single plus immediate
+class PowerPC_SPE_Vec_IntrinsicI<string GCCIntSuffix, LLVMType vec_type,
+                                 LLVMType imm>
+      : PowerPC_SPE_Intrinsic<GCCIntSuffix, [vec_type],
+                              [vec_type, imm], [IntrNoMem]>;
+
+class PowerPC_SPE_Vec_I1<string GCCIntSuffix>
+      : PowerPC_SPE_Vec_Intrinsic1<GCCIntSuffix, llvm_v2i32_ty>;
+
+class PowerPC_SPE_Vec_I2<string GCCIntSuffix>
+      : PowerPC_SPE_Vec_Intrinsic2<GCCIntSuffix, llvm_v2i32_ty>;
+
+class PowerPC_SPE_Vec_II<string GCCIntSuffix>
+      : PowerPC_SPE_Vec_IntrinsicI<GCCIntSuffix, llvm_v2i32_ty, llvm_i32_ty>;
+
+class PowerPC_SPE_Vec_F1<string GCCIntSuffix>
+      : PowerPC_SPE_Vec_Intrinsic1<GCCIntSuffix, llvm_v2f32_ty>;
+
+class PowerPC_SPE_Vec_F2<string GCCIntSuffix>
+      : PowerPC_SPE_Vec_Intrinsic2<GCCIntSuffix, llvm_v2f32_ty>;
+
+
+def int_ppc_spe_brinc :
+      PowerPC_SPE_Intrinsic<"brinc", [llvm_i32_ty],
+                            [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+
+def int_ppc_spe_evabs : PowerPC_SPE_Vec_I1<"evabs">;
+def int_ppc_spe_evneg : PowerPC_SPE_Vec_I1<"evneg">;
+def int_ppc_spe_evaddiw : PowerPC_SPE_Vec_II<"evaddiw">;
+def int_ppc_spe_evaddw : PowerPC_SPE_Vec_I2<"evaddw">;
+def int_ppc_spe_evsubifw : 
+      PowerPC_SPE_Intrinsic<"evsubifw", [llvm_v2i32_ty],
+                            [llvm_i32_ty, llvm_v2i32_ty], [IntrNoMem]>;
+
+def int_ppc_spe_evsubfw: PowerPC_SPE_Vec_I2<"evsubfw">;
+def int_ppc_spe_evand: PowerPC_SPE_Vec_I2<"evand">;
+def int_ppc_spe_evandc: PowerPC_SPE_Vec_I2<"evandc">;
+def int_ppc_spe_evnand: PowerPC_SPE_Vec_I2<"evnand">;
+def int_ppc_spe_evor: PowerPC_SPE_Vec_I2<"evor">;
+def int_ppc_spe_evorc: PowerPC_SPE_Vec_I2<"evorc">;
+def int_ppc_spe_evnor: PowerPC_SPE_Vec_I2<"evnor">;
+def int_ppc_spe_evextsb : PowerPC_SPE_Vec_I1<"evextsb">;
+def int_ppc_spe_evextsh : PowerPC_SPE_Vec_I1<"evextsh">;
+def int_ppc_spe_evrlw : PowerPC_SPE_Vec_I2<"evrlw">;
+def int_ppc_spe_evrlwi : PowerPC_SPE_Vec_II<"evrlwi">;
+def int_ppc_spe_evslw : PowerPC_SPE_Vec_I2<"evslw">;
+def int_ppc_spe_evslwi : PowerPC_SPE_Vec_II<"evslwi">;
+def int_ppc_spe_evsrws : PowerPC_SPE_Vec_I2<"evsrws">;
+def int_ppc_spe_evsrwis : PowerPC_SPE_Vec_II<"evsrwis">;
+def int_ppc_spe_evsrwu : PowerPC_SPE_Vec_I2<"evsrwu">;
+def int_ppc_spe_evsrwiu : PowerPC_SPE_Vec_II<"evsrwiu">;
+def int_ppc_spe_evcntlsw : PowerPC_SPE_Vec_I1<"evcntlsw">;
+def int_ppc_spe_evcntlzw : PowerPC_SPE_Vec_I1<"evcntlzw">;
+def int_ppc_spe_evrndw : PowerPC_SPE_Vec_I1<"evrndw">;
+def int_ppc_spe_evmergehi : PowerPC_SPE_Vec_I2<"evmergehi">;
+def int_ppc_spe_evmergelo : PowerPC_SPE_Vec_I2<"evmergelo">;
+def int_ppc_spe_evmergehilo : PowerPC_SPE_Vec_I2<"evmergehilo">;
+def int_ppc_spe_evmergelohi : PowerPC_SPE_Vec_I2<"evmergelohi">;
+def int_ppc_spe_evsplati : PowerPC_SPE_Intrinsic<"evsplati",
+                          [llvm_v2i32_ty], [llvm_i32_ty], [IntrNoMem]>;
+def int_ppc_spe_evsplatfi : PowerPC_SPE_Intrinsic<"evsplatfi",
+                          [llvm_v2i32_ty], [llvm_i32_ty], [IntrNoMem]>;
+def int_ppc_spe_evdivws : PowerPC_SPE_Vec_I2<"evdivws">;
+def int_ppc_spe_evdivwu : PowerPC_SPE_Vec_I2<"evdivwu">;
+
+def int_ppc_spe_evfsabs : PowerPC_SPE_Vec_F1<"evfsabs">;
+def int_ppc_spe_evfsnabs : PowerPC_SPE_Vec_F1<"evfsnabs">;
+def int_ppc_spe_evfsneg : PowerPC_SPE_Vec_F1<"evfsneg">;
+def int_ppc_spe_evfsadd : PowerPC_SPE_Vec_F2<"evfsadd">;
+def int_ppc_spe_evfssub : PowerPC_SPE_Vec_F2<"evfssub">;
+def int_ppc_spe_evfsmul : PowerPC_SPE_Vec_F2<"evfsmul">;
+def int_ppc_spe_evfsdiv : PowerPC_SPE_Vec_F2<"evfsdiv">;
+def int_ppc_spe_evfscfui : PowerPC_SPE_Vec_Intrinsic21<"evfscfui",
+                                                      llvm_v2f32_ty,
+                                                      llvm_v2i32_ty>;
+def int_ppc_spe_evfscfsi : PowerPC_SPE_Vec_Intrinsic21<"evfscfsi",
+                                                      llvm_v2f32_ty,
+                                                      llvm_v2i32_ty>;
+def int_ppc_spe_evfscfuf : PowerPC_SPE_Vec_Intrinsic21<"evfscfuf",
+                                                      llvm_v2f32_ty,
+                                                      llvm_v2i32_ty>;
+def int_ppc_spe_evfscfsf : PowerPC_SPE_Vec_Intrinsic21<"evfscfsf",
+                                                      llvm_v2f32_ty,
+                                                      llvm_v2i32_ty>;
+
+def int_ppc_spe_evfsctui : PowerPC_SPE_Vec_Intrinsic21<"evfsctui",
+                                                      llvm_v2i32_ty,
+                                                      llvm_v2f32_ty>;
+def int_ppc_spe_evfsctsi : PowerPC_SPE_Vec_Intrinsic21<"evfsctsi",
+                                                      llvm_v2i32_ty,
+                                                      llvm_v2f32_ty>;
+def int_ppc_spe_evfsctuf : PowerPC_SPE_Vec_Intrinsic21<"evfsctuf",
+                                                      llvm_v2i32_ty,
+                                                      llvm_v2f32_ty>;
+def int_ppc_spe_evfsctsf : PowerPC_SPE_Vec_Intrinsic21<"evfsctsf",
+                                                      llvm_v2i32_ty,
+                                                      llvm_v2f32_ty>;
+def int_ppc_spe_evfsctuiz : PowerPC_SPE_Vec_Intrinsic21<"evfsctuiz",
+                                                        llvm_v2i32_ty,
+                                                        llvm_v2f32_ty>;
+def int_ppc_spe_evfsctsiz : PowerPC_SPE_Vec_Intrinsic21<"evfsctsiz",
+                                                        llvm_v2i32_ty,
+                                                        llvm_v2f32_ty>;
+
+// Load/store
+def int_ppc_spe_evlddx : PowerPC_SPE_Intrinsic<"evlddx",
+                        [llvm_double_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evldwx : PowerPC_SPE_Intrinsic<"evldwx",
+                        [llvm_v2i32_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evldhx : PowerPC_SPE_Intrinsic<"evldhx",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhex : PowerPC_SPE_Intrinsic<"evlwhex",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhoux : PowerPC_SPE_Intrinsic<"evlwhoux",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhosx : PowerPC_SPE_Intrinsic<"evlwhosx",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwwsplatx : PowerPC_SPE_Intrinsic<"evlwwsplatx",
+                              [llvm_v2i32_ty], [llvm_i32_ty, llvm_i32_ty],
+                              [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhsplatx : PowerPC_SPE_Intrinsic<"evlwhsplatx",
+                                [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                                [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhesplatx : PowerPC_SPE_Intrinsic<"evlhhesplatx",
+                        [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhousplatx : PowerPC_SPE_Intrinsic<"evlhhousplatx",
+                                  [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                                  [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhossplatx : PowerPC_SPE_Intrinsic<"evlhhossplatx",
+                                  [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                                  [IntrArgMemOnly]>;
+def int_ppc_spe_evldd : PowerPC_SPE_Intrinsic<"evldd",
+                        [llvm_double_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evldw : PowerPC_SPE_Intrinsic<"evldw",
+                        [llvm_v2i32_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evldh : PowerPC_SPE_Intrinsic<"evldh",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhe : PowerPC_SPE_Intrinsic<"evlwhe",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhou : PowerPC_SPE_Intrinsic<"evlwhou",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhos : PowerPC_SPE_Intrinsic<"evlwhos",
+                        [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlwwsplat : PowerPC_SPE_Intrinsic<"evlwwsplat",
+                              [llvm_v2i32_ty], [llvm_i32_ty, llvm_i32_ty],
+                              [IntrArgMemOnly]>;
+def int_ppc_spe_evlwhsplat : PowerPC_SPE_Intrinsic<"evlwhsplat",
+                                [llvm_v4i16_ty], [llvm_i32_ty, llvm_i32_ty],
+                                [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhesplat : PowerPC_SPE_Intrinsic<"evlhhesplat",
+                        [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                        [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhousplat : PowerPC_SPE_Intrinsic<"evlhhousplat",
+                                  [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                                  [IntrArgMemOnly]>;
+def int_ppc_spe_evlhhossplat : PowerPC_SPE_Intrinsic<"evlhhossplat",
+                                  [llvm_v4i16_ty], [llvm_i16_ty, llvm_i32_ty],
+                                  [IntrArgMemOnly]>;
+
+def int_ppc_spe_evstddx : PowerPC_SPE_Intrinsic<"evstddx",
+                            [],
+                            [llvm_double_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstdwx : PowerPC_SPE_Intrinsic<"evstdwx",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstdhx : PowerPC_SPE_Intrinsic<"evstdhx",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwwex : PowerPC_SPE_Intrinsic<"evstwwex",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwwox : PowerPC_SPE_Intrinsic<"evstwwox",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwhex : PowerPC_SPE_Intrinsic<"evstwhex",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwhox : PowerPC_SPE_Intrinsic<"evstwhox",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+
+def int_ppc_spe_evstdd : PowerPC_SPE_Intrinsic<"evstdd",
+                            [],
+                            [llvm_double_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstdw : PowerPC_SPE_Intrinsic<"evstdw",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstdh : PowerPC_SPE_Intrinsic<"evstdh",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwwe : PowerPC_SPE_Intrinsic<"evstwwe",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwwo : PowerPC_SPE_Intrinsic<"evstwwo",
+                            [],
+                            [llvm_v2i32_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwhe : PowerPC_SPE_Intrinsic<"evstwhe",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+def int_ppc_spe_evstwho : PowerPC_SPE_Intrinsic<"evstwho",
+                            [],
+                            [llvm_v4i16_ty, llvm_i32_ty, llvm_i32_ty],
+                            [IntrWriteMem]>;
+
+// Vector accumulate instructions
+def int_ppc_spe_evmhegsmfaa     : PowerPC_SPE_Vec_I2<"evmhegsmfaa">;
+def int_ppc_spe_evmhegsmfan     : PowerPC_SPE_Vec_I2<"evmhegsmfan">;
+def int_ppc_spe_evmhegsmiaa     : PowerPC_SPE_Vec_I2<"evmhegsmiaa">;
+def int_ppc_spe_evmhegsmian     : PowerPC_SPE_Vec_I2<"evmhegsmian">;
+def int_ppc_spe_evmhegumiaa     : PowerPC_SPE_Vec_I2<"evmhegumiaa">;
+def int_ppc_spe_evmhegumian     : PowerPC_SPE_Vec_I2<"evmhegumian">;
+def int_ppc_spe_evmhesmf        : PowerPC_SPE_Vec_I2<"evmhesmf">;
+def int_ppc_spe_evmhesmfa       : PowerPC_SPE_Vec_I2<"evmhesmfa">;
+def int_ppc_spe_evmhesmfaaw     : PowerPC_SPE_Vec_I2<"evmhesmfaaw">;
+def int_ppc_spe_evmhesmfanw     : PowerPC_SPE_Vec_I2<"evmhesmfanw">;
+def int_ppc_spe_evmhesmi        : PowerPC_SPE_Vec_I2<"evmhesmi">;
+def int_ppc_spe_evmhesmia       : PowerPC_SPE_Vec_I2<"evmhesmia">;
+def int_ppc_spe_evmhesmiaaw     : PowerPC_SPE_Vec_I2<"evmhesmiaaw">;
+def int_ppc_spe_evmhesmianw     : PowerPC_SPE_Vec_I2<"evmhesmianw">;
+def int_ppc_spe_evmhessf        : PowerPC_SPE_Vec_I2<"evmhessf">;
+def int_ppc_spe_evmhessfa       : PowerPC_SPE_Vec_I2<"evmhessfa">;
+def int_ppc_spe_evmhessfaaw     : PowerPC_SPE_Vec_I2<"evmhessfaaw">;
+def int_ppc_spe_evmhessfanw     : PowerPC_SPE_Vec_I2<"evmhessfanw">;
+def int_ppc_spe_evmhessiaaw     : PowerPC_SPE_Vec_I2<"evmhessiaaw">;
+def int_ppc_spe_evmhessianw     : PowerPC_SPE_Vec_I2<"evmhessianw">;
+def int_ppc_spe_evmheumi        : PowerPC_SPE_Vec_I2<"evmheumi">;
+def int_ppc_spe_evmheumia       : PowerPC_SPE_Vec_I2<"evmheumia">;
+def int_ppc_spe_evmheumiaaw     : PowerPC_SPE_Vec_I2<"evmheumiaaw">;
+def int_ppc_spe_evmheumianw     : PowerPC_SPE_Vec_I2<"evmheumianw">;
+def int_ppc_spe_evmheusiaaw     : PowerPC_SPE_Vec_I2<"evmheusiaaw">;
+def int_ppc_spe_evmheusianw     : PowerPC_SPE_Vec_I2<"evmheusianw">;
+def int_ppc_spe_evmhogsmfaa     : PowerPC_SPE_Vec_I2<"evmhogsmfaa">;
+def int_ppc_spe_evmhogsmfan     : PowerPC_SPE_Vec_I2<"evmhogsmfan">;
+def int_ppc_spe_evmhogsmiaa     : PowerPC_SPE_Vec_I2<"evmhogsmiaa">;
+def int_ppc_spe_evmhogsmian     : PowerPC_SPE_Vec_I2<"evmhogsmian">;
+def int_ppc_spe_evmhogumiaa     : PowerPC_SPE_Vec_I2<"evmhogumiaa">;
+def int_ppc_spe_evmhogumian     : PowerPC_SPE_Vec_I2<"evmhogumian">;
+def int_ppc_spe_evmhosmf        : PowerPC_SPE_Vec_I2<"evmhosmf">;
+def int_ppc_spe_evmhosmfa       : PowerPC_SPE_Vec_I2<"evmhosmfa">;
+def int_ppc_spe_evmhosmfaaw     : PowerPC_SPE_Vec_I2<"evmhosmfaaw">;
+def int_ppc_spe_evmhosmfanw     : PowerPC_SPE_Vec_I2<"evmhosmfanw">;
+def int_ppc_spe_evmhosmi        : PowerPC_SPE_Vec_I2<"evmhosmi">;
+def int_ppc_spe_evmhosmia       : PowerPC_SPE_Vec_I2<"evmhosmia">;
+def int_ppc_spe_evmhosmiaaw     : PowerPC_SPE_Vec_I2<"evmhosmiaaw">;
+def int_ppc_spe_evmhosmianw     : PowerPC_SPE_Vec_I2<"evmhosmianw">;
+def int_ppc_spe_evmhossf        : PowerPC_SPE_Vec_I2<"evmhossf">;
+def int_ppc_spe_evmhossfa       : PowerPC_SPE_Vec_I2<"evmhossfa">;
+def int_ppc_spe_evmhossfaaw     : PowerPC_SPE_Vec_I2<"evmhossfaaw">;
+def int_ppc_spe_evmhossfanw     : PowerPC_SPE_Vec_I2<"evmhossfanw">;
+def int_ppc_spe_evmhossiaaw     : PowerPC_SPE_Vec_I2<"evmhossiaaw">;
+def int_ppc_spe_evmhossianw     : PowerPC_SPE_Vec_I2<"evmhossianw">;
+def int_ppc_spe_evmhoumi        : PowerPC_SPE_Vec_I2<"evmhoumi">;
+def int_ppc_spe_evmhoumia       : PowerPC_SPE_Vec_I2<"evmhoumia">;
+def int_ppc_spe_evmhoumiaaw     : PowerPC_SPE_Vec_I2<"evmhoumiaaw">;
+def int_ppc_spe_evmhoumianw     : PowerPC_SPE_Vec_I2<"evmhoumianw">;
+def int_ppc_spe_evmhousiaaw     : PowerPC_SPE_Vec_I2<"evmhousiaaw">;
+def int_ppc_spe_evmhousianw     : PowerPC_SPE_Vec_I2<"evmhousianw">;
+def int_ppc_spe_evmra           : PowerPC_SPE_Vec_I1<"evmra">;
+def int_ppc_spe_evmwhsmf        : PowerPC_SPE_Vec_I2<"evmwhsmf">;
+def int_ppc_spe_evmwhsmfa       : PowerPC_SPE_Vec_I2<"evmwhsmfa">;
+def int_ppc_spe_evmwhsmi        : PowerPC_SPE_Vec_I2<"evmwhsmi">;
+def int_ppc_spe_evmwhsmia       : PowerPC_SPE_Vec_I2<"evmwhsmia">;
+def int_ppc_spe_evmwhssf        : PowerPC_SPE_Vec_I2<"evmwhssf">;
+def int_ppc_spe_evmwhssfa       : PowerPC_SPE_Vec_I2<"evmwhssfa">;
+def int_ppc_spe_evmwhumi        : PowerPC_SPE_Vec_I2<"evmwhumi">;
+def int_ppc_spe_evmwhumia       : PowerPC_SPE_Vec_I2<"evmwhumia">;
+def int_ppc_spe_evmwlsmiaaw     : PowerPC_SPE_Vec_I2<"evmwlsmiaaw">;
+def int_ppc_spe_evmwlsmianw     : PowerPC_SPE_Vec_I2<"evmwlsmianw">;
+def int_ppc_spe_evmwlssiaaw     : PowerPC_SPE_Vec_I2<"evmwlssiaaw">;
+def int_ppc_spe_evmwlssianw     : PowerPC_SPE_Vec_I2<"evmwlssianw">;
+def int_ppc_spe_evmwlumi        : PowerPC_SPE_Vec_I2<"evmwlumi">;
+def int_ppc_spe_evmwlumia       : PowerPC_SPE_Vec_I2<"evmwlumia">;
+def int_ppc_spe_evmwlumiaaw     : PowerPC_SPE_Vec_I2<"evmwlumiaaw">;
+def int_ppc_spe_evmwlumianw     : PowerPC_SPE_Vec_I2<"evmwlumianw">;
+def int_ppc_spe_evmwlusiaaw     : PowerPC_SPE_Vec_I2<"evmwlusiaaw">;
+def int_ppc_spe_evmwlusianw     : PowerPC_SPE_Vec_I2<"evmwlusianw">;
+def int_ppc_spe_evmwsmf         : PowerPC_SPE_Vec_I2<"evmwsmf">;
+def int_ppc_spe_evmwsmfa        : PowerPC_SPE_Vec_I2<"evmwsmfa">;
+def int_ppc_spe_evmwsmfaa       : PowerPC_SPE_Vec_I2<"evmwsmfaa">;
+def int_ppc_spe_evmwsmfan       : PowerPC_SPE_Vec_I2<"evmwsmfan">;
+def int_ppc_spe_evmwsmi         : PowerPC_SPE_Vec_I2<"evmwsmi">;
+def int_ppc_spe_evmwsmia        : PowerPC_SPE_Vec_I2<"evmwsmia">;
+def int_ppc_spe_evmwsmiaa       : PowerPC_SPE_Vec_I2<"evmwsmiaa">;
+def int_ppc_spe_evmwsmian       : PowerPC_SPE_Vec_I2<"evmwsmian">;
+def int_ppc_spe_evmwssf         : PowerPC_SPE_Vec_I2<"evmwssf">;
+def int_ppc_spe_evmwssfa        : PowerPC_SPE_Vec_I2<"evmwssfa">;
+def int_ppc_spe_evmwssfaa       : PowerPC_SPE_Vec_I2<"evmwssfaa">;
+def int_ppc_spe_evmwssfan       : PowerPC_SPE_Vec_I2<"evmwssfan">;
+def int_ppc_spe_evmwumi         : PowerPC_SPE_Vec_I2<"evmwumi">;
+def int_ppc_spe_evmwumia        : PowerPC_SPE_Vec_I2<"evmwumia">;
+def int_ppc_spe_evmwumiaa       : PowerPC_SPE_Vec_I2<"evmwumiaa">;
+def int_ppc_spe_evmwumian       : PowerPC_SPE_Vec_I2<"evmwumian">;
+}

lib/Target/PowerPC/AsmParser/PPCAsmParser.cpp

static const MCPhysReg FRegs[32] = {
   PPC::F4,  PPC::F5,  PPC::F6,  PPC::F7,
   PPC::F8,  PPC::F9,  PPC::F10, PPC::F11,
   PPC::F12, PPC::F13, PPC::F14, PPC::F15,
   PPC::F16, PPC::F17, PPC::F18, PPC::F19,
   PPC::F20, PPC::F21, PPC::F22, PPC::F23,
   PPC::F24, PPC::F25, PPC::F26, PPC::F27,
   PPC::F28, PPC::F29, PPC::F30, PPC::F31
 };
+static const MCPhysReg SPERegs[32] = {
+  PPC::S0,  PPC::S1,  PPC::S2,  PPC::S3,
+  PPC::S4,  PPC::S5,  PPC::S6,  PPC::S7,
+  PPC::S8,  PPC::S9,  PPC::S10, PPC::S11,
+  PPC::S12, PPC::S13, PPC::S14, PPC::S15,
+  PPC::S16, PPC::S17, PPC::S18, PPC::S19,
+  PPC::S20, PPC::S21, PPC::S22, PPC::S23,
+  PPC::S24, PPC::S25, PPC::S26, PPC::S27,
+  PPC::S28, PPC::S29, PPC::S30, PPC::S31
+};
 static const MCPhysReg VFRegs[32] = {
   PPC::VF0,  PPC::VF1,  PPC::VF2,  PPC::VF3,
   PPC::VF4,  PPC::VF5,  PPC::VF6,  PPC::VF7,
   PPC::VF8,  PPC::VF9,  PPC::VF10, PPC::VF11,
   PPC::VF12, PPC::VF13, PPC::VF14, PPC::VF15,
   PPC::VF16, PPC::VF17, PPC::VF18, PPC::VF19,
   PPC::VF20, PPC::VF21, PPC::VF22, PPC::VF23,
   PPC::VF24, PPC::VF25, PPC::VF26, PPC::VF27,
void addRegQSRCOperands(MCInst &Inst, unsigned N) const {
     Inst.addOperand(MCOperand::createReg(QFRegs[getReg()]));
   }
 
   void addRegQBRCOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createReg(QFRegs[getReg()]));
   }
 
+  void addRegSPE4RCOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    Inst.addOperand(MCOperand::createReg(RRegs[getReg()]));
+  }
+
+  void addRegSPERCOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    Inst.addOperand(MCOperand::createReg(SPERegs[getReg()]));
+  }
+
   void addRegCRBITRCOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createReg(CRBITRegs[getCRBit()]));
   }
 
   void addRegCRRCOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createReg(CRRegs[getCCReg()]));

lib/Target/PowerPC/Disassembler/PPCDisassembler.cpp

static const unsigned QFRegs[] = {
   PPC::QF8, PPC::QF9, PPC::QF10, PPC::QF11,
   PPC::QF12, PPC::QF13, PPC::QF14, PPC::QF15,
   PPC::QF16, PPC::QF17, PPC::QF18, PPC::QF19,
   PPC::QF20, PPC::QF21, PPC::QF22, PPC::QF23,
   PPC::QF24, PPC::QF25, PPC::QF26, PPC::QF27,
   PPC::QF28, PPC::QF29, PPC::QF30, PPC::QF31
 };
 
+static const unsigned SPERegs[] = {
+  PPC::S0, PPC::S1, PPC::S2, PPC::S3,
+  PPC::S4, PPC::S5, PPC::S6, PPC::S7,
+  PPC::S8, PPC::S9, PPC::S10, PPC::S11,
+  PPC::S12, PPC::S13, PPC::S14, PPC::S15,
+  PPC::S16, PPC::S17, PPC::S18, PPC::S19,
+  PPC::S20, PPC::S21, PPC::S22, PPC::S23,
+  PPC::S24, PPC::S25, PPC::S26, PPC::S27,
+  PPC::S28, PPC::S29, PPC::S30, PPC::S31
+};
+
 template <std::size_t N>
 static DecodeStatus decodeRegisterClass(MCInst &Inst, uint64_t RegNo,
                                         const unsigned (&Regs)[N]) {
   assert(RegNo < N && "Invalid register number");
   Inst.addOperand(MCOperand::createReg(Regs[RegNo]));
   return MCDisassembler::Success;
 }
 

 #define DecodePointerLikeRegClass1 DecodeGPRC_NOR0RegisterClass
 
 static DecodeStatus DecodeQFRCRegisterClass(MCInst &Inst, uint64_t RegNo,
                                             uint64_t Address,
                                             const void *Decoder) {
   return decodeRegisterClass(Inst, RegNo, QFRegs);
 }
 
+static DecodeStatus DecodeSPE4RCRegisterClass(MCInst &Inst, uint64_t RegNo,
+                                            uint64_t Address,
+                                            const void *Decoder) {
+  return decodeRegisterClass(Inst, RegNo, GPRegs);
+}
+
+static DecodeStatus DecodeSPERCRegisterClass(MCInst &Inst, uint64_t RegNo,
+                                            uint64_t Address,
+                                            const void *Decoder) {
+  return decodeRegisterClass(Inst, RegNo, SPERegs);
+}
+
 #define DecodeQSRCRegisterClass DecodeQFRCRegisterClass
 #define DecodeQBRCRegisterClass DecodeQFRCRegisterClass
 
 template<unsigned N>
 static DecodeStatus decodeUImmOperand(MCInst &Inst, uint64_t Imm,
                                       int64_t Address, const void *Decoder) {
   assert(isUInt<N>(Imm) && "Invalid immediate");
   Inst.addOperand(MCOperand::createImm(Imm));

lib/Target/PowerPC/PPC.td

 def Directive604 : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_603", "">;
 def Directive620 : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_603", "">;
 def Directive7400: SubtargetFeature<"", "DarwinDirective", "PPC::DIR_7400", "">;
 def Directive750 : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_750", "">;
 def Directive970 : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_970", "">;
 def Directive32  : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_32", "">;
 def Directive64  : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_64", "">;
 def DirectiveA2  : SubtargetFeature<"", "DarwinDirective", "PPC::DIR_A2", "">;
+def DirectiveE500   : SubtargetFeature<"", "DarwinDirective",
+                                       "PPC::DIR_E500", "">;
 def DirectiveE500mc : SubtargetFeature<"", "DarwinDirective",
                                        "PPC::DIR_E500mc", "">;
 def DirectiveE5500  : SubtargetFeature<"", "DarwinDirective",
                                        "PPC::DIR_E5500", "">;
 def DirectivePwr3: SubtargetFeature<"", "DarwinDirective", "PPC::DIR_PWR3", "">;
 def DirectivePwr4: SubtargetFeature<"", "DarwinDirective", "PPC::DIR_PWR4", "">;
 def DirectivePwr5: SubtargetFeature<"", "DarwinDirective", "PPC::DIR_PWR5", "">;
 def DirectivePwr5x
def : ProcessorModel<"970", G5Model,
                    Feature64Bit /*, Feature64BitRegs */,
                    FeatureMFTB]>;
 def : ProcessorModel<"g5", G5Model,
                   [Directive970, FeatureAltivec,
                    FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX,
                    FeatureFRES, FeatureFRSQRTE,
                    Feature64Bit /*, Feature64BitRegs */,
                    FeatureMFTB, DeprecatedDST]>;
+def : ProcessorModel<"e500", PPCE500Model,
+                  [DirectiveE500,
+                   FeatureSPE, FeatureICBT, FeatureBookE, 
+                   FeatureISEL, FeatureMFTB]>;
 def : ProcessorModel<"e500mc", PPCE500mcModel,
                   [DirectiveE500mc,
                    FeatureSTFIWX, FeatureICBT, FeatureBookE, 
                    FeatureISEL, FeatureMFTB]>;
 def : ProcessorModel<"e5500", PPCE5500Model,
                   [DirectiveE5500, FeatureMFOCRF, Feature64Bit,
                    FeatureSTFIWX, FeatureICBT, FeatureBookE, 
                    FeatureISEL, FeatureMFTB]>;

lib/Target/PowerPC/PPCAsmPrinter.cpp

static const char *const CPUDirectives[] = {
     "ppc440",
     "ppc601",
     "ppc602",
     "ppc603",
     "ppc7400",
     "ppc750",
     "ppc970",
     "ppcA2",
+    "ppce500",
     "ppce500mc",
     "ppce5500",
     "power3",
     "power4",
     "power5",
     "power5x",
     "power6",
     "power6x",

lib/Target/PowerPC/PPCCallingConv.td

def RetCC_PPC : CallingConv<[
   CCIfType<[i1], CCIfNotSubtarget<"isPPC64()", CCPromoteToType<i32>>>,
 
   CCIfType<[i32], CCAssignToReg<[R3, R4, R5, R6, R7, R8, R9, R10]>>,
   CCIfType<[i64], CCAssignToReg<[X3, X4, X5, X6]>>,
   CCIfType<[i128], CCAssignToReg<[X3, X4, X5, X6]>>,
 
   // Floating point types returned as "direct" go into F1 .. F8; note that
   // only the ELFv2 ABI fully utilizes all these registers.
-  CCIfType<[f32], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
-  CCIfType<[f64], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
+  CCIfNotSubtarget<"hasSPE()",
+       CCIfType<[f32], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>>,
+  CCIfNotSubtarget<"hasSPE()",
+       CCIfType<[f64], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>>,
+  CCIfSubtarget<"hasSPE()",
+       CCIfType<[f32], CCAssignToReg<[R3, R4, R5, R6, R7, R8, R9, R10]>>>,
+  CCIfSubtarget<"hasSPE()",
+       CCIfType<[f64], CCAssignToReg<[S3, S4, S5, S6, S7, S8, S9, S10]>>>,
+  CCIfSubtarget<"hasSPE()",
+       CCIfType<[v2i32,v2f32],
+                CCAssignToReg<[S3, S4, S5, S6, S7, S8, S9, S10]>>>,
 
   // QPX vectors are returned in QF1 and QF2. 
   CCIfType<[v4f64, v4f32, v4i1],
            CCIfSubtarget<"hasQPX()", CCAssignToReg<[QF1, QF2]>>>,
  
   // Vector types returned as "direct" go into V2 .. V9; note that only the
   // ELFv2 ABI fully utilizes all these registers.
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v1i128, v4f32, v2f64],
def CC_PPC32_SVR4_Common : CallingConv<[
   // The first 8 integer arguments are passed in integer registers.
   CCIfType<[i32], CCAssignToReg<[R3, R4, R5, R6, R7, R8, R9, R10]>>,
 
   // Make sure the i64 words from a long double are either both passed in
   // registers or both passed on the stack.
   CCIfType<[f64], CCIfSplit<CCCustom<"CC_PPC32_SVR4_Custom_AlignFPArgRegs">>>,
   
   // FP values are passed in F1 - F8.
-  CCIfType<[f32, f64], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
+  CCIfType<[f32, f64],
+           CCIfNotSubtarget<"hasSPE()",
+                            CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>>,
+  CCIfType<[f64],
+           CCIfSubtarget<"hasSPE()",
+                         CCAssignToReg<[S3, S4, S5, S6, S7, S8, S9, S10]>>>,
+  CCIfType<[f32],
+           CCIfSubtarget<"hasSPE()",
+                         CCAssignToReg<[R3, R4, R5, R6, R7, R8, R9, R10]>>>,
+  CCIfType<[v2i32,v2f32],
+           CCIfSubtarget<"hasSPE()",
+                         CCAssignToReg<[S3, S4, S5, S6, S7, S8, S9, S10]>>>,
 
   // Split arguments have an alignment of 8 bytes on the stack.
   CCIfType<[i32], CCIfSplit<CCAssignToStack<4, 8>>>,
   
   CCIfType<[i32], CCAssignToStack<4, 4>>,
   
   // Floats are stored in double precision format, thus they have the same
   // alignment and size as doubles.
-  CCIfType<[f32,f64], CCAssignToStack<8, 8>>,  
+  CCIfType<[f32,f64], CCIfNotSubtarget<"hasSPE()", CCAssignToStack<8, 8>>>,  
+  CCIfType<[f32], CCIfSubtarget<"hasSPE()", CCAssignToStack<4, 4>>>,  
+  CCIfType<[f64, v2i32, v2i64],
+           CCIfSubtarget<"hasSPE()", CCAssignToStack<8, 8>>>,  
 
   // QPX vectors that are stored in double precision need 32-byte alignment.
   CCIfType<[v4f64, v4i1], CCAssignToStack<32, 32>>,
 
   // Vectors get 16-byte stack slots that are 16-byte aligned.
   CCIfType<[v16i8, v8i16, v4i32, v4f32, v2f64, v2i64], CCAssignToStack<16, 16>>
 ]>;
 
def CSR_Darwin32 : CalleeSavedRegs<(add R13, R14, R15, R16, R17, R18, R19, R20,
                                         F19, F20, F21, F22, F23, F24, F25, F26,
                                         F27, F28, F29, F30, F31, CR2, CR3, CR4
                                    )>;
 
 def CSR_Darwin32_Altivec : CalleeSavedRegs<(add CSR_Darwin32, CSR_Altivec)>;
 
 def CSR_SVR432   : CalleeSavedRegs<(add R14, R15, R16, R17, R18, R19, R20,
                                         R21, R22, R23, R24, R25, R26, R27, R28,
                                         R29, R30, R31, F14, F15, F16, F17, F18,
                                         F19, F20, F21, F22, F23, F24, F25, F26,
                                         F27, F28, F29, F30, F31, CR2, CR3, CR4
                                    )>;
 
 def CSR_SVR432_Altivec : CalleeSavedRegs<(add CSR_SVR432, CSR_Altivec)>;
 
 def CSR_Darwin64 : CalleeSavedRegs<(add X13, X14, X15, X16, X17, X18, X19, X20,
                                         X21, X22, X23, X24, X25, X26, X27, X28,

lib/Target/PowerPC/PPCFastISel.cpp

private:
     bool isValueAvailable(const Value *V) const;
     bool isVSFRCRegClass(const TargetRegisterClass *RC) const {
       return RC->getID() == PPC::VSFRCRegClassID;
     }
     bool isVSSRCRegClass(const TargetRegisterClass *RC) const {
       return RC->getID() == PPC::VSSRCRegClassID;
     }
     bool PPCEmitCmp(const Value *Src1Value, const Value *Src2Value,
-                    bool isZExt, unsigned DestReg);
+                    bool isZExt, unsigned DestReg,
+                    const PPC::Predicate Pred);
     bool PPCEmitLoad(MVT VT, unsigned &ResultReg, Address &Addr,
                      const TargetRegisterClass *RC, bool IsZExt = true,
                      unsigned FP64LoadOpc = PPC::LFD);
     bool PPCEmitStore(MVT VT, unsigned SrcReg, Address &Addr);
     bool PPCComputeAddress(const Value *Obj, Address &Addr);
     void PPCSimplifyAddress(Address &Addr, bool &UseOffset,
                             unsigned &IndexReg);
     bool PPCEmitIntExt(MVT SrcVT, unsigned SrcReg, MVT DestVT,
                            unsigned DestReg, bool IsZExt);
     unsigned PPCMaterializeFP(const ConstantFP *CFP, MVT VT);
     unsigned PPCMaterializeGV(const GlobalValue *GV, MVT VT);

 // Emit a load instruction if possible, returning true if we succeeded,
 // otherwise false.  See commentary below for how the register class of
 // the load is determined.
 bool PPCFastISel::PPCEmitLoad(MVT VT, unsigned &ResultReg, Address &Addr,
                               const TargetRegisterClass *RC,
                               bool IsZExt, unsigned FP64LoadOpc) {
   unsigned Opc;
   bool UseOffset = true;
+  bool UseSPE = PPCSubTarget->hasSPE();
 
   // If ResultReg is given, it determines the register class of the load.
   // Otherwise, RC is the register class to use.  If the result of the
   // load isn't anticipated in this block, both may be zero, in which
   // case we must make a conservative guess.  In particular, don't assign
   // R0 or X0 to the result register, as the result may be used in a load,
   // store, add-immediate, or isel that won't permit this.  (Though
   // perhaps the spill and reload of live-exit values would handle this?)
   const TargetRegisterClass *UseRC =
     (ResultReg ? MRI.getRegClass(ResultReg) :
      (RC ? RC :
-      (VT == MVT::f64 ? &PPC::F8RCRegClass :
-       (VT == MVT::f32 ? &PPC::F4RCRegClass :
+      (VT == MVT::f64 ? (UseSPE ? &PPC::SPERCRegClass : &PPC::F8RCRegClass) :
+       (VT == MVT::f32 ? (UseSPE ? &PPC::SPE4RCRegClass : &PPC::F4RCRegClass) :
         (VT == MVT::i64 ? &PPC::G8RC_and_G8RC_NOX0RegClass :
          &PPC::GPRC_and_GPRC_NOR0RegClass)))));
 
   bool Is32BitInt = UseRC->hasSuperClassEq(&PPC::GPRCRegClass);
 
   switch (VT.SimpleTy) {
     default: // e.g., vector types not handled
       return false;
     case MVT::i8:
       Opc = Is32BitInt ? PPC::LBZ : PPC::LBZ8;
case MVT::i32:
       break;
     case MVT::i64:
       Opc = PPC::LD;
       assert(UseRC->hasSuperClassEq(&PPC::G8RCRegClass) &&
              "64-bit load with 32-bit target??");
       UseOffset = ((Addr.Offset & 3) == 0);
       break;
     case MVT::f32:
-      Opc = PPC::LFS;
+      Opc = PPCSubTarget->hasSPE() ? PPC::SPELWZ : PPC::LFS;
       break;
     case MVT::f64:
       Opc = FP64LoadOpc;
       break;
   }
 
   // If necessary, materialize the offset into a register and use
   // the indexed form.  Also handle stack pointers with special needs.
if (Addr.BaseType == Address::FrameIndexBase) {
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
       .addImm(Addr.Offset).addFrameIndex(Addr.Base.FI).addMemOperand(MMO);
 
   // Base reg with offset in range.
   } else if (UseOffset) {
     // VSX only provides an indexed load.
     if (Is32VSXLoad || Is64VSXLoad) return false;
 
+    printf("Offset: %d, reg: %d\n", (int)Addr.Offset, Addr.Base.Reg);
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
       .addImm(Addr.Offset).addReg(Addr.Base.Reg);
 
   // Indexed form.
   } else {
     // Get the RR opcode corresponding to the RI one.  FIXME: It would be
     // preferable to use the ImmToIdxMap from PPCRegisterInfo.cpp, but it
     // is hard to get at.
     switch (Opc) {
       default:        llvm_unreachable("Unexpected opcode!");
       case PPC::LBZ:    Opc = PPC::LBZX;    break;
       case PPC::LBZ8:   Opc = PPC::LBZX8;   break;
       case PPC::LHZ:    Opc = PPC::LHZX;    break;
       case PPC::LHZ8:   Opc = PPC::LHZX8;   break;
       case PPC::LHA:    Opc = PPC::LHAX;    break;
       case PPC::LHA8:   Opc = PPC::LHAX8;   break;
       case PPC::LWZ:    Opc = PPC::LWZX;    break;
       case PPC::LWZ8:   Opc = PPC::LWZX8;   break;
       case PPC::LWA:    Opc = PPC::LWAX;    break;
       case PPC::LWA_32: Opc = PPC::LWAX_32; break;
       case PPC::LD:     Opc = PPC::LDX;     break;
       case PPC::LFS:    Opc = IsVSSRC ? PPC::LXSSPX : PPC::LFSX; break;
       case PPC::LFD:    Opc = IsVSFRC ? PPC::LXSDX : PPC::LFDX; break;
+      case PPC::EVLDD:  Opc = PPC::EVLDDX;  break;
+      case PPC::SPELWZ: Opc = PPC::SPELWZX;    break;
     }
 
     auto MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc),
                        ResultReg);
 
     // If we have an index register defined we use it in the store inst,
     // otherwise we use X0 as base as it makes the vector instructions to
     // use zero in the computation of the effective address regardless the
bool PPCFastISel::SelectLoad(const Instruction *I) {
   // Look at the currently assigned register for this instruction
   // to determine the required register class.  This is necessary
   // to constrain RA from using R0/X0 when this is not legal.
   unsigned AssignedReg = FuncInfo.ValueMap[I];
   const TargetRegisterClass *RC =
     AssignedReg ? MRI.getRegClass(AssignedReg) : nullptr;
 
   unsigned ResultReg = 0;
-  if (!PPCEmitLoad(VT, ResultReg, Addr, RC))
+  if (!PPCEmitLoad(VT, ResultReg, Addr, RC, true,
+      PPCSubTarget->hasSPE() ? PPC::EVLDD : PPC::LFD))
     return false;
   updateValueMap(I, ResultReg);
   return true;
 }
 
 // Emit a store instruction to store SrcReg at Addr.
 bool PPCFastISel::PPCEmitStore(MVT VT, unsigned SrcReg, Address &Addr) {
   assert(SrcReg && "Nothing to store!");
case MVT::i32:
       assert(Is32BitInt && "Not GPRC for i32??");
       Opc = PPC::STW;
       break;
     case MVT::i64:
       Opc = PPC::STD;
       UseOffset = ((Addr.Offset & 3) == 0);
       break;
     case MVT::f32:
+      if (PPCSubTarget->hasSPE())
+        Opc = PPC::SPESTW;
+      else
         Opc = PPC::STFS;
       break;
     case MVT::f64:
+      if (PPCSubTarget->hasSPE())
+        Opc = PPC::EVSTDD;
+      else
         Opc = PPC::STFD;
       break;
   }
 
   // If necessary, materialize the offset into a register and use
   // the indexed form.  Also handle stack pointers with special needs.
   unsigned IndexReg = 0;
   PPCSimplifyAddress(Addr, UseOffset, IndexReg);
 
switch (Opc) {
       case PPC::STH : Opc = PPC::STHX;  break;
       case PPC::STW : Opc = PPC::STWX;  break;
       case PPC::STB8: Opc = PPC::STBX8; break;
       case PPC::STH8: Opc = PPC::STHX8; break;
       case PPC::STW8: Opc = PPC::STWX8; break;
       case PPC::STD:  Opc = PPC::STDX;  break;
       case PPC::STFS: Opc = IsVSSRC ? PPC::STXSSPX : PPC::STFSX; break;
       case PPC::STFD: Opc = IsVSFRC ? PPC::STXSDX : PPC::STFDX; break;
+      case PPC::EVSTDD: Opc = PPC::EVSTDDX;
+      case PPC::SPESTW: Opc = PPC::SPESTWX;  break;
     }
 
     auto MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc))
         .addReg(SrcReg);
 
     // If we have an index register defined we use it in the store inst,
     // otherwise we use X0 as base as it makes the vector instructions to
     // use zero in the computation of the effective address regardless the
if (isValueAvailable(CI)) {
       if (FuncInfo.MBB->isLayoutSuccessor(TBB)) {
         std::swap(TBB, FBB);
         PPCPred = PPC::InvertPredicate(PPCPred);
       }
 
       unsigned CondReg = createResultReg(&PPC::CRRCRegClass);
 
       if (!PPCEmitCmp(CI->getOperand(0), CI->getOperand(1), CI->isUnsigned(),
-                      CondReg))
+                      CondReg, PPCPred))
         return false;
 
       BuildMI(*BrBB, FuncInfo.InsertPt, DbgLoc, TII.get(PPC::BCC))
         .addImm(PPCPred).addReg(CondReg).addMBB(TBB);
       finishCondBranch(BI->getParent(), TBB, FBB);
       return true;
     }
   } else if (const ConstantInt *CI =
              dyn_cast<ConstantInt>(BI->getCondition())) {
     uint64_t Imm = CI->getZExtValue();
     MachineBasicBlock *Target = (Imm == 0) ? FBB : TBB;
     fastEmitBranch(Target, DbgLoc);
     return true;
   }
 
   // FIXME: ARM looks for a case where the block containing the compare
   // has been split from the block containing the branch.  If this happens,
   // there is a vreg available containing the result of the compare.  I'm
   // not sure we can do much, as we've lost the predicate information with
   // the compare instruction -- we have a 4-bit CR but don't know which bit
   // to test here.
   return false;
 }
 
 // Attempt to emit a compare of the two source values.  Signed and unsigned
 // comparisons are supported.  Return false if we can't handle it.
 bool PPCFastISel::PPCEmitCmp(const Value *SrcValue1, const Value *SrcValue2,
-                             bool IsZExt, unsigned DestReg) {
+                             bool IsZExt, unsigned DestReg,
+                             const PPC::Predicate Pred) {
   Type *Ty = SrcValue1->getType();
   EVT SrcEVT = TLI.getValueType(DL, Ty, true);
   if (!SrcEVT.isSimple())
     return false;
   MVT SrcVT = SrcEVT.getSimpleVT();
 
   if (SrcVT == MVT::i1 && PPCSubTarget->useCRBits())
     return false;
 
   // See if operand 2 is an immediate encodeable in the compare.
   // FIXME: Operands are not in canonical order at -O0, so an immediate
   // operand in position 1 is a lost opportunity for now.  We are
   // similar to ARM in this regard.
   long Imm = 0;
   bool UseImm = false;
+  const bool UseSPE = PPCSubTarget->hasSPE();
 
   // Only 16-bit integer constants can be represented in compares for
   // PowerPC.  Others will be materialized into a register.
   if (const ConstantInt *ConstInt = dyn_cast<ConstantInt>(SrcValue2)) {
     if (SrcVT == MVT::i64 || SrcVT == MVT::i32 || SrcVT == MVT::i16 ||
         SrcVT == MVT::i8 || SrcVT == MVT::i1) {
       const APInt &CIVal = ConstInt->getValue();
       Imm = (IsZExt) ? (long)CIVal.getZExtValue() : (long)CIVal.getSExtValue();
       if ((IsZExt && isUInt<16>(Imm)) || (!IsZExt && isInt<16>(Imm)))
         UseImm = true;
     }
   }
 
   unsigned CmpOpc;
   bool NeedsExt = false;
   switch (SrcVT.SimpleTy) {
     default: return false;
     case MVT::f32:
+      if (UseSPE) {
+        switch (Pred) {
+          default: return false;
+          case PPC::PRED_EQ:
+            CmpOpc = PPC::EFSCMPEQ;
+            break;
+          case PPC::PRED_LT:
+            CmpOpc = PPC::EFSCMPLT;
+            break;
+          case PPC::PRED_GT:
+            CmpOpc = PPC::EFSCMPGT;
+            break;
+        }
+      } else
         CmpOpc = PPC::FCMPUS;
       break;
     case MVT::f64:
+      if (UseSPE) {
+        switch (Pred) {
+          default: return false;
+          case PPC::PRED_EQ:
+            CmpOpc = PPC::EFDCMPEQ;
+            break;
+          case PPC::PRED_LT:
+            CmpOpc = PPC::EFDCMPLT;
+            break;
+          case PPC::PRED_GT:
+            CmpOpc = PPC::EFDCMPGT;
+            break;
+        }
+      } else
         CmpOpc = PPC::FCMPUD;
       break;
     case MVT::i1:
     case MVT::i8:
     case MVT::i16:
       NeedsExt = true;
       // Intentional fall-through.
     case MVT::i32:
       if (!UseImm)
bool PPCFastISel::SelectFPTrunc(const Instruction *I) {
   if (SrcVT != MVT::f64 || DestVT != MVT::f32)
     return false;
 
   unsigned SrcReg = getRegForValue(Src);
   if (!SrcReg)
     return false;
 
   // Round the result to single precision.
-  unsigned DestReg = createResultReg(&PPC::F4RCRegClass);
+  unsigned DestReg;
+  
+  if (PPCSubTarget->hasSPE()) {
+    DestReg = createResultReg(&PPC::SPE4RCRegClass);
+    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(PPC::EFSCFD), DestReg)
+      .addReg(SrcReg);
+  } else {
+    DestReg = createResultReg(&PPC::F4RCRegClass);
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(PPC::FRSP), DestReg)
       .addReg(SrcReg);
+  }
 
   updateValueMap(I, DestReg);
   return true;
 }
 
 // Move an i32 or i64 value in a GPR to an f64 value in an FPR.
 // FIXME: When direct register moves are implemented (see PowerISA 2.07),
 // those should be used instead of moving via a stack slot when the
bool PPCFastISel::SelectIToFP(const Instruction *I, bool IsSigned) {
   if (SrcVT != MVT::i8  && SrcVT != MVT::i16 &&
       SrcVT != MVT::i32 && SrcVT != MVT::i64)
     return false;
 
   unsigned SrcReg = getRegForValue(Src);
   if (SrcReg == 0)
     return false;
 
+  // Shortcut for SPE.  Doesn't need to store/load, since it's all in the GPRs
+  if (PPCSubTarget->hasSPE()) {
+    unsigned Opc;
+    if (DstVT == MVT::f32)
+      Opc = IsSigned ? PPC::EFSCFSI : PPC::EFSCFUI;
+    else
+      Opc = IsSigned ? PPC::EFDCFSI : PPC::EFDCFUI;
+
+    unsigned DestReg = createResultReg(&PPC::SPERCRegClass);
+    // Generate the convert.
+    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), DestReg)
+      .addReg(SrcReg);
+    return true;
+  }
+
   // We can only lower an unsigned convert if we have the newer
   // floating-point conversion operations.
   if (!IsSigned && !PPCSubTarget->hasFPCVT())
     return false;
 
   // FIXME: For now we require the newer floating-point conversion operations
   // (which are present only on P7 and A2 server models) when converting
   // to single-precision float.  Otherwise we have to generate a lot of
bool PPCFastISel::SelectFPToI(const Instruction *I, bool IsSigned) {
   MVT DstVT, SrcVT;
   Type *DstTy = I->getType();
   if (!isTypeLegal(DstTy, DstVT))
     return false;
 
   if (DstVT != MVT::i32 && DstVT != MVT::i64)
     return false;
 
-  // If we don't have FCTIDUZ and we need it, punt to SelectionDAG.
-  if (DstVT == MVT::i64 && !IsSigned && !PPCSubTarget->hasFPCVT())
+  // If we don't have FCTIDUZ, or SPE, and we need it, punt to SelectionDAG.
+  if (DstVT == MVT::i64 && !IsSigned &&
+      !PPCSubTarget->hasFPCVT() && !PPCSubTarget->hasSPE())
     return false;
 
   Value *Src = I->getOperand(0);
   Type *SrcTy = Src->getType();
   if (!isTypeLegal(SrcTy, SrcVT))
     return false;
 
   if (SrcVT != MVT::f32 && SrcVT != MVT::f64)
if (InRC == &PPC::F4RCRegClass) {
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
             TII.get(TargetOpcode::COPY), TmpReg)
       .addReg(SrcReg);
     SrcReg = TmpReg;
   }
 
   // Determine the opcode for the conversion, which takes place
   // entirely within FPRs.
-  unsigned DestReg = createResultReg(&PPC::F8RCRegClass);
+  unsigned DestReg;
   unsigned Opc;
 
+  if (PPCSubTarget->hasSPE()) {
+    if (DstVT == MVT::i32) {
+      DestReg = createResultReg(&PPC::SPE4RCRegClass);
+      if (IsSigned)
+        Opc = InRC == &PPC::SPE4RCRegClass ? PPC::EFSCTSI : PPC::EFDCTSI;
+      else
+        Opc = InRC == &PPC::SPE4RCRegClass ? PPC::EFSCTUI : PPC::EFDCTUI;
+    } else {
+      // TODO: Convert single-precision to double if InReg SPE4RC
+      DestReg = createResultReg(&PPC::SPERCRegClass);
+      Opc = IsSigned ? PPC::EFDCTSI : PPC::EFDCTUI;
+    }
+  } else {
+    DestReg = createResultReg(&PPC::F8RCRegClass);
     if (DstVT == MVT::i32)
       if (IsSigned)
         Opc = PPC::FCTIWZ;
       else
         Opc = PPCSubTarget->hasFPCVT() ? PPC::FCTIWUZ : PPC::FCTIDZ;
     else
       Opc = IsSigned ? PPC::FCTIDZ : PPC::FCTIDUZ;
+  }
 
   // Generate the convert.
   BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), DestReg)
     .addReg(SrcReg);
 
   // Now move the integer value from a float register to an integer register.
-  unsigned IntReg = PPCMoveToIntReg(I, DstVT, DestReg, IsSigned);
+  unsigned IntReg;
+  
+  if (PPCSubTarget->hasSPE())
+    IntReg = DestReg;
+  else
+    IntReg = PPCMoveToIntReg(I, DstVT, DestReg, IsSigned);
+
   if (IntReg == 0)
     return false;
 
   updateValueMap(I, IntReg);
   return true;
 }
 
 // Attempt to fast-select a binary integer operation that isn't already
unsigned PPCFastISel::PPCMaterializeFP(const ConstantFP *CFP, MVT VT) {
   // No plans to handle long double here.
   if (VT != MVT::f32 && VT != MVT::f64)
     return 0;
 
   // All FP constants are loaded from the constant pool.
   unsigned Align = DL.getPrefTypeAlignment(CFP->getType());
   assert(Align > 0 && "Unexpectedly missing alignment information!");
   unsigned Idx = MCP.getConstantPoolIndex(cast<Constant>(CFP), Align);
+  const bool hasSPE = PPCSubTarget->hasSPE();
   const TargetRegisterClass *RC =
-    (VT == MVT::f32) ? &PPC::F4RCRegClass : &PPC::F8RCRegClass;
+    hasSPE ?
+    ((VT == MVT::f32) ? &PPC::SPE4RCRegClass : &PPC::SPERCRegClass) :
+    ((VT == MVT::f32) ? &PPC::F4RCRegClass : &PPC::F8RCRegClass);
   unsigned DestReg = createResultReg(RC);
   CodeModel::Model CModel = TM.getCodeModel();
 
   MachineMemOperand *MMO = FuncInfo.MF->getMachineMemOperand(
       MachinePointerInfo::getConstantPool(*FuncInfo.MF),
       MachineMemOperand::MOLoad, (VT == MVT::f32) ? 4 : 8, Align);
 
-  unsigned Opc = (VT == MVT::f32) ? PPC::LFS : PPC::LFD;
+  unsigned Opc = (hasSPE) ? ((VT == MVT::f32) ? PPC::SPELWZ : PPC::EVLDD) :
+    ((VT == MVT::f32) ? PPC::LFS : PPC::LFD);
   unsigned TmpReg = createResultReg(&PPC::G8RC_and_G8RC_NOX0RegClass);
 
   PPCFuncInfo->setUsesTOCBasePtr();
   // For small code model, generate a LF[SD](0, LDtocCPT(Idx, X2)).
   if (CModel == CodeModel::Small) {
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(PPC::LDtocCPT),
             TmpReg)
       .addConstantPoolIndex(Idx).addReg(PPC::X2);
bool PPCFastISel::tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
 
   // See if we can handle this address.
   Address Addr;
   if (!PPCComputeAddress(LI->getOperand(0), Addr))
     return false;
 
   unsigned ResultReg = MI->getOperand(0).getReg();
 
-  if (!PPCEmitLoad(VT, ResultReg, Addr, nullptr, IsZExt))
+  if (!PPCEmitLoad(VT, ResultReg, Addr, nullptr, IsZExt,
+        PPCSubTarget->hasSPE() ? PPC::EVLDD : PPC::LFD))
     return false;
 
   MI->eraseFromParent();
   return true;
 }
 
 // Attempt to lower call arguments in a faster way than done by
 // the selection DAG code.

lib/Target/PowerPC/PPCFrameLowering.cpp

static const SpillSlot Offsets[] = {
       {PPC::V28, -64},
       {PPC::V27, -80},
       {PPC::V26, -96},
       {PPC::V25, -112},
       {PPC::V24, -128},
       {PPC::V23, -144},
       {PPC::V22, -160},
       {PPC::V21, -176},
-      {PPC::V20, -192}};
+      {PPC::V20, -192},
+  
+      // SPE register save area (overlaps Vector save area
+      {PPC::S31, -8},
+      {PPC::S30, -16},
+      {PPC::S29, -24},
+      {PPC::S28, -32},
+      {PPC::S27, -40},
+      {PPC::S26, -48},
+      {PPC::S25, -56},
+      {PPC::S24, -64},
+      {PPC::S23, -72},
+      {PPC::S22, -80},
+      {PPC::S21, -88},
+      {PPC::S20, -96},
+      {PPC::S19, -104},
+      {PPC::S18, -112},
+      {PPC::S17, -120},
+      {PPC::S16, -128},
+      {PPC::S15, -136},
+      {PPC::S14, -144}};
 
   static const SpillSlot Offsets64[] = {
       // Floating-point register save area offsets.
       {PPC::F31, -8},
       {PPC::F30, -16},
       {PPC::F29, -24},
       {PPC::F28, -32},
       {PPC::F27, -40},
void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF,
 
   SmallVector<CalleeSavedInfo, 18> GPRegs;
   SmallVector<CalleeSavedInfo, 18> G8Regs;
   SmallVector<CalleeSavedInfo, 18> FPRegs;
   SmallVector<CalleeSavedInfo, 18> VRegs;
 
   for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
     unsigned Reg = CSI[i].getReg();
-    if (PPC::GPRCRegClass.contains(Reg)) {
+    if (PPC::GPRCRegClass.contains(Reg) ||
+        PPC::SPE4RCRegClass.contains(Reg)) {
       HasGPSaveArea = true;
 
       GPRegs.push_back(CSI[i]);
 
       if (Reg < MinGPR) {
         MinGPR = Reg;
       }
     } else if (PPC::G8RCRegClass.contains(Reg)) {
if (PPC::GPRCRegClass.contains(Reg) ||
       if (Reg < MinFPR) {
         MinFPR = Reg;
       }
     } else if (PPC::CRBITRCRegClass.contains(Reg) ||
                PPC::CRRCRegClass.contains(Reg)) {
       ; // do nothing, as we already know whether CRs are spilled
     } else if (PPC::VRSAVERCRegClass.contains(Reg)) {
       HasVRSAVESaveArea = true;
-    } else if (PPC::VRRCRegClass.contains(Reg)) {
+    } else if (PPC::VRRCRegClass.contains(Reg) ||
+               PPC::SPERCRegClass.contains(Reg)) {
       HasVRSaveArea = true;
 
       VRegs.push_back(CSI[i]);
 
       if (Reg < MinVR) {
         MinVR = Reg;
       }
     } else {
       llvm_unreachable("Unknown RegisterClass!");
     }
   }
 
   PPCFunctionInfo *PFI = MF.getInfo<PPCFunctionInfo>();
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
 
         MFI.setObjectOffset(FI, LowerBound + MFI.getObjectOffset(FI));
       }
     }
 
     LowerBound -= 4; // The VRSAVE save area is always 4 bytes long.
   }
 
   if (HasVRSaveArea) {
     // Insert alignment padding, we need 16-byte alignment. Note: for postive
     // number the alignment formula is : y = (x + (n-1)) & (~(n-1)). But since
     // we are using negative number here (the stack grows downward). We should
     // use formula : y = x & (~(n-1)). Where x is the size before aligning, n
     // is the alignment size ( n = 16 here) and y is the size after aligning.
     assert(LowerBound <= 0 && "Expect LowerBound have a non-positive value!");
     LowerBound &= ~(15);
 

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

if (CC == ISD::SETEQ || CC == ISD::SETNE) {
       int16_t SImm;
       if (isIntS16Immediate(RHS, SImm))
         return SDValue(CurDAG->getMachineNode(PPC::CMPDI, dl, MVT::i64, LHS,
                                               getI64Imm(SImm & 0xFFFF, dl)),
                          0);
       Opc = PPC::CMPD;
     }
   } else if (LHS.getValueType() == MVT::f32) {
+    if (PPCSubTarget->hasSPE()) {
+      switch (CC) {
+        default:
+        case ISD::SETEQ:
+        case ISD::SETNE:
+          Opc = PPC::EFSCMPEQ;
+          break;
+        case ISD::SETLT:
+        case ISD::SETGE:
+        case ISD::SETOLT:
+        case ISD::SETOGE:
+        case ISD::SETULT:
+        case ISD::SETUGE:
+          Opc = PPC::EFSCMPLT;
+          break;
+        case ISD::SETGT:
+        case ISD::SETLE:
+        case ISD::SETOGT:
+        case ISD::SETOLE:
+        case ISD::SETUGT:
+        case ISD::SETULE:
+          Opc = PPC::EFSCMPGT;
+          break;
+      }
+    } else
       Opc = PPC::FCMPUS;
   } else {
     assert(LHS.getValueType() == MVT::f64 && "Unknown vt!");
+    if (PPCSubTarget->hasSPE()) {
+      switch (CC) {
+        default:
+        case ISD::SETEQ:
+        case ISD::SETNE:
+          Opc = PPC::EFDCMPEQ;
+          break;
+        case ISD::SETLT:
+        case ISD::SETGE:
+        case ISD::SETOLT:
+        case ISD::SETOGE:
+        case ISD::SETULT:
+        case ISD::SETUGE:
+          Opc = PPC::EFDCMPLT;
+          break;
+        case ISD::SETGT:
+        case ISD::SETLE:
+        case ISD::SETOGT:
+        case ISD::SETOLE:
+        case ISD::SETUGT:
+        case ISD::SETULE:
+          Opc = PPC::EFDCMPGT;
+          break;
+      }
+    } else
       Opc = PPCSubTarget->hasVSX() ? PPC::XSCMPUDP : PPC::FCMPUD;
   }
   return SDValue(CurDAG->getMachineNode(Opc, dl, MVT::i32, LHS, RHS), 0);
 }
 
 static PPC::Predicate getPredicateForSetCC(ISD::CondCode CC) {
   switch (CC) {
   case ISD::SETUEQ:
   case ISD::SETONE:
if (!PPCSubTarget->useCRBits() &&
     }
   }
 
   SDValue LHS = N->getOperand(0);
   SDValue RHS = N->getOperand(1);
 
   // Altivec Vector compare instructions do not set any CR register by default and
   // vector compare operations return the same type as the operands.
-  if (LHS.getValueType().isVector()) {
+  if (LHS.getValueType().isVector() && PPCSubTarget->hasAltivec()) {
     if (PPCSubTarget->hasQPX())
       return false;
 
     EVT VecVT = LHS.getValueType();
     bool Swap, Negate;
     unsigned int VCmpInst = getVCmpInst(VecVT.getSimpleVT(), CC,
                                         PPCSubTarget->hasVSX(), Swap, Negate);
     if (Swap)
       std::swap(LHS, RHS);
 
     EVT ResVT = VecVT.changeVectorElementTypeToInteger();
bool PPCDAGToDAGISel::trySETCC(SDNode *N) {
   if (PPCSubTarget->useCRBits())
     return false;
 
   bool Inv;
   unsigned Idx = getCRIdxForSetCC(CC, Inv);
   SDValue CCReg = SelectCC(LHS, RHS, CC, dl);
   SDValue IntCR;
 
+  // SPE e*cmp* instructions only set the 'gt' bit, so hard-code that
+  // The correct compare instruction is already set by SelectCC()
+  if (PPCSubTarget->hasSPE() && (LHS.getValueType().isFloatingPoint())) {
+    Idx = 1;
+  }
+
   // Force the ccreg into CR7.
   SDValue CR7Reg = CurDAG->getRegister(PPC::CR7, MVT::i32);
 
   SDValue InFlag(nullptr, 0);  // Null incoming flag value.
   CCReg = CurDAG->getCopyToReg(CurDAG->getEntryNode(), dl, CR7Reg, CCReg,
                                InFlag).getValue(1);
 
   IntCR = SDValue(CurDAG->getMachineNode(PPC::MFOCRF, dl, MVT::i32, CR7Reg,
case ISD::SELECT_CC: {
 
     unsigned BROpc = getPredicateForSetCC(CC);
 
     unsigned SelectCCOp;
     if (N->getValueType(0) == MVT::i32)
       SelectCCOp = PPC::SELECT_CC_I4;
     else if (N->getValueType(0) == MVT::i64)
       SelectCCOp = PPC::SELECT_CC_I8;
     else if (N->getValueType(0) == MVT::f32)
       if (PPCSubTarget->hasP8Vector())
         SelectCCOp = PPC::SELECT_CC_VSSRC;
+      else if (PPCSubTarget->hasSPE())
+        SelectCCOp = PPC::SELECT_CC_SPE4;
       else
         SelectCCOp = PPC::SELECT_CC_F4;
     else if (N->getValueType(0) == MVT::f64)
       if (PPCSubTarget->hasVSX())
         SelectCCOp = PPC::SELECT_CC_VSFRC;
+      else if (PPCSubTarget->hasSPE())
+        SelectCCOp = PPC::SELECT_CC_SPE;
       else
         SelectCCOp = PPC::SELECT_CC_F8;
     else if (PPCSubTarget->hasQPX() && N->getValueType(0) == MVT::v4f64)
       SelectCCOp = PPC::SELECT_CC_QFRC;
     else if (PPCSubTarget->hasQPX() && N->getValueType(0) == MVT::v4f32)
       SelectCCOp = PPC::SELECT_CC_QSRC;
     else if (PPCSubTarget->hasQPX() && N->getValueType(0) == MVT::v4i1)
       SelectCCOp = PPC::SELECT_CC_QBRC;
     else if (N->getValueType(0) == MVT::v2f64 ||
              N->getValueType(0) == MVT::v2i64)
       SelectCCOp = PPC::SELECT_CC_VSRC;
     else

lib/Target/PowerPC/PPCISelLowering.h

TargetLoweringBase::LegalizeTypeAction getPreferredVectorAction(EVT VT)
       const override {
       if (VT.getScalarSizeInBits() % 8 == 0)
         return TypeWidenVector;
       return TargetLoweringBase::getPreferredVectorAction(VT);
     }
 
     bool useSoftFloat() const override;
 
+    bool hasSPE() const;
+
     MVT getScalarShiftAmountTy(const DataLayout &, EVT) const override {
       return MVT::i32;
     }
 
     bool isCheapToSpeculateCttz() const override {
       return true;
     }
 
public:
     unsigned getJumpTableEncoding() const override;
     bool isJumpTableRelative() const override;
     SDValue getPICJumpTableRelocBase(SDValue Table,
                                      SelectionDAG &DAG) const override;
     const MCExpr *getPICJumpTableRelocBaseExpr(const MachineFunction *MF,
                                                unsigned JTI,
                                                MCContext &Ctx) const override;
 
+    unsigned getNumRegistersForCallingConv(LLVMContext &Context,
+                                           EVT VT) const override;
+
+    MVT getRegisterTypeForCallingConv(MVT VT) const override;
+    MVT getRegisterTypeForCallingConv(LLVMContext &Context,
+                                      EVT VT) const override;
+
   private:
     struct ReuseLoadInfo {
       SDValue Ptr;
       SDValue Chain;
       SDValue ResChain;
       MachinePointerInfo MPI;
       bool IsDereferenceable = false;
       bool IsInvariant = false;

lib/Target/PowerPC/PPCISelLowering.cpp

 cl::desc("disable sibling call optimization on ppc"), cl::Hidden);
 
 STATISTIC(NumTailCalls, "Number of tail calls");
 STATISTIC(NumSiblingCalls, "Number of sibling calls");
 
 // FIXME: Remove this once the bug has been fixed!
 extern cl::opt<bool> ANDIGlueBug;
 
 PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM,
                                      const PPCSubtarget &STI)
     : TargetLowering(TM), Subtarget(STI) {
   // Use _setjmp/_longjmp instead of setjmp/longjmp.
   setUseUnderscoreSetJmp(true);
   setUseUnderscoreLongJmp(true);
 
   // On PPC32/64, arguments smaller than 4/8 bytes are extended, so all
   // arguments are at least 4/8 bytes aligned.
   bool isPPC64 = Subtarget.isPPC64();
   setMinStackArgumentAlignment(isPPC64 ? 8:4);
 
   // Set up the register classes.
   addRegisterClass(MVT::i32, &PPC::GPRCRegClass);
   if (!useSoftFloat()) {
+    if (hasSPE()) {
+      addRegisterClass(MVT::f32, &PPC::SPE4RCRegClass);
+      addRegisterClass(MVT::f64, &PPC::SPERCRegClass);
+      addRegisterClass(MVT::v2i32, &PPC::SPERCRegClass);
+      addRegisterClass(MVT::v2f32, &PPC::SPERCRegClass);
+    } else {
       addRegisterClass(MVT::f32, &PPC::F4RCRegClass);
       addRegisterClass(MVT::f64, &PPC::F8RCRegClass);
     }
+  }
 
   // Match BITREVERSE to customized fast code sequence in the td file.
   setOperationAction(ISD::BITREVERSE, MVT::i32, Legal);
   setOperationAction(ISD::BITREVERSE, MVT::i64, Legal);
 
   // PowerPC has an i16 but no i8 (or i1) SEXTLOAD.
   for (MVT VT : MVT::integer_valuetypes()) {
     setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote);
if (!Subtarget.useCRBits())
     setOperationAction(ISD::SETCC, MVT::i32, Custom);
 
   // PowerPC does not have BRCOND which requires SetCC
   if (!Subtarget.useCRBits())
     setOperationAction(ISD::BRCOND, MVT::Other, Expand);
 
   setOperationAction(ISD::BR_JT,  MVT::Other, Expand);
 
+  if (Subtarget.hasSPE()) {
+    // SPE has built-in conversions
+    setOperationAction(ISD::FP_TO_SINT, MVT::i32, Legal);
+    setOperationAction(ISD::SINT_TO_FP, MVT::i32, Legal);
+    setOperationAction(ISD::UINT_TO_FP, MVT::i32, Legal);
+  } else {
     // PowerPC turns FP_TO_SINT into FCTIWZ and some load/stores.
     setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom);
 
     // PowerPC does not have [U|S]INT_TO_FP
     setOperationAction(ISD::SINT_TO_FP, MVT::i32, Expand);
     setOperationAction(ISD::UINT_TO_FP, MVT::i32, Expand);
+  }
 
   if (Subtarget.hasDirectMove() && isPPC64) {
     setOperationAction(ISD::BITCAST, MVT::f32, Legal);
     setOperationAction(ISD::BITCAST, MVT::i32, Legal);
     setOperationAction(ISD::BITCAST, MVT::i64, Legal);
     setOperationAction(ISD::BITCAST, MVT::f64, Legal);
   } else {
     setOperationAction(ISD::BITCAST, MVT::f32, Expand);
if (Subtarget.has64BitSupport()) {
     // This is just the low 32 bits of a (signed) fp->i64 conversion.
     // We cannot do this with Promote because i64 is not a legal type.
     setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom);
 
     if (Subtarget.hasLFIWAX() || Subtarget.isPPC64())
       setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom);
   } else {
     // PowerPC does not have FP_TO_UINT on 32-bit implementations.
+    if (Subtarget.hasSPE())
+      setOperationAction(ISD::FP_TO_UINT, MVT::i32, Legal);
+    else
       setOperationAction(ISD::FP_TO_UINT, MVT::i32, Expand);
   }
 
   // With the instructions enabled under FPCVT, we can do everything.
   if (Subtarget.hasFPCVT()) {
     if (Subtarget.has64BitSupport()) {
       setOperationAction(ISD::FP_TO_SINT, MVT::i64, Custom);
       setOperationAction(ISD::FP_TO_UINT, MVT::i64, Custom);
       setOperationAction(ISD::SINT_TO_FP, MVT::i64, Custom);
if (Subtarget.use64BitRegs()) {
     setOperationAction(ISD::SRL_PARTS, MVT::i64, Custom);
   } else {
     // 32-bit PowerPC wants to expand i64 shifts itself.
     setOperationAction(ISD::SHL_PARTS, MVT::i32, Custom);
     setOperationAction(ISD::SRA_PARTS, MVT::i32, Custom);
     setOperationAction(ISD::SRL_PARTS, MVT::i32, Custom);
   }
 
+  if (Subtarget.hasSPE()) {
+    setOperationAction(ISD::ADD, MVT::v2i32, Legal);
+    setOperationAction(ISD::ADD, MVT::v2f32, Legal);
+    setOperationAction(ISD::SUB, MVT::v2i32, Legal);
+    setOperationAction(ISD::SUB, MVT::v2f32, Legal);
+
+    setOperationAction(ISD::AND, MVT::v2i32, Legal);
+    setOperationAction(ISD::OR, MVT::v2i32, Legal);
+    setOperationAction(ISD::XOR, MVT::v2i32, Legal);
+
+    setOperationAction(ISD::SETCC, MVT::v2i32, Legal);
+    setOperationAction(ISD::SETCC, MVT::v2f32, Legal);
+  }
   if (Subtarget.hasAltivec()) {
     // First set operation action for all vector types to expand. Then we
     // will selectively turn on ones that can be effectively codegen'd.
     for (MVT VT : MVT::vector_valuetypes()) {
       // add/sub are legal for all supported vector VT's.
       setOperationAction(ISD::ADD, VT, Legal);
       setOperationAction(ISD::SUB, VT, Legal);
 
PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM,
   setMinFunctionAlignment(2);
   if (Subtarget.isDarwin())
     setPrefFunctionAlignment(4);
 
   switch (Subtarget.getDarwinDirective()) {
   default: break;
   case PPC::DIR_970:
   case PPC::DIR_A2:
+  case PPC::DIR_E500:
   case PPC::DIR_E500mc:
   case PPC::DIR_E5500:
   case PPC::DIR_PWR4:
   case PPC::DIR_PWR5:
   case PPC::DIR_PWR5X:
   case PPC::DIR_PWR6:
   case PPC::DIR_PWR6X:
   case PPC::DIR_PWR7:
unsigned PPCTargetLowering::getByValTypeAlignment(Type *Ty,
   // 16byte and wider vectors are passed on 16byte boundary.
   // The rest is 8 on PPC64 and 4 on PPC32 boundary.
   unsigned Align = Subtarget.isPPC64() ? 8 : 4;
   if (Subtarget.hasAltivec() || Subtarget.hasQPX())
     getMaxByValAlign(Ty, Align, Subtarget.hasQPX() ? 32 : 16);
   return Align;
 }
 
+unsigned PPCTargetLowering::getNumRegistersForCallingConv(LLVMContext &Context,
+                                                          EVT VT) const {
+  if (Subtarget.hasSPE() && VT == MVT::f64)
+    return 2;
+  return PPCTargetLowering::getNumRegisters(Context, VT);
+}
+
+MVT PPCTargetLowering::getRegisterTypeForCallingConv(LLVMContext &Context,
+                                                     EVT VT) const {
+  if (Subtarget.hasSPE() && VT == MVT::f64)
+    return MVT::i32;
+  return PPCTargetLowering::getRegisterType(Context, VT);
+}
+
+MVT PPCTargetLowering::getRegisterTypeForCallingConv(MVT VT) const {
+  if (Subtarget.hasSPE() && VT == MVT::f64)
+    return MVT::i32;
+  return PPCTargetLowering::getRegisterType(VT);
+}
+
 bool PPCTargetLowering::useSoftFloat() const {
   return Subtarget.useSoftFloat();
 }
 
+bool PPCTargetLowering::hasSPE() const {
+  return Subtarget.hasSPE();
+}
+
 const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const {
   switch ((PPCISD::NodeType)Opcode) {
   case PPCISD::FIRST_NUMBER:    break;
   case PPCISD::FSEL:            return "PPCISD::FSEL";
   case PPCISD::FCFID:           return "PPCISD::FCFID";
   case PPCISD::FCFIDU:          return "PPCISD::FCFIDU";
   case PPCISD::FCFIDS:          return "PPCISD::FCFIDS";
   case PPCISD::FCFIDUS:         return "PPCISD::FCFIDUS";
SDValue PPCTargetLowering::LowerFormalArguments_32SVR4(
   // Assign locations to all of the incoming arguments.
   SmallVector<CCValAssign, 16> ArgLocs;
   PPCCCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
                  *DAG.getContext());
 
   // Reserve space for the linkage area on the stack.
   unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize();
   CCInfo.AllocateStack(LinkageSize, PtrByteSize);
-  if (useSoftFloat())
+  if (useSoftFloat() || hasSPE())
     CCInfo.PreAnalyzeFormalArguments(Ins);
 
   CCInfo.AnalyzeFormalArguments(Ins, CC_PPC32_SVR4);
   CCInfo.clearWasPPCF128();
 
   for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
     CCValAssign &VA = ArgLocs[i];
 
     // Arguments stored in registers.
     if (VA.isRegLoc()) {
       const TargetRegisterClass *RC;
       EVT ValVT = VA.getValVT();
 
       switch (ValVT.getSimpleVT().SimpleTy) {
         default:
           llvm_unreachable("ValVT not supported by formal arguments Lowering");
         case MVT::i1:
         case MVT::i32:
           RC = &PPC::GPRCRegClass;
           break;
         case MVT::f32:
           if (Subtarget.hasP8Vector())
             RC = &PPC::VSSRCRegClass;
+          else if (Subtarget.hasSPE())
+            RC = &PPC::SPE4RCRegClass;
           else
             RC = &PPC::F4RCRegClass;
           break;
         case MVT::f64:
           if (Subtarget.hasVSX())
             RC = &PPC::VSFRCRegClass;
+          else if (Subtarget.hasSPE())
+            RC = &PPC::SPERCRegClass;
           else
             RC = &PPC::F8RCRegClass;
           break;
         case MVT::v16i8:
         case MVT::v8i16:
         case MVT::v4i32:
           RC = &PPC::VRRCRegClass;
           break;
         case MVT::v4f32:
           RC = Subtarget.hasQPX() ? &PPC::QSRCRegClass : &PPC::VRRCRegClass;
           break;
         case MVT::v2f64:
         case MVT::v2i64:
           RC = &PPC::VRRCRegClass;
           break;
         case MVT::v4f64:
           RC = &PPC::QFRCRegClass;
           break;
         case MVT::v4i1:
           RC = &PPC::QBRCRegClass;
           break;
+        case MVT::v2i32:
+        case MVT::v2f32:
+          RC= &PPC::SPERCRegClass;
+          break;
       }
 
       // Transform the arguments stored in physical registers into virtual ones.
       unsigned Reg = MF.addLiveIn(VA.getLocReg(), RC);
       SDValue ArgValue = DAG.getCopyFromReg(Chain, dl, Reg,
                                             ValVT == MVT::i1 ? MVT::i32 : ValVT);
 
       if (ValVT == MVT::i1)
if (isVarArg) {
     const unsigned NumGPArgRegs = array_lengthof(GPArgRegs);
 
     static const MCPhysReg FPArgRegs[] = {
       PPC::F1, PPC::F2, PPC::F3, PPC::F4, PPC::F5, PPC::F6, PPC::F7,
       PPC::F8
     };
     unsigned NumFPArgRegs = array_lengthof(FPArgRegs);
 
-    if (useSoftFloat())
+    if (useSoftFloat() || hasSPE())
        NumFPArgRegs = 0;
 
     FuncInfo->setVarArgsNumGPR(CCInfo.getFirstUnallocated(GPArgRegs));
     FuncInfo->setVarArgsNumFPR(CCInfo.getFirstUnallocated(FPArgRegs));
 
     // Make room for NumGPArgRegs and NumFPArgRegs.
     int Depth = NumGPArgRegs * PtrVT.getSizeInBits()/8 +
                 NumFPArgRegs * MVT(MVT::f64).getSizeInBits()/8;
SDValue PPCTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
 
   if (IntrinsicID == Intrinsic::thread_pointer) {
     // Reads the thread pointer register, used for __builtin_thread_pointer.
     if (Subtarget.isPPC64())
       return DAG.getRegister(PPC::X13, MVT::i64);
     return DAG.getRegister(PPC::R2, MVT::i32);
   }
 
+  if (IntrinsicID == Intrinsic::ppc_spe_evabs) {
+    SDValue V = Op.getOperand(1);
+    return DAG.getNode(ISD::ABS, dl, V.getValueType(), V);
+  }
+
   // We are looking for absolute values here.
   // The idea is to try to fit one of two patterns:
   //  max (a, (0-a))  OR  max ((0-a), a)
   if (Subtarget.hasP9Vector() &&
       (IntrinsicID == Intrinsic::ppc_altivec_vmaxsw ||
        IntrinsicID == Intrinsic::ppc_altivec_vmaxsh ||
        IntrinsicID == Intrinsic::ppc_altivec_vmaxsb)) {
     SDValue V1 = Op.getOperand(1);
TII->insertSelect(*BB, MI, dl, MI.getOperand(0).getReg(), Cond,
              MI.getOpcode() == PPC::SELECT_CC_F8 ||
              MI.getOpcode() == PPC::SELECT_CC_QFRC ||
              MI.getOpcode() == PPC::SELECT_CC_QSRC ||
              MI.getOpcode() == PPC::SELECT_CC_QBRC ||
              MI.getOpcode() == PPC::SELECT_CC_VRRC ||
              MI.getOpcode() == PPC::SELECT_CC_VSFRC ||
              MI.getOpcode() == PPC::SELECT_CC_VSSRC ||
              MI.getOpcode() == PPC::SELECT_CC_VSRC ||
+             MI.getOpcode() == PPC::SELECT_CC_SPE4 ||
+             MI.getOpcode() == PPC::SELECT_CC_SPE ||
              MI.getOpcode() == PPC::SELECT_I4 ||
              MI.getOpcode() == PPC::SELECT_I8 ||
              MI.getOpcode() == PPC::SELECT_F4 ||
              MI.getOpcode() == PPC::SELECT_F8 ||
+             MI.getOpcode() == PPC::SELECT_SPE4 ||
+             MI.getOpcode() == PPC::SELECT_SPE ||
              MI.getOpcode() == PPC::SELECT_QFRC ||
              MI.getOpcode() == PPC::SELECT_QSRC ||
              MI.getOpcode() == PPC::SELECT_QBRC ||
              MI.getOpcode() == PPC::SELECT_VRRC ||
              MI.getOpcode() == PPC::SELECT_VSFRC ||
              MI.getOpcode() == PPC::SELECT_VSSRC ||
              MI.getOpcode() == PPC::SELECT_VSRC) {
     // The incoming instruction knows the destination vreg to set, the
unsigned PPCTargetLowering::combineRepeatedFPDivisors() const {
   // Combine multiple FDIVs with the same divisor into multiple FMULs by the
   // reciprocal if there are two or more FDIVs (for embedded cores with only
   // one FP pipeline) for three or more FDIVs (for generic OOO cores).
   switch (Subtarget.getDarwinDirective()) {
   default:
     return 3;
   case PPC::DIR_440:
   case PPC::DIR_A2:
+  case PPC::DIR_E500:
   case PPC::DIR_E500mc:
   case PPC::DIR_E5500:
     return 2;
   }
 }
 
 // isConsecutiveLSLoc needs to work even if all adds have not yet been
 // collapsed, and so we need to look through chains of them.
case 'r':   // R0-R31
       if (VT == MVT::i64 && Subtarget.isPPC64())
         return std::make_pair(0U, &PPC::G8RCRegClass);
       return std::make_pair(0U, &PPC::GPRCRegClass);
     // 'd' and 'f' constraints are both defined to be "the floating point
     // registers", where one is for 32-bit and the other for 64-bit. We don't
     // really care overly much here so just give them all the same reg classes.
     case 'd':
     case 'f':
+      if (Subtarget.hasSPE()) {
+        if (VT == MVT::f32 || VT == MVT::i32)
+          return std::make_pair(0U, &PPC::SPE4RCRegClass);
+        if (VT == MVT::f64 || VT == MVT::i64)
+          return std::make_pair(0U, &PPC::SPERCRegClass);
+      } else {
         if (VT == MVT::f32 || VT == MVT::i32)
           return std::make_pair(0U, &PPC::F4RCRegClass);
         if (VT == MVT::f64 || VT == MVT::i64)
           return std::make_pair(0U, &PPC::F8RCRegClass);
         if (VT == MVT::v4f64 && Subtarget.hasQPX())
           return std::make_pair(0U, &PPC::QFRCRegClass);
         if (VT == MVT::v4f32 && Subtarget.hasQPX())
           return std::make_pair(0U, &PPC::QSRCRegClass);
+      }
       break;
     case 'v':
       if (VT == MVT::v4f64 && Subtarget.hasQPX())
         return std::make_pair(0U, &PPC::QFRCRegClass);
       if (VT == MVT::v4f32 && Subtarget.hasQPX())
         return std::make_pair(0U, &PPC::QSRCRegClass);
       if (Subtarget.hasAltivec())
         return std::make_pair(0U, &PPC::VRRCRegClass);

lib/Target/PowerPC/PPCInstrInfo.cpp

 unsigned PPCInstrInfo::isLoadFromStackSlot(const MachineInstr &MI,
                                            int &FrameIndex) const {
   // Note: This list must be kept consistent with LoadRegFromStackSlot.
   switch (MI.getOpcode()) {
   default: break;
   case PPC::LD:
   case PPC::LWZ:
   case PPC::LFS:
   case PPC::LFD:
   case PPC::RESTORE_CR:
   case PPC::RESTORE_CRBIT:
   case PPC::LVX:
   case PPC::LXVD2X:
   case PPC::LXVX:
   case PPC::QVLFDX:
   case PPC::QVLFSXs:
   case PPC::QVLFDXb:
if (PPC::CRBITRCRegClass.contains(SrcReg) &&
     NumGPRtoVSRSpill++;
     getKillRegState(KillSrc);
     return;
   } else if (PPC::VSFRCRegClass.contains(SrcReg) &&
              PPC::G8RCRegClass.contains(DestReg)) {
     BuildMI(MBB, I, DL, get(PPC::MFVSRD), DestReg).addReg(SrcReg);
     getKillRegState(KillSrc);
     return;
+  } else if (PPC::SPERCRegClass.contains(SrcReg) &&
+             PPC::SPE4RCRegClass.contains(DestReg)) {
+    BuildMI(MBB, I, DL, get(PPC::EFSCFD), DestReg).addReg(SrcReg);
+    getKillRegState(KillSrc);
+    return;
+  } else if (PPC::SPE4RCRegClass.contains(SrcReg) &&
+             PPC::SPERCRegClass.contains(DestReg)) {
+    BuildMI(MBB, I, DL, get(PPC::EFDCFS), DestReg).addReg(SrcReg);
+    getKillRegState(KillSrc);
+    return;
   }
 
+
   unsigned Opc;
   if (PPC::GPRCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::OR;
   else if (PPC::G8RCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::OR8;
   else if (PPC::F4RCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::FMR;
   else if (PPC::CRRCRegClass.contains(DestReg, SrcReg))
void PPCInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
   else if (PPC::QFRCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::QVFMR;
   else if (PPC::QSRCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::QVFMRs;
   else if (PPC::QBRCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::QVFMRb;
   else if (PPC::CRBITRCRegClass.contains(DestReg, SrcReg))
     Opc = PPC::CROR;
+  else if (PPC::SPERCRegClass.contains(DestReg, SrcReg))
+    Opc = PPC::EVOR;
+  else if (PPC::SPE4RCRegClass.contains(DestReg, SrcReg))
+    Opc = PPC::OR;
   else
     llvm_unreachable("Impossible reg-to-reg copy");
 
   const MCInstrDesc &MCID = get(Opc);
   if (MCID.getNumOperands() == 3)
     BuildMI(MBB, I, DL, MCID, DestReg)
       .addReg(SrcReg).addReg(SrcReg, getKillRegState(KillSrc));
   else
NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STFD))
                                        .addReg(SrcReg,
                                                getKillRegState(isKill)),
                                        FrameIdx));
   } else if (PPC::F4RCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STFS))
                                        .addReg(SrcReg,
                                                getKillRegState(isKill)),
                                        FrameIdx));
+  } else if (PPC::SPERCRegClass.hasSubClassEq(RC)) {
+    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::EVSTDD))
+                                       .addReg(SrcReg,
+                                               getKillRegState(isKill)),
+                                       FrameIdx));
+  } else if (PPC::SPE4RCRegClass.hasSubClassEq(RC)) {
+    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPESTW))
+                                       .addReg(SrcReg,
+                                               getKillRegState(isKill)),
+                                       FrameIdx));
   } else if (PPC::CRRCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPILL_CR))
                                        .addReg(SrcReg,
                                                getKillRegState(isKill)),
                                        FrameIdx));
     return true;
   } else if (PPC::CRBITRCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPILL_CRBIT))
if (PPC::GPRCRegClass.hasSubClassEq(RC) ||
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD), DestReg),
                                        FrameIdx));
   } else if (PPC::F8RCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LFD), DestReg),
                                        FrameIdx));
   } else if (PPC::F4RCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LFS), DestReg),
                                        FrameIdx));
+  } else if (PPC::SPERCRegClass.hasSubClassEq(RC)) {
+    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::EVLDD), DestReg),
+                                       FrameIdx));
+  } else if (PPC::SPE4RCRegClass.hasSubClassEq(RC)) {
+    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPELWZ), DestReg),
+                                       FrameIdx));
   } else if (PPC::CRRCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL,
                                                get(PPC::RESTORE_CR), DestReg),
                                        FrameIdx));
     return true;
   } else if (PPC::CRBITRCRegClass.hasSubClassEq(RC)) {
     NewMIs.push_back(addFrameReference(BuildMI(MF, DL,
                                                get(PPC::RESTORE_CRBIT), DestReg),

lib/Target/PowerPC/PPCInstrInfo.td

 }
 def crrc : RegisterOperand<CRRC> {
   let ParserMatchClass = PPCRegCRRCAsmOperand;
 }
 def crrc0 : RegisterOperand<CRRC0> {
   let ParserMatchClass = PPCRegCRRCAsmOperand;
 }
 
+def PPCRegSPERCAsmOperand : AsmOperandClass {
+  let Name = "RegSPERC"; let PredicateMethod = "isRegNumber";
+}
+def sperc : RegisterOperand<SPERC> {
+  let ParserMatchClass = PPCRegSPERCAsmOperand;
+}
+def PPCRegSPE4RCAsmOperand : AsmOperandClass {
+  let Name = "RegSPE4RC"; let PredicateMethod = "isRegNumber";
+}
+def spe4rc : RegisterOperand<SPE4RC> {
+  let ParserMatchClass = PPCRegSPE4RCAsmOperand;
+}
+
 def PPCU1ImmAsmOperand : AsmOperandClass {
   let Name = "U1Imm"; let PredicateMethod = "isU1Imm";
   let RenderMethod = "addImmOperands";
 }
 def u1imm   : Operand<i32> {
   let PrintMethod = "printU1ImmOperand";
   let ParserMatchClass = PPCU1ImmAsmOperand;
 }

 def In64BitMode  : Predicate<"PPCSubTarget->isPPC64()">;
 def IsBookE  : Predicate<"PPCSubTarget->isBookE()">;
 def IsNotBookE  : Predicate<"!PPCSubTarget->isBookE()">;
 def HasOnlyMSYNC : Predicate<"PPCSubTarget->hasOnlyMSYNC()">;
 def HasSYNC   : Predicate<"!PPCSubTarget->hasOnlyMSYNC()">;
 def IsPPC4xx  : Predicate<"PPCSubTarget->isPPC4xx()">;
 def IsPPC6xx  : Predicate<"PPCSubTarget->isPPC6xx()">;
 def IsE500  : Predicate<"PPCSubTarget->isE500()">;
-def HasSPE  : Predicate<"PPCSubTarget->HasSPE()">;
+def HasSPE  : Predicate<"PPCSubTarget->hasSPE()">;
 def HasICBT : Predicate<"PPCSubTarget->hasICBT()">;
 def HasPartwordAtomics : Predicate<"PPCSubTarget->hasPartwordAtomics()">;
 def NoNaNsFPMath : Predicate<"TM.Options.NoNaNsFPMath">;
 def NaNsFPMath   : Predicate<"!TM.Options.NoNaNsFPMath">;
 def HasBPERMD : Predicate<"PPCSubTarget->hasBPERMD()">;
 def HasExtDiv : Predicate<"PPCSubTarget->hasExtDiv()">;
 def IsISA3_0 : Predicate<"PPCSubTarget->isISA3_0()">;
+def HasTraditionalFPU : Predicate<"!PPCSubTarget->hasSPE()">;
 
 //===----------------------------------------------------------------------===//
 // PowerPC Multiclass Definitions.
 
 multiclass XForm_6r<bits<6> opcode, bits<10> xo, dag OOL, dag IOL,
                     string asmbase, string asmstr, InstrItinClass itin,
                     list<dag> pattern> {
   let BaseName = asmbase in {
let usesCustomInserter = 1,    // Expanded after instruction selection.
   // SELECT_* pseudo instructions, like SELECT_CC_* but taking condition
   // register bit directly.
   def SELECT_I4 : Pseudo<(outs gprc:$dst), (ins crbitrc:$cond,
                           gprc_nor0:$T, gprc_nor0:$F), "#SELECT_I4",
                           [(set i32:$dst, (select i1:$cond, i32:$T, i32:$F))]>;
   def SELECT_I8 : Pseudo<(outs g8rc:$dst), (ins crbitrc:$cond,
                           g8rc_nox0:$T, g8rc_nox0:$F), "#SELECT_I8",
                           [(set i64:$dst, (select i1:$cond, i64:$T, i64:$F))]>;
+let Predicates = [HasTraditionalFPU] in {
   def SELECT_F4  : Pseudo<(outs f4rc:$dst), (ins crbitrc:$cond,
                           f4rc:$T, f4rc:$F), "#SELECT_F4",
                           [(set f32:$dst, (select i1:$cond, f32:$T, f32:$F))]>;
   def SELECT_F8  : Pseudo<(outs f8rc:$dst), (ins crbitrc:$cond,
                           f8rc:$T, f8rc:$F), "#SELECT_F8",
                           [(set f64:$dst, (select i1:$cond, f64:$T, f64:$F))]>;
+}
   def SELECT_VRRC: Pseudo<(outs vrrc:$dst), (ins crbitrc:$cond,
                           vrrc:$T, vrrc:$F), "#SELECT_VRRC",
                           [(set v4i32:$dst,
                                 (select i1:$cond, v4i32:$T, v4i32:$F))]>;
 }
 
 // SPILL_CR - Indicate that we're dumping the CR register, so we'll need to
 // scavenge a register for it.
def LHA : DForm_1<42, (outs gprc:$rD), (ins memri:$src),
                   PPC970_DGroup_Cracked;
 def LHZ : DForm_1<40, (outs gprc:$rD), (ins memri:$src),
                   "lhz $rD, $src", IIC_LdStLoad,
                   [(set i32:$rD, (zextloadi16 iaddr:$src))]>;
 def LWZ : DForm_1<32, (outs gprc:$rD), (ins memri:$src),
                   "lwz $rD, $src", IIC_LdStLoad,
                   [(set i32:$rD, (load iaddr:$src))]>;
 
+let Predicates = [HasTraditionalFPU] in {
 def LFS : DForm_1<48, (outs f4rc:$rD), (ins memri:$src),
                   "lfs $rD, $src", IIC_LdStLFD,
                   [(set f32:$rD, (load iaddr:$src))]>;
 def LFD : DForm_1<50, (outs f8rc:$rD), (ins memri:$src),
                   "lfd $rD, $src", IIC_LdStLFD,
                   [(set f64:$rD, (load iaddr:$src))]>;
+}
 
 
 // Unindexed (r+i) Loads with Update (preinc).
 let mayLoad = 1, mayStore = 0, hasSideEffects = 0 in {
 def LBZU : DForm_1<35, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                    "lbzu $rD, $addr", IIC_LdStLoadUpd,
                    []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
 
 def LHAU : DForm_1<43, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                    "lhau $rD, $addr", IIC_LdStLHAU,
                    []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
 
 def LHZU : DForm_1<41, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                    "lhzu $rD, $addr", IIC_LdStLoadUpd,
                    []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
 
 def LWZU : DForm_1<33, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                    "lwzu $rD, $addr", IIC_LdStLoadUpd,
                    []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
 
+let Predicates = [HasTraditionalFPU] in {
 def LFSU : DForm_1<49, (outs f4rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                   "lfsu $rD, $addr", IIC_LdStLFDU,
                   []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
 
 def LFDU : DForm_1<51, (outs f8rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr),
                   "lfdu $rD, $addr", IIC_LdStLFDU,
                   []>, RegConstraint<"$addr.reg = $ea_result">,
                    NoEncode<"$ea_result">;
+}
 
 
 // Indexed (r+r) Loads with Update (preinc).
 def LBZUX : XForm_1<31, 119, (outs gprc:$rD, ptr_rc_nor0:$ea_result),
                    (ins memrr:$addr),
                    "lbzux $rD, $addr", IIC_LdStLoadUpdX,
                    []>, RegConstraint<"$addr.ptrreg = $ea_result">,
                    NoEncode<"$ea_result">;
def LHZUX : XForm_1<31, 311, (outs gprc:$rD, ptr_rc_nor0:$ea_result),
                    NoEncode<"$ea_result">;
 
 def LWZUX : XForm_1<31, 55, (outs gprc:$rD, ptr_rc_nor0:$ea_result),
                    (ins memrr:$addr),
                    "lwzux $rD, $addr", IIC_LdStLoadUpdX,
                    []>, RegConstraint<"$addr.ptrreg = $ea_result">,
                    NoEncode<"$ea_result">;
 
+let Predicates = [HasTraditionalFPU] in {
 def LFSUX : XForm_1<31, 567, (outs f4rc:$rD, ptr_rc_nor0:$ea_result),
                    (ins memrr:$addr),
                    "lfsux $rD, $addr", IIC_LdStLFDUX,
                    []>, RegConstraint<"$addr.ptrreg = $ea_result">,
                    NoEncode<"$ea_result">;
 
 def LFDUX : XForm_1<31, 631, (outs f8rc:$rD, ptr_rc_nor0:$ea_result),
                    (ins memrr:$addr),
                    "lfdux $rD, $addr", IIC_LdStLFDUX,
                    []>, RegConstraint<"$addr.ptrreg = $ea_result">,
                    NoEncode<"$ea_result">;
 }
 }
+}
 
 // Indexed (r+r) Loads.
 //
 let PPC970_Unit = 2, mayLoad = 1, mayStore = 0 in {
 def LBZX : XForm_1<31,  87, (outs gprc:$rD), (ins memrr:$src),
                    "lbzx $rD, $src", IIC_LdStLoad,
                    [(set i32:$rD, (zextloadi8 xaddr:$src))]>;
 def LHAX : XForm_1<31, 343, (outs gprc:$rD), (ins memrr:$src),
                    "lhax $rD, $src", IIC_LdStLHA,
                    [(set i32:$rD, (sextloadi16 xaddr:$src))]>,
                    PPC970_DGroup_Cracked;
 def LHZX : XForm_1<31, 279, (outs gprc:$rD), (ins memrr:$src),
                    "lhzx $rD, $src", IIC_LdStLoad,
                    [(set i32:$rD, (zextloadi16 xaddr:$src))]>;
 def LWZX : XForm_1<31,  23, (outs gprc:$rD), (ins memrr:$src),
                    "lwzx $rD, $src", IIC_LdStLoad,
                    [(set i32:$rD, (load xaddr:$src))]>;
 def LHBRX : XForm_1<31, 790, (outs gprc:$rD), (ins memrr:$src),
                    "lhbrx $rD, $src", IIC_LdStLoad,
                    [(set i32:$rD, (PPClbrx xoaddr:$src, i16))]>;
 def LWBRX : XForm_1<31,  534, (outs gprc:$rD), (ins memrr:$src),
                    "lwbrx $rD, $src", IIC_LdStLoad,
                    [(set i32:$rD, (PPClbrx xoaddr:$src, i32))]>;
 
+let Predicates = [HasTraditionalFPU] in {
 def LFSX   : XForm_25<31, 535, (outs f4rc:$frD), (ins memrr:$src),
                       "lfsx $frD, $src", IIC_LdStLFD,
                       [(set f32:$frD, (load xaddr:$src))]>;
 def LFDX   : XForm_25<31, 599, (outs f8rc:$frD), (ins memrr:$src),
                       "lfdx $frD, $src", IIC_LdStLFD,
                       [(set f64:$frD, (load xaddr:$src))]>;
 
 def LFIWAX : XForm_25<31, 855, (outs f8rc:$frD), (ins memrr:$src),
                       "lfiwax $frD, $src", IIC_LdStLFD,
                       [(set f64:$frD, (PPClfiwax xoaddr:$src))]>;
 def LFIWZX : XForm_25<31, 887, (outs f8rc:$frD), (ins memrr:$src),
                       "lfiwzx $frD, $src", IIC_LdStLFD,
                       [(set f64:$frD, (PPClfiwzx xoaddr:$src))]>;
 }
+}
 
 // Load Multiple
 def LMW : DForm_1<46, (outs gprc:$rD), (ins memri:$src),
                   "lmw $rD, $src", IIC_LdStLMW, []>;
 
 //===----------------------------------------------------------------------===//
 // PPC32 Store Instructions.
 //
 
 // Unindexed (r+i) Stores.
 let PPC970_Unit = 2, mayStore = 1, mayLoad = 0 in {
 def STB  : DForm_1<38, (outs), (ins gprc:$rS, memri:$src),
                    "stb $rS, $src", IIC_LdStStore,
                    [(truncstorei8 i32:$rS, iaddr:$src)]>;
 def STH  : DForm_1<44, (outs), (ins gprc:$rS, memri:$src),
                    "sth $rS, $src", IIC_LdStStore,
                    [(truncstorei16 i32:$rS, iaddr:$src)]>;
 def STW  : DForm_1<36, (outs), (ins gprc:$rS, memri:$src),
                    "stw $rS, $src", IIC_LdStStore,
                    [(store i32:$rS, iaddr:$src)]>;
+let Predicates = [HasTraditionalFPU] in {
 def STFS : DForm_1<52, (outs), (ins f4rc:$rS, memri:$dst),
                    "stfs $rS, $dst", IIC_LdStSTFD,
                    [(store f32:$rS, iaddr:$dst)]>;
 def STFD : DForm_1<54, (outs), (ins f8rc:$rS, memri:$dst),
                    "stfd $rS, $dst", IIC_LdStSTFD,
                    [(store f64:$rS, iaddr:$dst)]>;
 }
+}
 
 // Unindexed (r+i) Stores with Update (preinc).
 let PPC970_Unit = 2, mayStore = 1, mayLoad = 0 in {
 def STBU  : DForm_1<39, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst),
                     "stbu $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
 def STHU  : DForm_1<45, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst),
                     "sthu $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
 def STWU  : DForm_1<37, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst),
                     "stwu $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
+let Predicates = [HasTraditionalFPU] in {
 def STFSU : DForm_1<53, (outs ptr_rc_nor0:$ea_res), (ins f4rc:$rS, memri:$dst),
                     "stfsu $rS, $dst", IIC_LdStSTFDU, []>,
                     RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
 def STFDU : DForm_1<55, (outs ptr_rc_nor0:$ea_res), (ins f8rc:$rS, memri:$dst),
                     "stfdu $rS, $dst", IIC_LdStSTFDU, []>,
                     RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">;
 }
+}
 
 // Patterns to match the pre-inc stores.  We can't put the patterns on
 // the instruction definitions directly as ISel wants the address base
 // and offset to be separate operands, not a single complex operand.
 def : Pat<(pre_truncsti8 i32:$rS, iPTR:$ptrreg, iaddroff:$ptroff),
           (STBU $rS, iaddroff:$ptroff, $ptrreg)>;
 def : Pat<(pre_truncsti16 i32:$rS, iPTR:$ptrreg, iaddroff:$ptroff),
           (STHU $rS, iaddroff:$ptroff, $ptrreg)>;
def STHBRX: XForm_8<31, 918, (outs), (ins gprc:$rS, memrr:$dst),
                    "sthbrx $rS, $dst", IIC_LdStStore,
                    [(PPCstbrx i32:$rS, xoaddr:$dst, i16)]>,
                    PPC970_DGroup_Cracked;
 def STWBRX: XForm_8<31, 662, (outs), (ins gprc:$rS, memrr:$dst),
                    "stwbrx $rS, $dst", IIC_LdStStore,
                    [(PPCstbrx i32:$rS, xoaddr:$dst, i32)]>,
                    PPC970_DGroup_Cracked;
 
+let Predicates = [HasTraditionalFPU] in {
 def STFIWX: XForm_28<31, 983, (outs), (ins f8rc:$frS, memrr:$dst),
                      "stfiwx $frS, $dst", IIC_LdStSTFD,
                      [(PPCstfiwx f64:$frS, xoaddr:$dst)]>;
                      
 def STFSX : XForm_28<31, 663, (outs), (ins f4rc:$frS, memrr:$dst),
                      "stfsx $frS, $dst", IIC_LdStSTFD,
                      [(store f32:$frS, xaddr:$dst)]>;
 def STFDX : XForm_28<31, 727, (outs), (ins f8rc:$frS, memrr:$dst),
                      "stfdx $frS, $dst", IIC_LdStSTFD,
                      [(store f64:$frS, xaddr:$dst)]>;
 }
+}
 
 // Indexed (r+r) Stores with Update (preinc).
 let PPC970_Unit = 2, mayStore = 1, mayLoad = 0 in {
 def STBUX : XForm_8<31, 247, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst),
                     "stbux $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">,
                     PPC970_DGroup_Cracked;
 def STHUX : XForm_8<31, 439, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst),
                     "sthux $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">,
                     PPC970_DGroup_Cracked;
 def STWUX : XForm_8<31, 183, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst),
                     "stwux $rS, $dst", IIC_LdStStoreUpd, []>,
                     RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">,
                     PPC970_DGroup_Cracked;
+let Predicates = [HasTraditionalFPU] in {
 def STFSUX: XForm_8<31, 695, (outs ptr_rc_nor0:$ea_res), (ins f4rc:$rS, memrr:$dst),
                     "stfsux $rS, $dst", IIC_LdStSTFDU, []>,
                     RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">,
                     PPC970_DGroup_Cracked;
 def STFDUX: XForm_8<31, 759, (outs ptr_rc_nor0:$ea_res), (ins f8rc:$rS, memrr:$dst),
                     "stfdux $rS, $dst", IIC_LdStSTFDU, []>,
                     RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">,
                     PPC970_DGroup_Cracked;
 }
+}
 
 // Patterns to match the pre-inc stores.  We can't put the patterns on
 // the instruction definitions directly as ISel wants the address base
 // and offset to be separate operands, not a single complex operand.
 def : Pat<(pre_truncsti8 i32:$rS, iPTR:$ptrreg, iPTR:$ptroff),
           (STBUX $rS, $ptrreg, $ptroff)>;
 def : Pat<(pre_truncsti16 i32:$rS, iPTR:$ptrreg, iPTR:$ptroff),
           (STHUX $rS, $ptrreg, $ptroff)>;
 def : Pat<(pre_store i32:$rS, iPTR:$ptrreg, iPTR:$ptroff),
           (STWUX $rS, $ptrreg, $ptroff)>;
+let Predicates = [HasTraditionalFPU] in {
 def : Pat<(pre_store f32:$rS, iPTR:$ptrreg, iPTR:$ptroff),
           (STFSUX $rS, $ptrreg, $ptroff)>;
 def : Pat<(pre_store f64:$rS, iPTR:$ptrreg, iPTR:$ptroff),
           (STFDUX $rS, $ptrreg, $ptroff)>;
+}
 
 // Store Multiple
 def STMW : DForm_1<47, (outs), (ins gprc:$rS, memri:$dst),
                    "stmw $rS, $dst", IIC_LdStLMW, []>;
 
 def SYNC : XForm_24_sync<31, 598, (outs), (ins i32imm:$L),
                         "sync $L", IIC_LdStSync, []>;
 

 }
 let isCompare = 1, hasSideEffects = 0 in {
   def CMPW   : XForm_16_ext<31, 0, (outs crrc:$crD), (ins gprc:$rA, gprc:$rB),
                             "cmpw $crD, $rA, $rB", IIC_IntCompare>;
   def CMPLW  : XForm_16_ext<31, 32, (outs crrc:$crD), (ins gprc:$rA, gprc:$rB),
                             "cmplw $crD, $rA, $rB", IIC_IntCompare>;
 }
 }
-let PPC970_Unit = 3 in {  // FPU Operations.
+let PPC970_Unit = 3, Predicates = [HasTraditionalFPU] in {  // FPU Operations.
 //def FCMPO  : XForm_17<63, 32, (outs CRRC:$crD), (ins FPRC:$fA, FPRC:$fB),
 //                      "fcmpo $crD, $fA, $fB", IIC_FPCompare>;
 let isCompare = 1, hasSideEffects = 0 in {
   def FCMPUS : XForm_17<63, 0, (outs crrc:$crD), (ins f4rc:$fA, f4rc:$fB),
                         "fcmpu $crD, $fA, $fB", IIC_FPCompare>;
   let Interpretation64Bit = 1, isCodeGenOnly = 1 in
   def FCMPUD : XForm_17<63, 0, (outs crrc:$crD), (ins f8rc:$fA, f8rc:$fB),
                         "fcmpu $crD, $fA, $fB", IIC_FPCompare>;
let Uses = [RM] in {
   }
   }
 }
 
 /// Note that FMR is defined as pseudo-ops on the PPC970 because they are
 /// often coalesced away and we don't want the dispatch group builder to think
 /// that they will fill slots (which could cause the load of a LSU reject to
 /// sneak into a d-group with a store).
-let hasSideEffects = 0 in
+let hasSideEffects = 0, Predicates = [HasTraditionalFPU] in
 defm FMR   : XForm_26r<63, 72, (outs f4rc:$frD), (ins f4rc:$frB),
                        "fmr", "$frD, $frB", IIC_FPGeneral,
                        []>,  // (set f32:$frD, f32:$frB)
                        PPC970_Unit_Pseudo;
 
-let PPC970_Unit = 3, hasSideEffects = 0 in {  // FPU Operations.
+let PPC970_Unit = 3, hasSideEffects = 0, Predicates = [HasTraditionalFPU] in {  // FPU Operations.
 // These are artificially split into two different forms, for 4/8 byte FP.
 defm FABSS  : XForm_26r<63, 264, (outs f4rc:$frD), (ins f4rc:$frB),
                         "fabs", "$frD, $frB", IIC_FPGeneral,
                         [(set f32:$frD, (fabs f32:$frB))]>;
 let Interpretation64Bit = 1, isCodeGenOnly = 1 in
 defm FABSD  : XForm_26r<63, 264, (outs f8rc:$frD), (ins f8rc:$frB),
                         "fabs", "$frD, $frB", IIC_FPGeneral,
                         [(set f64:$frD, (fabs f64:$frB))]>;
def MFCR : XFXForm_3<31, 19, (outs gprc:$rT), (ins),
                      "mfcr $rT", IIC_SprMFCR>,
                      PPC970_MicroCode, PPC970_Unit_CRU;
 } // hasExtraSrcRegAllocReq = 1
 
 def MCRXRX : X_BF3<31, 576, (outs crrc:$BF), (ins),
                    "mcrxrx $BF", IIC_BrMCRX>, Requires<[IsISA3_0]>;
 } // hasSideEffects = 0
 
+let Predicates = [HasTraditionalFPU] in {
 // Pseudo instruction to perform FADD in round-to-zero mode.
 let usesCustomInserter = 1, Uses = [RM] in {
   def FADDrtz: Pseudo<(outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB), "",
                       [(set f64:$FRT, (PPCfaddrtz f64:$FRA, f64:$FRB))]>;
 }
 
 // The above pseudo gets expanded to make use of the following instructions
 // to manipulate FPSCR.  Note that FPSCR is not modeled at the DAG level.
def MFFSCRNI : X_FRT5_XO2_XO3_RM2_X10<63, 2, 7, 583, (outs f8rc:$rT),
                                        (ins u2imm:$RM), "mffscrni $rT, $RM",
                                        IIC_IntMFFS, []>,
                  PPC970_DGroup_Single, PPC970_Unit_FPU;
 
   def MFFSL  : X_FRT5_XO2_XO3_XO10<63, 3, 0, 583, (outs f8rc:$rT), (ins),
                                   "mffsl $rT", IIC_IntMFFS, []>,
                PPC970_DGroup_Single, PPC970_Unit_FPU;
 }
+}
 
 let Predicates = [IsISA3_0] in {
 def MODSW : XForm_8<31, 779, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB),
                         "modsw $rT, $rA, $rB", IIC_IntDivW,
                         [(set i32:$rT, (srem i32:$rA, i32:$rB))]>;
 def MODUW : XForm_8<31, 267, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB),
                         "moduw $rT, $rA, $rB", IIC_IntDivW,
                         [(set i32:$rT, (urem i32:$rA, i32:$rB))]>;
defm SUBFZE : XOForm_3rc<31, 200, 0, (outs gprc:$rT), (ins gprc:$rA),
                          "subfze", "$rT, $rA", IIC_IntGeneral,
                          [(set i32:$rT, (sube 0, i32:$rA))]>;
 }
 }
 
 // A-Form instructions.  Most of the instructions executed in the FPU are of
 // this type.
 //
-let PPC970_Unit = 3, hasSideEffects = 0 in {  // FPU Operations.
+let PPC970_Unit = 3, hasSideEffects = 0, Predicates = [HasTraditionalFPU] in {  // FPU Operations.
 let Uses = [RM] in {
 let isCommutable = 1 in {
   defm FMADD : AForm_1r<63, 29, 
                       (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB),
                       "fmadd", "$FRT, $FRA, $FRC, $FRB", IIC_FPFused,
                       [(set f64:$FRT, (fma f64:$FRA, f64:$FRC, f64:$FRB))]>;
   defm FMADDS : AForm_1r<59, 29,
                       (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC, f4rc:$FRB),

 def : Pat<(extloadi8 iaddr:$src),
           (LBZ iaddr:$src)>;
 def : Pat<(extloadi8 xaddr:$src),
           (LBZX xaddr:$src)>;
 def : Pat<(extloadi16 iaddr:$src),
           (LHZ iaddr:$src)>;
 def : Pat<(extloadi16 xaddr:$src),
           (LHZX xaddr:$src)>;
+let Predicates = [HasTraditionalFPU] in {
 def : Pat<(f64 (extloadf32 iaddr:$src)),
           (COPY_TO_REGCLASS (LFS iaddr:$src), F8RC)>;
 def : Pat<(f64 (extloadf32 xaddr:$src)),
           (COPY_TO_REGCLASS (LFSX xaddr:$src), F8RC)>;
 
 def : Pat<(f64 (fpextend f32:$src)),
           (COPY_TO_REGCLASS $src, F8RC)>;
+}
 
 // Only seq_cst fences require the heavyweight sync (SYNC 0).
 // All others can use the lightweight sync (SYNC 1).
 // source: http://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html
 // The rule for seq_cst is duplicated to work with both 64 bits and 32 bits
 // versions of Power.
 def : Pat<(atomic_fence (i64 7), (imm)), (SYNC 0)>, Requires<[HasSYNC]>;
 def : Pat<(atomic_fence (i32 7), (imm)), (SYNC 0)>, Requires<[HasSYNC]>;
 def : Pat<(atomic_fence (imm),   (imm)), (SYNC 1)>, Requires<[HasSYNC]>;
 def : Pat<(atomic_fence (imm), (imm)), (MSYNC)>, Requires<[HasOnlyMSYNC]>;
 
+let Predicates = [HasTraditionalFPU] in {
 // Additional FNMSUB patterns: -a*c + b == -(a*c - b)
 def : Pat<(fma (fneg f64:$A), f64:$C, f64:$B),
           (FNMSUB $A, $C, $B)>;
 def : Pat<(fma f64:$A, (fneg f64:$C), f64:$B),
           (FNMSUB $A, $C, $B)>;
 def : Pat<(fma (fneg f32:$A), f32:$C, f32:$B),
           (FNMSUBS $A, $C, $B)>;
 def : Pat<(fma f32:$A, (fneg f32:$C), f32:$B),
           (FNMSUBS $A, $C, $B)>;
 
 // FCOPYSIGN's operand types need not agree.
 def : Pat<(fcopysign f64:$frB, f32:$frA),
           (FCPSGND (COPY_TO_REGCLASS $frA, F8RC), $frB)>;
 def : Pat<(fcopysign f32:$frB, f64:$frA),
           (FCPSGNS (COPY_TO_REGCLASS $frA, F4RC), $frB)>;
+}
 
 include "PPCInstrAltivec.td"
 include "PPCInstrSPE.td"
 include "PPCInstr64Bit.td"
 include "PPCInstrVSX.td"
 include "PPCInstrQPX.td"
 include "PPCInstrHTM.td"
 

 defm : CRNotPat<(i1 (setcc i64:$s1, i64:$s2, SETULE)),
                 (EXTRACT_SUBREG (CMPLD $s1, $s2), sub_gt)>;
 defm : CRNotPat<(i1 (setcc i64:$s1, i64:$s2, SETLE)),
                 (EXTRACT_SUBREG (CMPD $s1, $s2), sub_gt)>;
 defm : CRNotPat<(i1 (setcc i64:$s1, i64:$s2, SETNE)),
                 (EXTRACT_SUBREG (CMPD $s1, $s2), sub_eq)>;
 
 // SETCC for f32.
+let Predicates = [HasTraditionalFPU] in {
 def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETOLT)),
           (EXTRACT_SUBREG (FCMPUS $s1, $s2), sub_lt)>;
 def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETLT)),
           (EXTRACT_SUBREG (FCMPUS $s1, $s2), sub_lt)>;
 def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETOGT)),
           (EXTRACT_SUBREG (FCMPUS $s1, $s2), sub_gt)>;
 def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETGT)),
           (EXTRACT_SUBREG (FCMPUS $s1, $s2), sub_gt)>;

 defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETLE)),
                 (EXTRACT_SUBREG (FCMPUD $s1, $s2), sub_gt)>;
 defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETUNE)),
                 (EXTRACT_SUBREG (FCMPUD $s1, $s2), sub_eq)>;
 defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETNE)),
                 (EXTRACT_SUBREG (FCMPUD $s1, $s2), sub_eq)>;
 defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETO)),
                 (EXTRACT_SUBREG (FCMPUD $s1, $s2), sub_un)>;
+}
+
+let Predicates = [HasSPE] in {
+def SELECT_CC_SPE4 : Pseudo<(outs spe4rc:$dst),
+                            (ins crrc:$cond, spe4rc:$T, spe4rc:$F,
+                            i32imm:$BROPC), "#SELECT_CC_SPE4",
+                            []>;
+def SELECT_CC_SPE  : Pseudo<(outs sperc:$dst),
+                            (ins crrc:$cond, sperc:$T, sperc:$F, i32imm:$BROPC),
+                            "#SELECT_CC_SPE",
+                            []>;
+def SELECT_SPE4  : Pseudo<(outs spe4rc:$dst), (ins crbitrc:$cond,
+                          spe4rc:$T, spe4rc:$F), "#SELECT_F4",
+                          [(set f32:$dst, (select i1:$cond, f32:$T, f32:$F))]>;
+def SELECT_SPE   : Pseudo<(outs sperc:$dst), (ins crbitrc:$cond,
+                          sperc:$T, sperc:$F), "#SELECT_F8",
+                          [(set f64:$dst, (select i1:$cond, f64:$T, f64:$F))]>;
+
+// SETCC for f32.
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETOLT)),
+          (EXTRACT_SUBREG (EFSCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETLT)),
+          (EXTRACT_SUBREG (EFSCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETOGT)),
+          (EXTRACT_SUBREG (EFSCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETGT)),
+          (EXTRACT_SUBREG (EFSCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETOEQ)),
+          (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETEQ)),
+          (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f32:$s1, f32:$s2, SETUO)),
+          (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_un)>;
+
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETUGE)),
+                (EXTRACT_SUBREG (EFSCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETGE)),
+                (EXTRACT_SUBREG (EFSCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETULE)),
+                (EXTRACT_SUBREG (EFSCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETLE)),
+                (EXTRACT_SUBREG (EFSCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETUNE)),
+                (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETNE)),
+                (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f32:$s1, f32:$s2, SETO)),
+                (EXTRACT_SUBREG (EFSCMPEQ $s1, $s2), sub_gt)>;
 
+// SETCC for f64.
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETOLT)),
+          (EXTRACT_SUBREG (EFDCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETLT)),
+          (EXTRACT_SUBREG (EFDCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETOGT)),
+          (EXTRACT_SUBREG (EFDCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETGT)),
+          (EXTRACT_SUBREG (EFDCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETOEQ)),
+          (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETEQ)),
+          (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc f64:$s1, f64:$s2, SETUO)),
+          (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_un)>;
+
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETUGE)),
+                (EXTRACT_SUBREG (EFDCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETGE)),
+                (EXTRACT_SUBREG (EFDCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETULE)),
+                (EXTRACT_SUBREG (EFDCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETLE)),
+                (EXTRACT_SUBREG (EFDCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETUNE)),
+                (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETNE)),
+                (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc f64:$s1, f64:$s2, SETO)),
+                (EXTRACT_SUBREG (EFDCMPEQ $s1, $s2), sub_gt)>;
+
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETOLT)),
+          (EXTRACT_SUBREG (EVFSCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETLT)),
+          (EXTRACT_SUBREG (EVFSCMPLT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETOGT)),
+          (EXTRACT_SUBREG (EVFSCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETGT)),
+          (EXTRACT_SUBREG (EVFSCMPGT $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETOEQ)),
+          (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETEQ)),
+          (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETUO)),
+          (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_un)>;
+
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETUGE)),
+                (EXTRACT_SUBREG (EVFSCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETGE)),
+                (EXTRACT_SUBREG (EVFSCMPLT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETULE)),
+                (EXTRACT_SUBREG (EVFSCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETLE)),
+                (EXTRACT_SUBREG (EVFSCMPGT $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETUNE)),
+                (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETNE)),
+                (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2f32:$s1, v2f32:$s2, SETO)),
+                (EXTRACT_SUBREG (EVFSCMPEQ $s1, $s2), sub_gt)>;
+
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETOLT)),
+          (EXTRACT_SUBREG (EVCMPLTS $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETLT)),
+          (EXTRACT_SUBREG (EVCMPLTS $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETOGT)),
+          (EXTRACT_SUBREG (EVCMPGTS $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETGT)),
+          (EXTRACT_SUBREG (EVCMPGTS $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETOEQ)),
+          (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETEQ)),
+          (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_gt)>;
+def : Pat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETUO)),
+          (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_un)>;
+
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETUGE)),
+                (EXTRACT_SUBREG (EVCMPLTU $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETGE)),
+                (EXTRACT_SUBREG (EVCMPLTS $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETULE)),
+                (EXTRACT_SUBREG (EVCMPGTU $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETLE)),
+                (EXTRACT_SUBREG (EVCMPGTS $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETUNE)),
+                (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETNE)),
+                (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_gt)>;
+defm : CRNotPat<(i1 (setcc v2i32:$s1, v2i32:$s2, SETO)),
+                (EXTRACT_SUBREG (EVCMPEQ $s1, $s2), sub_gt)>;
+}
 // match select on i1 variables:
 def : Pat<(i1 (select i1:$cond, i1:$tval, i1:$fval)),
           (CROR (CRAND        $cond , $tval),
                 (CRAND (crnot $cond), $fval))>;
 
 // match selectcc on i1 variables:
 //   select (lhs == rhs), tval, fval is:
 //   ((lhs == rhs) & tval) | (!(lhs == rhs) & fval)
def : Pat<(i64 (selectcc i1:$lhs, i1:$rhs, i64:$tval, i64:$fval, SETUGE)),
           (SELECT_I8 (CRORC  $lhs, $rhs), $tval, $fval)>;
 def : Pat<(i64 (selectcc i1:$lhs, i1:$rhs, i64:$tval, i64:$fval, SETGT)),
           (SELECT_I8 (CRANDC $rhs, $lhs), $tval, $fval)>;
 def : Pat<(i64 (selectcc i1:$lhs, i1:$rhs, i64:$tval, i64:$fval, SETUGT)),
           (SELECT_I8 (CRANDC $lhs, $rhs), $tval, $fval)>;
 def : Pat<(i64 (selectcc i1:$lhs, i1:$rhs, i64:$tval, i64:$fval, SETNE)),
           (SELECT_I8 (CRXOR $lhs, $rhs), $tval, $fval)>;
 
+let Predicates = [HasTraditionalFPU] in {
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETLT)),
           (SELECT_F4 (CRANDC $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETULT)),
           (SELECT_F4 (CRANDC $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETLE)),
           (SELECT_F4 (CRORC  $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETULE)),
           (SELECT_F4 (CRORC  $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETEQ)),
           (SELECT_F4 (CREQV $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETGE)),
           (SELECT_F4 (CRORC  $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETUGE)),
           (SELECT_F4 (CRORC  $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETGT)),
           (SELECT_F4 (CRANDC $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETUGT)),
           (SELECT_F4 (CRANDC $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETNE)),
           (SELECT_F4 (CRXOR $lhs, $rhs), $tval, $fval)>;
+}
 
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETLT)),
           (SELECT_F8 (CRANDC $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETULT)),
           (SELECT_F8 (CRANDC $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETLE)),
           (SELECT_F8 (CRORC  $lhs, $rhs), $tval, $fval)>;
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETULE)),
           (SELECT_F8 (CRORC  $rhs, $lhs), $tval, $fval)>;
 def : Pat<(f64 (selectcc i1:$lhs, i1:$rhs, f64:$tval, f64:$fval, SETEQ)),
def MTFSFI : XLForm_4<63, 134, (outs crrc:$BF), (ins i32imm:$U, i32imm:$W),
                       "mtfsfi $BF, $U, $W", IIC_IntMFFS>;
 
 def MTFSFIo : XLForm_4<63, 134, (outs crrc:$BF), (ins i32imm:$U, i32imm:$W),
                        "mtfsfi. $BF, $U, $W", IIC_IntMFFS>, isDOT;
 
 def : InstAlias<"mtfsfi $BF, $U", (MTFSFI crrc:$BF, i32imm:$U, 0)>;
 def : InstAlias<"mtfsfi. $BF, $U", (MTFSFIo crrc:$BF, i32imm:$U, 0)>;
 
+let Predicates = [HasTraditionalFPU] in {
 def MTFSF : XFLForm_1<63, 711, (outs),
                       (ins i32imm:$FLM, f8rc:$FRB, i32imm:$L, i32imm:$W),
                       "mtfsf $FLM, $FRB, $L, $W", IIC_IntMFFS, []>;
 def MTFSFo : XFLForm_1<63, 711, (outs),
                        (ins i32imm:$FLM, f8rc:$FRB, i32imm:$L, i32imm:$W),
                        "mtfsf. $FLM, $FRB, $L, $W", IIC_IntMFFS, []>, isDOT;
 
 def : InstAlias<"mtfsf $FLM, $FRB", (MTFSF i32imm:$FLM, f8rc:$FRB, 0, 0)>;
 def : InstAlias<"mtfsf. $FLM, $FRB", (MTFSFo i32imm:$FLM, f8rc:$FRB, 0, 0)>;
+}
 
 def SLBIE : XForm_16b<31, 434, (outs), (ins gprc:$RB),
                         "slbie $RB", IIC_SprSLBIE, []>;
 
 def SLBMTE : XForm_26<31, 402, (outs), (ins gprc:$RS, gprc:$RB),
                     "slbmte $RS, $RB", IIC_SprSLBMTE, []>;
 
 def SLBMFEE : XForm_26<31, 915, (outs gprc:$RT), (ins gprc:$RB),

lib/Target/PowerPC/PPCInstrSPE.td

 //=======-- PPCInstrSPE.td - The PowerPC SPE Extension -*- tablegen -*-=======//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
 // This file describes the Signal Processing Engine extension to
 // the PowerPC instruction set.
 //
 //===----------------------------------------------------------------------===//
 
+class EFXForm_1<bits<11> xo, dag OOL, dag IOL, string asmstr,
+               InstrItinClass itin, list<dag> pattern> :
+               I<4, OOL, IOL, asmstr, itin> {
+  bits<5> RT;
+  bits<5> RA;
+  bits<5> RB;
+
+  let Pattern = pattern;
+
+  let Inst{6-10}  = RT;
+  let Inst{11-15} = RA;
+  let Inst{16-20} = RB;
+  let Inst{21-31} = xo;
+}
+
+def imm32SExt5  : Operand<i32>, ImmLeaf<i32, [{
+  // imm32SExt5 predicate - True if the i32 immediate fits in a 5-bit
+  // sign extended field.  Used by instructions like 'evsplati'.
+  return ((int32_t)Imm >= -16 && (int32_t)Imm <= 15);
+}]>;
+
+class EFXForm_2<bits<11> xo, dag OOL, dag IOL, string asmstr,
+               InstrItinClass itin, list<dag> pattern> :
+               EFXForm_1<xo, OOL, IOL, asmstr, itin, pattern> {
+  let RB = 0;
+}
+
+class EFXForm_2a<bits<11> xo, dag OOL, dag IOL, string asmstr,
+               InstrItinClass itin, list<dag> pattern> :
+               EFXForm_1<xo, OOL, IOL, asmstr, itin, pattern> {
+  let RA = 0;
+}
+
+class EFXForm_3<bits<11> xo, dag OOL, dag IOL, string asmstr,
+              InstrItinClass itin> :
+              I<4, OOL, IOL, asmstr, itin> {
+  bits<3> crD;
+  bits<5> RA;
+  bits<5> RB;
+
+  let Inst{6-8}  = crD;
+  let Inst{9-10}  = 0;
+  let Inst{11-15} = RA;
+  let Inst{16-20} = RB;
+  let Inst{21-31} = xo;
+}
+
 class EVXForm_1<bits<11> xo, dag OOL, dag IOL, string asmstr,
-               InstrItinClass itin> : I<4, OOL, IOL, asmstr, itin> {
+               InstrItinClass itin, list<dag> pattern> :
+               I<4, OOL, IOL, asmstr, itin> {
   bits<5> RT;
   bits<5> RA;
   bits<5> RB;
 
-  let Pattern = [];
-  
+  let Pattern = pattern;
   let Inst{6-10}  = RT;
   let Inst{11-15} = RA;
   let Inst{16-20} = RB;
   let Inst{21-31} = xo;
 }
 
 class EVXForm_2<bits<11> xo, dag OOL, dag IOL, string asmstr,
-               InstrItinClass itin> : EVXForm_1<xo, OOL, IOL, asmstr, itin> {
+               InstrItinClass itin, list<dag> pattern> :
+               EVXForm_1<xo, OOL, IOL, asmstr, itin, pattern> {
   let RB = 0;
 }
 
+class EVXForm_2a<bits<11> xo, dag OOL, dag IOL, string asmstr,
+               InstrItinClass itin, list<dag> pattern> :
+               EVXForm_1<xo, OOL, IOL, asmstr, itin, pattern> {
+  let RA = 0;
+}
+
 class EVXForm_3<bits<11> xo, dag OOL, dag IOL, string asmstr,
-              InstrItinClass itin> : I<4, OOL, IOL, asmstr, itin> {
+              InstrItinClass itin, list<dag> pattern> :
+              I<4, OOL, IOL, asmstr, itin> {
   bits<3> crD;
   bits<5> RA;
   bits<5> RB;
 
-  let Pattern = [];
-  
+  let Pattern = pattern;
   let Inst{6-8}  = crD;
   let Inst{9-10}  = 0;
   let Inst{11-15} = RA;
   let Inst{16-20} = RB;
   let Inst{21-31} = xo;
 }
 
+class EVXForm_4<bits<8> xo, dag OOL, dag IOL, string asmstr,
+               InstrItinClass itin, list<dag> pattern> :
+               I<4, OOL, IOL, asmstr, itin> {
+  bits<3> crD;
+  bits<5> RA;
+  bits<5> RB;
+  bits<5> RT;
+
+  let Pattern = pattern;
+
+  let Inst{6-10}  = RT;
+  let Inst{11-15} = RA;
+  let Inst{16-20} = RB;
+  let Inst{21-28} = xo;
+  let Inst{29-31} = crD;
+}
+
 class EVXForm_D<bits<11> xo, dag OOL, dag IOL, string asmstr,
-               InstrItinClass itin> : I<4, OOL, IOL, asmstr, itin> {
+               InstrItinClass itin, list<dag> pattern> :
+               I<4, OOL, IOL, asmstr, itin> {
   bits<5> RT;
   bits<21> D;
 
-  let Pattern = [];
+  let Pattern = pattern;
 
   let Inst{6-10}  = RT;
   let Inst{20} = D{0};
   let Inst{19} = D{1};
   let Inst{18} = D{2};
   let Inst{17} = D{3};
   let Inst{16} = D{4};
   let Inst{15} = D{5};
   let Inst{14} = D{6};
   let Inst{13} = D{7};
   let Inst{12} = D{8};
   let Inst{11} = D{9};
   let Inst{11-20} = D{0-9};
   let Inst{21-31} = xo;
 }
 
-let Predicates = [HasSPE], isAsmParserOnly = 1 in {
+let DecoderNamespace = "SPE", Predicates = [HasSPE] in {
+
+def BRINC          : EVXForm_1<527, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
+                               "brinc $RT, $RA, $RB", IIC_IntSimple, []>;
+
+// Double-precision floating point
+def EFDABS         : EFXForm_2<740, (outs sperc:$RT), (ins sperc:$RA),
+                                "efdabs $RT, $RA", IIC_FPDGeneral,
+                                [(set f64:$RT, (fabs f64:$RA))]>;
+
+def EFDADD         : EFXForm_1<736, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                                "efdadd $RT, $RA, $RB", IIC_FPAddSub,
+                                [(set f64:$RT, (fadd f64:$RA, f64:$RB))]>;
+
+def EFDCFS         : EFXForm_2a<751, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efdcfs $RT, $RB", IIC_FPDGeneral,
+                                [(set f64:$RT, (fpextend f32:$RB))]>;
+
+def EFDCFSF        : EFXForm_2a<755, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efdcfsf $RT, $RB", IIC_FPDGeneral, []>;
+
+def EFDCFSI        : EFXForm_2a<753, (outs sperc:$RT), (ins gprc:$RB),
+                                "efdcfsi $RT, $RB", IIC_FPDGeneral,
+                                [(set f64:$RT, (sint_to_fp i32:$RB))]>;
+
+def EFDCFSID       : EFXForm_2a<739, (outs sperc:$RT), (ins gprc:$RB),
+                                "efdcfsid $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+def EFDCFUF        : EFXForm_2a<754, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efdcfuf $RT, $RB", IIC_FPDGeneral, []>;
+
+def EFDCFUI        : EFXForm_2a<752, (outs sperc:$RT), (ins gprc:$RB),
+                                "efdcfui $RT, $RB", IIC_FPDGeneral,
+                                [(set f64:$RT, (uint_to_fp i32:$RB))]>;
+
+def EFDCFUID       : EFXForm_2a<738, (outs sperc:$RT), (ins gprc:$RB),
+                                "efdcfuid $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+let isCompare = 1 in {
+def EFDCMPEQ       : EFXForm_3<750, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdcmpeq $crD, $RA, $RB", IIC_FPDGeneral>;
+def EFDCMPGT       : EFXForm_3<748, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdcmpgt $crD, $RA, $RB", IIC_FPDGeneral>;
+def EFDCMPLT       : EFXForm_3<749, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdcmplt $crD, $RA, $RB", IIC_FPDGeneral>;
+}
+
+def EFDCTSF        : EFXForm_2a<759, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efdctsf $RT, $RB", IIC_FPDGeneral, []>;
 
-def EVLDD          : EVXForm_D<769, (outs gprc:$RT), (ins spe8dis:$dst),
-                               "evldd $RT, $dst", IIC_VecFP>;
-def EVLDW          : EVXForm_D<771, (outs gprc:$RT), (ins spe8dis:$dst),
-                               "evldw $RT, $dst", IIC_VecFP>;
-def EVLDH          : EVXForm_D<773, (outs gprc:$RT), (ins spe8dis:$dst),
-                               "evldh $RT, $dst", IIC_VecFP>;
-def EVLHHESPLAT    : EVXForm_D<777, (outs gprc:$RT), (ins spe2dis:$dst),
-                               "evlhhesplat $RT, $dst", IIC_VecFP>;
-def EVLHHOUSPLAT   : EVXForm_D<781, (outs gprc:$RT), (ins spe2dis:$dst),
-                               "evlhhousplat $RT, $dst", IIC_VecFP>;
-def EVLHHOSSPLAT   : EVXForm_D<783, (outs gprc:$RT), (ins spe2dis:$dst),
-                               "evlhhossplat $RT, $dst", IIC_VecFP>;
-def EVLWHE         : EVXForm_D<785, (outs gprc:$RT), (ins spe4dis:$dst),
-                               "evlwhe $RT, $dst", IIC_VecFP>;
-def EVLWHOU        : EVXForm_D<789, (outs gprc:$RT), (ins spe4dis:$dst),
-                               "evlwhou $RT, $dst", IIC_VecFP>;
-def EVLWHOS        : EVXForm_D<791, (outs gprc:$RT), (ins spe4dis:$dst),
-                               "evlwhos $RT, $dst", IIC_VecFP>;
-def EVLWWSPLAT     : EVXForm_D<793, (outs gprc:$RT), (ins spe4dis:$dst),
-                               "evlwwsplat $RT, $dst", IIC_VecFP>;
-def EVLWHSPLAT     : EVXForm_D<797, (outs gprc:$RT), (ins spe4dis:$dst),
-                               "evlwhsplat $RT, $dst", IIC_VecFP>;
-
-def EVSTDD         : EVXForm_D<801, (outs), (ins gprc:$RT, spe8dis:$dst),
-                               "evstdd $RT, $dst", IIC_VecFP>;
-def EVSTDH         : EVXForm_D<805, (outs), (ins gprc:$RT, spe8dis:$dst),
-                               "evstdh $RT, $dst", IIC_VecFP>;
-def EVSTDW         : EVXForm_D<803, (outs), (ins gprc:$RT, spe8dis:$dst),
-                               "evstdw $RT, $dst", IIC_VecFP>;
-def EVSTWHE        : EVXForm_D<817, (outs), (ins gprc:$RT, spe4dis:$dst),
-                               "evstwhe $RT, $dst", IIC_VecFP>;
-def EVSTWHO        : EVXForm_D<821, (outs), (ins gprc:$RT, spe4dis:$dst),
-                               "evstwho $RT, $dst", IIC_VecFP>;
-def EVSTWWE        : EVXForm_D<825, (outs), (ins gprc:$RT, spe4dis:$dst),
-                               "evstwwe $RT, $dst", IIC_VecFP>;
-def EVSTWWO        : EVXForm_D<829, (outs), (ins gprc:$RT, spe4dis:$dst),
-                               "evstwwo $RT, $dst", IIC_VecFP>;
+def EFDCTSI        : EFXForm_2a<757, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctsi $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+def EFDCTSIDZ      : EFXForm_2a<747, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctsidz $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+def EFDCTSIZ       : EFXForm_2a<762, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctsiz $RT, $RB", IIC_FPDGeneral,
+                                [(set i32:$RT, (fp_to_sint f64:$RB))]>;
+
+def EFDCTUF        : EFXForm_2a<758, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efdctuf $RT, $RB", IIC_FPDGeneral, []>;
+
+def EFDCTUI        : EFXForm_2a<756, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctui $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+def EFDCTUIDZ      : EFXForm_2a<746, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctuidz $RT, $RB", IIC_FPDGeneral,
+                                []>;
+
+def EFDCTUIZ       : EFXForm_2a<760, (outs gprc:$RT), (ins sperc:$RB),
+                                "efdctuiz $RT, $RB", IIC_FPDGeneral,
+                                [(set i32:$RT, (fp_to_uint f64:$RB))]>;
+
+def EFDDIV         : EFXForm_1<745, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "efddiv $RT, $RA, $RB", IIC_FPDivD,
+                               [(set f64:$RT, (fdiv f64:$RA, f64:$RB))]>;
+
+def EFDMUL         : EFXForm_1<744, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "efdmul $RT, $RA, $RB", IIC_FPDGeneral,
+                               [(set f64:$RT, (fmul f64:$RA, f64:$RB))]>;
+
+def EFDNABS        : EFXForm_2<741, (outs sperc:$RT), (ins sperc:$RA),
+                               "efdnabs $RT, $RA", IIC_FPDGeneral,
+                               [(set f64:$RT, (fneg (fabs f64:$RA)))]>;
+
+def EFDNEG         : EFXForm_2<742, (outs sperc:$RT), (ins sperc:$RA),
+                               "efdneg $RT, $RA", IIC_FPDGeneral,
+                               [(set f64:$RT, (fneg f64:$RA))]>;
+
+def EFDSUB         : EFXForm_1<737, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "efdsub $RT, $RA, $RB", IIC_FPDGeneral,
+                               [(set f64:$RT, (fsub f64:$RA, f64:$RB))]>;
+
+let isCompare = 1 in {
+def EFDTSTEQ       : EFXForm_3<766, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdtsteq $crD, $RA, $RB", IIC_FPDGeneral>;
+def EFDTSTGT       : EFXForm_3<764, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdtstgt $crD, $RA, $RB", IIC_FPDGeneral>;
+def EFDTSTLT       : EFXForm_3<765, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efdtstlt $crD, $RA, $RB", IIC_FPDGeneral>;
+}
 
-def EVMRA : EVXForm_1<1220, (outs gprc:$RT), (ins gprc:$RA),
-                      "evmra $RT, $RA", IIC_VecFP> {
-  let RB = 0;
+// Single-precision floating point
+def EFSABS         : EFXForm_2<708, (outs spe4rc:$RT), (ins spe4rc:$RA),
+                                "efsabs $RT, $RA", IIC_FPSGeneral,
+                                [(set f32:$RT, (fabs f32:$RA))]>;
+
+def EFSADD         : EFXForm_1<704, (outs spe4rc:$RT), (ins spe4rc:$RA, spe4rc:$RB),
+                                "efsadd $RT, $RA, $RB", IIC_FPAddSub,
+                                [(set f32:$RT, (fadd f32:$RA, f32:$RB))]>;
+
+def EFSCFD         : EFXForm_2a<719, (outs spe4rc:$RT), (ins sperc:$RB),
+                                "efscfd $RT, $RB", IIC_FPSGeneral,
+                                [(set f32:$RT, (fpround f64:$RB))]>;
+
+def EFSCFSF        : EFXForm_2a<723, (outs spe4rc:$RT), (ins spe4rc:$RB),
+                                "efscfsf $RT, $RB", IIC_FPSGeneral, []>;
+
+def EFSCFSI        : EFXForm_2a<721, (outs spe4rc:$RT), (ins gprc:$RB),
+                                "efscfsi $RT, $RB", IIC_FPSGeneral,
+                                [(set f32:$RT, (sint_to_fp i32:$RB))]>;
+
+def EFSCFUF        : EFXForm_2a<722, (outs spe4rc:$RT), (ins spe4rc:$RB),
+                                "efscfuf $RT, $RB", IIC_FPSGeneral, []>;
+
+def EFSCFUI        : EFXForm_2a<720, (outs spe4rc:$RT), (ins gprc:$RB),
+                                "efscfui $RT, $RB", IIC_FPSGeneral,
+                                [(set f32:$RT, (uint_to_fp i32:$RB))]>;
+
+let isCompare = 1 in {
+def EFSCMPEQ       : EFXForm_3<718, (outs crrc:$crD), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efscmpeq $crD, $RA, $RB", IIC_FPCompare>;
+def EFSCMPGT       : EFXForm_3<716, (outs crrc:$crD), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efscmpgt $crD, $RA, $RB", IIC_FPCompare>;
+def EFSCMPLT       : EFXForm_3<717, (outs crrc:$crD), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efscmplt $crD, $RA, $RB", IIC_FPCompare>;
 }
 
-def BRINC          : EVXForm_1<527, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "brinc $RT, $RA, $RB", IIC_VecFP>;
-def EVABS          : EVXForm_2<520, (outs gprc:$RT), (ins gprc:$RA),
-                               "evabs $RT, $RA", IIC_VecFP>;
-
-def EVADDIW        : EVXForm_1<514, (outs gprc:$RT), (ins gprc:$RA, u5imm:$RB),
-                               "evaddiw $RT, $RB, $RA", IIC_VecFP>;
-def EVADDSMIAAW    : EVXForm_2<1225, (outs gprc:$RT), (ins gprc:$RA),
-                               "evaddsmiaaw $RT, $RA", IIC_VecFP>;
-def EVADDSSIAAW    : EVXForm_2<1217, (outs gprc:$RT), (ins gprc:$RA),
-                               "evaddssiaaw $RT, $RA", IIC_VecFP>;
-def EVADDUSIAAW    : EVXForm_2<1216, (outs gprc:$RT), (ins gprc:$RA),
-                               "evaddusiaaw $RT, $RA", IIC_VecFP>;
-def EVADDUMIAAW    : EVXForm_2<1224, (outs gprc:$RT), (ins gprc:$RA),
-                               "evaddumiaaw $RT, $RA", IIC_VecFP>;
-def EVADDW         : EVXForm_1<512, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evaddw $RT, $RA, $RB", IIC_VecFP>;
-
-def EVAND          : EVXForm_1<529, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evand $RT, $RA, $RB", IIC_VecFP>;
-def EVANDC         : EVXForm_1<530, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evandc $RT, $RA, $RB", IIC_VecFP>;
-
-def EVCMPEQ        : EVXForm_3<564, (outs crrc:$crD), (ins gprc:$RA, gprc:$RB),
-                               "evcmpeq $crD, $RA, $RB", IIC_VecFP>;
-def EVCMPGTS       : EVXForm_3<561, (outs crrc:$crD), (ins gprc:$RA, gprc:$RB),
-                               "evcmpgts $crD, $RA, $RB", IIC_VecFP>;
-def EVCMPGTU       : EVXForm_3<560, (outs crrc:$crD), (ins gprc:$RA, gprc:$RB),
-                               "evcmpgtu $crD, $RA, $RB", IIC_VecFP>;
-def EVCMPLTS       : EVXForm_3<563, (outs crrc:$crD), (ins gprc:$RA, gprc:$RB),
-                               "evcmplts $crD, $RA, $RB", IIC_VecFP>;
-def EVCMPLTU       : EVXForm_3<562, (outs crrc:$crD), (ins gprc:$RA, gprc:$RB),
-                               "evcmpltu $crD, $RA, $RB", IIC_VecFP>;
-
-def EVCNTLSW       : EVXForm_2<526, (outs gprc:$RT), (ins gprc:$RA),
-                               "evcntlsw $RT, $RA", IIC_VecFP>;
-def EVCNTLZW       : EVXForm_2<525, (outs gprc:$RT), (ins gprc:$RA),
-                               "evcntlzw $RT, $RA", IIC_VecFP>;
-
-def EVDIVWS        : EVXForm_1<1222, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evdivws $RT, $RA, $RB", IIC_VecFP>;
-def EVDIVWU        : EVXForm_1<1223, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evdivwu $RT, $RA, $RB", IIC_VecFP>;
-
-def EVEQV          : EVXForm_1<537, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "eveqv $RT, $RA, $RB", IIC_VecFP>;
-
-def EVEXTSB        : EVXForm_2<522, (outs gprc:$RT), (ins gprc:$RA),
-                               "evextsb $RT, $RA", IIC_VecFP>;
-def EVEXTSH        : EVXForm_2<523, (outs gprc:$RT), (ins gprc:$RA),
-                               "evextsh $RT, $RA", IIC_VecFP>;
-
-def EVLDDX         : EVXForm_1<768, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlddx $RT, $RA, $RB", IIC_VecFP>;
-def EVLDWX         : EVXForm_1<770, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evldwx $RT, $RA, $RB", IIC_VecFP>;
-def EVLDHX         : EVXForm_1<772, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evldhx $RT, $RA, $RB", IIC_VecFP>;
-def EVLHHESPLATX   : EVXForm_1<776, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlhhesplatx $RT, $RA, $RB", IIC_VecFP>;
-def EVLHHOUSPLATX  : EVXForm_1<780, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlhhousplatx $RT, $RA, $RB", IIC_VecFP>;
-def EVLHHOSSPLATX  : EVXForm_1<782, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlhhossplatx $RT, $RA, $RB", IIC_VecFP>;
-def EVLWHEX        : EVXForm_1<784, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlwhex $RT, $RA, $RB", IIC_VecFP>;
-def EVLWHOUX       : EVXForm_1<788, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlwhoux $RT, $RA, $RB", IIC_VecFP>;
-def EVLWHOSX       : EVXForm_1<790, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlwhosx $RT, $RA, $RB", IIC_VecFP>;
-def EVLWWSPLATX    : EVXForm_1<792, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlwwsplatx $RT, $RA, $RB", IIC_VecFP>;
-def EVLWHSPLATX    : EVXForm_1<796, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evlwhsplatx $RT, $RA, $RB", IIC_VecFP>;
-
-def EVMERGEHI      : EVXForm_1<556, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmergehi $RT, $RA, $RB", IIC_VecFP>;
-def EVMERGELO      : EVXForm_1<557, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmergelo $RT, $RA, $RB", IIC_VecFP>;
-def EVMERGEHILO    : EVXForm_1<558, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmergehilo $RT, $RA, $RB", IIC_VecFP>;
-def EVMERGELOHI    : EVXForm_1<559, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmergelohi $RT, $RA, $RB", IIC_VecFP>;
-
-def EVMHEGSMFAA    : EVXForm_1<1323, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegsmfaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEGSMFAN    : EVXForm_1<1451, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegsmfan $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEGSMIAA    : EVXForm_1<1321, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegsmiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEGSMIAN    : EVXForm_1<1449, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegsmian $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEGUMIAA    : EVXForm_1<1320, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegumiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEGUMIAN    : EVXForm_1<1448, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhegumian $RT, $RA, $RB", IIC_VecFP>;
-
-def EVMHESMF       : EVXForm_1<1035, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmf $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMFA      : EVXForm_1<1067, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMFAAW    : EVXForm_1<1291, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmfaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMFANW    : EVXForm_1<1419, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmfanw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMI       : EVXForm_1<1033, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmi $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMIA      : EVXForm_1<1065, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmia $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMIAAW    : EVXForm_1<1289, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESMIANW    : EVXForm_1<1417, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhesmianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSF       : EVXForm_1<1027, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessf $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSFA      : EVXForm_1<1059, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSFAAW    : EVXForm_1<1283, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessfaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSFANW    : EVXForm_1<1411, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessfanw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSIAAW    : EVXForm_1<1281, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHESSIANW    : EVXForm_1<1409, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhessianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUMI       : EVXForm_1<1032, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheumi $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUMIA      : EVXForm_1<1064, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheumia $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUMIAAW    : EVXForm_1<1288, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheumiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUMIANW    : EVXForm_1<1416, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheumianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUSIAAW    : EVXForm_1<1280, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheusiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHEUSIANW    : EVXForm_1<1408, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmheusianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGSMFAA    : EVXForm_1<1327, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogsmfaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGSMFAN    : EVXForm_1<1455, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogsmfan $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGSMIAA    : EVXForm_1<1325, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogsmiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGSMIAN    : EVXForm_1<1453, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogsmian $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGUMIAA    : EVXForm_1<1324, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogumiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOGUMIAN    : EVXForm_1<1452, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhogumian $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMF       : EVXForm_1<1039, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmf $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMFA      : EVXForm_1<1071, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMFAAW    : EVXForm_1<1295, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmfaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMFANW    : EVXForm_1<1423, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmfanw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMI       : EVXForm_1<1037, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmi $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMIA      : EVXForm_1<1069, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmia $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMIAAW    : EVXForm_1<1293, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSMIANW    : EVXForm_1<1421, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhosmianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSF       : EVXForm_1<1031, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossf $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSFA      : EVXForm_1<1063, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSFAAW    : EVXForm_1<1287, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossfaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSFANW    : EVXForm_1<1415, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossfanw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSIAAW    : EVXForm_1<1285, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOSSIANW    : EVXForm_1<1413, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhossianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUMI       : EVXForm_1<1036, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhoumi $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUMIA      : EVXForm_1<1068, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhoumia $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUMIAAW    : EVXForm_1<1292, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhoumiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUMIANW    : EVXForm_1<1420, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhoumianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUSIAAW    : EVXForm_1<1284, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhousiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMHOUSIANW    : EVXForm_1<1412, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmhousianw $RT, $RA, $RB", IIC_VecFP>;
-
-
-def EVMWHSMF       : EVXForm_1<1103, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhsmf $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHSMFA      : EVXForm_1<1135, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhsmfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHSMI       : EVXForm_1<1101, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhsmi $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHSMIA      : EVXForm_1<1133, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhsmia $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHSSF       : EVXForm_1<1095, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhssf $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHSSFA      : EVXForm_1<1127, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhssfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHUMI       : EVXForm_1<1100, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhumi $RT, $RA, $RB", IIC_VecFP>;
-def EVMWHUMIA      : EVXForm_1<1132, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwhumia $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLSMIAAW    : EVXForm_1<1353, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlsmiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLSMIANW    : EVXForm_1<1481, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlsmianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLSSIAAW    : EVXForm_1<1345, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlssiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLSSIANW    : EVXForm_1<1473, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlssianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUMI       : EVXForm_1<1096, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlumi $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUMIA      : EVXForm_1<1128, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlumia $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUMIAAW    : EVXForm_1<1352, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlumiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUMIANW    : EVXForm_1<1480, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlumianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUSIAAW    : EVXForm_1<1344, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlusiaaw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWLUSIANW    : EVXForm_1<1472, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwlusianw $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMF        : EVXForm_1<1115, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmf $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMFA       : EVXForm_1<1147, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMFAA      : EVXForm_1<1371, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmfaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMFAN      : EVXForm_1<1499, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmfan $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMI        : EVXForm_1<1113, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmi $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMIA       : EVXForm_1<1145, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmia $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMIAA      : EVXForm_1<1369, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSMIAN      : EVXForm_1<1497, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwsmian $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSSF        : EVXForm_1<1107, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwssf $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSSFA       : EVXForm_1<1139, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwssfa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSSFAA      : EVXForm_1<1363, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwssfaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWSSFAN      : EVXForm_1<1491, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwssfan $RT, $RA, $RB", IIC_VecFP>;
-def EVMWUMI        : EVXForm_1<1112, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwumi $RT, $RA, $RB", IIC_VecFP>;
-def EVMWUMIA       : EVXForm_1<1144, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwumia $RT, $RA, $RB", IIC_VecFP>;
-def EVMWUMIAA      : EVXForm_1<1368, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwumiaa $RT, $RA, $RB", IIC_VecFP>;
-def EVMWUMIAN      : EVXForm_1<1496, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evmwumian $RT, $RA, $RB", IIC_VecFP>;
-
-
-def EVNAND         : EVXForm_1<542, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evnand $RT, $RA, $RB", IIC_VecFP>;
-
-def EVNEG          : EVXForm_2<521, (outs gprc:$RT), (ins gprc:$RA),
-                               "evneg $RT, $RA", IIC_VecFP>;
-
-def EVNOR          : EVXForm_1<536, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evnor $RT, $RA, $RB", IIC_VecFP>;
-def EVOR           : EVXForm_1<535, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evor $RT, $RA, $RB", IIC_VecFP>;
-def EVORC          : EVXForm_1<539, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evorc $RT, $RA, $RB", IIC_VecFP>;
-
-def EVRLWI         : EVXForm_1<554, (outs gprc:$RT), (ins gprc:$RA, u5imm:$RB),
-                               "evrlwi $RT, $RA, $RB", IIC_VecFP>;
-def EVRLW          : EVXForm_1<552, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evrlw $RT, $RA, $RB", IIC_VecFP>;
-
-def EVRNDW         : EVXForm_2<524, (outs gprc:$RT), (ins gprc:$RA),
-                               "evrndw $RT, $RA", IIC_VecFP>;
-
-def EVSLWI         : EVXForm_1<550, (outs gprc:$RT), (ins gprc:$RA, u5imm:$RB),
-                               "evslwi $RT, $RA, $RB", IIC_VecFP>;
-def EVSLW          : EVXForm_1<548, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evslw $RT, $RA, $RB", IIC_VecFP>;
-
-def EVSPLATFI      : EVXForm_2<555, (outs gprc:$RT), (ins i32imm:$RA),
-                               "evsplatfi $RT, $RA", IIC_VecFP>;
-def EVSPLATI       : EVXForm_2<553, (outs gprc:$RT), (ins i32imm:$RA),
-                               "evsplati $RT, $RA", IIC_VecFP>;
-
-def EVSRWIS        : EVXForm_1<547, (outs gprc:$RT), (ins gprc:$RA, u5imm:$RB),
-                               "evsrwis $RT, $RA, $RB", IIC_VecFP>;
-def EVSRWIU        : EVXForm_1<546, (outs gprc:$RT), (ins gprc:$RA, u5imm:$RB),
-                               "evsrwiu $RT, $RA, $RB", IIC_VecFP>;
-def EVSRWS         : EVXForm_1<545, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evsrws $RT, $RA, $RB", IIC_VecFP>;
-def EVSRWU         : EVXForm_1<544, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evsrwu $RT, $RA, $RB", IIC_VecFP>;
-
-def EVSTDDX        : EVXForm_1<800, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstddx $RT, $RA, $RB", IIC_VecFP>;
-def EVSTDHX        : EVXForm_1<804, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstdhx $RT, $RA, $RB", IIC_VecFP>;
-def EVSTDWX        : EVXForm_1<802, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstdwx $RT, $RA, $RB", IIC_VecFP>;
-def EVSTWHEX       : EVXForm_1<816, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstwhex $RT, $RA, $RB", IIC_VecFP>;
-def EVSTWHOX       : EVXForm_1<820, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstwhox $RT, $RA, $RB", IIC_VecFP>;
-def EVSTWWEX       : EVXForm_1<824, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstwwex $RT, $RA, $RB", IIC_VecFP>;
-def EVSTWWOX       : EVXForm_1<828, (outs), (ins gprc:$RT, gprc:$RA, gprc:$RB),
-                               "evstwwox $RT, $RA, $RB", IIC_VecFP>;
-
-def EVSUBFSSIAAW   : EVXForm_2<1219, (outs gprc:$RT), (ins gprc:$RA),
-                               "evsubfssiaaw $RT, $RA", IIC_VecFP>;
-def EVSUBFSMIAAW   : EVXForm_2<1227, (outs gprc:$RT), (ins gprc:$RA),
-                               "evsubfsmiaaw $RT, $RA", IIC_VecFP>;
-def EVSUBFUMIAAW   : EVXForm_2<1226, (outs gprc:$RT), (ins gprc:$RA),
-                               "evsubfumiaaw $RT, $RA", IIC_VecFP>;
-def EVSUBFUSIAAW   : EVXForm_2<1218, (outs gprc:$RT), (ins gprc:$RA),
-                               "evsubfusiaaw $RT, $RA", IIC_VecFP>;
-def EVSUBFW        : EVXForm_1<516, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evsubfw $RT, $RA, $RB", IIC_VecFP>;
-def EVSUBIFW       : EVXForm_1<518, (outs gprc:$RT), (ins u5imm:$RA, gprc:$RB),
-                               "evsubifw $RT, $RA, $RB", IIC_VecFP>;
-def EVXOR          : EVXForm_1<534, (outs gprc:$RT), (ins gprc:$RA, gprc:$RB),
-                               "evxor $RT, $RA, $RB", IIC_VecFP>;
+def EFSCTSF        : EFXForm_2a<727, (outs spe4rc:$RT), (ins spe4rc:$RB),
+                                "efsctsf $RT, $RB", IIC_FPSGeneral, []>;
+
+def EFSCTSI        : EFXForm_2a<725, (outs gprc:$RT), (ins spe4rc:$RB),
+                                "efsctsi $RT, $RB", IIC_FPSGeneral,
+                                [(set i32:$RT, (fp_to_sint f32:$RB))]>;
+
+def EFSCTSIZ       : EFXForm_2a<730, (outs gprc:$RT), (ins spe4rc:$RB),
+                                "efsctsiz $RT, $RB", IIC_FPSGeneral,
+                                []>;
+
+def EFSCTUF        : EFXForm_2a<726, (outs sperc:$RT), (ins spe4rc:$RB),
+                                "efsctuf $RT, $RB", IIC_FPSGeneral, []>;
+
+def EFSCTUI        : EFXForm_2a<724, (outs gprc:$RT), (ins spe4rc:$RB),
+                                "efsctui $RT, $RB", IIC_FPSGeneral,
+                                [(set i32:$RT, (fp_to_uint f32:$RB))]>;
+
+def EFSCTUIZ       : EFXForm_2a<728, (outs gprc:$RT), (ins spe4rc:$RB),
+                                "efsctuiz $RT, $RB", IIC_FPSGeneral,
+                                []>;
+
+def EFSDIV         : EFXForm_1<713, (outs spe4rc:$RT), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efsdiv $RT, $RA, $RB", IIC_FPDivD,
+                               [(set f32:$RT, (fdiv f32:$RA, f32:$RB))]>;
+
+def EFSMUL         : EFXForm_1<712, (outs spe4rc:$RT), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efsmul $RT, $RA, $RB", IIC_FPGeneral,
+                               [(set f32:$RT, (fmul f32:$RA, f32:$RB))]>;
+
+def EFSNABS        : EFXForm_2<709, (outs spe4rc:$RT), (ins spe4rc:$RA),
+                               "efsnabs $RT, $RA", IIC_FPGeneral,
+                               [(set f32:$RT, (fneg (fabs f32:$RA)))]>;
+
+def EFSNEG         : EFXForm_2<710, (outs spe4rc:$RT), (ins spe4rc:$RA),
+                               "efsneg $RT, $RA", IIC_FPGeneral,
+                               [(set f32:$RT, (fneg f32:$RA))]>;
+
+def EFSSUB         : EFXForm_1<705, (outs spe4rc:$RT), (ins spe4rc:$RA, spe4rc:$RB),
+                               "efssub $RT, $RA, $RB", IIC_FPSGeneral,
+                               [(set f32:$RT, (fsub f32:$RA, f32:$RB))]>;
+
+let isCompare = 1 in {
+def EFSTSTEQ       : EFXForm_3<734, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efststeq $crD, $RA, $RB", IIC_FPCompare>;
+def EFSTSTGT       : EFXForm_3<732, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efststgt $crD, $RA, $RB", IIC_FPCompare>;
+def EFSTSTLT       : EFXForm_3<733, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "efststlt $crD, $RA, $RB", IIC_FPCompare>;
+}
+
+// SPE Vector operations
+
+def EVABS          : EVXForm_2<520, (outs sperc:$RT), (ins sperc:$RA),
+                               "evabs $RT, $RA", IIC_VecGeneral,
+                               [(set v2i32:$RT, (abs v2i32:$RA))]>;
+
+def EVADDIW        : EVXForm_1<514, (outs sperc:$RT), (ins sperc:$RA, u5imm:$RB),
+                               "evaddiw $RT, $RB, $RA", IIC_VecGeneral, []>;
+def EVADDSMIAAW    : EVXForm_2<1225, (outs sperc:$RT), (ins sperc:$RA),
+                               "evaddsmiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVADDSSIAAW    : EVXForm_2<1217, (outs sperc:$RT), (ins sperc:$RA),
+                               "evaddssiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVADDUSIAAW    : EVXForm_2<1216, (outs sperc:$RT), (ins sperc:$RA),
+                               "evaddusiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVADDUMIAAW    : EVXForm_2<1224, (outs sperc:$RT), (ins sperc:$RA),
+                               "evaddumiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVADDW         : EVXForm_1<512, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evaddw $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (add v2i32:$RA, v2i32:$RB))]>;
+
+def EVAND          : EVXForm_1<529, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evand $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (and v2i32:$RA, v2i32:$RB))]>;
+def EVANDC         : EVXForm_1<530, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evandc $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (and v2i32:$RA,
+                                (xor (v2i32 immAllOnesV), v2i32:$RB)))]>;
+
+let isCompare = 1 in {
+def EVCMPEQ        : EVXForm_3<564, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evcmpeq $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVCMPGTS       : EVXForm_3<561, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evcmpgts $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVCMPGTU       : EVXForm_3<560, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evcmpgtu $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVCMPLTS       : EVXForm_3<563, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evcmplts $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVCMPLTU       : EVXForm_3<562, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evcmpltu $crD, $RA, $RB", IIC_VecGeneral, []>;
+}
+
+def EVCNTLSW       : EVXForm_2<526, (outs sperc:$RT), (ins sperc:$RA),
+                               "evcntlsw $RT, $RA", IIC_VecGeneral, []>;
+def EVCNTLZW       : EVXForm_2<525, (outs sperc:$RT), (ins sperc:$RA),
+                               "evcntlzw $RT, $RA", IIC_VecGeneral,
+                               [(set v2i32:$RT, (ctlz v2i32:$RA))]>;
+
+def EVDIVWS        : EVXForm_1<1222, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evdivws $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2i32:$RT, (sdiv v2i32:$RA, v2i32:$RB))]>;
+def EVDIVWU        : EVXForm_1<1223, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evdivwu $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2i32:$RT, (udiv v2i32:$RA, v2i32:$RB))]>;
+
+def EVEQV          : EVXForm_1<537, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "eveqv $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT,
+                                 (not (xor v2i32:$RA, v2i32:$RB)))]>;
+
+def EVEXTSB        : EVXForm_2<522, (outs sperc:$RT), (ins sperc:$RA),
+                               "evextsb $RT, $RA", IIC_VecGeneral,
+                               []>;
+def EVEXTSH        : EVXForm_2<523, (outs sperc:$RT), (ins sperc:$RA),
+                               "evextsh $RT, $RA", IIC_VecGeneral,
+                               []>;
+
+def EVFSABS        : EVXForm_2<644, (outs sperc:$RT), (ins sperc:$RA),
+                               "evfsabs $RT, $RA", IIC_VecGeneral,
+                               [(set v2f32:$RT, (fabs v2f32:$RA))]>;
+def EVFSADD        : EVXForm_1<640, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evfsadd $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2f32:$RT, (fadd v2f32:$RA, v2f32:$RB))]>;
+def EVFSCFSF       : EVXForm_2a<659, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfscfsf $RT, $RB", IIC_VecComplex, []>;
+def EVFSCFSI       : EVXForm_2a<657, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfscfsi $RT, $RB", IIC_VecComplex,
+                                [(set v2f32:$RT, (sint_to_fp v2i32:$RB))]>;
+def EVFSCFUF       : EVXForm_2a<658, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfscfuf $RT, $RB", IIC_VecComplex, []>;
+def EVFSCFUI       : EVXForm_2a<650, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfscfui $RT, $RB", IIC_VecComplex,
+                                [(set v2f32:$RT, (uint_to_fp v2i32:$RB))]>;
+let isCompare = 1 in {
+def EVFSCMPEQ      : EVXForm_3<654, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfscmpeq $crD, $RA, $RB", IIC_FPSGeneral, []>;
+def EVFSCMPGT      : EVXForm_3<652, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfscmpgt $crD, $RA, $RB", IIC_FPSGeneral, []>;
+def EVFSCMPLT      : EVXForm_3<653, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfscmplt $crD, $RA, $RB", IIC_FPSGeneral, []>;
+}
+
+def EVFSCTSF        : EVXForm_2a<663, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctsf $RT, $RB", IIC_VecComplex, []>;
+def EVFSCTSI        : EVXForm_2a<661, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctsi $RT, $RB", IIC_VecComplex,
+                                [(set v2i32:$RT, (fp_to_sint v2f32:$RB))]>;
+def EVFSCTSIZ       : EVXForm_2a<666, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctsiz $RT, $RB", IIC_VecComplex,
+                                []>;
+def EVFSCTUF        : EVXForm_2a<662, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctsf $RT, $RB", IIC_VecComplex, []>;
+def EVFSCTUI        : EVXForm_2a<660, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctui $RT, $RB", IIC_VecComplex,
+                                [(set v2i32:$RT, (fp_to_uint v2f32:$RB))]>;
+def EVFSCTUIZ       : EVXForm_2a<664, (outs sperc:$RT), (ins sperc:$RB),
+                                "evfsctsiz $RT, $RB", IIC_VecComplex,
+                                []>;
+def EVFSDIV         : EVXForm_1<649, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evfsdiv $RT, $RA, $RB", IIC_FPDivD,
+                               [(set v2f32:$RT, (fdiv v2f32:$RA, v2f32:$RB))]>;
+def EVFSMUL         : EVXForm_1<648, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evfsmul $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2f32:$RT, (fmul v2f32:$RA, v2f32:$RB))]>;
+def EVFSNABS        : EVXForm_2<645, (outs sperc:$RT), (ins sperc:$RA),
+                               "evfsnabs $RT, $RA", IIC_VecGeneral,
+                               [(set v2f32:$RT, (fneg (fabs v2f32:$RA)))]>;
+def EVFSNEG         : EVXForm_2<646, (outs sperc:$RT), (ins sperc:$RA),
+                               "evfsneg $RT, $RA", IIC_VecGeneral,
+                               [(set v2f32:$RT, (fneg v2f32:$RA))]>;
+def EVFSSUB         : EVXForm_1<641, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evfssub $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2f32:$RT, (fsub v2f32:$RA, v2f32:$RB))]>;
+
+let isCompare = 1 in {
+def EVFSTSTEQ       : EVXForm_3<670, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfststeq $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVFSTSTGT       : EVXForm_3<668, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfststgt $crD, $RA, $RB", IIC_VecGeneral, []>;
+def EVFSTSTLT       : EVXForm_3<669, (outs crrc:$crD), (ins sperc:$RA, sperc:$RB),
+                               "evfststlt $crD, $RA, $RB", IIC_VecGeneral, []>;
+}
+
+def EVLDD          : EVXForm_D<769, (outs sperc:$RT), (ins spe8dis:$dst),
+                               "evldd $RT, $dst", IIC_LdStLoad,
+                               [(set f64:$RT, (load ixaddr:$dst))]>;
+def EVLDDX         : EVXForm_1<768, (outs sperc:$RT), (ins memrr:$src),
+                               "evlddx $RT, $src", IIC_LdStLoad,
+                               [(set f64:$RT, (load xaddr:$src))]>;
+def EVLDH          : EVXForm_D<773, (outs sperc:$RT), (ins spe8dis:$dst),
+                               "evldh $RT, $dst", IIC_LdStLoad, []>;
+def EVLDHX         : EVXForm_1<772, (outs sperc:$RT), (ins memrr:$src),
+                               "evldhx $RT, $src", IIC_LdStLoad, []>;
+def EVLDW          : EVXForm_D<771, (outs sperc:$RT), (ins spe8dis:$dst),
+                               "evldw $RT, $dst", IIC_LdStLoad,
+                               [(set v2i32:$RT, (load ixaddr:$dst))]>;
+def EVLDWX         : EVXForm_1<770, (outs sperc:$RT), (ins memrr:$src),
+                               "evldwx $RT, $src", IIC_LdStLoad,
+                               [(set v2i32:$RT, (load xaddr:$src))]>;
+let isAsmParserOnly = 1 in {
+def EVLDWFS        : EVXForm_D<771, (outs sperc:$RT), (ins spe8dis:$dst),
+                               "evldw $RT, $dst", IIC_LdStLoad,
+                               [(set v2f32:$RT, (load ixaddr:$dst))]>;
+def EVLDWXFS       : EVXForm_1<770, (outs sperc:$RT), (ins memrr:$src),
+                               "evldwx $RT, $src", IIC_LdStLoad,
+                               [(set v2f32:$RT, (load xaddr:$src))]>;
+}
+def EVLHHESPLAT    : EVXForm_D<777, (outs sperc:$RT), (ins spe2dis:$dst),
+                               "evlhhesplat $RT, $dst", IIC_LdStLoad, []>;
+def EVLHHESPLATX   : EVXForm_1<776, (outs sperc:$RT), (ins memrr:$src),
+                               "evlhhesplatx $RT, $src", IIC_LdStLoad, []>;
+def EVLHHOUSPLAT   : EVXForm_D<781, (outs sperc:$RT), (ins spe2dis:$dst),
+                               "evlhhousplat $RT, $dst", IIC_LdStLoad, []>;
+def EVLHHOUSPLATX  : EVXForm_1<780, (outs sperc:$RT), (ins memrr:$src),
+                               "evlhhousplatx $RT, $src", IIC_LdStLoad, []>;
+def EVLHHOSSPLAT   : EVXForm_D<783, (outs sperc:$RT), (ins spe2dis:$dst),
+                               "evlhhossplat $RT, $dst", IIC_LdStLoad, []>;
+def EVLHHOSSPLATX  : EVXForm_1<782, (outs sperc:$RT), (ins memrr:$src),
+                               "evlhhossplatx $RT, $src", IIC_LdStLoad, []>;
+def EVLWHE         : EVXForm_D<785, (outs sperc:$RT), (ins spe4dis:$dst),
+                               "evlwhe $RT, $dst", IIC_LdStLoad, []>;
+def EVLWHEX        : EVXForm_1<784, (outs sperc:$RT), (ins memrr:$src),
+                               "evlwhex $RT, $src", IIC_LdStLoad, []>;
+def EVLWHOS        : EVXForm_D<791, (outs sperc:$RT), (ins spe4dis:$dst),
+                               "evlwhos $RT, $dst", IIC_LdStLoad, []>;
+def EVLWHOSX       : EVXForm_1<790, (outs sperc:$RT), (ins memrr:$src),
+                               "evlwhosx $RT, $src", IIC_LdStLoad, []>;
+def EVLWHOU        : EVXForm_D<789, (outs sperc:$RT), (ins spe4dis:$dst),
+                               "evlwhou $RT, $dst", IIC_LdStLoad, []>;
+def EVLWHOUX       : EVXForm_1<788, (outs sperc:$RT), (ins memrr:$src),
+                               "evlwhoux $RT, $src", IIC_LdStLoad, []>;
+def EVLWHSPLAT     : EVXForm_D<797, (outs sperc:$RT), (ins spe4dis:$dst),
+                               "evlwhsplat $RT, $dst", IIC_LdStLoad, []>;
+def EVLWHSPLATX    : EVXForm_1<796, (outs sperc:$RT), (ins memrr:$src),
+                               "evlwhsplatx $RT, $src", IIC_LdStLoad, []>;
+def EVLWWSPLAT     : EVXForm_D<793, (outs sperc:$RT), (ins spe4dis:$dst),
+                               "evlwwsplat $RT, $dst", IIC_LdStLoad, []>;
+def EVLWWSPLATX    : EVXForm_1<792, (outs sperc:$RT), (ins memrr:$src),
+                               "evlwwsplatx $RT, $src", IIC_LdStLoad, []>;
+
+def EVMERGEHI      : EVXForm_1<556, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmergehi $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVMERGELO      : EVXForm_1<557, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmergelo $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVMERGEHILO    : EVXForm_1<558, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmergehilo $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVMERGELOHI    : EVXForm_1<559, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmergelohi $RT, $RA, $RB", IIC_VecGeneral, []>;
+
+def EVMHEGSMFAA    : EVXForm_1<1323, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegsmfaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEGSMFAN    : EVXForm_1<1451, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegsmfan $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEGSMIAA    : EVXForm_1<1321, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegsmiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEGSMIAN    : EVXForm_1<1449, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegsmian $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEGUMIAA    : EVXForm_1<1320, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegumiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEGUMIAN    : EVXForm_1<1448, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhegumian $RT, $RA, $RB", IIC_VecComplex, []>;
+
+def EVMHESMF       : EVXForm_1<1035, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMFA      : EVXForm_1<1067, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMFAAW    : EVXForm_1<1291, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmfaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMFANW    : EVXForm_1<1419, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmfanw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMI       : EVXForm_1<1033, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMIA      : EVXForm_1<1065, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMIAAW    : EVXForm_1<1289, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESMIANW    : EVXForm_1<1417, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhesmianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSF       : EVXForm_1<1027, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSFA      : EVXForm_1<1059, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSFAAW    : EVXForm_1<1283, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessfaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSFANW    : EVXForm_1<1411, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessfanw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSIAAW    : EVXForm_1<1281, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHESSIANW    : EVXForm_1<1409, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhessianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUMI       : EVXForm_1<1032, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheumi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUMIA      : EVXForm_1<1064, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheumia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUMIAAW    : EVXForm_1<1288, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheumiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUMIANW    : EVXForm_1<1416, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheumianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUSIAAW    : EVXForm_1<1280, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheusiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHEUSIANW    : EVXForm_1<1408, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmheusianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGSMFAA    : EVXForm_1<1327, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogsmfaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGSMFAN    : EVXForm_1<1455, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogsmfan $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGSMIAA    : EVXForm_1<1325, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogsmiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGSMIAN    : EVXForm_1<1453, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogsmian $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGUMIAA    : EVXForm_1<1324, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogumiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOGUMIAN    : EVXForm_1<1452, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhogumian $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMF       : EVXForm_1<1039, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMFA      : EVXForm_1<1071, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMFAAW    : EVXForm_1<1295, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmfaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMFANW    : EVXForm_1<1423, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmfanw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMI       : EVXForm_1<1037, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMIA      : EVXForm_1<1069, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMIAAW    : EVXForm_1<1293, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSMIANW    : EVXForm_1<1421, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhosmianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSF       : EVXForm_1<1031, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSFA      : EVXForm_1<1063, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSFAAW    : EVXForm_1<1287, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossfaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSFANW    : EVXForm_1<1415, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossfanw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSIAAW    : EVXForm_1<1285, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOSSIANW    : EVXForm_1<1413, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhossianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUMI       : EVXForm_1<1036, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhoumi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUMIA      : EVXForm_1<1068, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhoumia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUMIAAW    : EVXForm_1<1292, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhoumiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUMIANW    : EVXForm_1<1420, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhoumianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUSIAAW    : EVXForm_1<1284, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhousiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMHOUSIANW    : EVXForm_1<1412, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmhousianw $RT, $RA, $RB", IIC_VecComplex, []>;
+
+def EVMRA          : EVXForm_2<1220, (outs sperc:$RT), (ins sperc:$RA),
+                               "evmra $RT, $RA", IIC_VecComplex, []>;
+
+def EVMWHSMF       : EVXForm_1<1103, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhsmf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHSMFA      : EVXForm_1<1135, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhsmfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHSMI       : EVXForm_1<1101, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhsmi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHSMIA      : EVXForm_1<1133, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhsmia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHSSF       : EVXForm_1<1095, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhssf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHSSFA      : EVXForm_1<1127, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhssfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHUMI       : EVXForm_1<1100, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhumi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWHUMIA      : EVXForm_1<1132, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwhumia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLSMIAAW    : EVXForm_1<1353, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlsmiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLSMIANW    : EVXForm_1<1481, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlsmianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLSSIAAW    : EVXForm_1<1345, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlssiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLSSIANW    : EVXForm_1<1473, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlssianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLUMI       : EVXForm_1<1096, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlumi $RT, $RA, $RB", IIC_VecComplex,
+                               [(set v2i32:$RT, (mul v2i32:$RA, v2i32:$RB))]>;
+def EVMWLUMIA      : EVXForm_1<1128, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlumia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLUMIAAW    : EVXForm_1<1352, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlumiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLUMIANW    : EVXForm_1<1480, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlumianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLUSIAAW    : EVXForm_1<1344, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlusiaaw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWLUSIANW    : EVXForm_1<1472, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwlusianw $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMF        : EVXForm_1<1115, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMFA       : EVXForm_1<1147, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMFAA      : EVXForm_1<1371, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmfaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMFAN      : EVXForm_1<1499, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmfan $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMI        : EVXForm_1<1113, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMIA       : EVXForm_1<1145, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMIAA      : EVXForm_1<1369, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSMIAN      : EVXForm_1<1497, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwsmian $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSSF        : EVXForm_1<1107, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwssf $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSSFA       : EVXForm_1<1139, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwssfa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSSFAA      : EVXForm_1<1363, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwssfaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWSSFAN      : EVXForm_1<1491, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwssfan $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWUMI        : EVXForm_1<1112, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwumi $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWUMIA       : EVXForm_1<1144, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwumia $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWUMIAA      : EVXForm_1<1368, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwumiaa $RT, $RA, $RB", IIC_VecComplex, []>;
+def EVMWUMIAN      : EVXForm_1<1496, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evmwumian $RT, $RA, $RB", IIC_VecComplex, []>;
+
+
+def EVNAND         : EVXForm_1<542, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evnand $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, 
+                                  (xor (v2i32 immAllOnesV),
+                                    (and v2i32:$RA, v2i32:$RB)))]>;
+
+def EVNEG          : EVXForm_2<521, (outs sperc:$RT), (ins sperc:$RA),
+                               "evneg $RT, $RA", IIC_VecGeneral,
+                               [(set v2i32:$RT,
+                                  (sub (v2i32 immAllZerosV), v2i32:$RA))]>;
+
+def EVNOR          : EVXForm_1<536, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evnor $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, 
+                                  (xor (v2i32 immAllOnesV),
+                                    (or v2i32:$RA, v2i32:$RB)))]>;
+def EVOR           : EVXForm_1<535, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evor $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (or v2i32:$RA, v2i32:$RB))]>;
+def EVORC          : EVXForm_1<539, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evorc $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (or v2i32:$RA,
+                                (xor (v2i32 immAllOnesV), v2i32:$RB)))]>;
+
+def EVRLWI         : EVXForm_1<554, (outs sperc:$RT), (ins sperc:$RA, u5imm:$RB),
+                               "evrlwi $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVRLW          : EVXForm_1<552, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evrlw $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (rotl v2i32:$RA, v2i32:$RB))]>;
+
+def EVRNDW         : EVXForm_2<524, (outs sperc:$RT), (ins sperc:$RA),
+                               "evrndw $RT, $RA", IIC_VecGeneral, []>;
+
+def EVSEL          : EVXForm_4<79, (outs sperc:$RT),
+                               (ins sperc:$RA, sperc:$RB, crrc:$crD),
+                               "evsel crD,$RT,$RA,$RB", IIC_VecGeneral, []>;
+
+def EVSLWI         : EVXForm_1<550, (outs sperc:$RT), (ins sperc:$RA, u5imm:$RB),
+                               "evslwi $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVSLW          : EVXForm_1<548, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evslw $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (shl v2i32:$RA, v2i32:$RB))]>;
+
+def EVSPLATFI      : EVXForm_2<555, (outs sperc:$RT), (ins i32imm:$RA),
+                               "evsplatfi $RT, $RA", IIC_VecGeneral, []>;
+def EVSPLATI       : EVXForm_2<553, (outs sperc:$RT), (ins s5imm:$RA),
+                               "evsplati $RT, $RA", IIC_VecGeneral, []>;
+
+def EVSRWIS        : EVXForm_1<547, (outs sperc:$RT), (ins sperc:$RA, u5imm:$RB),
+                               "evsrwis $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVSRWIU        : EVXForm_1<546, (outs sperc:$RT), (ins sperc:$RA, u5imm:$RB),
+                               "evsrwiu $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVSRWS         : EVXForm_1<545, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evsrws $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (sra v2i32:$RA, v2i32:$RB))]>;
+def EVSRWU         : EVXForm_1<544, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evsrwu $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (srl v2i32:$RA, v2i32:$RB))]>;
+
+def EVSTDD         : EVXForm_D<801, (outs), (ins sperc:$RT, spe8dis:$dst),
+                               "evstdd $RT, $dst", IIC_LdStStore,
+                               [(store f64:$RT, ixaddr:$dst)]>;
+def EVSTDDX        : EVXForm_1<800, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstddx $RT, $dst", IIC_LdStStore,
+                               [(store f64:$RT, xaddr:$dst)]>;
+def EVSTDH         : EVXForm_D<805, (outs), (ins sperc:$RT, spe8dis:$dst),
+                               "evstdh $RT, $dst", IIC_LdStStore, []>;
+def EVSTDHX        : EVXForm_1<804, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstdhx $RT, $dst", IIC_LdStStore, []>;
+def EVSTDW         : EVXForm_D<803, (outs), (ins sperc:$RT, spe8dis:$dst),
+                               "evstdw $RT, $dst", IIC_LdStStore,
+                               [(store v2i32:$RT, ixaddr:$dst)]>;
+def EVSTDWX        : EVXForm_1<802, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstdwx $RT, $dst", IIC_LdStStore,
+                               [(store v2i32:$RT, xaddr:$dst)]>;
+let isAsmParserOnly = 1 in {
+def EVSTDWFS       : EVXForm_D<803, (outs), (ins sperc:$RT, spe8dis:$dst),
+                               "evstdw $RT, $dst", IIC_LdStStore,
+                               [(store v2f32:$RT, ixaddr:$dst)]>;
+def EVSTDWXFS      : EVXForm_1<802, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstdwx $RT, $dst", IIC_LdStStore,
+                               [(store v2f32:$RT, xaddr:$dst)]>;
+}
+def EVSTWHE        : EVXForm_D<817, (outs), (ins sperc:$RT, spe4dis:$dst),
+                               "evstwhe $RT, $dst", IIC_LdStStore, []>;
+def EVSTWHEX       : EVXForm_1<816, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstwhex $RT, $dst", IIC_LdStStore, []>;
+def EVSTWHO        : EVXForm_D<821, (outs), (ins sperc:$RT, spe4dis:$dst),
+                               "evstwho $RT, $dst", IIC_LdStStore, []>;
+def EVSTWHOX       : EVXForm_1<820, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstwhox $RT, $dst", IIC_LdStStore, []>;
+def EVSTWWE        : EVXForm_D<825, (outs), (ins sperc:$RT, spe4dis:$dst),
+                               "evstwwe $RT, $dst", IIC_LdStStore, []>;
+def EVSTWWEX       : EVXForm_1<824, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstwwex $RT, $dst", IIC_LdStStore, []>;
+def EVSTWWO        : EVXForm_D<829, (outs), (ins sperc:$RT, spe4dis:$dst),
+                               "evstwwo $RT, $dst", IIC_LdStStore, []>;
+def EVSTWWOX       : EVXForm_1<828, (outs), (ins sperc:$RT, memrr:$dst),
+                               "evstwwox $RT, $dst", IIC_LdStStore, []>;
+
+def EVSUBFSSIAAW   : EVXForm_2<1219, (outs sperc:$RT), (ins sperc:$RA),
+                               "evsubfssiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVSUBFSMIAAW   : EVXForm_2<1227, (outs sperc:$RT), (ins sperc:$RA),
+                               "evsubfsmiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVSUBFUMIAAW   : EVXForm_2<1226, (outs sperc:$RT), (ins sperc:$RA),
+                               "evsubfumiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVSUBFUSIAAW   : EVXForm_2<1218, (outs sperc:$RT), (ins sperc:$RA),
+                               "evsubfusiaaw $RT, $RA", IIC_VecComplex, []>;
+def EVSUBFW        : EVXForm_1<516, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evsubfw $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (sub v2i32:$RB, v2i32:$RA))]>;
+def EVSUBIFW       : EVXForm_1<518, (outs sperc:$RT), (ins u5imm:$RA, sperc:$RB),
+                               "evsubifw $RT, $RA, $RB", IIC_VecGeneral, []>;
+def EVXOR          : EVXForm_1<534, (outs sperc:$RT), (ins sperc:$RA, sperc:$RB),
+                               "evxor $RT, $RA, $RB", IIC_VecGeneral,
+                               [(set v2i32:$RT, (xor v2i32:$RA, v2i32:$RB))]>;
+
+let isAsmParserOnly = 1 in {
+// Identical to the integer Load/Stores, but to handle floats
+def SPELWZ        : DForm_1<32, (outs spe4rc:$rD), (ins memri:$src),
+                            "lwz $rD, $src", IIC_LdStLoad,
+                            [(set f32:$rD, (load iaddr:$src))]>;
+def SPELWZX       : XForm_1<31,  23, (outs spe4rc:$rD), (ins memrr:$src),
+                            "lwzx $rD, $src", IIC_LdStLoad,
+                            [(set f32:$rD, (load xaddr:$src))]>;
+def SPESTW        : DForm_1<36, (outs), (ins spe4rc:$rS, memri:$src),
+                            "stw $rS, $src", IIC_LdStStore,
+                            [(store f32:$rS, iaddr:$src)]>;
+def SPESTWX       : XForm_8<31, 151, (outs), (ins spe4rc:$rS, memrr:$dst),
+                           "stwx $rS, $dst", IIC_LdStStore,
+                           [(store f32:$rS, xaddr:$dst)]>;
+}
 
 } // HasSPE
+
+let Predicates = [HasSPE] in {
+def : Pat<(f64 (extloadf32 iaddr:$src)),
+          (COPY_TO_REGCLASS (SPELWZ iaddr:$src), SPERC)>;
+def : Pat<(f64 (extloadf32 xaddr:$src)),
+          (COPY_TO_REGCLASS (SPELWZX xaddr:$src), SPERC)>;
+
+def : Pat<(f64 (fpextend f32:$src)),
+          (COPY_TO_REGCLASS $src, SPERC)>;
+}
+
+// SPE Intrinsics
+def : Pat<(int_ppc_spe_brinc i32:$A, i32:$B), (BRINC $A, $B)>;
+def : Pat<(int_ppc_spe_evabs v2i32:$A), (EVABS $A)>;
+def : Pat<(int_ppc_spe_evaddiw v2i32:$A, imm32SExt5:$B), (EVADDIW $A, $B)>;
+def : Pat<(int_ppc_spe_evaddw v2i32:$A, v2i32:$B), (EVADDW $A, $B)>;
+def : Pat<(int_ppc_spe_evsubifw imm32SExt5:$A, v2i32:$B), (EVSUBIFW $A, $B)>;
+def : Pat<(int_ppc_spe_evsubfw v2i32:$A, v2i32:$B), (EVSUBFW $A, $B)>;
+
+def : Pat<(int_ppc_spe_evneg v2i32:$A), (EVNEG $A)>;
+def : Pat<(int_ppc_spe_evand v2i32:$A, v2i32:$B), (EVAND $A, $B)>;
+def : Pat<(int_ppc_spe_evandc v2i32:$A, v2i32:$B), (EVANDC $A, $B)>;
+def : Pat<(int_ppc_spe_evnand v2i32:$A, v2i32:$B), (EVNAND $A, $B)>;
+def : Pat<(int_ppc_spe_evor v2i32:$A, v2i32:$B), (EVOR $A, $B)>;
+def : Pat<(int_ppc_spe_evorc v2i32:$A, v2i32:$B), (EVORC $A, $B)>;
+def : Pat<(int_ppc_spe_evnor v2i32:$A, v2i32:$B), (EVNOR $A, $B)>;
+def : Pat<(int_ppc_spe_evextsb v2i32:$A), (EVEXTSB $A)>;
+def : Pat<(int_ppc_spe_evextsh v2i32:$A), (EVEXTSH $A)>;
+def : Pat<(int_ppc_spe_evrlw v2i32:$A, v2i32:$B), (EVRLW $A, $B)>;
+def : Pat<(int_ppc_spe_evrlwi v2i32:$A, imm32SExt5:$B), (EVRLWI $A, $B)>;
+def : Pat<(int_ppc_spe_evslw v2i32:$A, v2i32:$B), (EVSLW $A, $B)>;
+def : Pat<(int_ppc_spe_evslwi v2i32:$A, imm32SExt5:$B), (EVSLWI $A, $B)>;
+def : Pat<(int_ppc_spe_evsrws v2i32:$A, v2i32:$B), (EVSRWS $A, $B)>;
+def : Pat<(int_ppc_spe_evsrwis v2i32:$A, imm32SExt5:$B), (EVSRWIS $A, $B)>;
+def : Pat<(int_ppc_spe_evsrwu v2i32:$A, v2i32:$B), (EVSRWU $A, $B)>;
+def : Pat<(int_ppc_spe_evsrwiu v2i32:$A, imm32SExt5:$B), (EVSRWIU $A, $B)>;
+def : Pat<(int_ppc_spe_evcntlsw v2i32:$A), (EVCNTLSW $A)>;
+def : Pat<(int_ppc_spe_evcntlzw v2i32:$A), (EVCNTLZW $A)>;
+def : Pat<(int_ppc_spe_evrndw v2i32:$A), (EVRNDW $A)>;
+def : Pat<(int_ppc_spe_evmergehi v2i32:$A, v2i32:$B), (EVMERGEHI $A, $B)>;
+def : Pat<(int_ppc_spe_evmergelo v2i32:$A, v2i32:$B), (EVMERGELO $A, $B)>;
+def : Pat<(int_ppc_spe_evmergehilo v2i32:$A, v2i32:$B), (EVMERGEHILO $A, $B)>;
+def : Pat<(int_ppc_spe_evmergelohi v2i32:$A, v2i32:$B), (EVMERGELOHI $A, $B)>;
+def : Pat<(int_ppc_spe_evsplati imm32SExt5:$A), (EVSPLATI $A)>;
+def : Pat<(int_ppc_spe_evsplatfi imm32SExt5:$A), (EVSPLATFI $A)>;
+def : Pat<(int_ppc_spe_evdivws v2i32:$A, v2i32:$B), (EVDIVWS $A, $B)>;
+def : Pat<(int_ppc_spe_evdivwu v2i32:$A, v2i32:$B), (EVDIVWU $A, $B)>;
+def : Pat<(int_ppc_spe_evmra v2i32:$A), (EVMRA $A)>;
+
+def : Pat<(int_ppc_spe_evfsabs v2f32:$A), (EVFSABS $A)>;
+def : Pat<(int_ppc_spe_evfsnabs v2f32:$A), (EVFSNABS $A)>;
+def : Pat<(int_ppc_spe_evfsneg v2f32:$A), (EVFSNEG $A)>;
+def : Pat<(int_ppc_spe_evfsadd v2f32:$A, v2f32:$B), (EVFSADD $A, $B)>;
+def : Pat<(int_ppc_spe_evfssub v2f32:$A, v2f32:$B), (EVFSSUB $A, $B)>;
+def : Pat<(int_ppc_spe_evfsmul v2f32:$A, v2f32:$B), (EVFSMUL $A, $B)>;
+def : Pat<(int_ppc_spe_evfsdiv v2f32:$A, v2f32:$B), (EVFSDIV $A, $B)>;
+def : Pat<(int_ppc_spe_evfscfui v2i32:$A), (EVFSCFUI $A)>;
+def : Pat<(int_ppc_spe_evfscfsi v2i32:$A), (EVFSCFSI $A)>;
+def : Pat<(int_ppc_spe_evfscfuf v2i32:$A), (EVFSCFUF $A)>;
+def : Pat<(int_ppc_spe_evfscfsf v2i32:$A), (EVFSCFSF $A)>;
+def : Pat<(int_ppc_spe_evfsctui v2f32:$A), (EVFSCTUI $A)>;
+def : Pat<(int_ppc_spe_evfsctsi v2f32:$A), (EVFSCTSI $A)>;
+def : Pat<(int_ppc_spe_evfsctuf v2f32:$A), (EVFSCTUF $A)>;
+def : Pat<(int_ppc_spe_evfsctsf v2f32:$A), (EVFSCTSF $A)>;
+def : Pat<(int_ppc_spe_evfsctuiz v2f32:$A), (EVFSCTUIZ $A)>;
+def : Pat<(int_ppc_spe_evfsctsiz v2f32:$A), (EVFSCTSIZ $A)>;
+def : Pat<(int_ppc_spe_evmhegsmfaa v2i32:$A, v2i32:$B), (EVMHEGSMFAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhegsmfan v2i32:$A, v2i32:$B), (EVMHEGSMFAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhegsmiaa v2i32:$A, v2i32:$B), (EVMHEGSMIAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhegsmian v2i32:$A, v2i32:$B), (EVMHEGSMIAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhegumiaa v2i32:$A, v2i32:$B), (EVMHEGUMIAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhegumian v2i32:$A, v2i32:$B), (EVMHEGUMIAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmf    v2i32:$A, v2i32:$B), (EVMHESMF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmfa   v2i32:$A, v2i32:$B), (EVMHESMFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmfaaw v2i32:$A, v2i32:$B), (EVMHESMFAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmfanw v2i32:$A, v2i32:$B), (EVMHESMFANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmi    v2i32:$A, v2i32:$B), (EVMHESMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmia   v2i32:$A, v2i32:$B), (EVMHESMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmiaaw v2i32:$A, v2i32:$B), (EVMHESMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhesmianw v2i32:$A, v2i32:$B), (EVMHESMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessf    v2i32:$A, v2i32:$B), (EVMHESSF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessfa   v2i32:$A, v2i32:$B), (EVMHESSFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessfaaw v2i32:$A, v2i32:$B), (EVMHESSFAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessfanw v2i32:$A, v2i32:$B), (EVMHESSFANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessiaaw v2i32:$A, v2i32:$B), (EVMHESSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhessianw v2i32:$A, v2i32:$B), (EVMHESSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmheumi    v2i32:$A, v2i32:$B), (EVMHEUMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmheumia   v2i32:$A, v2i32:$B), (EVMHEUMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmheumiaaw v2i32:$A, v2i32:$B), (EVMHEUMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmheumianw v2i32:$A, v2i32:$B), (EVMHEUMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmheusiaaw v2i32:$A, v2i32:$B), (EVMHEUSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmheusianw v2i32:$A, v2i32:$B), (EVMHEUSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogsmfaa v2i32:$A, v2i32:$B), (EVMHOGSMFAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogsmfan v2i32:$A, v2i32:$B), (EVMHOGSMFAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogsmiaa v2i32:$A, v2i32:$B), (EVMHOGSMIAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogsmian v2i32:$A, v2i32:$B), (EVMHOGSMIAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogumiaa v2i32:$A, v2i32:$B), (EVMHOGUMIAA $A, $B)>;
+def : Pat<(int_ppc_spe_evmhogumian v2i32:$A, v2i32:$B), (EVMHOGUMIAN $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmf    v2i32:$A, v2i32:$B), (EVMHOSMF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmfa   v2i32:$A, v2i32:$B), (EVMHOSMFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmfaaw v2i32:$A, v2i32:$B), (EVMHOSMFAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmfanw v2i32:$A, v2i32:$B), (EVMHOSMFANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmi    v2i32:$A, v2i32:$B), (EVMHOSMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmia   v2i32:$A, v2i32:$B), (EVMHOSMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmiaaw v2i32:$A, v2i32:$B), (EVMHOSMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhosmianw v2i32:$A, v2i32:$B), (EVMHOSMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossf    v2i32:$A, v2i32:$B), (EVMHOSSF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossfa   v2i32:$A, v2i32:$B), (EVMHOSSFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossfaaw v2i32:$A, v2i32:$B), (EVMHOSSFAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossfanw v2i32:$A, v2i32:$B), (EVMHOSSFANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossiaaw v2i32:$A, v2i32:$B), (EVMHOSSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhossianw v2i32:$A, v2i32:$B), (EVMHOSSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhoumi    v2i32:$A, v2i32:$B), (EVMHOUMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmhoumia   v2i32:$A, v2i32:$B), (EVMHOUMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmhoumiaaw v2i32:$A, v2i32:$B), (EVMHOUMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhoumianw v2i32:$A, v2i32:$B), (EVMHOUMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhousiaaw v2i32:$A, v2i32:$B), (EVMHOUSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmhousianw v2i32:$A, v2i32:$B), (EVMHOUSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhsmf    v2i32:$A, v2i32:$B), (EVMWHSMF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhsmfa   v2i32:$A, v2i32:$B), (EVMWHSMFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhsmi    v2i32:$A, v2i32:$B), (EVMWHSMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhsmia   v2i32:$A, v2i32:$B), (EVMWHSMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhssf    v2i32:$A, v2i32:$B), (EVMWHSSF    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhssfa   v2i32:$A, v2i32:$B), (EVMWHSSFA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhumi    v2i32:$A, v2i32:$B), (EVMWHUMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwhumia   v2i32:$A, v2i32:$B), (EVMWHUMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlsmiaaw v2i32:$A, v2i32:$B), (EVMWLSMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlsmianw v2i32:$A, v2i32:$B), (EVMWLSMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlssiaaw v2i32:$A, v2i32:$B), (EVMWLSSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlssianw v2i32:$A, v2i32:$B), (EVMWLSSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlumi    v2i32:$A, v2i32:$B), (EVMWLUMI    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlumia   v2i32:$A, v2i32:$B), (EVMWLUMIA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlumiaaw v2i32:$A, v2i32:$B), (EVMWLUMIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlumianw v2i32:$A, v2i32:$B), (EVMWLUMIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlusiaaw v2i32:$A, v2i32:$B), (EVMWLUSIAAW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwlusianw v2i32:$A, v2i32:$B), (EVMWLUSIANW $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmf     v2i32:$A, v2i32:$B), (EVMWSMF     $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmfa    v2i32:$A, v2i32:$B), (EVMWSMFA    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmfaa   v2i32:$A, v2i32:$B), (EVMWSMFAA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmfan   v2i32:$A, v2i32:$B), (EVMWSMFAN   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmi     v2i32:$A, v2i32:$B), (EVMWSMI     $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmia    v2i32:$A, v2i32:$B), (EVMWSMIA    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmiaa   v2i32:$A, v2i32:$B), (EVMWSMIAA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwsmian   v2i32:$A, v2i32:$B), (EVMWSMIAN   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwssf     v2i32:$A, v2i32:$B), (EVMWSSF     $A, $B)>;
+def : Pat<(int_ppc_spe_evmwssfa    v2i32:$A, v2i32:$B), (EVMWSSFA    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwssfaa   v2i32:$A, v2i32:$B), (EVMWSSFAA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwssfan   v2i32:$A, v2i32:$B), (EVMWSSFAN   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwumi     v2i32:$A, v2i32:$B), (EVMWUMI     $A, $B)>;
+def : Pat<(int_ppc_spe_evmwumia    v2i32:$A, v2i32:$B), (EVMWUMIA    $A, $B)>;
+def : Pat<(int_ppc_spe_evmwumiaa   v2i32:$A, v2i32:$B), (EVMWUMIAA   $A, $B)>;
+def : Pat<(int_ppc_spe_evmwumian   v2i32:$A, v2i32:$B), (EVMWUMIAN   $A, $B)>;

lib/Target/PowerPC/PPCRegisterInfo.cpp

PPCRegisterInfo::PPCRegisterInfo(const PPCTargetMachine &TM)
   ImmToIdxMap[PPC::DFSTOREf32] = PPC::STXSSPX;
   ImmToIdxMap[PPC::DFSTOREf64] = PPC::STXSDX;
   ImmToIdxMap[PPC::LXV] = PPC::LXVX;
   ImmToIdxMap[PPC::LXSD] = PPC::LXSDX;
   ImmToIdxMap[PPC::LXSSP] = PPC::LXSSPX;
   ImmToIdxMap[PPC::STXV] = PPC::STXVX;
   ImmToIdxMap[PPC::STXSD] = PPC::STXSDX;
   ImmToIdxMap[PPC::STXSSP] = PPC::STXSSPX;
+
+  // SPE
+  ImmToIdxMap[PPC::EVLDD] = PPC::EVLDDX;
+  ImmToIdxMap[PPC::EVSTDD] = PPC::EVSTDDX;
+  ImmToIdxMap[PPC::SPESTW] = PPC::SPESTWX;
+  ImmToIdxMap[PPC::SPELWZ] = PPC::SPELWZX;
 }
 
 /// getPointerRegClass - Return the register class to use to hold pointers.
 /// This is used for addressing modes.
 const TargetRegisterClass *
 PPCRegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind)
                                                                        const {
   // Note that PPCInstrInfo::FoldImmediate also directly uses this Kind value
unsigned PPCRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC,
   const PPCFrameLowering *TFI = getFrameLowering(MF);
   const unsigned DefaultSafety = 1;
 
   switch (RC->getID()) {
   default:
     return 0;
   case PPC::G8RC_NOX0RegClassID:
   case PPC::GPRC_NOR0RegClassID:
+  case PPC::SPERCRegClassID:
+  case PPC::SPE4RCRegClassID:
   case PPC::G8RCRegClassID:
   case PPC::GPRCRegClassID: {
     unsigned FP = TFI->hasFP(MF) ? 1 : 0;
     return 32 - FP - DefaultSafety;
   }
   case PPC::F8RCRegClassID:
   case PPC::F4RCRegClassID:
   case PPC::QFRCRegClassID:

lib/Target/PowerPC/PPCRegisterInfo.td

 
 // GP8 - One of the 32 64-bit general-purpose registers
 class GP8<GPR SubReg, string n> : PPCReg<n> {
   let HWEncoding = SubReg.HWEncoding;
   let SubRegs = [SubReg];
   let SubRegIndices = [sub_32];
 }
 
+// SPE - One of the 32 64-bit general-purpose registers (SPE)
+class SPE<GPR SubReg, string n> : PPCReg<n> {
+  let HWEncoding = SubReg.HWEncoding;
+  let SubRegs = [SubReg];
+  let SubRegIndices = [sub_32];
+}
+
 // SPR - One of the 32-bit special-purpose registers
 class SPR<bits<10> num, string n> : PPCReg<n> {
   let HWEncoding{9-0} = num;
 }
 
 // FPR - One of the 32 64-bit floating-point registers
 class FPR<bits<5> num, string n> : PPCReg<n> {
   let HWEncoding{4-0} = num;

 }
 
 // 64-bit General-purpose registers
 foreach Index = 0-31 in {
   def X#Index : GP8<!cast<GPR>("R"#Index), "r"#Index>,
                     DwarfRegNum<[Index, -2]>;
 }
 
+// SPE registers
+foreach Index = 0-31 in {
+  def S#Index : SPE<!cast<GPR>("R"#Index), "r"#Index>,
+                    DwarfRegNum<[!add(Index, 1200), !add(Index, 1200)]>;
+}
+
 // Floating-point registers
 foreach Index = 0-31 in {
   def F#Index : FPR<Index, "f"#Index>,
                 DwarfRegNum<[!add(Index, 32), !add(Index, 32)]>;
 }
 
 // 64-bit Floating-point subregisters of Altivec registers
 // Note: the register names are v0-v31 or vs32-vs63 depending on the use.

 
 // Count register
 def CTR  : SPR<9, "ctr">, DwarfRegNum<[-2, 66]>;
 def CTR8 : SPR<9, "ctr">, DwarfRegNum<[66, -2]>;
 
 // VRsave register
 def VRSAVE: SPR<256, "vrsave">, DwarfRegNum<[109]>;
 
+// SPE extra registers
+def SPEACC: DwarfRegNum<[99, 111]>;
+def SPEFSCR: SPR<512, "spefscr">, DwarfRegNum<[612, 112]>;
 // Carry bit.  In the architecture this is really bit 0 of the XER register
 // (which really is SPR register 1);  this is the only bit interesting to a
 // compiler.
 def CARRY: SPR<1, "ca">, DwarfRegNum<[76]>;
 
 // FP rounding mode:  bits 30 and 31 of the FP status and control register
 // This is not allocated as a normal register; it appears only in
 // Uses and Defs.  The ABI says it needs to be preserved by a function,
def G8RC_NOX0 : RegisterClass<"PPC", [i64], 64, (add (sub G8RC, X0), ZERO8)> {
   // put it at the end of the list.
   let AltOrders = [(add (sub G8RC_NOX0, X2), X2)];
   let AltOrderSelect = [{
     const PPCSubtarget &S = MF.getSubtarget<PPCSubtarget>();
     return S.isPPC64() && S.isSVR4ABI();
   }];
 }
 
+def SPERC : RegisterClass<"PPC", [f64,v2i32,v2f32], 64, (add (sequence "S%u", 2, 12),
+                                                (sequence "S%u", 30, 13),
+                                                S31, S0, S1)> {
+  // On non-Darwin PPC64 systems, R2 can be allocated, but must be restored, so
+  // put it at the end of the list.
+  let AltOrders = [(add (sub SPERC, S2), S2)];
+  let AltOrderSelect = [{
+    const PPCSubtarget &S = MF.getSubtarget<PPCSubtarget>();
+    return S.isPPC64() && S.isSVR4ABI();
+  }];
+}
+
+def SPE4RC : RegisterClass<"PPC", [f32], 32, (add GPRC)>;
+
 // Allocate volatiles first, then non-volatiles in reverse order. With the SVR4
 // ABI the size of the Floating-point register save area is determined by the
 // allocated non-volatile register with the lowest register number, as FP
 // register N is spilled to offset 8 * (32 - N) below the back chain word of the
 // previous stack frame. By allocating non-volatiles in reverse order we make
 // sure that the Floating-point register save area is always as small as
 // possible because there aren't any unused spill slots.
 def F8RC : RegisterClass<"PPC", [f64], 64, (add (sequence "F%u", 0, 13),

lib/Target/PowerPC/PPCSchedule.td

 def IIC_SprMFMSR     : InstrItinClass;
 def IIC_SprMFSPR     : InstrItinClass;
 def IIC_SprMFTB      : InstrItinClass;
 def IIC_SprMTSPR     : InstrItinClass;
 def IIC_SprMTSRIN    : InstrItinClass;
 def IIC_SprRFI       : InstrItinClass;
 def IIC_SprSC        : InstrItinClass;
 def IIC_FPGeneral    : InstrItinClass;
+def IIC_FPDGeneral   : InstrItinClass;
+def IIC_FPSGeneral   : InstrItinClass;
 def IIC_FPAddSub     : InstrItinClass;
 def IIC_FPCompare    : InstrItinClass;
 def IIC_FPDivD       : InstrItinClass;
 def IIC_FPDivS       : InstrItinClass;
 def IIC_FPFused      : InstrItinClass;
 def IIC_FPRes        : InstrItinClass;
 def IIC_FPSqrtD      : InstrItinClass;
 def IIC_FPSqrtS      : InstrItinClass;

 include "PPCSchedule440.td"
 include "PPCScheduleG4.td"
 include "PPCScheduleG4Plus.td"
 include "PPCScheduleG5.td"
 include "PPCScheduleP7.td"
 include "PPCScheduleP8.td"
 include "PPCScheduleP9.td"
 include "PPCScheduleA2.td"
+include "PPCScheduleE500.td"
 include "PPCScheduleE500mc.td"
 include "PPCScheduleE5500.td"

lib/Target/PowerPC/PPCScheduleE500.td

-//===-- PPCScheduleE500mc.td - e500mc Scheduling Defs ------*- tablegen -*-===//
+//===-- PPCScheduleE500.td - e500 Scheduling Defs ------*- tablegen -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
-// This file defines the itinerary class data for the Freescale e500mc 32-bit
+// This file defines the itinerary class data for the Freescale e500 32-bit
 // Power processor.
 //
-// All information is derived from the "e500mc Core Reference Manual",
+// All information is derived from the "e500 Core Reference Manual",
 // Freescale Document Number E500MCRM, Rev. 1, 03/2012.
 //
 //===----------------------------------------------------------------------===//
-// Relevant functional units in the Freescale e500mc core:
+// Relevant functional units in the Freescale e500 core:
 //
 //  * Decode & Dispatch
 //    Can dispatch up to 2 instructions per clock cycle to either the GPR Issue
-//    queues (GIQx), FP Issue Queue (FIQ), or Branch issue queue (BIQ).
+//    queues (GIQx) or Branch issue queue (BIQ).
 def E500_DIS0 : FuncUnit; // Dispatch stage - insn 1
 def E500_DIS1 : FuncUnit; // Dispatch stage - insn 2
 
 //  * Execute
-//    6 pipelined execution units: SFX0, SFX1, BU, FPU, LSU, CFX.
-//    Some instructions can only execute in SFX0 but not SFX1.
-//    The CFX has a bypass path, allowing non-divide instructions to execute
-//    while a divide instruction is executed.
-def E500_SFX0  : FuncUnit; // Simple unit 0
-def E500_SFX1  : FuncUnit; // Simple unit 1
+//    6 pipelined execution units: SU0, SU1, BU, LSU, MU.
+//    Some instructions can only execute in SU0 but not SU1.
+def E500_SU0  : FuncUnit; // Simple unit 0
+def E500_SU1  : FuncUnit; // Simple unit 1
 def E500_BU    : FuncUnit; // Branch unit
-def E500_CFX_DivBypass
-               : FuncUnit; // CFX divide bypass path
-def E500_CFX_0 : FuncUnit; // CFX pipeline
+def E500_MU    : FuncUnit; // MU pipeline
 def E500_LSU_0 : FuncUnit; // LSU pipeline
-def E500_FPU_0 : FuncUnit; // FPU pipeline
 
 def E500_GPR_Bypass : Bypass;
-def E500_FPR_Bypass : Bypass;
 def E500_CR_Bypass  : Bypass;
+def E500_DivBypass  : Bypass;
 
-def PPCE500mcItineraries : ProcessorItineraries<
-  [E500_DIS0, E500_DIS1, E500_SFX0, E500_SFX1, E500_BU, E500_CFX_DivBypass,
-   E500_CFX_0, E500_LSU_0, E500_FPU_0],
-  [E500_CR_Bypass, E500_GPR_Bypass, E500_FPR_Bypass], [
+def PPCE500Itineraries : ProcessorItineraries<
+  [E500_DIS0, E500_DIS1, E500_SU0, E500_SU1, E500_BU,
+   E500_MU, E500_LSU_0],
+  [E500_CR_Bypass, E500_GPR_Bypass, E500_DivBypass], [
   InstrItinData<IIC_IntSimple,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntGeneral,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntISEL,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1, 1], // Latency = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass,
                                   E500_CR_Bypass]>,
   InstrItinData<IIC_IntCompare,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [5, 1, 1], // Latency = 1 or 2
                                  [E500_CR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntDivW,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0], 0>,
-                                  InstrStage<14, [E500_CFX_DivBypass]>],
+                                  InstrStage<1, [E500_MU], 0>,
+                                  InstrStage<14, [E500_MU]>],
                                  [17, 1, 1], // Latency=4..35, Repeat= 4..35
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntMFFS,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<8, [E500_FPU_0]>],
-                                 [11], // Latency = 8
-                                 [E500_FPR_Bypass]>,
-  InstrItinData<IIC_IntMTFSB0,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<8, [E500_FPU_0]>],
-                                 [11, 1, 1], // Latency = 8
-                                 [NoBypass, NoBypass, NoBypass]>,
   InstrItinData<IIC_IntMulHW,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+                                  InstrStage<4, [E500_MU]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntMulHWU,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+                                  InstrStage<4, [E500_MU]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntMulLI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+                                  InstrStage<4, [E500_MU]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntRotate,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntShift,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_IntTrapW,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_SFX0]>],
+                                  InstrStage<2, [E500_SU0]>],
                                  [5, 1], // Latency = 2, Repeat rate = 2
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_BrB,         [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_BU]>],
                                  [4, 1], // Latency = 1
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_BrCR,        [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_BU]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_CR_Bypass,
                                   E500_CR_Bypass, E500_CR_Bypass]>,
   InstrItinData<IIC_BrMCR,       [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_BU]>],
                                  [4, 1], // Latency = 1
                                  [E500_CR_Bypass, E500_CR_Bypass]>,
   InstrItinData<IIC_BrMCRX,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1, 1], // Latency = 1
                                  [E500_CR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStDCBA,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3, Repeat rate = 1
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStDCBF,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStDCBI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLoad,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLoadUpd, [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
+                                  InstrStage<1, [E500_SU0, E500_SU1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass],
                                  2>, // 2 micro-ops
   InstrItinData<IIC_LdStLoadUpdX,[InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
+                                  InstrStage<1, [E500_SU0, E500_SU1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass],
                                  2>, // 2 micro-ops
   InstrItinData<IIC_LdStStore,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStStoreUpd,[InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
+                                  InstrStage<1, [E500_SU0, E500_SU1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [NoBypass, E500_GPR_Bypass],
                                  2>, // 2 micro-ops
   InstrItinData<IIC_LdStICBI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSTFD,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1, 1], // Latency = 3
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSTFDU,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1, 1], // Latency = 3
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
-                                 2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLFD,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLFDU,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
-                                 2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLFDUX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
-                                 2>, // 2 micro-ops
   InstrItinData<IIC_LdStLHA,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLHAU,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
+                                  InstrStage<1, [E500_SU0, E500_SU1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLHAUX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
+                                  InstrStage<1, [E500_SU0, E500_SU1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLMW,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [7, 1], // Latency = r+3
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStLWARX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<3, [E500_LSU_0]>],
                                  [6, 1, 1], // Latency = 3, Repeat rate = 3
                                  [E500_GPR_Bypass,
                                   E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStSTWCX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>],
                                  [6, 1], // Latency = 3
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_LdStSync,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0]>]>,
   InstrItinData<IIC_SprMFSR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                  InstrStage<4, [E500_SU0]>],
                                  [7, 1],
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMTMSR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<2, [E500_SU0, E500_SU1]>],
                                  [5, 1], // Latency = 2, Repeat rate = 4
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMTSR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                  InstrStage<1, [E500_SU0]>],
                                  [5, 1],
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprTLBSYNC,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
                                   InstrStage<1, [E500_LSU_0], 0>]>,
   InstrItinData<IIC_SprMFCR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<5, [E500_SFX0]>],
+                                  InstrStage<5, [E500_SU0]>],
                                  [8, 1],
                                  [E500_GPR_Bypass, E500_CR_Bypass]>,
   InstrItinData<IIC_SprMFCRF,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<5, [E500_SFX0]>],
+                                  InstrStage<5, [E500_SU0]>],
                                  [8, 1],
                                  [E500_GPR_Bypass, E500_CR_Bypass]>,
   InstrItinData<IIC_SprMFPMR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                  InstrStage<4, [E500_SU0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMFMSR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                  InstrStage<4, [E500_SU0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
                                  [E500_GPR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMFSPR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
                                  [E500_GPR_Bypass, E500_CR_Bypass]>,
   InstrItinData<IIC_SprMTPMR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                  InstrStage<1, [E500_SU0]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
                                  [E500_CR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMFTB,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                  InstrStage<4, [E500_SU0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
                                  [NoBypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMTSPR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                  InstrStage<1, [E500_SU0, E500_SU1]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
                                  [E500_CR_Bypass, E500_GPR_Bypass]>,
   InstrItinData<IIC_SprMTSRIN,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                  InstrStage<1, [E500_SU0]>],
                                  [4, 1],
                                  [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_FPGeneral,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_FPU_0]>],
-                                 [11, 1, 1], // Latency = 8, Repeat rate = 2
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPAddSub,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_FPU_0]>],
-                                 [13, 1, 1], // Latency = 10, Repeat rate = 4
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPCompare,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_FPU_0]>],
-                                 [11, 1, 1], // Latency = 8, Repeat rate = 2
-                                 [E500_CR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
+  InstrItinData<IIC_FPDGeneral,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
+                                  InstrStage<6, [E500_MU]>],
+                                 [9, 1, 1],  // Latency = 6, Repeat rate = 1
+                                 [NoBypass]>,
+  InstrItinData<IIC_FPSGeneral,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
+                                  InstrStage<4, [E500_MU]>],
+                                 [7, 1, 1],  // Latency = 4, Repeat rate = 1
+                                 [NoBypass]>,
   InstrItinData<IIC_FPDivD,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<68, [E500_FPU_0]>],
-                                 [71, 1, 1], // Latency = 68, Repeat rate = 68
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
+                                  InstrStage<32, [E500_MU]>],
+                                 [35, 1, 1], // Latency = 32, Repeat rate = 32
+                                 [E500_DivBypass]>,
   InstrItinData<IIC_FPDivS,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<38, [E500_FPU_0]>],
-                                 [41, 1, 1], // Latency = 38, Repeat rate = 38
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPFused,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_FPU_0]>],
-                                 [13, 1, 1, 1], // Latency = 10, Repeat rate = 4
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass,
-                                  E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPRes,       [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<38, [E500_FPU_0]>],
-                                 [41, 1], // Latency = 38, Repeat rate = 38
-                                 [E500_FPR_Bypass, E500_FPR_Bypass]>
+                                  InstrStage<29, [E500_MU]>],
+                                 [32, 1, 1], // Latency = 29, Repeat rate = 29
+                                 [E500_DivBypass]>,
+  InstrItinData<IIC_VecGeneral,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
+                                  InstrStage<1, [E500_SU0]>],
+                                 [4, 1, 1], // Latency = 1, Repeat rate = 1
+                                 [NoBypass]>,
+  InstrItinData<IIC_VecComplex,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
+                                  InstrStage<4, [E500_MU]>],
+                                 [7, 1, 1], // Latency = 4, Repeat rate = 1
+                                 [NoBypass]>
 ]>;
 
 // ===---------------------------------------------------------------------===//
-// e500mc machine model for scheduling and other instruction cost heuristics.
+// e500 machine model for scheduling and other instruction cost heuristics.
 
-def PPCE500mcModel : SchedMachineModel {
+def PPCE500Model : SchedMachineModel {
   let IssueWidth = 2;  // 2 micro-ops are dispatched per cycle.
   let LoadLatency = 5; // Optimistic load latency assuming bypass.
                        // This is overriden by OperandCycles if the
                        // Itineraries are queried instead.
 
   let CompleteModel = 0;
 
-  let Itineraries = PPCE500mcItineraries;
+  let Itineraries = PPCE500Itineraries;
 }

lib/Target/PowerPC/PPCScheduleE500mc.td

 // Freescale Document Number E500MCRM, Rev. 1, 03/2012.
 //
 //===----------------------------------------------------------------------===//
 // Relevant functional units in the Freescale e500mc core:
 //
 //  * Decode & Dispatch
 //    Can dispatch up to 2 instructions per clock cycle to either the GPR Issue
 //    queues (GIQx), FP Issue Queue (FIQ), or Branch issue queue (BIQ).
-def E500_DIS0 : FuncUnit; // Dispatch stage - insn 1
-def E500_DIS1 : FuncUnit; // Dispatch stage - insn 2
+def E500mc_DIS0 : FuncUnit; // Dispatch stage - insn 1
+def E500mc_DIS1 : FuncUnit; // Dispatch stage - insn 2
 
 //  * Execute
 //    6 pipelined execution units: SFX0, SFX1, BU, FPU, LSU, CFX.
 //    Some instructions can only execute in SFX0 but not SFX1.
 //    The CFX has a bypass path, allowing non-divide instructions to execute
 //    while a divide instruction is executed.
-def E500_SFX0  : FuncUnit; // Simple unit 0
-def E500_SFX1  : FuncUnit; // Simple unit 1
-def E500_BU    : FuncUnit; // Branch unit
-def E500_CFX_DivBypass
+def E500mc_SFX0  : FuncUnit; // Simple unit 0
+def E500mc_SFX1  : FuncUnit; // Simple unit 1
+def E500mc_BU    : FuncUnit; // Branch unit
+def E500mc_CFX_DivBypass
                : FuncUnit; // CFX divide bypass path
-def E500_CFX_0 : FuncUnit; // CFX pipeline
-def E500_LSU_0 : FuncUnit; // LSU pipeline
-def E500_FPU_0 : FuncUnit; // FPU pipeline
+def E500mc_CFX_0 : FuncUnit; // CFX pipeline
+def E500mc_LSU_0 : FuncUnit; // LSU pipeline
+def E500mc_FPU_0 : FuncUnit; // FPU pipeline
 
-def E500_GPR_Bypass : Bypass;
-def E500_FPR_Bypass : Bypass;
-def E500_CR_Bypass  : Bypass;
+def E500mc_GPR_Bypass : Bypass;
+def E500mc_FPR_Bypass : Bypass;
+def E500mc_CR_Bypass  : Bypass;
 
 def PPCE500mcItineraries : ProcessorItineraries<
-  [E500_DIS0, E500_DIS1, E500_SFX0, E500_SFX1, E500_BU, E500_CFX_DivBypass,
-   E500_CFX_0, E500_LSU_0, E500_FPU_0],
-  [E500_CR_Bypass, E500_GPR_Bypass, E500_FPR_Bypass], [
-  InstrItinData<IIC_IntSimple,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+  [E500mc_DIS0, E500mc_DIS1, E500mc_SFX0, E500mc_SFX1, E500mc_BU, E500mc_CFX_DivBypass,
+   E500mc_CFX_0, E500mc_LSU_0, E500mc_FPU_0],
+  [E500mc_CR_Bypass, E500mc_GPR_Bypass, E500mc_FPR_Bypass], [
+  InstrItinData<IIC_IntSimple,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntGeneral,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntGeneral,  [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntISEL,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntISEL,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1, 1], // Latency = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass,
-                                  E500_CR_Bypass]>,
-  InstrItinData<IIC_IntCompare,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass,
+                                  E500mc_CR_Bypass]>,
+  InstrItinData<IIC_IntCompare,  [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [5, 1, 1], // Latency = 1 or 2
-                                 [E500_CR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntDivW,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0], 0>,
-                                  InstrStage<14, [E500_CFX_DivBypass]>],
+                                 [E500mc_CR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntDivW,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_CFX_0], 0>,
+                                  InstrStage<14, [E500mc_CFX_DivBypass]>],
                                  [17, 1, 1], // Latency=4..35, Repeat= 4..35
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntMFFS,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<8, [E500_FPU_0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntMFFS,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<8, [E500mc_FPU_0]>],
                                  [11], // Latency = 8
-                                 [E500_FPR_Bypass]>,
-  InstrItinData<IIC_IntMTFSB0,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<8, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_IntMTFSB0,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<8, [E500mc_FPU_0]>],
                                  [11, 1, 1], // Latency = 8
                                  [NoBypass, NoBypass, NoBypass]>,
-  InstrItinData<IIC_IntMulHW,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+  InstrItinData<IIC_IntMulHW,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_CFX_0]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntMulHWU,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntMulHWU,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_CFX_0]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntMulLI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_CFX_0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntMulLI,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_CFX_0]>],
                                  [7, 1, 1], // Latency = 4, Repeat rate = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntRotate,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntRotate,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntShift,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntShift,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_IntTrapW,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_IntTrapW,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<2, [E500mc_SFX0]>],
                                  [5, 1], // Latency = 2, Repeat rate = 2
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_BrB,         [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_BU]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_BrB,         [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_BU]>],
                                  [4, 1], // Latency = 1
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_BrCR,        [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_BU]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_BrCR,        [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_BU]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_CR_Bypass,
-                                  E500_CR_Bypass, E500_CR_Bypass]>,
-  InstrItinData<IIC_BrMCR,       [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_BU]>],
+                                 [E500mc_CR_Bypass,
+                                  E500mc_CR_Bypass, E500mc_CR_Bypass]>,
+  InstrItinData<IIC_BrMCR,       [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_BU]>],
                                  [4, 1], // Latency = 1
-                                 [E500_CR_Bypass, E500_CR_Bypass]>,
-  InstrItinData<IIC_BrMCRX,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_CR_Bypass, E500mc_CR_Bypass]>,
+  InstrItinData<IIC_BrMCRX,      [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1, 1], // Latency = 1
-                                 [E500_CR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStDCBA,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [E500mc_CR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStDCBA,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3, Repeat rate = 1
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStDCBF,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStDCBI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLoad,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLoadUpd, [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStDCBF,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStDCBI,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLoad,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLoadUpd, [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLoadUpdX,[InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStLoadUpdX,[InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStStore,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStStore,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStStoreUpd,[InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStStoreUpd,[InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [NoBypass, E500_GPR_Bypass],
+                                 [NoBypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStICBI,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStICBI,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSTFD,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStSTFD,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1, 1], // Latency = 3
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSTFDU,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStSTFDU,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1, 1], // Latency = 3
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLFD,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStLFD,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLFDU,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLFDU,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLFDUX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStLFDUX,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [7, 1, 1], // Latency = 4
-                                 [E500_FPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass],
                                  2>, // 2 micro-ops
-  InstrItinData<IIC_LdStLHA,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+  InstrItinData<IIC_LdStLHA,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLHAU,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLHAUX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLMW,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLHAU,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLHAUX,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLMW,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
                                  [7, 1], // Latency = r+3
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStLWARX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<3, [E500_LSU_0]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStLWARX,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<3, [E500mc_LSU_0]>],
                                  [6, 1, 1], // Latency = 3, Repeat rate = 3
-                                 [E500_GPR_Bypass,
-                                  E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSTWCX,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>],
-                                 [6, 1], // Latency = 3
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_LdStSync,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0]>]>,
-  InstrItinData<IIC_SprMFSR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass,
+                                  E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStSTWCX,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>],
+                                 [6, 1], // Latency = 3
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_LdStSync,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0]>]>,
+  InstrItinData<IIC_SprMFSR,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_SFX0]>],
                                  [7, 1],
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMTMSR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMTMSR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<2, [E500mc_SFX0, E500mc_SFX1]>],
                                  [5, 1], // Latency = 2, Repeat rate = 4
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMTSR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMTSR,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0]>],
                                  [5, 1],
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprTLBSYNC,  [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_LSU_0], 0>]>,
-  InstrItinData<IIC_SprMFCR,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<5, [E500_SFX0]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprTLBSYNC,  [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_LSU_0], 0>]>,
+  InstrItinData<IIC_SprMFCR,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<5, [E500mc_SFX0]>],
                                  [8, 1],
-                                 [E500_GPR_Bypass, E500_CR_Bypass]>,
-  InstrItinData<IIC_SprMFCRF,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<5, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass, E500mc_CR_Bypass]>,
+  InstrItinData<IIC_SprMFCRF,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<5, [E500mc_SFX0]>],
                                  [8, 1],
-                                 [E500_GPR_Bypass, E500_CR_Bypass]>,
-  InstrItinData<IIC_SprMFPMR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass, E500mc_CR_Bypass]>,
+  InstrItinData<IIC_SprMFPMR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_SFX0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMFMSR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMFMSR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_SFX0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
-                                 [E500_GPR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMFSPR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [E500mc_GPR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMFSPR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
-                                 [E500_GPR_Bypass, E500_CR_Bypass]>,
-  InstrItinData<IIC_SprMTPMR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                 [E500mc_GPR_Bypass, E500mc_CR_Bypass]>,
+  InstrItinData<IIC_SprMTPMR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
-                                 [E500_CR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMFTB,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_SFX0]>],
+                                 [E500mc_CR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMFTB,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_SFX0]>],
                                  [7, 1], // Latency = 4, Repeat rate = 4
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMTSPR,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0, E500_SFX1]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMTSPR,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0, E500mc_SFX1]>],
                                  [4, 1], // Latency = 1, Repeat rate = 1
-                                 [E500_CR_Bypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_SprMTSRIN,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<1, [E500_SFX0]>],
+                                 [E500mc_CR_Bypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_SprMTSRIN,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<1, [E500mc_SFX0]>],
                                  [4, 1],
-                                 [NoBypass, E500_GPR_Bypass]>,
-  InstrItinData<IIC_FPGeneral,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_FPU_0]>],
+                                 [NoBypass, E500mc_GPR_Bypass]>,
+  InstrItinData<IIC_FPGeneral,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<2, [E500mc_FPU_0]>],
                                  [11, 1, 1], // Latency = 8, Repeat rate = 2
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPAddSub,    [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPAddSub,    [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_FPU_0]>],
                                  [13, 1, 1], // Latency = 10, Repeat rate = 4
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPCompare,   [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<2, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPCompare,   [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<2, [E500mc_FPU_0]>],
                                  [11, 1, 1], // Latency = 8, Repeat rate = 2
-                                 [E500_CR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPDivD,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<68, [E500_FPU_0]>],
+                                 [E500mc_CR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPDivD,      [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<68, [E500mc_FPU_0]>],
                                  [71, 1, 1], // Latency = 68, Repeat rate = 68
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPDivS,      [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<38, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPDivS,      [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<38, [E500mc_FPU_0]>],
                                  [41, 1, 1], // Latency = 38, Repeat rate = 38
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPFused,     [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<4, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPFused,     [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<4, [E500mc_FPU_0]>],
                                  [13, 1, 1, 1], // Latency = 10, Repeat rate = 4
-                                 [E500_FPR_Bypass,
-                                  E500_FPR_Bypass, E500_FPR_Bypass,
-                                  E500_FPR_Bypass]>,
-  InstrItinData<IIC_FPRes,       [InstrStage<1, [E500_DIS0, E500_DIS1], 0>,
-                                  InstrStage<38, [E500_FPU_0]>],
+                                 [E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass, E500mc_FPR_Bypass,
+                                  E500mc_FPR_Bypass]>,
+  InstrItinData<IIC_FPRes,       [InstrStage<1, [E500mc_DIS0, E500mc_DIS1], 0>,
+                                  InstrStage<38, [E500mc_FPU_0]>],
                                  [41, 1], // Latency = 38, Repeat rate = 38
-                                 [E500_FPR_Bypass, E500_FPR_Bypass]>
+                                 [E500mc_FPR_Bypass, E500mc_FPR_Bypass]>
 ]>;
 
 // ===---------------------------------------------------------------------===//
 // e500mc machine model for scheduling and other instruction cost heuristics.
 
 def PPCE500mcModel : SchedMachineModel {
   let IssueWidth = 2;  // 2 micro-ops are dispatched per cycle.
   let LoadLatency = 5; // Optimistic load latency assuming bypass.
                        // This is overriden by OperandCycles if the
                        // Itineraries are queried instead.
 
   let CompleteModel = 0;
 
   let Itineraries = PPCE500mcItineraries;
 }

lib/Target/PowerPC/PPCSubtarget.h

enum {
     DIR_440,
     DIR_601,
     DIR_602,
     DIR_603,
     DIR_7400,
     DIR_750,
     DIR_970,
     DIR_A2,
+    DIR_E500,
     DIR_E500mc,
     DIR_E5500,
     DIR_PWR3,
     DIR_PWR4,
     DIR_PWR5,
     DIR_PWR5X,
     DIR_PWR6,
     DIR_PWR6X,

test/CodeGen/PowerPC/spe.ll

+; RUN: llc -verify-machineinstrs < %s -mtriple=powerpc-unknown-linux-gnu \
+; RUN:          -mattr=+spe |  FileCheck %s
+
+declare float @llvm.fabs.float(float)
+define float @test_float_abs(float %a) #0 {
+; CHECK-LABEL: test_float_abs
+  entry:
+    %0 = tail call float @llvm.fabs.float(float %a)
+    ret float %0
+}
+; CHECK: efsabs 3, 3
+; CHECK: blr
+
+define float @test_fnabs(float %a) #0 {
+  entry:
+    %0 = tail call float @llvm.fabs.float(float %a)
+    %sub = fsub float -0.000000e+00, %0
+    ret float %sub
+}
+; CHECK-LABEL: @test_fnabs
+; CHECK-NO: efsnabs
+; CHECK: blr
+
+define float @test_fdiv(float %a, float %b) {
+entry:
+  %v = fdiv float %a, %b
+  ret float %v
+
+; CHECK-LABEL: test_fdiv
+; CHECK: efsdiv 
+; CHECK: blr
+}
+
+define float @test_fmul(float %a, float %b) {
+  entry:
+  %v = fmul float %a, %b
+  ret float %v
+; CHECK-LABEL @test_fmul
+; CHECK: efsmul
+; CHECK: blr
+}
+
+define float @test_fadd(float %a, float %b) {
+  entry:
+  %v = fadd float %a, %b
+  ret float %v
+; CHECK-LABEL @test_fadd
+; CHECK: efsadd
+; CHECK: blr
+}
+
+define float @test_fsub(float %a, float %b) {
+  entry:
+  %v = fsub float %a, %b
+  ret float %v
+; CHECK-LABEL @test_fsub
+; CHECK: efssub
+; CHECK: blr
+}
+
+define float @test_fneg(float %a) {
+  entry:
+  %v = fsub float -0.0, %a
+  ret float %v
+
+; CHECK-LABEL @test_fneg
+; CHECK: efsneg
+; CHECK: blr
+}
+
+define float @test_dtos(double %a) {
+  entry:
+  %v = fptrunc double %a to float
+  ret float %v
+; CHECK-LABEL: test_dtos
+; CHECK: efscfd
+; CHECK: blr
+}
+
+define i1 @test_fcmpgt(float %a, float %b) {
+  entry:
+  %r = fcmp ogt float %a, %b
+  ret i1 %r
+}
+
+define i1 @test_fcmpeq(float %a, float %b) {
+  entry:
+  %r = fcmp oeq float %a, %b
+  ret i1 %r
+}
+
+define i1 @test_fcmplt(float %a, float %b) {
+  entry:
+  %r = fcmp olt float %a, %b
+  ret i1 %r
+}
+
+define i32 @test_ftoui(float %a) {
+  %v = fptoui float %a to i32
+  ret i32 %v
+}
+
+define i32 @test_ftosi(float %a) {
+  %v = fptosi float %a to i32
+  ret i32 %v
+}
+
+define float @test_ffromui(i32 %a) {
+  %v = uitofp i32 %a to float
+  ret float %v
+}
+
+define float @test_ffromsi(i32 %a) {
+  %v = sitofp i32 %a to float
+  ret float %v
+}
+
+; Double tests
+
+define double @test_double_abs(double %aa) #0 {
+
+; CHECK-LABEL: test_double_abs
+
+  entry:
+    %0 = tail call double @llvm.fabs.f64(double %aa) #2
+    ret double %0
+}
+; CHECK: blr
+
+; Function Attrs: nounwind readnone
+declare double @llvm.fabs.f64(double) #1
+
+define double @test_dnabs(double %aa) #0 {
+  entry:
+    %0 = tail call double @llvm.fabs.f64(double %aa) #2
+    %sub = fsub double -0.000000e+00, %0
+    ;%add2 = fadd double %aa, %sub
+    ;ret double %add2
+    %add = fadd double %sub, 1.0
+    ret double %add
+}
+; CHECK-LABEL: @test_dnabs
+; CHECK-NO: efdnabs
+; CHECK: blr
+
+define double @test_ddiv(double %a, double %b) {
+entry:
+  %v = fdiv double %a, %b
+  ret double %v
+
+; CHECK-LABEL: test_ddiv
+; CHECK: efddiv 
+; CHECK: blr
+}
+
+define double @test_dmul(double %a, double %b) {
+  entry:
+  %v = fmul double %a, %b
+  ret double %v
+; CHECK-LABEL @test_dmul
+; CHECK: efdmul
+; CHECK: blr
+}
+
+define double @test_dadd(double %a, double %b) {
+  entry:
+  %v = fadd double %a, %b
+  ret double %v
+; CHECK-LABEL @test_dadd
+; CHECK: efdadd
+; CHECK: blr
+}
+
+define double @test_dsub(double %a, double %b) {
+  entry:
+  %v = fsub double %a, %b
+  ret double %v
+; CHECK-LABEL @test_dsub
+; CHECK: efdsub
+; CHECK: blr
+}
+
+define double @test_dneg(double %a) {
+  entry:
+  %v = fsub double -0.0, %a
+  ret double %v
+
+; CHECK-LABEL @test_dneg
+; CHECK: blr
+}
+
+define double @test_stod(float %a) {
+  entry:
+  %v = fpext float %a to double
+  ret double %v
+; CHECK-LABEL: test_stod
+; CHECK: efdcfs
+; CHECK: blr
+}
+
+define i1 @test_dcmpgt(double %a, double %b) {
+  entry:
+  %r = fcmp ogt double %a, %b
+  ret i1 %r
+}
+
+define i1 @test_dcmpeq(double %a, double %b) {
+  entry:
+  %r = fcmp oeq double %a, %b
+  ret i1 %r
+}
+
+define i1 @test_dcmplt(double %a, double %b) {
+  entry:
+  %r = fcmp olt double %a, %b
+  ret i1 %r
+}
+
+define i32 @test_dtoui(double %a) {
+  %v = fptoui double %a to i32
+  ret i32 %v
+}
+
+define i32 @test_dtosi(double %a) {
+  %v = fptosi double %a to i32
+  ret i32 %v
+}
+
+define double @test_dfromui(i32 %a) {
+  %v = uitofp i32 %a to double
+  ret double %v
+}
+
+define double @test_dfromsi(i32 %a) {
+  %v = sitofp i32 %a to double
+  ret double %v
+}
+
+; Vector float tests
+
+define <2 x float> @test_float_abs_v(<2 x float> %aa) #0 {
+
+; CHECK-LABEL: test_float_abs_v
+
+  entry:
+    %0 = tail call <2 x float> @llvm.fabs.v2f32(<2 x float> %aa) #2
+    ret <2 x float> %0
+}
+; Function Attrs: nounwind readnone
+declare <2 x float> @llvm.fabs.v2f32(<2 x float>) #1
+
+; CHECK: evfsabs 3, 3
+; CHECK: blr
+
+define <2 x float> @test2_float_abs_v(<2 x float> %aa) #0 {
+
+; CHECK-LABEL: test2_float_abs_v
+
+  entry:
+    %0 = tail call <2 x float> @llvm.fabs.v2f32(<2 x float> %aa) #2
+    %sub = fsub <2 x float> <float -0.000000e+00, float -0.000000e+00>, %0
+    ret <2 x float> %sub
+}
+
+; CHECK: evfsnabs 3, 3
+; CHECK: blr
+
+define <2 x float> @test_fneg_v(<2 x float> %a) {
+  entry:
+  %v = fsub <2 x float> <float-0.0, float -0.0>, %a
+  ret <2 x float> %v
+
+; CHECK-LABEL @test_fneg_v
+; CHECK: evfsneg
+; CHECK: blr
+}
+
+define <2 x float> @test_fdiv_v(<2 x float> %a, <2 x float> %b) {
+entry:
+  %v = fdiv <2 x float> %a, %b
+  ret <2 x float> %v
+
+; CHECK-LABEL: test_fdiv_v
+; CHECK: evfsdiv 
+; CHECK: blr
+}
+
+define <2 x float> @test_fmul_v(<2 x float> %a, <2 x float> %b) {
+  entry:
+  %v = fmul <2 x float> %a, %b
+  ret <2 x float> %v
+; CHECK-LABEL @test_fmul_v
+; CHECK: evfsmul
+; CHECK: blr
+}
+
+define <2 x float> @test_fadd_v(<2 x float> %a, <2 x float> %b) {
+  entry:
+  %v = fadd <2 x float> %a, %b
+  ret <2 x float> %v
+; CHECK-LABEL @test_fadd_v
+; CHECK: evfsadd
+; CHECK: blr
+}
+
+define <2 x float> @test_fsub_v(<2 x float> %a, <2 x float> %b) {
+  entry:
+  %v = fsub <2 x float> %a, %b
+  ret <2 x float> %v
+; CHECK-LABEL @test_fsub_v
+; CHECK: evfssub
+; CHECK: blr
+}
+
+;define <2 x i1> @test_fcmpgt_v(<2 x float> %a, <2 x float> %b) {
+;  entry:
+;  %r = fcmp ogt <2 x float> %a, %b
+;  ret <2 x i1> %r
+;}
+;
+;define <2 x i1> @test_fcmpeq_v(<2 x float> %a, <2 x float> %b) {
+;  entry:
+;  %r = fcmp oeq <2 x float> %a, %b
+;  ret <2 x i1> %r
+;}
+;
+;define <2 x i1> @test_fcmplt_v(<2 x float> %a, <2 x float> %b) {
+;  entry:
+;  %r = fcmp olt <2 x float> %a, %b
+;  ret <2 x i1> %r
+;}
+;
+define <2 x i32> @test_ftoui_v(<2 x float> %a) {
+  %v = fptoui <2 x float> %a to <2 x i32>
+  ret <2 x i32> %v
+}
+
+define <2 x i32> @test_ftosi_v(<2 x float> %a) {
+  %v = fptosi <2 x float> %a to <2 x i32>
+  ret <2 x i32> %v
+}
+
+define <2 x float> @test_ffromui_v(<2 x i32> %a) {
+  %v = uitofp <2 x i32> %a to <2 x float>
+  ret <2 x float> %v
+}
+
+define <2 x float> @test_ffromsi_v(<2 x i32> %a) {
+  %v = sitofp <2 x i32> %a to <2 x float>
+  ret <2 x float> %v
+}
+
+; Vector int tests
+
+define <2 x i32> @test_i32_abs_v(<2 x i32> %aa) #0 {
+
+; CHECK-LABEL: test_i32_abs_v
+
+  entry:
+    %0 = tail call <2 x i32> @llvm.ppc.spe.evabs(<2 x i32> %aa) #2
+    ret <2 x i32> %0
+}
+
+declare <2 x i32> @llvm.ppc.spe.evabs(<2 x i32>) #1
+
+; CHECK: evabs 3, 3
+; CHECK: blr
+
+define <2 x i32> @test_neg_v(<2 x i32> %a) {
+  entry:
+  %v = sub <2 x i32> zeroinitializer, %a
+  ret <2 x i32> %v
+
+; CHECK-LABEL @test_neg_v
+; CHECK: evneg
+; CHECK: blr
+}
+
+define <2 x i32> @test_nor_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = or <2 x i32> %a, %b
+  %r = xor <2 x i32> %v, <i32 -1, i32 -1>
+  ret <2 x i32> %r
+
+; CHECK-LABEL @test_nor_v
+; CHECK: evnor
+; CHECK: blr
+}
+
+define <2 x i32> @test_orc_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = xor <2 x i32> %b, <i32 -1, i32 -1>
+  %r = or <2 x i32> %a, %v
+  ret <2 x i32> %r
+
+; CHECK-LABEL @test_orc_v
+; CHECK: evorc
+; CHECK: blr
+}
+
+define <2 x i32> @test_nand_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = and <2 x i32> %a, %b
+  %r = xor <2 x i32> %v, <i32 -1, i32 -1>
+  ret <2 x i32> %r
+
+; CHECK-LABEL @test_nand_v
+; CHECK: evnand
+; CHECK: blr
+}
+
+define <2 x i32> @test_andc_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = xor <2 x i32> %b, <i32 -1, i32 -1>
+  %r = and <2 x i32> %a, %v
+  ret <2 x i32> %r
+
+; CHECK-LABEL @test_andc_v
+; CHECK: evandc
+; CHECK: blr
+}
+
+define <2 x i32> @test_xor_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = xor <2 x i32> %a, %b
+  ret <2 x i32> %v
+
+; CHECK-LABEL @test_xor_v
+; CHECK: evxor
+; CHECK: blr
+}
+
+define <2 x i32> @test_slw_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = shl <2 x i32> %a, %b
+  ret <2 x i32> %v
+}
+
+define <2 x i32> @test_srws_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = ashr <2 x i32> %a, %b
+  ret <2 x i32> %v
+}
+
+define <2 x i32> @test_srwu_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = lshr <2 x i32> %a, %b
+  ret <2 x i32> %v
+}
+
+define <2 x i32> @test_divs_v(<2 x i32> %a, <2 x i32> %b) {
+entry:
+  %v = sdiv <2 x i32> %a, %b
+  ret <2 x i32> %v
+
+; CHECK-LABEL: test_divs_v
+; CHECK: evdivws
+; CHECK: blr
+}
+
+define <2 x i32> @test_divu_v(<2 x i32> %a, <2 x i32> %b) {
+entry:
+  %v = udiv <2 x i32> %a, %b
+  ret <2 x i32> %v
+
+; CHECK-LABEL: test_divu_v
+; CHECK: evdivwu
+; CHECK: blr
+}
+
+define <2 x i32> @test_mul_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = mul <2 x i32> %a, %b
+  ret <2 x i32> %v
+; CHECK-LABEL @test_mul_v
+; CHECK: evmwlumi
+; CHECK: blr
+}
+
+define <2 x i32> @test_add_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = add <2 x i32> %a, %b
+  ret <2 x i32> %v
+; CHECK-LABEL @test_add_v
+; CHECK: evadd
+; CHECK: blr
+}
+
+define <2 x i32> @test_sub_v(<2 x i32> %a, <2 x i32> %b) {
+  entry:
+  %v = sub <2 x i32> %a, %b
+  ret <2 x i32> %v
+; CHECK-LABEL @test_sub_v
+; CHECK: evsubf
+; CHECK: blr
+}
+
+;define <2 x i1> @test_cmpgt_v(<2 x i32> %a, <2 x i32> %b) {
+;  entry:
+;  %r = cmp ogt <2 x i32> %a, %b
+;  ret <2 x i1> %r
+;}
+;
+;define <2 x i1> @test_cmpeq_v(<2 x i32> %a, <2 x i32> %b) {
+;  entry:
+;  %r = cmp oeq <2 x i32> %a, %b
+;  ret <2 x i1> %r
+;}
+;
+;define <2 x i1> @test_cmplt_v(<2 x i32> %a, <2 x i32> %b) {
+;  entry:
+;  %r = cmp olt <2 x i32> %a, %b
+;  ret <2 x i1> %r
+;}
+;

test/MC/PowerPC/ppc64-encoding-spe.s

-# RUN: llvm-mc -triple powerpc64-unknown-unknown --show-encoding %s | FileCheck -check-prefix=CHECK-BE %s
-# RUN: llvm-mc -triple powerpc64le-unknown-unknown --show-encoding %s | FileCheck -check-prefix=CHECK-LE %s
+# RUN: llvm-mc -triple powerpc64-unknown-unknown --show-encoding %s | FileCheck  -check-prefix=CHECK-BE %s
+# RUN: llvm-mc -triple powerpc64le-unknown-unknown --show-encoding %s | FileCheck  -check-prefix=CHECK-LE %s
 
 # Instructions from the Signal Processing Engine extension:
 
 # CHECK-BE: evlddx 14, 21, 28               # encoding: [0x11,0xd5,0xe3,0x00]
 # CHECK-LE: evlddx 14, 21, 28               # encoding: [0x00,0xe3,0xd5,0x11]
             evlddx %r14, %r21, %r28
 # CHECK-BE: evldwx 14, 21, 28               # encoding: [0x11,0xd5,0xe3,0x02]
 # CHECK-LE: evldwx 14, 21, 28               # encoding: [0x02,0xe3,0xd5,0x11]

 # CHECK-LE: evstwwo 14, 0(27)               # encoding: [0x3d,0x03,0xdb,0x11]
             evstwwo %r14, 0(%r27)
 # CHECK-BE: evstwwo 14, 124(27)             # encoding: [0x11,0xdb,0xfb,0x3d]
 # CHECK-LE: evstwwo 14, 124(27)             # encoding: [0x3d,0xfb,0xdb,0x11]
             evstwwo %r14, 124(%r27)
 # CHECK-BE: evstwwo 14, 124(9)              # encoding: [0x11,0xc9,0xfb,0x3d]
 # CHECK-LE: evstwwo 14, 124(9)              # encoding: [0x3d,0xfb,0xc9,0x11]
             evstwwo %r14, 124(%r9)
+
+# CHECK-BE: efdabs 3, 4                     # encoding: [0x10,0x64,0x02,0xe4]
+# CHECK-LE: efdabs 3, 4                     # encoding: [0xe4,0x02,0x64,0x10]
+            efdabs %r3, %r4
+# CHECK-BE: efdadd 3, 4, 5                  # encoding: [0x10,0x64,0x2a,0xe0]
+# CHECK-LE: efdadd 3, 4, 5                  # encoding: [0xe0,0x2a,0x64,0x10]
+            efdadd %r3, %r4, %r5
+# CHECK-BE: efdcfs 3, 4                     # encoding: [0x10,0x60,0x22,0xef]
+# CHECK-LE: efdcfs 3, 4                     # encoding: [0xef,0x22,0x60,0x10]
+            efdcfs %r3, %r4
+# CHECK-BE: efdcfsf 5, 6                    # encoding: [0x10,0xa0,0x32,0xf3]
+# CHECK-LE: efdcfsf 5, 6                    # encoding: [0xf3,0x32,0xa0,0x10]
+            efdcfsf %r5, %r6
+# CHECK-BE: efdcfsi 5, 6                    # encoding: [0x10,0xa0,0x32,0xf1]
+# CHECK-LE: efdcfsi 5, 6                    # encoding: [0xf1,0x32,0xa0,0x10]
+            efdcfsi %r5, %r6
+# CHECK-BE: efdcfsid 10, 14                 # encoding: [0x11,0x40,0x72,0xe3]
+# CHECK-LE: efdcfsid 10, 14                 # encoding: [0xe3,0x72,0x40,0x11]
+            efdcfsid %r10, %r14
+# CHECK-BE: efdcfuf 5, 8                    # encoding: [0x10,0xa0,0x42,0xf2]
+# CHECK-LE: efdcfuf 5, 8                    # encoding: [0xf2,0x42,0xa0,0x10]
+            efdcfuf %r5, %r8
+# CHECK-BE: efdcfui 6, 9                    # encoding: [0x10,0xc0,0x4a,0xf0]
+# CHECK-LE: efdcfui 6, 9                    # encoding: [0xf0,0x4a,0xc0,0x10]
+            efdcfui %r6, %r9
+# CHECK-BE: efdcfuid 7, 10                  # encoding: [0x10,0xe0,0x52,0xe2]
+# CHECK-LE: efdcfuid 7, 10                  # encoding: [0xe2,0x52,0xe0,0x10]
+            efdcfuid %r7, %r10
+# CHECK-BE: efdcmpeq 3, 3, 8                # encoding: [0x11,0x83,0x42,0xee]
+# CHECK-LE: efdcmpeq 3, 3, 8                # encoding: [0xee,0x42,0x83,0x11]
+            efdcmpeq %cr3, %r3, %r8
+# CHECK-BE: efdcmpgt 4, 7, 3                # encoding: [0x12,0x07,0x1a,0xec]
+# CHECK-LE: efdcmpgt 4, 7, 3                # encoding: [0xec,0x1a,0x07,0x12]
+            efdcmpgt %cr4, %r7, %r3
+# CHECK-BE: efdcmplt 2, 3, 4                # encoding: [0x11,0x03,0x22,0xed]
+# CHECK-LE: efdcmplt 2, 3, 4                # encoding: [0xed,0x22,0x03,0x11]
+            efdcmplt %cr2, %r3, %r4
+# CHECK-BE: efdctsf 5, 3                    # encoding: [0x10,0xa0,0x1a,0xf7]
+# CHECK-LE: efdctsf 5, 3                    # encoding: [0xf7,0x1a,0xa0,0x10]
+            efdctsf %r5, %r3
+# CHECK-BE: efdctsi 6, 4                    # encoding: [0x10,0xc0,0x22,0xf5]
+# CHECK-LE: efdctsi 6, 4                    # encoding: [0xf5,0x22,0xc0,0x10]
+            efdctsi %r6, %r4
+# CHECK-BE: efdctsidz 3, 4                  # encoding: [0x10,0x60,0x22,0xeb]
+# CHECK-LE: efdctsidz 3, 4                  # encoding: [0xeb,0x22,0x60,0x10]
+            efdctsidz %r3, %r4
+# CHECK-BE: efdctsiz 3, 4                   # encoding: [0x10,0x60,0x22,0xfa]
+# CHECK-LE: efdctsiz 3, 4                   # encoding: [0xfa,0x22,0x60,0x10]
+            efdctsiz %r3, %r4
+# CHECK-BE: efdctuf 5, 8                    # encoding: [0x10,0xa0,0x42,0xf6]
+# CHECK-LE: efdctuf 5, 8                    # encoding: [0xf6,0x42,0xa0,0x10]
+            efdctuf %r5, %r8
+# CHECK-BE: efdctui 9, 10                   # encoding: [0x11,0x20,0x52,0xf4]
+# CHECK-LE: efdctui 9, 10                   # encoding: [0xf4,0x52,0x20,0x11]
+            efdctui %r9, %r10
+# CHECK-BE: efdctuidz 3, 8                  # encoding: [0x10,0x60,0x42,0xea]
+# CHECK-LE: efdctuidz 3, 8                  # encoding: [0xea,0x42,0x60,0x10]
+            efdctuidz %r3, %r8
+# CHECK-BE: efdctuiz 5, 17                  # encoding: [0x10,0xa0,0x8a,0xf8]
+# CHECK-LE: efdctuiz 5, 17                  # encoding: [0xf8,0x8a,0xa0,0x10]
+            efdctuiz %r5, %r17
+# CHECK-BE: efddiv 3, 4, 5                  # encoding: [0x10,0x64,0x2a,0xe9]
+# CHECK-LE: efddiv 3, 4, 5                  # encoding: [0xe9,0x2a,0x64,0x10]
+            efddiv %r3, %r4, %r5
+# CHECK-BE: efdmul 0, 3, 8                  # encoding: [0x10,0x03,0x42,0xe8]
+# CHECK-LE: efdmul 0, 3, 8                  # encoding: [0xe8,0x42,0x03,0x10]
+            efdmul %r0, %r3, %r8
+# CHECK-BE: efdnabs 3, 23                   # encoding: [0x10,0x77,0x02,0xe5]
+# CHECK-LE: efdnabs 3, 23                   # encoding: [0xe5,0x02,0x77,0x10]
+            efdnabs %r3, %r23
+# CHECK-BE: efdneg 3, 22                    # encoding: [0x10,0x76,0x02,0xe6]
+# CHECK-LE: efdneg 3, 22                    # encoding: [0xe6,0x02,0x76,0x10]
+            efdneg %r3, %r22
+# CHECK-BE: efdsub 3, 4, 6                  # encoding: [0x10,0x64,0x32,0xe1]
+# CHECK-LE: efdsub 3, 4, 6                  # encoding: [0xe1,0x32,0x64,0x10]
+            efdsub %r3, %r4, %r6
+# CHECK-BE: efdtsteq 3, 4, 5                # encoding: [0x11,0x84,0x2a,0xfe]
+# CHECK-LE: efdtsteq 3, 4, 5                # encoding: [0xfe,0x2a,0x84,0x11]
+            efdtsteq %cr3, %r4, %r5
+# CHECK-BE: efdtstgt 3, 3, 6                # encoding: [0x11,0x83,0x32,0xfc]
+# CHECK-LE: efdtstgt 3, 3, 6                # encoding: [0xfc,0x32,0x83,0x11]
+            efdtstgt %cr3, %r3, %r6
+# CHECK-BE: efdtstlt 4, 0, 3                # encoding: [0x12,0x00,0x1a,0xfd]
+# CHECK-LE: efdtstlt 4, 0, 3                # encoding: [0xfd,0x1a,0x00,0x12]
+            efdtstlt %cr4, %r0, %r3
+# CHECK-BE: efsabs 3, 4                     # encoding: [0x10,0x64,0x02,0xc4]
+# CHECK-LE: efsabs 3, 4                     # encoding: [0xc4,0x02,0x64,0x10]
+            efsabs %r3, %r4
+# CHECK-BE: efsadd 3, 4, 5                  # encoding: [0x10,0x64,0x2a,0xc0]
+# CHECK-LE: efsadd 3, 4, 5                  # encoding: [0xc0,0x2a,0x64,0x10]
+            efsadd %r3, %r4, %r5
+# CHECK-BE: efscfsf 5, 6                    # encoding: [0x10,0xa0,0x32,0xd3]
+# CHECK-LE: efscfsf 5, 6                    # encoding: [0xd3,0x32,0xa0,0x10]
+            efscfsf %r5, %r6
+# CHECK-BE: efscfsi 5, 6                    # encoding: [0x10,0xa0,0x32,0xd1]
+# CHECK-LE: efscfsi 5, 6                    # encoding: [0xd1,0x32,0xa0,0x10]
+            efscfsi %r5, %r6
+# CHECK-BE: efscfuf 5, 8                    # encoding: [0x10,0xa0,0x42,0xd2]
+# CHECK-LE: efscfuf 5, 8                    # encoding: [0xd2,0x42,0xa0,0x10]
+            efscfuf %r5, %r8
+# CHECK-BE: efscfui 6, 9                    # encoding: [0x10,0xc0,0x4a,0xd0]
+# CHECK-LE: efscfui 6, 9                    # encoding: [0xd0,0x4a,0xc0,0x10]
+            efscfui %r6, %r9
+# CHECK-BE: efscmpeq 3, 3, 8                # encoding: [0x11,0x83,0x42,0xce]
+# CHECK-LE: efscmpeq 3, 3, 8                # encoding: [0xce,0x42,0x83,0x11]
+            efscmpeq %cr3, %r3, %r8
+# CHECK-BE: efscmpgt 4, 7, 3                # encoding: [0x12,0x07,0x1a,0xcc]
+# CHECK-LE: efscmpgt 4, 7, 3                # encoding: [0xcc,0x1a,0x07,0x12]
+            efscmpgt %cr4, %r7, %r3
+# CHECK-BE: efscmplt 2, 3, 4                # encoding: [0x11,0x03,0x22,0xcd]
+# CHECK-LE: efscmplt 2, 3, 4                # encoding: [0xcd,0x22,0x03,0x11]
+            efscmplt %cr2, %r3, %r4
+# CHECK-BE: efsctsf 5, 3                    # encoding: [0x10,0xa0,0x1a,0xd7]
+# CHECK-LE: efsctsf 5, 3                    # encoding: [0xd7,0x1a,0xa0,0x10]
+            efsctsf %r5, %r3
+# CHECK-BE: efsctsi 6, 4                    # encoding: [0x10,0xc0,0x22,0xd5]
+# CHECK-LE: efsctsi 6, 4                    # encoding: [0xd5,0x22,0xc0,0x10]
+            efsctsi %r6, %r4
+# CHECK-BE: efsctsiz 3, 4                   # encoding: [0x10,0x60,0x22,0xda]
+# CHECK-LE: efsctsiz 3, 4                   # encoding: [0xda,0x22,0x60,0x10]
+            efsctsiz %r3, %r4
+# CHECK-BE: efsctuf 5, 8                    # encoding: [0x10,0xa0,0x42,0xd6]
+# CHECK-LE: efsctuf 5, 8                    # encoding: [0xd6,0x42,0xa0,0x10]
+            efsctuf %r5, %r8
+# CHECK-BE: efsctui 9, 10                   # encoding: [0x11,0x20,0x52,0xd4]
+# CHECK-LE: efsctui 9, 10                   # encoding: [0xd4,0x52,0x20,0x11]
+            efsctui %r9, %r10
+# CHECK-BE: efsctuiz 5, 17                  # encoding: [0x10,0xa0,0x8a,0xd8]
+# CHECK-LE: efsctuiz 5, 17                  # encoding: [0xd8,0x8a,0xa0,0x10]
+            efsctuiz %r5, %r17
+# CHECK-BE: efsdiv 3, 4, 5                  # encoding: [0x10,0x64,0x2a,0xc9]
+# CHECK-LE: efsdiv 3, 4, 5                  # encoding: [0xc9,0x2a,0x64,0x10]
+            efsdiv %r3, %r4, %r5
+# CHECK-BE: efsmul 0, 3, 8                  # encoding: [0x10,0x03,0x42,0xc8]
+# CHECK-LE: efsmul 0, 3, 8                  # encoding: [0xc8,0x42,0x03,0x10]
+            efsmul %r0, %r3, %r8
+# CHECK-BE: efsnabs 3, 23                   # encoding: [0x10,0x77,0x02,0xc5]
+# CHECK-LE: efsnabs 3, 23                   # encoding: [0xc5,0x02,0x77,0x10]
+            efsnabs %r3, %r23
+# CHECK-BE: efsneg 3, 22                    # encoding: [0x10,0x76,0x02,0xc6]
+# CHECK-LE: efsneg 3, 22                    # encoding: [0xc6,0x02,0x76,0x10]
+            efsneg %r3, %r22
+# CHECK-BE: efssub 3, 4, 6                  # encoding: [0x10,0x64,0x32,0xc1]
+# CHECK-LE: efssub 3, 4, 6                  # encoding: [0xc1,0x32,0x64,0x10]
+            efssub %r3, %r4, %r6
+# CHECK-BE: efststeq 3, 4, 5                # encoding: [0x11,0x84,0x2a,0xde]
+# CHECK-LE: efststeq 3, 4, 5                # encoding: [0xde,0x2a,0x84,0x11]
+            efststeq %cr3, %r4, %r5
+# CHECK-BE: efststgt 3, 3, 6                # encoding: [0x11,0x83,0x32,0xdc]
+# CHECK-LE: efststgt 3, 3, 6                # encoding: [0xdc,0x32,0x83,0x11]
+            efststgt %cr3, %r3, %r6
+# CHECK-BE: efststlt 4, 0, 3                # encoding: [0x12,0x00,0x1a,0xdd]
+# CHECK-LE: efststlt 4, 0, 3                # encoding: [0xdd,0x1a,0x00,0x12]
+            efststlt %cr4, %r0, %r3