Add Marvell ThunderX3T110 support

[wxz2020]

Summary

This is the first checkin to support Marvell ThunderX3T110.

Initial definition of the micro-ops of the instructions in ThunderX3T110 is included.

[wxz2020]

The newly added two files were not accepted by the pre-merge check. Re worked on the diff file and uploaded it now.

[wxz2020]

Resubmit as the previous one was rejected by test plan changes

[wxz2020]

fix a format issus.

[wxz2020]

Fix a format issue.

[ktkachov]

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

IIRC these instructions are deliberately allowed in pre-armv8.3 targets because they are encoded in the NOP-space and can be deployed on pre-armv8.3 targets

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

10

Typo in the processor name?

[wxz2020]

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

I will do some research on this.

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

10

Sure, thanks for spotting it. I will have it fixed.

[wxz2020]

Remove non-official target options. We now only support -mcpu=thunderx3t110

Typo fix.

[wxz2020]

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

According to the documents, pointer authenticatoin got officially supporoted starting from armv8.3.

[chill]

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

These instructions are executed as NOP on pre v8.3-A architectures. It allows you to have a single compatible binary that works correctly on pre v8.3-a (ofc, without pointer authentication), as well as on
8.3-a and later cores, with pointer authentication.

Please, remove the predicates.

[DavidSpickett]

clang/lib/Driver/ToolChains/Arch/AArch64.cpp

89 ↗

(On Diff #260680)

Remove stray change.

llvm/include/llvm/Support/AArch64TargetParser.def

167–171

Is this change also correcting the options for the x2t99? I think that should be a separate patch if so.

[DavidSpickett]

llvm/include/llvm/Support/AArch64TargetParser.def

168

Add a tests for this in llvm/unittests/Support/TargetParserTest.cpp (there are existing ones for "thunderx2t99").
Might need to update "NumAArch64CPUArchs" too.

[SjoerdMeijer]

llvm/lib/Target/AArch64/AArch64.td

856

HasV8_3aOps implies HasV8_2aOps, which implies HasV8_1aOps.
So you can just remove HasV8_1aOps?

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

14

I don't intend to check the numbers here, but just curious if there's an optimisation guide if people are curious?

llvm/test/CodeGen/AArch64/loop-micro-op-buffer-size-t110.ll

1 ↗

(On Diff #260680)

Couple of question about this test.
Looks like you're both testing output on stdout and stderr, and your checks for this are interleaved. If I am not mistaken, the behaviour of this can vary on different platforms, and thus may fail on different platforms.

But more importantly, what is this test exactly testing? The file name gives away a hint of course, but I don't see yet how this interact with loop unrolling. Is the loop unroller looking at MaxInterleaveFactor that is set here, but is that what we are testing here?

25 ↗

(On Diff #260680)

I guess there won't be another define %val, it will be %val6, so this CHECK-NOT will never match even if there's another add?

[wxz2020]

clang/lib/Driver/ToolChains/Arch/AArch64.cpp

89 ↗

(On Diff #260680)

Removed

llvm/include/llvm/Support/AArch64TargetParser.def

167–171

I will have it removed. irrelevant to this change.

168

OK, we will have it fixed

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

Agree. Will remove it.

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

14

We will release a document about the details soon.

llvm/test/CodeGen/AArch64/loop-micro-op-buffer-size-t110.ll

25 ↗

(On Diff #260680)

This test is for our internal use. Sorry, Will have it removed.

[wxz2020]

fixed all the feedback suggestions. Thanks,

[joelkevinjones]

In email Wei asked for help about he following error message:

error message from tblgen
Included from /home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64.td:439:
/home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64InstrInfo.td:961:5: error: 'CycloneModel' lacks information for 'AUTDZA'

def DZA  : SignAuthZero<prefix_z, 0b10, !strconcat(asm, "dza")>;
^
In the end:
Incomplete schedule models found.

·         Consider setting 'CompleteModel = 0' while developing new models.
·         Pseudo instructions can be marked with 'hasNoSchedulingInfo = 1'.
·         Instructions should usually have Sched<[...]> as a superclass, you may temporarily use an empty list.
·         Instructions related to unsupported features can be excluded with list<Predicate> UnsupportedFeatures = [HasA,..,HasY]; in the processor model.
error: Incomplete schedule model

And the comment from the person on the email was to define the instruction. The instruction is defined, as evidenced when table gen is run to produce the record-list. The error message can be suppressed by defining CompleteModele = 0, but that isn't correct, as there are models for those instructions.

However, I now note that at least In the output that Wei captured, it isn't complaining about the ThunderX3T110 model, but about Cyclone.

[SjoerdMeijer]

In D78129#2022307, @joelkevinjones wrote:

In email Wei asked for help about he following error message:

error message from tblgen
Included from /home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64.td:439:
/home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64InstrInfo.td:961:5: error: 'CycloneModel' lacks information for 'AUTDZA'

def DZA  : SignAuthZero<prefix_z, 0b10, !strconcat(asm, "dza")>;
^
In the end:
Incomplete schedule models found.

·         Consider setting 'CompleteModel = 0' while developing new models.
·         Pseudo instructions can be marked with 'hasNoSchedulingInfo = 1'.
·         Instructions should usually have Sched<[...]> as a superclass, you may temporarily use an empty list.
·         Instructions related to unsupported features can be excluded with list<Predicate> UnsupportedFeatures = [HasA,..,HasY]; in the processor model.
error: Incomplete schedule model

And the comment from the person on the email was to define the instruction. The instruction is defined, as evidenced when table gen is run to produce the record-list. The error message can be suppressed by defining CompleteModele = 0, but that isn't correct, as there are models for those instructions.

However, I now note that at least In the output that Wei captured, it isn't complaining about the ThunderX3T110 model, but about Cyclone.

Hi Joel,
Thanks for your message, and apologies for not replying earlier to the updates; missed the notification.
I am now taking a closer at some changes I hadn't looked at. In the mean, some drive-by comments:

It is okay'ish to set CompleteModel = 0 if you're not interested in describing all instructions. The other way is of course to add the missing instruction to the model.

But first things first. @chill commented earlier about predicates. I also don't think you need to add HasV8_3a for the reason Momchill described, and there is still one left AArch64InstrInfo.td.

And while I am still looking into this, let me ask the question what I am wondering: why do you need to change the other scheduling models? These instructions are always supported, as a NOP, or otherwise.

[joelkevinjones]

It is okay'ish to set CompleteModel = 0 if you're not interested in describing all instructions. The other way is of course to add the missing instruction to the model.

We are trying to describe all instructions. The issue Wei is having is that the pointer auth instructions are being called out in the error message, but their timing information is defined—see the very end of AArch64SchedThunderX3T110.td for AUTDZA.

But first things first. @chill commented earlier about predicates. I also don't think you need to add HasV8_3a for the reason Momchill described, and there is still one left AArch64InstrInfo.td.

And while I am still looking into this, let me ask the question what I am wondering: why do you need to change the other scheduling models? These instructions are always supported, as a NOP, or otherwise.

The changes to UnsupportedFeatures that was in every model has been taken out. I'm not sure why UnsupportedFeatures exists at all, but it isn't just AArch64 that has this.

Should the latest version be put up even if it fails on table-gen? That would make things easier to review.

[wxz2020]

I tried to set CompleteModel = 0 on line 25 in file llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

The same error is still there.

llvm/lib/Target/AArch64/AArch64InstrInfo.td

849–857 ↗

(On Diff #259737)

One questions here, we want to model these instructions in our machine model, scroll down to line 1992-1997, in
llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td below.

However, if we remove this predicate here, tblgen will complain the following errors for all the instructions defined in line 1992-1997:

error message from tblgen

Included from /home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64.td:439:
/home/wei/project/tx3/llvm-project/llvm/lib/Target/AArch64/AArch64InstrInfo.td:961:5: error: 'CycloneModel' lacks information for 'AUTDZA'

def DZA  : SignAuthZero<prefix_z, 0b10, !strconcat(asm, "dza")>;
^

In the end:
Incomplete schedule models found.

• Consider setting 'CompleteModel = 0' while developing new models.
• Pseudo instructions can be marked with 'hasNoSchedulingInfo = 1'.
• Instructions should usually have Sched<[...]> as a superclass, you may temporarily use an empty list.
• Instructions related to unsupported features can be excluded with list<Predicate> UnsupportedFeatures = [HasA,..,HasY]; in the processor model.

error: Incomplete schedule model

Anyone has clue why we have this problem? How should I model these instructions in the .md file.

Thanks,

[SjoerdMeijer]

Since debugging on phabricator is a bit difficult, I took the patch and had a little play. Now I actually remember seeing this before. I think the way this works is that when you describe new instructions (PA in this case) that other models don't have, they start complaining about incomplete models when they have set let CompleteModel = 1

I got things compiling with these changes:

1. leave all existing models untouched, revert those changes,

2. remove that predicate HasPA that you added,

3. Change this:

def SVEUnsupported : AArch64Unsupported {
let F = [HasSVE, HasSVE2, HasSVE2AES, HasSVE2SM4, HasSVE2SHA3,

• HasSVE2BitPerm]; + HasSVE2BitPerm, HasPA]; }

4. In your new model set:

let CompleteModel = 0;
list<Predicate> UnsupportedFeatures = SVEUnsupported.F;

5. This is where the problem is, comment out these descriptions in your model:

def : InstRW<[THX3T110Write_7Cyc_I1],
(instrs AUTDA, AUTDB, AUTDZA, AUTDZB, AUTIA, AUTIA1716, AUTIASP,
AUTIAZ, AUTIB, AUTIB1716, AUTIBSP, AUTIBZ, AUTIZA, AUTIZB,
PACDA, PACDB, PACDZA, PACDZB, PACGA, PACIA, PACIA1716,
PACIASP, PACIAZ, PACIB, PACIB1716, PACIBSP, PACIBZ,
PACIZA, PACIZB, XPACD, XPACI, XPACLRI)>;

I have never really looked into these instructions, and quick grep doesn't show any descriptions for some. For example, AUTDA doesn't exist. The others look aliases, perhaps they need a different treatment, or a regexp, I don't know, but perhaps you can take it from here.

[wxz2020]

I can confirm your findings, basically as long as we comment out the new instructions, it will be OK to build.

Agree with your changes 1), 2) and 3).

4. this seems does not matter to build, as long as we comment out the new instructions in 5), setting it "1" or "0", both are OK. they all can be compiled

5. if we comment out these instructions, then we will not be able to schedule them efficiently in the instruction scheduler. I know in other models or earlier versions, they could be treated as NOP, but in our new machine, they do have a timing effect, 7 cycles to finish. Any way to represent this in .md?

Thanks a lot for the help,

[joelkevinjones]

I don't think it makes sense to combine two unrelated things SVE and PA support into a combined thing. Since we already have UnsupportedFeatures in every sub-target .td file, I think it would be better to instead have:

def PAUnsupported : AArch64Unsupported {
  let F = [HasPA];
}

and modify each .td file to have

list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F, PAUnsupported.F);

[evandro]

In D78129#2023772, @joelkevinjones wrote:

list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F, PAUnsupported.F);

Agree. Methinks that this is easier to read.

[SjoerdMeijer]

Agreed about the UnsupportedFeatures list. But I just wanted to unblock this work, create a first version that compiles, so you can pick it up and clean things further up.

[wxz2020]

I put every people's feedbacks into the code and upload them here.

I think we can use it as the start point for TX3, and add more when we move on.

Thank you all for great help,

[wxz2020]

I will put the "PAUnsupported" predicate later once this got passed.

[SjoerdMeijer]

Let's separate out HasPA from SVEFeatures now while we are at it, probably it's more work to do this as a follow up, and after that this looks good to me.

Bonus points for adding some llvm-mca tests, see the llvm-project/llvm/test/tools/llvm-mca/ directory for some inspiration how this latencies can be checked, but that seems like something for follow-up.

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

1993

just remove them then instead of commenting them out.

[wxz2020]

Removed the PA related instructions from the .md file as suggested.

[wxz2020]

I think I made all the changes mentioned in the feedbacks. Please take sometime to review the current version. We are eager to get the first thunderx3t110 checked in.

There are a few things we plan to do after the 1st version:

1. add more tests as suggested in the review
2. make changes to the predicate for clarity and easy readability
3. solve the PA instruction issue

Let me know if there is anything missing. Thanks,

[SjoerdMeijer]

In D78129#2023772, @joelkevinjones wrote:

I don't think it makes sense to combine two unrelated things SVE and PA support into a combined thing. Since we already have UnsupportedFeatures in every sub-target .td file, I think it would be better to instead have:

def PAUnsupported : AArch64Unsupported {
  let F = [HasPA];
}

and modify each .td file to have

list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F, PAUnsupported.F);

Can you add what Joel suggested? I don't see the point of doing this as a follow up. This very simple thing is done half right here, so is best fixed here as there is no reason to leave this for another time; following up on this and talking about this is more work than just doing it.

[wxz2020]

Sure I will add Joel's suggestion and other fixes later.

Thanks,

[wxz2020]

1. Added the predicates as suggested
2. brought back the TargetParserTest.cpp which was missed last time

[wxz2020]

Fix the format issue in TargetParserTest.cpp

[wxz2020]

fix a typo, thunderx3t110, not thunderx3t100 in TargetParserTest.cpp

[SjoerdMeijer]

This looks good now, but sorry, one more request: I've just noticed a Clang driver test is missing. Can you add a test for this to clang/test/Driver/aarch64-cpus.c? And related to this, the relevant tests to llvm/unittests/Support/TargetParserTest.cpp?

[wxz2020]

1. in clang/test/Driver/aarch64-cpus.c, expanded the the same tests for thunderx2t99 to thunderx3t110
2. in llvm/unittest/Support/TargetParserTest.cpp, fix the missing target features

So far all the existing issues are fixed according to review feedbacks

Thanks

[wxz2020]

Correct the wrong version of llvm/unittests/Support/TargetParserTest.cpp which causes the test failure.

[SjoerdMeijer]

Thanks, LGTM

[SjoerdMeijer]

Forgot to ask/add: can you commit this, do you have commit rights?

[wxz2020]

Joel will help me to commit it. Thanks,

[wxz2020]

I have a question here. Our customers need the TX3 support urgently, they asked me to put this TX3 support to LLVM 10.0.1 and if there is 9.0.2 release planned, they also need it there.

How can I do this? To put this checkin to LLVM 10.0.1 branch and 9.0.2 branch?

Is it possible? What steps I should follow?

Thanks,

[SjoerdMeijer]

Best to ask the release manager. Today Tom Stellard posted a message to the llvm dev list with subject:

[llvm-dev] LLVM 10.0.1-rc1 release update

So best to reply to this(*) asap with your query if it's possible to get this onto the branch.

(*) http://lists.llvm.org/pipermail/llvm-dev/2020-May/141670.html

[wxz2020]

Thanks, Just sent Tom an email.

clang/test/Driver/aarch64-cpus.c

 // RUN: %clang -target aarch64_be -mlittle-endian -mcpu=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99 %s
 // RUN: %clang -target aarch64 -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-TUNE %s
 // RUN: %clang -target aarch64 -mlittle-endian -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-TUNE %s
 // RUN: %clang -target aarch64_be -mlittle-endian -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-TUNE %s
 // THUNDERX2T99: "-cc1"{{.*}} "-triple" "aarch64{{(--)?}}"{{.*}} "-target-cpu" "thunderx2t99" "-target-feature" "+v8.1a"
 // THUNDERX2T99-TUNE: "-cc1"{{.*}} "-triple" "aarch64{{(--)?}}"{{.*}} "-target-cpu" "generic"
 // THUNDERX2T99-TUNE-NOT: +v8.1a
 
+// RUN: %clang -target aarch64 -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110 %s
+// RUN: %clang -target aarch64 -mlittle-endian -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110 %s
+// RUN: %clang -target aarch64_be -mlittle-endian -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110 %s
+// RUN: %clang -target aarch64 -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-TUNE %s
+// RUN: %clang -target aarch64 -mlittle-endian -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-TUNE %s
+// RUN: %clang -target aarch64_be -mlittle-endian -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-TUNE %s
+// THUNDERX3T110: "-cc1"{{.*}} "-triple" "aarch64{{(--)?}}"{{.*}} "-target-cpu" "thunderx3t110" "-target-feature" "+v8.3a"
+// THUNDERX3T110-TUNE: "-cc1"{{.*}} "-triple" "aarch64{{(--)?}}"{{.*}} "-target-cpu" "generic"
+
 // RUN: %clang -target arm64 -mcpu=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX2T99 %s
 // RUN: %clang -target arm64 -mlittle-endian -mcpu=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX2T99 %s
 // RUN: %clang -target arm64 -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX2T99-TUNE %s
 // RUN: %clang -target arm64 -mlittle-endian -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX2T99-TUNE %s
 // ARM64-THUNDERX2T99: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "thunderx2t99" "-target-feature" "+v8.1a"
 // ARM64-THUNDERX2T99-TUNE: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "generic"
 // ARM64-THUNDERX2T99-TUNE-NOT: +v8.1a
 
+// RUN: %clang -target arm64 -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX3T110 %s
+// RUN: %clang -target arm64 -mlittle-endian -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX3T110 %s
+// RUN: %clang -target arm64 -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX3T110-TUNE %s
+// RUN: %clang -target arm64 -mlittle-endian -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-THUNDERX3T110-TUNE %s
+// ARM64-THUNDERX3T110: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "thunderx3t110" "-target-feature" "+v8.3a"
+// ARM64-THUNDERX3T110-TUNE: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "generic"
+
 // RUN: %clang -target aarch64 -mcpu=a64fx -### -c %s 2>&1 | FileCheck -check-prefix=A64FX %s
 // RUN: %clang -target aarch64 -mlittle-endian -mcpu=a64fx -### -c %s 2>&1 | FileCheck -check-prefix=A64FX %s
 // RUN: %clang -target aarch64 -mtune=a64fx -### -c %s 2>&1 | FileCheck -check-prefix=A64FX-TUNE %s
 // RUN: %clang -target aarch64 -mlittle-endian -mtune=a64fx -### -c %s 2>&1 | FileCheck -check-prefix=A64FX-TUNE %s
 // A64FX: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "a64fx"
 // A64FX-TUNE: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "generic"
 
 // RUN: %clang -target arm64 -mcpu=a64fx -### -c %s 2>&1 | FileCheck -check-prefix=ARM64-A64FX %s

 // RUN: %clang -target aarch64 -mbig-endian -mcpu=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-BE %s
 // RUN: %clang -target aarch64_be -mbig-endian -mcpu=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-BE %s
 // RUN: %clang -target aarch64_be -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-BE-TUNE %s
 // RUN: %clang -target aarch64 -mbig-endian -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-BE-TUNE %s
 // RUN: %clang -target aarch64_be -mbig-endian -mtune=thunderx2t99 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX2T99-BE-TUNE %s
 // THUNDERX2T99-BE: "-cc1"{{.*}} "-triple" "aarch64_be{{.*}}" "-target-cpu" "thunderx2t99"
 // THUNDERX2T99-BE-TUNE: "-cc1"{{.*}} "-triple" "aarch64_be{{.*}}" "-target-cpu" "generic"
 
+// RUN: %clang -target aarch64_be -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE %s
+// RUN: %clang -target aarch64 -mbig-endian -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE %s
+// RUN: %clang -target aarch64_be -mbig-endian -mcpu=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE %s
+// RUN: %clang -target aarch64_be -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE-TUNE %s
+// RUN: %clang -target aarch64 -mbig-endian -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE-TUNE %s
+// RUN: %clang -target aarch64_be -mbig-endian -mtune=thunderx3t110 -### -c %s 2>&1 | FileCheck -check-prefix=THUNDERX3T110-BE-TUNE %s
+// THUNDERX3T110-BE: "-cc1"{{.*}} "-triple" "aarch64_be{{.*}}" "-target-cpu" "thunderx3t110"
+// THUNDERX3T110-BE-TUNE: "-cc1"{{.*}} "-triple" "aarch64_be{{.*}}" "-target-cpu" "generic"
+
 // RUN: %clang -target aarch64 -mcpu=cortex-a57 -mtune=cortex-a53 -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-A57 %s
 // RUN: %clang -target aarch64 -mtune=cortex-a53 -mcpu=cortex-a57  -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-A57 %s
 // RUN: %clang -target aarch64 -mcpu=cortex-a72 -mtune=cortex-a53 -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-A72 %s
 // RUN: %clang -target aarch64 -mtune=cortex-a53 -mcpu=cortex-a72  -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-A72 %s
 // RUN: %clang -target aarch64 -mtune=cortex-a53 -mcpu=cortex-a73     -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-A73 %s
 // RUN: %clang -target aarch64 -mcpu=thunderx2t99 -mtune=cortex-a53 -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-THUNDERX2T99 %s
 // RUN: %clang -target aarch64 -mtune=cortex-a53 -mcpu=thunderx2t99  -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-THUNDERX2T99 %s
+// RUN: %clang -target aarch64 -mcpu=thunderx3t110 -mtune=cortex-a53 -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-THUNDERX3T110 %s
+// RUN: %clang -target aarch64 -mtune=cortex-a53 -mcpu=thunderx3t110  -### -c %s 2>&1 | FileCheck -check-prefix=MCPU-MTUNE-THUNDERX3T110 %s
 // MCPU-MTUNE-A57: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "cortex-a57"
 // MCPU-MTUNE-A72: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "cortex-a72"
 // MCPU-MTUNE-A73: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "cortex-a73"
 // MCPU-MTUNE-THUNDERX2T99: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "thunderx2t99"
+// MCPU-MTUNE-THUNDERX3T110: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "thunderx3t110"
 
 // RUN: %clang -target aarch64 -march=armv8.1a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // RUN: %clang -target aarch64 -march=armv8.1-a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // RUN: %clang -target aarch64 -mlittle-endian -march=armv8.1a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // RUN: %clang -target aarch64 -mlittle-endian -march=armv8.1-a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // RUN: %clang -target aarch64_be -mlittle-endian -march=armv8.1a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // RUN: %clang -target aarch64_be -mlittle-endian -march=armv8.1-a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV81A %s
 // GENERICV81A: "-cc1"{{.*}} "-triple" "aarch64{{(--)?}}"{{.*}} "-target-cpu" "generic" "-target-feature" "+neon" "-target-feature" "+v8.1a"

llvm/include/llvm/Support/AArch64TargetParser.def

AARCH64_CPU_NAME("exynos-m5", ARMV8_2A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_DOTPROD | AArch64::AEK_FP16))
 AARCH64_CPU_NAME("falkor", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC | AArch64::AEK_RDM))
 AARCH64_CPU_NAME("saphira", ARMV8_3A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_PROFILE))
 AARCH64_CPU_NAME("kryo", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC))
 AARCH64_CPU_NAME("thunderx2t99", ARMV8_1A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_NONE))
+AARCH64_CPU_NAME("thunderx3t110", ARMV8_3A, FK_CRYPTO_NEON_FP_ARMV8, false,
+                 (AArch64::AEK_CRC | AEK_CRYPTO | AEK_FP | AEK_SIMD |
+                  AEK_LSE | AEK_RAND | AArch64::AEK_PROFILE |
+                  AArch64::AEK_RAS))
 AARCH64_CPU_NAME("thunderx", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC | AArch64::AEK_PROFILE))
 AARCH64_CPU_NAME("thunderxt88", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC | AArch64::AEK_PROFILE))
 AARCH64_CPU_NAME("thunderxt81", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC | AArch64::AEK_PROFILE))
 AARCH64_CPU_NAME("thunderxt83", ARMV8A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_CRC | AArch64::AEK_PROFILE))

llvm/lib/Target/AArch64/AArch64.td

 
 class AArch64Unsupported { list<Predicate> F; }
 
 def SVEUnsupported : AArch64Unsupported {
   let F = [HasSVE, HasSVE2, HasSVE2AES, HasSVE2SM4, HasSVE2SHA3,
            HasSVE2BitPerm];
 }
 
+def PAUnsupported : AArch64Unsupported {
+  let F = [HasPA];
+}
+
 include "AArch64SchedA53.td"
 include "AArch64SchedA57.td"
 include "AArch64SchedCyclone.td"
 include "AArch64SchedFalkor.td"
 include "AArch64SchedKryo.td"
 include "AArch64SchedExynosM3.td"
 include "AArch64SchedExynosM4.td"
 include "AArch64SchedExynosM5.td"
 include "AArch64SchedThunderX.td"
 include "AArch64SchedThunderX2T99.td"
+include "AArch64SchedThunderX3T110.td"
 
 def ProcA35     : SubtargetFeature<"a35", "ARMProcFamily", "CortexA35",
                                    "Cortex-A35 ARM processors", [
                                    FeatureCRC,
                                    FeatureCrypto,
                                    FeatureFPARMv8,
                                    FeatureNEON,
                                    FeaturePerfMon
def ProcThunderX2T99  : SubtargetFeature<"thunderx2t99", "ARMProcFamily",
                                           FeatureFPARMv8,
                                           FeatureArithmeticBccFusion,
                                           FeatureNEON,
                                           FeaturePostRAScheduler,
                                           FeaturePredictableSelectIsExpensive,
                                           FeatureLSE,
                                           HasV8_1aOps]>;
 
+def ProcThunderX3T110  : SubtargetFeature<"thunderx3t110", "ARMProcFamily",
+                                         "ThunderX3T110",
+                                         "Marvell ThunderX3 processors", [
+                                          FeatureAggressiveFMA,
+                                          FeatureCRC,
+                                          FeatureCrypto,
+                                          FeatureFPARMv8,
+                                          FeatureArithmeticBccFusion,
+                                          FeatureNEON,
+                                          FeaturePostRAScheduler,
+                                          FeaturePredictableSelectIsExpensive,
+                                          FeatureLSE,
+                                          FeaturePA,
+                                          FeatureUseAA,
+                                          FeatureBalanceFPOps,
+                                          FeaturePerfMon,
+                                          FeatureStrictAlign,
+                                          HasV8_3aOps]>;
+
 def ProcThunderX : SubtargetFeature<"thunderx", "ARMProcFamily", "ThunderX",
                                     "Cavium ThunderX processors", [
                                     FeatureCRC,
                                     FeatureCrypto,
                                     FeatureFPARMv8,
                                     FeaturePerfMon,
                                     FeaturePostRAScheduler,
                                     FeaturePredictableSelectIsExpensive,

 def : ProcessorModel<"kryo", KryoModel, [ProcKryo]>;
 // Cavium ThunderX/ThunderX T8X  Processors
 def : ProcessorModel<"thunderx", ThunderXT8XModel,  [ProcThunderX]>;
 def : ProcessorModel<"thunderxt88", ThunderXT8XModel,  [ProcThunderXT88]>;
 def : ProcessorModel<"thunderxt81", ThunderXT8XModel,  [ProcThunderXT81]>;
 def : ProcessorModel<"thunderxt83", ThunderXT8XModel,  [ProcThunderXT83]>;
 // Cavium ThunderX2T9X  Processors. Formerly Broadcom Vulcan.
 def : ProcessorModel<"thunderx2t99", ThunderX2T99Model, [ProcThunderX2T99]>;
+// Marvell ThunderX3T110 Processors.
+def : ProcessorModel<"thunderx3t110", ThunderX3T110Model, [ProcThunderX3T110]>;
 // FIXME: HiSilicon TSV110 is currently modeled as a Cortex-A57.
 def : ProcessorModel<"tsv110", CortexA57Model, [ProcTSV110]>;
 
 // Support cyclone as an alias for apple-a7 so we can still LTO old bitcode.
 def : ProcessorModel<"cyclone", CycloneModel, [ProcAppleA7]>;
 
 // iPhone and iPad CPUs
 def : ProcessorModel<"apple-a7", CycloneModel, [ProcAppleA7]>;

llvm/lib/Target/AArch64/AArch64SchedA53.td

def CortexA53Model : SchedMachineModel {
   let LoadLatency = 3;       // Optimistic load latency assuming bypass.
                              // This is overriden by OperandCycles if the
                              // Itineraries are queried instead.
   let MispredictPenalty = 9; // Based on "Cortex-A53 Software Optimisation
                              // Specification - Instruction Timings"
                              // v 1.0 Spreadsheet
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available.
 
 // Modeling each pipeline as a ProcResource using the BufferSize = 0 since
 // Cortex-A53 is in-order.

llvm/lib/Target/AArch64/AArch64SchedA57.td

def CortexA57Model : SchedMachineModel {
   let LoadLatency       =   4; // Optimistic load latency
   let MispredictPenalty =  14; // Fetch + Decode/Rename/Dispatch + Branch
 
   // Enable partial & runtime unrolling. The magic number is chosen based on
   // experiments and benchmarking data.
   let LoopMicroOpBufferSize = 16;
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on Cortex-A57.
 // Cortex A-57 has 8 pipelines that each has its own 8-entry queue where
 // micro-ops wait for their operands and then issue out-of-order.
 
 def A57UnitB : ProcResource<1>;  // Type B micro-ops

llvm/lib/Target/AArch64/AArch64SchedCyclone.td

 
 def CycloneModel : SchedMachineModel {
   let IssueWidth = 6; // 6 micro-ops are dispatched per cycle.
   let MicroOpBufferSize = 192; // Based on the reorder buffer.
   let LoadLatency = 4; // Optimistic load latency.
   let MispredictPenalty = 16; // 14-19 cycles are typical.
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on Cyclone.
 
 // 4 integer pipes
 def CyUnitI : ProcResource<4> {
   let BufferSize = 48;

llvm/lib/Target/AArch64/AArch64SchedExynosM3.td

 def ExynosM3Model : SchedMachineModel {
   let IssueWidth            =   6; // Up to 6 uops per cycle.
   let MicroOpBufferSize     = 228; // ROB size.
   let LoopMicroOpBufferSize =  40; // Based on the instruction queue size.
   let LoadLatency           =   4; // Optimistic load cases.
   let MispredictPenalty     =  16; // Minimum branch misprediction penalty.
   let CompleteModel         =   1; // Use the default model otherwise.
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on the Exynos-M3,
 // which has 12 pipelines, each with its own queue with out-of-order dispatch.
 
 let SchedModel = ExynosM3Model in {
 

llvm/lib/Target/AArch64/AArch64SchedExynosM4.td

 def ExynosM4Model : SchedMachineModel {
   let IssueWidth            =   6; // Up to 6 uops per cycle.
   let MicroOpBufferSize     = 228; // ROB size.
   let LoopMicroOpBufferSize =  48; // Based on the instruction queue size.
   let LoadLatency           =   4; // Optimistic load cases.
   let MispredictPenalty     =  16; // Minimum branch misprediction penalty.
   let CompleteModel         =   1; // Use the default model otherwise.
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on the Exynos-M4.
 
 let SchedModel = ExynosM4Model in {
 
 def M4UnitA  : ProcResource<2>; // Simple integer

llvm/lib/Target/AArch64/AArch64SchedExynosM5.td

 def ExynosM5Model : SchedMachineModel {
   let IssueWidth            =   6; // Up to 6 uops per cycle.
   let MicroOpBufferSize     = 228; // ROB size.
   let LoopMicroOpBufferSize =  60; // Based on the instruction queue size.
   let LoadLatency           =   4; // Optimistic load cases.
   let MispredictPenalty     =  15; // Minimum branch misprediction penalty.
   let CompleteModel         =   1; // Use the default model otherwise.
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on the Exynos-M5.
 
 let SchedModel = ExynosM5Model in {
 
 def M5UnitA  : ProcResource<2>; // Simple integer

llvm/lib/Target/AArch64/AArch64SchedFalkor.td

 def FalkorModel : SchedMachineModel {
   let IssueWidth = 8;          // 8 uops are dispatched per cycle.
   let MicroOpBufferSize = 128; // Out-of-order with temporary unified issue buffer.
   let LoopMicroOpBufferSize = 16;
   let LoadLatency = 3;         // Optimistic load latency.
   let MispredictPenalty = 11;  // Minimum branch misprediction penalty.
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
-
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
   // FIXME: Remove when all errors have been fixed.
   let FullInstRWOverlapCheck = 0;
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on Falkor.
 
 let SchedModel = FalkorModel in {

llvm/lib/Target/AArch64/AArch64SchedKryo.td

def KryoModel : SchedMachineModel {
   let LoadLatency       =   4; // Optimistic load latency
   let MispredictPenalty =  14; // Fetch + Decode/Rename/Dispatch + Branch
 
   // Enable partial & runtime unrolling. The magic number is chosen based on
   // experiments and benchmarking data.
   let LoopMicroOpBufferSize = 16;
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
-
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
   // FIXME: Remove when all errors have been fixed.
   let FullInstRWOverlapCheck = 0;
 }
 
 //===----------------------------------------------------------------------===//
 // Define each kind of processor resource and number available on Kryo.
 
 let SchedModel = KryoModel in {

llvm/lib/Target/AArch64/AArch64SchedThunderX.td

 def ThunderXT8XModel : SchedMachineModel {
   let IssueWidth = 2;         // 2 micro-ops dispatched per cycle.
   let MicroOpBufferSize = 0;  // ThunderX T88/T81/T83 are in-order.
   let LoadLatency = 3;        // Optimistic load latency.
   let MispredictPenalty = 8;  // Branch mispredict penalty.
   let PostRAScheduler = 1;    // Use PostRA scheduler.
   let CompleteModel = 1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
-
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
   // FIXME: Remove when all errors have been fixed.
   let FullInstRWOverlapCheck = 0;
 }
 
 // Modeling each pipeline with BufferSize == 0 since T8X is in-order.
 def THXT8XUnitALU    : ProcResource<2> { let BufferSize = 0; } // Int ALU
 def THXT8XUnitMAC    : ProcResource<1> { let BufferSize = 0; } // Int MAC
 def THXT8XUnitDiv    : ProcResource<1> { let BufferSize = 0; } // Int Division

llvm/lib/Target/AArch64/AArch64SchedThunderX2T99.td

def ThunderX2T99Model : SchedMachineModel {
   let MicroOpBufferSize     = 180; // 180 entries in micro-op re-order buffer.
   let LoadLatency           =   4; // Optimistic load latency.
   let MispredictPenalty     =  12; // Extra cycles for mispredicted branch.
   // Determined via a mix of micro-arch details and experimentation.
   let LoopMicroOpBufferSize = 128;
   let PostRAScheduler       =   1; // Using PostRA sched.
   let CompleteModel         =   1;
 
-  list<Predicate> UnsupportedFeatures = SVEUnsupported.F;
-
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
   // FIXME: Remove when all errors have been fixed.
   let FullInstRWOverlapCheck = 0;
 }
 
 let SchedModel = ThunderX2T99Model in {
 
 // Define the issue ports.
 

llvm/lib/Target/AArch64/AArch64SchedThunderX3T110.td

+//=- AArch64SchedThunderX3T110.td - Marvell ThunderX3 T110 ---*- tablegen -*-=//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the scheduling model for Marvell ThunderX3T110
+// family of processors.
+//
+//===----------------------------------------------------------------------===//
+
+//===----------------------------------------------------------------------===//
+// Pipeline Description.
+
+def ThunderX3T110Model : SchedMachineModel {
+  let IssueWidth            =   4; // 4 micro-ops dispatched at a time.
+  let MicroOpBufferSize     =  70; // 70 entries in micro-op re-order buffer.
+  let LoadLatency           =   4; // Optimistic load latency.
+  let MispredictPenalty     =  12; // Extra cycles for mispredicted branch.
+  // Determined via a mix of micro-arch details and experimentation.
+  let LoopMicroOpBufferSize = 128; // FIXME: might be much bigger in TX3.
+  let PostRAScheduler       =   1; // Using PostRA sched.
+  let CompleteModel         =   1;
+
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    PAUnsupported.F);
+  // FIXME: Remove when all errors have been fixed.
+  let FullInstRWOverlapCheck = 0;
+}
+
+let SchedModel = ThunderX3T110Model in {
+
+// Issue ports.
+
+// Port 0: ALU.
+def THX3T110P0 : ProcResource<1>;
+
+// Port 1: ALU.
+def THX3T110P1 : ProcResource<1>;
+
+// Port 2: ALU/Branch.
+def THX3T110P2 : ProcResource<1>;
+
+// Port 3: ALU/Branch.
+def THX3T110P3 : ProcResource<1>;
+
+// Port 4: Load/Store.
+def THX3T110P4 : ProcResource<1>;
+
+// Port 5: Load/store.
+def THX3T110P5 : ProcResource<1>;
+
+// Port 6: FP/Neon/SIMD/Crypto.
+def THX3T110P6FP0 : ProcResource<1>;
+
+// Port 7: FP/Neon/SIMD/Crypto.
+def THX3T110P7FP1 : ProcResource<1>;
+
+// Port 8: FP/Neon/SIMD/Crypto.
+def THX3T110P8FP2 : ProcResource<1>;
+
+// Port 9: FP/Neon/SIMD/Crypto.
+def THX3T110P9FP3 : ProcResource<1>;
+
+// Port 10: Store Data Unit.
+def THX3T110SD0 : ProcResource<1>;
+
+// Define groups for the functional units on each issue port.  Each group
+// created will be used by a WriteRes.
+
+// Integer divide/mulhi micro-ops only on port I1.
+def THX3T110I1 : ProcResGroup<[THX3T110P1]>;
+
+// Branch micro-ops on ports I2/I3.
+def THX3T110I23 : ProcResGroup<[THX3T110P2, THX3T110P3]>;
+
+// Branch micro-ops on ports I1/I2/I3.
+def THX3T110I123 : ProcResGroup<[THX3T110P1, THX3T110P2, THX3T110P3]>;
+
+// Integer micro-ops on ports I0/I1/I2.
+def THX3T110I012 : ProcResGroup<[THX3T110P0, THX3T110P1, THX3T110P2]>;
+
+// Integer micro-ops on ports I0/I1/I2/I3.
+def THX3T110I0123 : ProcResGroup<[THX3T110P0, THX3T110P1,
+                                  THX3T110P2, THX3T110P3]>;
+
+// FP micro-ops on ports FP0/FP1/FP2/FP3.
+def THX3T110FP0123 : ProcResGroup<[THX3T110P6FP0, THX3T110P7FP1,
+                                   THX3T110P8FP2, THX3T110P9FP3]>;
+
+// FP micro-ops on ports FP2/FP3.
+def THX3T110FP23 : ProcResGroup<[THX3T110P8FP2, THX3T110P9FP3]>;
+
+// ASIMD micro-ops on ports FP0/FP1/FP2/FP3.
+def THX3T110SIMD : ProcResGroup<[THX3T110P6FP0, THX3T110P7FP1,
+                                 THX3T110P8FP2, THX3T110P9FP3]>;
+
+// Store data micro-ops only on port 10.
+def THX3T110SD : ProcResGroup<[THX3T110SD0]>;
+
+// Load/store micro-ops on ports P4/P5.
+def THX3T110LS : ProcResGroup<[THX3T110P4, THX3T110P5]>;
+
+// 70 entry unified scheduler.
+def THX3T110ANY: ProcResGroup<[THX3T110P0, THX3T110P1, THX3T110P2,
+                               THX3T110P3, THX3T110P4, THX3T110P5,
+                               THX3T110P6FP0, THX3T110P7FP1,
+                               THX3T110P8FP2, THX3T110P9FP3]> {
+  let BufferSize = 70;
+}
+
+// Define commonly used write types for InstRW specializations.
+// All definitions follow the format: THX3T110Write_<NumCycles>Cyc_<Resources>.
+
+// 3 cycles on I1.
+def THX3T110Write_3Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 3;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on I1.
+def THX3T110Write_4Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+// 5 cycles on I1.
+def THX3T110Write_5Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+// 7 cycles on I1.
+def THX3T110Write_7Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 7;
+  let NumMicroOps = 3;
+}
+
+// 23 cycles on I1.
+def THX3T110Write_23Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 23;
+  let ResourceCycles = [13, 23];
+  let NumMicroOps = 4;
+}
+
+// 39 cycles on I1.
+def THX3T110Write_39Cyc_I1 : SchedWriteRes<[THX3T110I1]> {
+  let Latency = 39;
+  let ResourceCycles = [13, 39];
+  let NumMicroOps = 4;
+}
+
+// 1 cycle on I2/I3
+def THX3T110Write_1Cyc_I23 : SchedWriteRes<[THX3T110I23]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 8 cycles on I2/I3
+def THX3T110Write_8Cyc_I23 : SchedWriteRes<[THX3T110I23]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+// 1 cycle on I1/I2/I3
+def THX3T110Write_1Cyc_I123 : SchedWriteRes<[THX3T110I123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 8 cycles on I1/I2/I3
+def THX3T110Write_8Cyc_I123 : SchedWriteRes<[THX3T110I123]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+// 1 cycle on I0/I1/I2/I3.
+def THX3T110Write_1Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 2 cycles on I0/I1/I2/I3.
+def THX3T110Write_2Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+// 3 cycles on I0/I1/I2/I3.
+def THX3T110Write_3Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 3;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on I0/I1/I2/I3.
+def THX3T110Write_4Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+// 5 cycles on I0/I1/I2/I3.
+def THX3T110Write_5Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// 6 cycles on I0/I1/I2/I3.
+def THX3T110Write_6Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+// 8 cycles on I0/I1/I2/I3.
+def THX3T110Write_8Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 8;
+  let NumMicroOps = 4;
+}
+
+// 13 cycles on I0/I1/I2/I3.
+def THX3T110Write_13Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 13;
+  let NumMicroOps = 3;
+}
+
+// 23 cycles on I0/I1/I2/I3.
+def THX3T110Write_23Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 23;
+  let NumMicroOps = 3;
+}
+
+// 39 cycles on I0/I1/I2/I3.
+def THX3T110Write_39Cyc_I0123 : SchedWriteRes<[THX3T110I0123]> {
+  let Latency = 39;
+  let NumMicroOps = 3;
+}
+
+// 4 cycles on F2/F3.
+def THX3T110Write_4Cyc_F23 : SchedWriteRes<[THX3T110FP23]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+// 5 cycles on F0/F1/F2/F3.
+def THX3T110Write_5Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+// 6 cycles on F0/F1/F2/F3.
+def THX3T110Write_6Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+// 7 cycles on F0/F1/F2/F3.
+def THX3T110Write_7Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 7;
+  let NumMicroOps = 3;
+}
+
+// 8 cycles on F0/F1/F2/F3.
+def THX3T110Write_8Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+// 10 cycles on F0/F1/F2/F3.
+def THX3T110Write_10Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 10;
+  let NumMicroOps = 3;
+}
+
+// 16 cycles on F0/F1/F2/F3.
+def THX3T110Write_16Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 16;
+  let NumMicroOps = 3;
+  let ResourceCycles = [8];
+}
+
+// 23 cycles on F0/F1/F2/F3.
+def THX3T110Write_23Cyc_F01 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 23;
+  let NumMicroOps = 3;
+  let ResourceCycles = [11];
+}
+
+// 1 cycle on LS0/LS1.
+def THX3T110Write_1Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+// 2 cycles on LS0/LS1.
+def THX3T110Write_2Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on LS0/LS1.
+def THX3T110Write_4Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+  let ResourceCycles = [2];
+}
+
+// 5 cycles on LS0/LS1.
+def THX3T110Write_5Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// 6 cycles on LS0/LS1.
+def THX3T110Write_6Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+// 4 + 5 cycles on LS0/LS1.
+// First resource is available after 4 cycles.
+// Second resource is available after 5 cycles.
+// Load vector pair, immed offset, Q-form [LDP/LDNP].
+def THX3T110Write_4_5Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+  let ResourceCycles = [4, 5];
+}
+
+// 4 + 8 cycles on LS0/LS1.
+// First resource is available after 4 cycles.
+// Second resource is available after 8 cycles.
+// Load vector pair, immed offset, S/D-form [LDP/LDNP].
+def THX3T110Write_4_8Cyc_LS01 : SchedWriteRes<[THX3T110LS]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+  let ResourceCycles = [4, 8];
+}
+
+// 11 cycles on LS0/LS1 and I1.
+def THX3T110Write_11Cyc_LS01_I1 :
+  SchedWriteRes<[THX3T110LS, THX3T110I1]> {
+  let Latency = 11;
+  let NumMicroOps = 4;
+}
+
+// 1 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_1Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 1 cycles on LS0/LS1 and 2 of I0/I1/I2/I3.
+def THX3T110Write_1Cyc_LS01_I0123_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123, THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 3;
+}
+
+// 4 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_4Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+// 4 cycles on LS0/LS1 and 2 of I0/I1/I2/I3.
+def THX3T110Write_4Cyc_LS01_I0123_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123, THX3T110I0123]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+// 5 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_5Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// 5 cycles on LS0/LS1 and 2 of I0/I1/I2/I3.
+def THX3T110Write_5Cyc_LS01_I0123_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123, THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// 6 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_6Cyc_LS01_I012 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 6;
+  let NumMicroOps = 4;
+}
+
+// 6 cycles on LS0/LS1 and 2 of I0/I1/I2/I3.
+def THX3T110Write_6Cyc_LS01_I0123_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123, THX3T110I0123]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+// 1 cycle on LS0/LS1 and SD.
+def THX3T110Write_1Cyc_LS01_SD :
+  SchedWriteRes<[THX3T110LS, THX3T110SD]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 2 cycles on LS0/LS1 and SD.
+def THX3T110Write_2Cyc_LS01_SD :
+  SchedWriteRes<[THX3T110LS, THX3T110SD]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on LS0/LS1 and SD.
+def THX3T110Write_4Cyc_LS01_SD :
+  SchedWriteRes<[THX3T110LS, THX3T110SD]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+// 5 cycles on LS0/LS1 and SD.
+def THX3T110Write_5Cyc_LS01_SD :
+  SchedWriteRes<[THX3T110LS, THX3T110SD]> {
+  let Latency = 5;
+  let NumMicroOps = 4;
+}
+
+// 6 cycles on LS0/LS1 and SD.
+def THX3T110Write_6Cyc_LS01_SD :
+  SchedWriteRes<[THX3T110LS, THX3T110SD]> {
+  let Latency = 6;
+  let NumMicroOps = 5;
+}
+
+// 1 cycle on LS0/LS1, SD and I0/I1/I2/I3.
+def THX3T110Write_1Cyc_LS01_SD_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 2 cycles on LS0/LS1, SD and I0/I1/I2/I3.
+def THX3T110Write_2Cyc_LS01_SD_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on LS0/LS1, SD and I0/I1/I2/I3.
+def THX3T110Write_4Cyc_LS01_SD_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+// 5 cycles on LS0/LS1, SD and I0/I1/I2/I3.
+def THX3T110Write_5Cyc_LS01_SD_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 4;
+}
+
+// 6 cycles on LS0/LS1, SD and I0/I1/I2/I3.
+def THX3T110Write_6Cyc_LS01_SD_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 6;
+  let NumMicroOps = 5;
+}
+
+// 1 cycles on LS0/LS1 and F0/F1/F2/F3.
+def THX3T110Write_1Cyc_LS01_F0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110FP0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// 5 cycles on LS0/LS1 and F0/F1/F2/F3.
+def THX3T110Write_5Cyc_LS01_F0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// 6 cycles on LS0/LS1 and F0/F1/F2/F3.
+def THX3T110Write_6Cyc_LS01_F0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110FP0123]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+// 7 cycles on LS0/LS1 and F0/F1/F2/F3.
+def THX3T110Write_7Cyc_LS01_F0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110FP0123]> {
+  let Latency = 7;
+  let NumMicroOps = 3;
+}
+
+// 8 cycles on LS0/LS1 and F0/F1/F2/F3.
+def THX3T110Write_8Cyc_LS01_F0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110FP0123]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+// 8 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_8Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+// 12 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_12Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 12;
+  let NumMicroOps = 4;
+}
+
+// 16 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_16Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 16;
+  let NumMicroOps = 5;
+}
+
+// 24 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_24Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 24;
+  let NumMicroOps = 10;
+}
+
+// 32 cycles on LS0/LS1 and I0/I1/I2/I3.
+def THX3T110Write_32Cyc_LS01_I0123 :
+  SchedWriteRes<[THX3T110LS, THX3T110I0123]> {
+  let Latency = 32;
+  let NumMicroOps = 14;
+}
+
+// 3 cycles on F0/F1/F2/F3.
+def THX3T110Write_3Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 3;
+  let NumMicroOps = 2;
+}
+
+// 4 cycles on F0/F1/F2/F3.
+def THX3T110Write_4Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+// 5 cycles on F0/F1/F2/F3.
+def THX3T110Write_5Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+// 10 cycles on F0/F1/F2/F3.
+def THX3T110Write_10Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 10;
+  let NumMicroOps = 4;
+}
+
+// 15 cycles on F0/F1/F2/F3.
+def THX3T110Write_15Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 15;
+  let NumMicroOps = 7;
+}
+
+// 16 cycles on F0/F1/F2/F3.
+def THX3T110Write_16Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 16;
+  let NumMicroOps = 3;
+}
+
+// 18 cycles on F0/F1/F2/F3.
+def THX3T110Write_18Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 18;
+  let NumMicroOps = 3;
+}
+
+// 19 cycles on F0/F1/F2/F3.
+def THX3T110Write_19Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 19;
+  let NumMicroOps = 4;
+}
+
+// 20 cycles on F0/F1/F2/F3.
+def THX3T110Write_20Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 20;
+  let NumMicroOps = 4;
+}
+
+// 23 cycles on F0/F1/F2/F3.
+def THX3T110Write_23Cyc_F0123 : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 23;
+  let NumMicroOps = 4;
+}
+
+// 3 cycles on F2/F3 and 4 cycles on F0/F1/F2/F3.
+def THX3T110Write_3_4Cyc_F23_F0123 :
+  SchedWriteRes<[THX3T110FP23, THX3T110FP0123]> {
+  let Latency = 3;
+  let NumMicroOps = 2;
+  let ResourceCycles = [3, 4];
+}
+
+
+// Define commonly used read types.
+
+// No forwarding is provided for these types.
+def : ReadAdvance<ReadI,       0>;
+def : ReadAdvance<ReadISReg,   0>;
+def : ReadAdvance<ReadIEReg,   0>;
+def : ReadAdvance<ReadIM,      0>;
+def : ReadAdvance<ReadIMA,     0>;
+def : ReadAdvance<ReadID,      0>;
+def : ReadAdvance<ReadExtrHi,  0>;
+def : ReadAdvance<ReadAdrBase, 0>;
+def : ReadAdvance<ReadVLD,     0>;
+
+//===----------------------------------------------------------------------===//
+// 3. Instruction Tables.
+
+//---
+// 3.1 Branch Instructions
+//---
+
+// Branch, immed
+// Branch and link, immed
+// Compare and branch
+def : WriteRes<WriteBr,      [THX3T110I23]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Branch, register
+// Branch and link, register != LR
+// Branch and link, register = LR
+def : WriteRes<WriteBrReg,   [THX3T110I23]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def : WriteRes<WriteSys,     []> { let Latency = 1; }
+def : WriteRes<WriteBarrier, []> { let Latency = 1; }
+def : WriteRes<WriteHint,    []> { let Latency = 1; }
+
+def : WriteRes<WriteAtomic,  []> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+//---
+// Branch
+//---
+def : InstRW<[THX3T110Write_1Cyc_I23], (instrs B, BL, BR, BLR)>;
+def : InstRW<[THX3T110Write_1Cyc_I23], (instrs Bcc)>;
+def : InstRW<[THX3T110Write_1Cyc_I23], (instrs RET)>;
+def : InstRW<[THX3T110Write_1Cyc_I23],
+            (instrs CBZW, CBZX, CBNZW, CBNZX, TBZW, TBZX, TBNZW, TBNZX)>;
+
+//---
+// 3.2 Arithmetic and Logical Instructions
+// 3.3 Move and Shift Instructions
+//---
+
+
+// ALU, basic
+// Conditional compare
+// Conditional select
+// Address generation
+def : WriteRes<WriteI,       [THX3T110I0123]> {
+  let Latency = 1;
+  let ResourceCycles = [1];
+  let NumMicroOps = 2;
+}
+
+def : InstRW<[WriteI],
+            (instregex "ADD?(W|X)r(i|r|s|x)",   "ADDS?(W|X)r(i|r|s|x)(64)?",
+                       "AND?(W|X)r(i|r|s|x)",   "ANDS?(W|X)r(i|r|s|x)",
+                       "ADC(W|X)r",
+                       "BIC?(W|X)r(i|r|s|x)",   "BICS?(W|X)r(i|r|s|x)",
+                       "EON?(W|X)r(i|r|s|x)",   "ORN?(W|X)r(i|r|s|x)",
+                       "ORR?(W|X)r(i|r|s|x)",   "SUB?(W|X)r(i|r|s|x)",
+                       "SUBS?(W|X)r(i|r|s|x)",  "SBC(W|X)r",
+                       "SBCS(W|X)r",            "CCMN(W|X)(i|r)",
+                       "CCMP(W|X)(i|r)",        "CSEL(W|X)r",
+                       "CSINC(W|X)r",           "CSINV(W|X)r",
+                       "CSNEG(W|X)r")>;
+
+def : InstRW<[WriteI], (instrs COPY)>;
+
+// ALU, extend and/or shift
+def : WriteRes<WriteISReg,   [THX3T110I0123]> {
+  let Latency = 2;
+  let ResourceCycles = [2];
+  let NumMicroOps = 2;
+}
+
+def : InstRW<[WriteISReg],
+            (instregex "ADD?(W|X)r(i|r|s|x)",   "ADDS?(W|X)r(i|r|s|x)(64)?",
+                       "AND?(W|X)r(i|r|s|x)",   "ANDS?(W|X)r(i|r|s|x)",
+                       "ADC(W|X)r",
+                       "BIC?(W|X)r(i|r|s|x)",   "BICS?(W|X)r(i|r|s|x)",
+                       "EON?(W|X)r(i|r|s|x)",   "ORN?(W|X)r(i|r|s|x)",
+                       "ORR?(W|X)r(i|r|s|x)",   "SUB?(W|X)r(i|r|s|x)",
+                       "SUBS?(W|X)r(i|r|s|x)",  "SBC(W|X)r",
+                       "SBCS(W|X)r",            "CCMN(W|X)(i|r)",
+                       "CCMP(W|X)(i|r)",        "CSEL(W|X)r",
+                       "CSINC(W|X)r",           "CSINV(W|X)r",
+                       "CSNEG(W|X)r")>;
+
+def : WriteRes<WriteIEReg,   [THX3T110I0123]> {
+  let Latency = 1;
+  let ResourceCycles = [1];
+  let NumMicroOps = 2;
+}
+
+def : InstRW<[WriteIEReg],
+            (instregex "ADD?(W|X)r(i|r|s|x)",   "ADDS?(W|X)r(i|r|s|x)(64)?",
+                       "AND?(W|X)r(i|r|s|x)",   "ANDS?(W|X)r(i|r|s|x)",
+                       "ADC(W|X)r",
+                       "BIC?(W|X)r(i|r|s|x)",   "BICS?(W|X)r(i|r|s|x)",
+                       "EON?(W|X)r(i|r|s|x)",   "ORN?(W|X)r(i|r|s|x)",
+                       "ORR?(W|X)r(i|r|s|x)",   "SUB?(W|X)r(i|r|s|x)",
+                       "SUBS?(W|X)r(i|r|s|x)",  "SBC(W|X)r",
+                       "SBCS(W|X)r",            "CCMN(W|X)(i|r)",
+                       "CCMP(W|X)(i|r)",        "CSEL(W|X)r",
+                       "CSINC(W|X)r",           "CSINV(W|X)r",
+                       "CSNEG(W|X)r")>;
+
+// Move immed
+def : WriteRes<WriteImm,     [THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def : InstRW<[THX3T110Write_1Cyc_I0123],
+            (instrs MOVKWi, MOVKXi, MOVNWi, MOVNXi, MOVZWi, MOVZXi)>;
+
+def : InstRW<[THX3T110Write_1Cyc_I0123],
+            (instrs ASRVWr, ASRVXr, LSLVWr, LSLVXr, RORVWr, RORVXr)>;
+
+// Variable shift
+def : WriteRes<WriteIS,      [THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+//---
+// 3.4 Divide and Multiply Instructions
+//---
+
+// Divide, W-form
+// Latency range of 13-23/13-39.
+def : WriteRes<WriteID32,    [THX3T110I1]> {
+  let Latency = 39;
+  let ResourceCycles = [39];
+  let NumMicroOps = 4;
+}
+
+// Divide, X-form
+def : WriteRes<WriteID64,    [THX3T110I1]> {
+  let Latency = 23;
+  let ResourceCycles = [23];
+  let NumMicroOps = 4;
+}
+
+// Multiply accumulate, W-form
+def : WriteRes<WriteIM32,    [THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+// Multiply accumulate, X-form
+def : WriteRes<WriteIM64,    [THX3T110I0123]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+//def : InstRW<[WriteIM32, ReadIM, ReadIM, ReadIMA, THX3T110Write_5Cyc_I012],
+//             (instrs MADDWrrr, MSUBWrrr)>;
+def : InstRW<[WriteIM32], (instrs MADDWrrr, MSUBWrrr)>;
+def : InstRW<[WriteIM32], (instrs MADDXrrr, MSUBXrrr)>;
+def : InstRW<[THX3T110Write_5Cyc_I0123],
+            (instregex "(S|U)(MADDL|MSUBL)rrr")>;
+
+def : InstRW<[WriteID32], (instrs SDIVWr, UDIVWr)>;
+def : InstRW<[WriteID64], (instrs SDIVXr, UDIVXr)>;
+
+// Bitfield extract, two reg
+def : WriteRes<WriteExtr,    [THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Multiply high
+def : InstRW<[THX3T110Write_4Cyc_I1], (instrs SMULHrr, UMULHrr)>;
+
+// Miscellaneous Data-Processing Instructions
+// Bitfield extract
+def : InstRW<[THX3T110Write_1Cyc_I0123], (instrs EXTRWrri, EXTRXrri)>;
+
+// Bitifield move - basic
+def : InstRW<[THX3T110Write_1Cyc_I0123],
+            (instrs SBFMWri, SBFMXri, UBFMWri, UBFMXri)>;
+
+// Bitfield move, insert
+def : InstRW<[THX3T110Write_1Cyc_I0123], (instregex "^BFM")>;
+def : InstRW<[THX3T110Write_1Cyc_I0123], (instregex "(S|U)?BFM.*")>;
+
+// Count leading
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123],
+            (instregex "^CLS(W|X)r$", "^CLZ(W|X)r$")>;
+
+// Reverse bits
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123], (instrs RBITWr, RBITXr)>;
+
+// Cryptography Extensions
+def : InstRW<[THX3T110Write_4Cyc_F0123], (instregex "^AES[DE]")>;
+def : InstRW<[THX3T110Write_4Cyc_F0123], (instregex "^AESI?MC")>;
+def : InstRW<[THX3T110Write_4Cyc_F0123], (instregex "^PMULL")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SHA1SU0")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SHA1(H|SU1)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SHA1[CMP]")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SHA256SU0")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SHA256(H|H2|SU1)")>;
+
+// CRC Instructions
+// def : InstRW<[THX3T110Write_4Cyc_I1], (instregex "^CRC32", "^CRC32C")>;
+def : InstRW<[THX3T110Write_4Cyc_I1],
+            (instrs CRC32Brr, CRC32Hrr, CRC32Wrr, CRC32Xrr)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I1],
+            (instrs CRC32CBrr, CRC32CHrr, CRC32CWrr, CRC32CXrr)>;
+
+// Reverse bits/bytes
+// NOTE: Handled by WriteI.
+
+//---
+// 3.6 Load Instructions
+// 3.10 FP Load Instructions
+//---
+
+// Load register, literal
+// Load register, unscaled immed
+// Load register, immed unprivileged
+// Load register, unsigned immed
+def : WriteRes<WriteLD,      [THX3T110LS]> {
+  let Latency = 4;
+  let NumMicroOps = 4;
+}
+
+// Load register, immed post-index
+// NOTE: Handled by WriteLD, WriteI.
+// Load register, immed pre-index
+// NOTE: Handled by WriteLD, WriteAdr.
+def : WriteRes<WriteAdr,     [THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Load pair, immed offset, normal
+// Load pair, immed offset, signed words, base != SP
+// Load pair, immed offset signed words, base = SP
+// LDP only breaks into *one* LS micro-op.  Thus
+// the resources are handled by WriteLD.
+def : WriteRes<WriteLDHi,    []> {
+  let Latency = 4;
+  let NumMicroOps = 4;
+}
+
+// Load register offset, basic
+// Load register, register offset, scale by 4/8
+// Load register, register offset, scale by 2
+// Load register offset, extend
+// Load register, register offset, extend, scale by 4/8
+// Load register, register offset, extend, scale by 2
+def THX3T110WriteLDIdx : SchedWriteVariant<[
+  SchedVar<ScaledIdxPred, [THX3T110Write_4Cyc_LS01_I0123_I0123]>,
+  SchedVar<NoSchedPred,   [THX3T110Write_4Cyc_LS01_I0123]>]>;
+def : SchedAlias<WriteLDIdx, THX3T110WriteLDIdx>;
+
+def THX3T110ReadAdrBase : SchedReadVariant<[
+  SchedVar<ScaledIdxPred, [ReadDefault]>,
+  SchedVar<NoSchedPred,   [ReadDefault]>]>;
+def : SchedAlias<ReadAdrBase, THX3T110ReadAdrBase>;
+
+// Load pair, immed pre-index, normal
+// Load pair, immed pre-index, signed words
+// Load pair, immed post-index, normal
+// Load pair, immed post-index, signed words
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDNPDi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDNPQi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDNPSi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDNPWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDNPXi)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPDi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPQi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPSi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPSWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, WriteLDHi], (instrs LDPXi)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRBui)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRDui)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRHui)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01], (instrs LDRQui)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01], (instrs LDRSui)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRDl)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRQl)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRWl)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDRXl)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRBi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRHi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRXi)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRSBWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRSBXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRSHWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRSHXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDTRSWi)>;
+
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPDpre)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPQpre)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPSpre)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPWpre)>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPWpre)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01, WriteAdr],
+            (instrs LDRBpre, LDRDpre, LDRHpre, LDRQpre,
+                    LDRSpre, LDRWpre, LDRXpre,
+                    LDRSBWpre, LDRSBXpre, LDRSBWpost, LDRSBXpost,
+                    LDRSHWpre, LDRSHXpre, LDRSHWpost, LDRSHXpost,
+                    LDRBBpre, LDRBBpost, LDRHHpre, LDRHHpost)>;
+
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPDpost, LDPQpost, LDPSpost, LDPWpost, LDPXpost)>;
+
+def : InstRW<[THX3T110Write_5Cyc_LS01_I0123, WriteI],
+            (instrs LDRBpost, LDRDpost, LDRHpost,
+                    LDRQpost, LDRSpost, LDRWpost, LDRXpost)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPDpre, LDPQpre, LDPSpre, LDPWpre, LDPXpre)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123_I0123, WriteAdr],
+            (instrs LDRBpre, LDRDpre, LDRHpre, LDRQpre,
+                    LDRSpre, LDRWpre, LDRXpre)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123_I0123, WriteLDHi, WriteAdr],
+            (instrs LDPDpost, LDPQpost, LDPSpost, LDPWpost, LDPXpost)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123_I0123, WriteI],
+            (instrs LDRBpost, LDRDpost, LDRHpost, LDRQpost,
+                    LDRSpost, LDRWpost, LDRXpost)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRBroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRDroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRHroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRHHroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRQroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSHWroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSHXroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRWroW)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRXroW)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRBroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRDroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRHHroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRHroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRQroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSHWroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRSHXroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRWroX)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_I0123, ReadAdrBase], (instrs LDRXroX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURBi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURBBi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURDi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURHi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURHHi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURQi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSBWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSBXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSHWi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSHXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instrs LDURSWi)>;
+
+// Load exclusive
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instregex "^LDAR(B|H|W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instregex "^LDAXR(B|H|W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instregex "^LDXR(B|H|W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instregex "^LDAXP(W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01], (instregex "^LDXP(W|X)$")>;
+
+//---
+// Prefetch
+//---
+def : InstRW<[THX3T110Write_6Cyc_LS01_I012], (instrs PRFMl)>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_I012], (instrs PRFUMi)>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_I012], (instrs PRFMui)>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_I012], (instrs PRFMroW)>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_I012], (instrs PRFMroX)>;
+
+//--
+// 3.7 Store Instructions
+// 3.11 FP Store Instructions
+//--
+
+// Store register, unscaled immed
+// Store register, immed unprivileged
+// Store register, unsigned immed
+def : WriteRes<WriteST,      [THX3T110LS, THX3T110SD]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Store register, immed post-index
+// NOTE: Handled by WriteAdr, WriteST, ReadAdrBase
+
+// Store register, immed pre-index
+// NOTE: Handled by WriteAdr, WriteST
+
+// Store register, register offset, basic
+// Store register, register offset, scaled by 4/8
+// Store register, register offset, scaled by 2
+// Store register, register offset, extend
+// Store register, register offset, extend, scale by 4/8
+// Store register, register offset, extend, scale by 1
+def : WriteRes<WriteSTIdx, [THX3T110LS, THX3T110SD, THX3T110I0123]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Store pair, immed offset, W-form
+// Store pair, immed offset, X-form
+def : WriteRes<WriteSTP,     [THX3T110LS, THX3T110SD]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+// Store pair, immed post-index, W-form
+// Store pair, immed post-index, X-form
+// Store pair, immed pre-index, W-form
+// Store pair, immed pre-index, X-form
+// NOTE: Handled by WriteAdr, WriteSTP.
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURBi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURBBi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURDi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURHi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURHHi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURQi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURSi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURWi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STURXi)>;
+
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_SD], (instrs STTRBi)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_SD], (instrs STTRHi)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_SD], (instrs STTRWi)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_SD], (instrs STTRXi)>;
+
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STNPDi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STNPQi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STNPXi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STNPWi)>;
+
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STPDi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STPQi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STPXi)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_SD], (instrs STPWi)>;
+
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRBui)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRDui)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRHui)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRQui)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRXui)>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_I0123], (instrs STRWui)>;
+
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRBui)>;
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRDui)>;
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRHui)>;
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRQui)>;
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRXui)>;
+def : InstRW<[WriteSTP, THX3T110Write_1Cyc_LS01_SD], (instrs STRWui)>;
+
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRBui)>;
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRDui)>;
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRHui)>;
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRQui)>;
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRXui)>;
+def : InstRW<[WriteSTIdx, THX3T110Write_1Cyc_LS01_SD_I0123], (instrs STRWui)>;
+
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STPDpre, STPDpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STPDpre, STPDpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STPQpre, STPQpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STPQpre, STPQpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STPSpre, STPSpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STPSpre, STPSpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STPWpre, STPWpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STPWpre, STPWpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STPXpre, STPXpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STPXpre, STPXpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRBpre, STRBpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRBpre, STRBpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRBBpre, STRBBpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRBBpre, STRBBpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRDpre, STRDpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRDpre, STRDpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRHpre, STRHpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRHpre, STRHpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRHHpre, STRHHpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRHHpre, STRHHpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRQpre, STRQpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRQpre, STRQpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRSpre, STRSpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRSpre, STRSpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRWpre, STRWpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRWpre, STRWpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123],
+            (instrs STRXpre, STRXpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRXpre, STRXpost)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRBroW, STRBroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRBBroW, STRBBroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRDroW, STRDroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRHroW, STRHroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRHHroW, STRHHroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRQroW, STRQroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRSroW, STRSroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRWroW, STRWroX)>;
+def : InstRW<[WriteAdr, THX3T110Write_1Cyc_LS01_I0123, ReadAdrBase],
+            (instrs STRXroW, STRXroX)>;
+
+// Store exclusive
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instrs STNPWi, STNPXi)>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instregex "^STLR(B|H|W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instregex "^STXP(W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instregex "^STXR(B|H|W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instregex "^STLXP(W|X)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01_SD], (instregex "^STLXR(B|H|W|X)$")>;
+
+//---
+// 3.8 FP Data Processing Instructions
+//---
+
+// FP absolute value
+// FP min/max
+// FP negate
+def : WriteRes<WriteF,       [THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+// FP arithmetic
+def : InstRW<[THX3T110Write_6Cyc_F01], (instregex "^FADD", "^FSUB")>;
+
+// FP compare
+def : WriteRes<WriteFCmp,    [THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+// FP Mul, Div, Sqrt
+def : WriteRes<WriteFDiv, [THX3T110FP0123]> {
+  let Latency = 22;
+  let ResourceCycles = [19];
+}
+
+def THX3T110XWriteFDiv : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 16;
+  let ResourceCycles = [8];
+  let NumMicroOps = 4;
+}
+
+def THX3T110XWriteFDivSP : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 16;
+  let ResourceCycles = [8];
+  let NumMicroOps = 4;
+}
+
+def THX3T110XWriteFDivDP : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 23;
+  let ResourceCycles = [12];
+  let NumMicroOps = 4;
+}
+
+def THX3T110XWriteFSqrtSP : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 16;
+  let ResourceCycles = [8];
+  let NumMicroOps = 4;
+}
+
+def THX3T110XWriteFSqrtDP : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 23;
+  let ResourceCycles = [12];
+  let NumMicroOps = 4;
+}
+
+// FP divide, S-form
+// FP square root, S-form
+def : InstRW<[THX3T110XWriteFDivSP], (instrs FDIVSrr)>;
+def : InstRW<[THX3T110XWriteFSqrtSP], (instrs FSQRTSr)>;
+def : InstRW<[THX3T110XWriteFDivSP], (instregex "^FDIVv.*32$")>;
+def : InstRW<[THX3T110XWriteFSqrtSP], (instregex "^.*SQRT.*32$")>;
+def : InstRW<[THX3T110Write_16Cyc_F01], (instregex "^FDIVSrr", "^FSQRTSr")>;
+
+// FP divide, D-form
+// FP square root, D-form
+def : InstRW<[THX3T110XWriteFDivDP], (instrs FDIVDrr)>;
+def : InstRW<[THX3T110XWriteFSqrtDP], (instrs FSQRTDr)>;
+def : InstRW<[THX3T110XWriteFDivDP], (instregex "^FDIVv.*64$")>;
+def : InstRW<[THX3T110XWriteFSqrtDP], (instregex "^.*SQRT.*64$")>;
+def : InstRW<[THX3T110Write_23Cyc_F01], (instregex "^FDIVDrr", "^FSQRTDr")>;
+
+// FP multiply
+// FP multiply accumulate
+def : WriteRes<WriteFMul, [THX3T110FP0123]> {
+  let Latency = 6;
+  let ResourceCycles = [2];
+  let NumMicroOps = 3;
+}
+
+def THX3T110XWriteFMul : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 6;
+  let ResourceCycles = [2];
+  let NumMicroOps = 3;
+}
+
+def THX3T110XWriteFMulAcc : SchedWriteRes<[THX3T110FP0123]> {
+  let Latency = 6;
+  let ResourceCycles = [2];
+  let NumMicroOps = 3;
+}
+
+def : InstRW<[THX3T110XWriteFMul], (instregex "^FMUL", "^FNMUL")>;
+def : InstRW<[THX3T110XWriteFMulAcc],
+            (instregex "^FMADD", "^FMSUB", "^FNMADD", "^FNMSUB")>;
+
+// FP round to integral
+def : InstRW<[THX3T110Write_7Cyc_F01],
+            (instregex "^FRINT(A|I|M|N|P|X|Z)(Sr|Dr)")>;
+
+// FP select
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123], (instregex "^FCSEL")>;
+
+//---
+// 3.9 FP Miscellaneous Instructions
+//---
+
+// FP convert, from vec to vec reg
+// FP convert, from gen to vec reg
+// FP convert, from vec to gen reg
+def : WriteRes<WriteFCvt, [THX3T110FP0123]> {
+  let Latency = 7;
+  let NumMicroOps = 3;
+}
+
+// FP move, immed
+// FP move, register
+def : WriteRes<WriteFImm, [THX3T110FP0123]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+// FP transfer, from gen to vec reg
+// FP transfer, from vec to gen reg
+def : WriteRes<WriteFCopy, [THX3T110FP0123]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def : InstRW<[THX3T110Write_5Cyc_F01], (instrs FMOVXDHighr, FMOVDXHighr)>;
+
+//---
+// 3.12 ASIMD Integer Instructions
+//---
+
+// ASIMD absolute diff, D-form
+// ASIMD absolute diff, Q-form
+// ASIMD absolute diff accum, D-form
+// ASIMD absolute diff accum, Q-form
+// ASIMD absolute diff accum long
+// ASIMD absolute diff long
+// ASIMD arith, basic
+// ASIMD arith, complex
+// ASIMD compare
+// ASIMD logical (AND, BIC, EOR)
+// ASIMD max/min, basic
+// ASIMD max/min, reduce, 4H/4S
+// ASIMD max/min, reduce, 8B/8H
+// ASIMD max/min, reduce, 16B
+// ASIMD multiply, D-form
+// ASIMD multiply, Q-form
+// ASIMD multiply accumulate long
+// ASIMD multiply accumulate saturating long
+// ASIMD multiply long
+// ASIMD pairwise add and accumulate
+// ASIMD shift accumulate
+// ASIMD shift by immed, basic
+// ASIMD shift by immed and insert, basic, D-form
+// ASIMD shift by immed and insert, basic, Q-form
+// ASIMD shift by immed, complex
+// ASIMD shift by register, basic, D-form
+// ASIMD shift by register, basic, Q-form
+// ASIMD shift by register, complex, D-form
+// ASIMD shift by register, complex, Q-form
+def : WriteRes<WriteV, [THX3T110FP0123]> {
+  let Latency = 5;
+  let NumMicroOps = 4;
+  let ResourceCycles = [4];
+}
+
+// ASIMD arith, reduce, 4H/4S
+// ASIMD arith, reduce, 8B/8H
+// ASIMD arith, reduce, 16B
+
+// ASIMD logical (MVN (alias for NOT), ORN, ORR)
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^ANDv", "^BICv", "^EORv", "^ORRv", "^ORNv", "^NOTv")>;
+
+// ASIMD arith, reduce
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^ADDVv", "^SADDLVv", "^UADDLVv")>;
+
+// ASIMD polynomial (8x8) multiply long
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^(S|U|SQD)MULL")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "(S|U|SQD)(MLAL|MLSL|MULL)v.*")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^PMULL(v8i8|v16i8)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^PMULL(v1i64|v2i64)")>;
+
+// ASIMD absolute diff accum, D-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU]ABA(v8i8|v4i16|v2i32)$")>;
+// ASIMD absolute diff accum, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU]ABA(v16i8|v8i16|v4i32)$")>;
+// ASIMD absolute diff accum long
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU]ABAL")>;
+// ASIMD arith, reduce, 4H/4S
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU]?ADDL?V(v8i8|v4i16|v2i32)v$")>;
+// ASIMD arith, reduce, 8B
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU]?ADDL?V(v8i16|v4i32)v$")>;
+// ASIMD arith, reduce, 16B/16H
+def : InstRW<[THX3T110Write_10Cyc_F0123],
+            (instregex "^[SU]?ADDL?Vv16i8v$")>;
+// ASIMD max/min, reduce, 4H/4S
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU](MIN|MAX)V(v4i16|v4i32)v$")>;
+// ASIMD max/min, reduce, 8B/8H
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU](MIN|MAX)V(v8i8|v8i16)v$")>;
+// ASIMD max/min, reduce, 16B/16H
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU](MIN|MAX)Vv16i8v$")>;
+// ASIMD multiply, D-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^(P?MUL|SQR?DMULH)" #
+                       "(v8i8|v4i16|v2i32|v1i8|v1i16|v1i32|v1i64)" #
+                       "(_indexed)?$")>;
+// ASIMD multiply, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^(P?MUL|SQR?DMULH)(v16i8|v8i16|v4i32)(_indexed)?$")>;
+// ASIMD multiply accumulate, D-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^ML[AS](v8i8|v4i16|v2i32)(_indexed)?$")>;
+// ASIMD multiply accumulate, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^ML[AS](v16i8|v8i16|v4i32)(_indexed)?$")>;
+// ASIMD shift accumulate
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "SRSRAv","SSRAv","URSRAv","USRAv")>;
+
+// ASIMD shift by immed, basic
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "RSHRNv","SHRNv", "SQRSHRNv","SQRSHRUNv",
+                       "SQSHRNv","SQSHRUNv", "UQRSHRNv",
+                       "UQSHRNv","SQXTNv","SQXTUNv","UQXTNv")>;
+// ASIMD shift by immed, complex
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^[SU]?(Q|R){1,2}SHR")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SQSHLU")>;
+// ASIMD shift by register, basic, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F01],
+            (instregex "^[SU]SHL(v16i8|v8i16|v4i32|v2i64)")>;
+// ASIMD shift by register, complex, D-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU][QR]{1,2}SHL" #
+                       "(v1i8|v1i16|v1i32|v1i64|v8i8|v4i16|v2i32|b|d|h|s)")>;
+// ASIMD shift by register, complex, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^[SU][QR]{1,2}SHL(v16i8|v8i16|v4i32|v2i64)")>;
+
+// ASIMD Arithmetic
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "(ADD|SUB)(v8i8|v4i16|v2i32|v1i64)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "(ADD|SUB)(v16i8|v8i16|v4i32|v2i64)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "(ADD|SUB)HNv.*")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "(RADD|RSUB)HNv.*")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^SQADD", "^SQNEG", "^SQSUB", "^SRHADD",
+                       "^SUQADD", "^UQADD", "^UQSUB", "^URHADD", "^USQADD")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "ADDP(v16i8|v8i16|v4i32|v2i64)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "((AND|ORN|EOR|EON)S?(Xr[rsi]|v16i8|v8i16|v4i32)|" #
+                       "(ORR|BIC)S?(Xr[rs]|v16i8|v8i16|v4i32))")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "(CLS|CLZ|CNT)(v4i32|v8i16|v16i8)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SADALP","^UADALP")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SADDLPv","^UADDLPv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SADDLV","^UADDLV")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+             (instregex "^ADDVv","^SMAXVv","^UMAXVv","^SMINVv","^UMINVv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+             (instregex "^SABAv","^UABAv","^SABALv","^UABALv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^SQADDv","^SQSUBv","^UQADDv","^UQSUBv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^SUQADDv","^USQADDv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^ADDHNv","^RADDHNv", "^RSUBHNv",
+                       "^SQABS", "^SQADD", "^SQNEG", "^SQSUB",
+                       "^SRHADD", "^SUBHNv", "^SUQADD",
+                       "^UQADD", "^UQSUB", "^URHADD", "^USQADD")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^CMEQv","^CMGEv","^CMGTv",
+                       "^CMLEv","^CMLTv", "^CMHIv","^CMHSv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^SMAXv","^SMINv","^UMAXv","^UMINv",
+                       "^SMAXPv","^SMINPv","^UMAXPv","^UMINPv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^SABDv","^UABDv", "^SABDLv","^UABDLv")>;
+
+//---
+// 3.13 ASIMD Floating-point Instructions
+//---
+
+// ASIMD FP absolute value
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FABSv")>;
+
+// ASIMD FP arith, normal, D-form
+// ASIMD FP arith, normal, Q-form
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123],
+            (instregex "^FABDv", "^FADDv", "^FSUBv")>;
+
+// ASIMD FP arith,pairwise, D-form
+// ASIMD FP arith, pairwise, Q-form
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123], (instregex "^FADDPv")>;
+
+// ASIMD FP compare, D-form
+// ASIMD FP compare, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FACGEv", "^FACGTv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FCMEQv", "^FCMGEv",
+                                                 "^FCMGTv", "^FCMLEv",
+                                                 "^FCMLTv")>;
+
+// ASIMD FP round, D-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FRINT[AIMNPXZ](v2f32)")>;
+// ASIMD FP round, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FRINT[AIMNPXZ](v4f32|v2f64)")>;
+
+// ASIMD FP convert, long
+// ASIMD FP convert, narrow
+// ASIMD FP convert, other, D-form
+// ASIMD FP convert, other, Q-form
+// NOTE: Handled by WriteV.
+
+// ASIMD FP convert, long and narrow
+def : InstRW<[THX3T110Write_5Cyc_F01], (instregex "^FCVT(L|N|XN)v")>;
+// ASIMD FP convert, other, D-form
+def : InstRW<[THX3T110Write_5Cyc_F01],
+      (instregex "^[FVSU]CVT([AMNPZ][SU])?(_Int)?(v2f32|v1i32|v2i32|v1i64)")>;
+// ASIMD FP convert, other, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F01],
+      (instregex "^[FVSU]CVT([AMNPZ][SU])?(_Int)?(v4f32|v2f64|v4i32|v2i64)")>;
+
+// ASIMD FP divide, D-form, F32
+def : InstRW<[THX3T110Write_16Cyc_F0123], (instrs FDIVv2f32)>;
+def : InstRW<[THX3T110Write_16Cyc_F0123], (instregex "FDIVv2f32")>;
+
+// ASIMD FP divide, Q-form, F32
+def : InstRW<[THX3T110Write_16Cyc_F0123], (instrs FDIVv4f32)>;
+def : InstRW<[THX3T110Write_16Cyc_F0123], (instregex "FDIVv4f32")>;
+
+// ASIMD FP divide, Q-form, F64
+def : InstRW<[THX3T110Write_23Cyc_F0123], (instrs FDIVv2f64)>;
+def : InstRW<[THX3T110Write_23Cyc_F0123], (instregex "FDIVv2f64")>;
+
+// ASIMD FP max/min, normal, D-form
+// ASIMD FP max/min, normal, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FMAXv", "^FMAXNMv",
+                                                "^FMINv", "^FMINNMv")>;
+
+// ASIMD FP max/min, pairwise, D-form
+// ASIMD FP max/min, pairwise, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FMAXPv", "^FMAXNMPv",
+                                                "^FMINPv", "^FMINNMPv")>;
+
+// ASIMD FP max/min, reduce
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FMAXVv", "^FMAXNMVv",
+                                                "^FMINVv", "^FMINNMVv")>;
+
+// ASIMD FP multiply, D-form, FZ
+// ASIMD FP multiply, D-form, no FZ
+// ASIMD FP multiply, Q-form, FZ
+// ASIMD FP multiply, Q-form, no FZ
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FMULv", "^FMULXv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FMULX?(v2f32|v1i32|v2i32|v1i64|32|64)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FMULX?(v4f32|v2f64|v4i32|v2i64)")>;
+
+// ASIMD FP multiply accumulate, Dform, FZ
+// ASIMD FP multiply accumulate, Dform, no FZ
+// ASIMD FP multiply accumulate, Qform, FZ
+// ASIMD FP multiply accumulate, Qform, no FZ
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FMLAv", "^FMLSv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FML[AS](v2f32|v1i32|v2i32|v1i64)")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FML[AS](v4f32|v2f64|v4i32|v2i64)")>;
+
+// ASIMD FP negate
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^FNEGv")>;
+
+//--
+// 3.14 ASIMD Miscellaneous Instructions
+//--
+
+// ASIMD bit reverse
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^RBITv")>;
+
+// ASIMD bitwise insert, D-form
+// ASIMD bitwise insert, Q-form
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123],
+            (instregex "^BIFv", "^BITv", "^BSLv")>;
+
+// ASIMD count, D-form
+// ASIMD count, Q-form
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123],
+            (instregex "^CLSv", "^CLZv", "^CNTv")>;
+
+// ASIMD duplicate, gen reg
+// ASIMD duplicate, element
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^DUPv")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^CPY")>;
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^DUPv.+gpr")>;
+
+// ASIMD extract
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^EXTv")>;
+
+// ASIMD extract narrow
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^XTNv")>;
+
+// ASIMD extract narrow, saturating
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^SQXTNv", "^SQXTUNv", "^UQXTNv")>;
+
+// ASIMD insert, element to element
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^INSv")>;
+
+// ASIMD transfer, element to gen reg
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^[SU]MOVv")>;
+
+// ASIMD move, integer immed
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123], (instregex "^MOVIv")>;
+
+// ASIMD move, FP immed
+def : InstRW<[THX3T110Write_3_4Cyc_F23_F0123], (instregex "^FMOVv")>;
+
+// ASIMD transpose
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^TRN1", "^TRN2")>;
+
+// ASIMD unzip/zip
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^UZP1", "^UZP2", "^ZIP1", "^ZIP2")>;
+
+// ASIMD reciprocal estimate, D-form
+// ASIMD reciprocal estimate, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^FRECPEv", "^FRECPXv", "^URECPEv",
+                       "^FRSQRTEv", "^URSQRTEv")>;
+
+// ASIMD reciprocal step, D-form, FZ
+// ASIMD reciprocal step, D-form, no FZ
+// ASIMD reciprocal step, Q-form, FZ
+// ASIMD reciprocal step, Q-form, no FZ
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+              (instregex "^FRECPSv", "^FRSQRTSv")>;
+
+// ASIMD reverse
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instregex "^REV16v", "^REV32v", "^REV64v")>;
+
+// ASIMD table lookup, D-form
+// ASIMD table lookup, Q-form
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+            (instrs TBLv8i8One, TBLv16i8One, TBXv8i8One, TBXv16i8One)>;
+def : InstRW<[THX3T110Write_10Cyc_F0123],
+            (instrs TBLv8i8Two, TBLv16i8Two, TBXv8i8Two, TBXv16i8Two)>;
+def : InstRW<[THX3T110Write_15Cyc_F0123],
+            (instrs TBLv8i8Three, TBLv16i8Three, TBXv8i8Three, TBXv16i8Three)>;
+def : InstRW<[THX3T110Write_20Cyc_F0123],
+            (instrs TBLv8i8Four, TBLv16i8Four, TBXv8i8Four, TBXv16i8Four)>;
+
+// ASIMD transfer, element to word or word
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^[SU]MOVv")>;
+
+// ASIMD transfer, element to gen reg
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "(S|U)MOVv.*")>;
+
+// ASIMD transfer gen reg to element
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^INSv")>;
+
+// ASIMD transpose
+def : InstRW<[THX3T110Write_5Cyc_F0123],
+              (instregex "^TRN1v", "^TRN2v", "^UZP1v", "^UZP2v")>;
+
+// ASIMD unzip/zip
+def : InstRW<[THX3T110Write_5Cyc_F0123], (instregex "^ZIP1v", "^ZIP2v")>;
+
+//--
+// 3.15 ASIMD Load Instructions
+//--
+
+// ASIMD load, 1 element, multiple, 1 reg, D-form
+// ASIMD load, 1 element, multiple, 1 reg, Q-form
+def : InstRW<[THX3T110Write_4Cyc_LS01],
+            (instregex "^LD1Onev(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01, WriteAdr],
+            (instregex "^LD1Onev(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 1 element, multiple, 2 reg, D-form
+// ASIMD load, 1 element, multiple, 2 reg, Q-form
+def : InstRW<[THX3T110Write_4Cyc_LS01],
+            (instregex "^LD1Twov(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_4Cyc_LS01, WriteAdr],
+            (instregex "^LD1Twov(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 1 element, multiple, 3 reg, D-form
+// ASIMD load, 1 element, multiple, 3 reg, Q-form
+def : InstRW<[THX3T110Write_5Cyc_LS01],
+            (instregex "^LD1Threev(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01, WriteAdr],
+            (instregex "^LD1Threev(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 1 element, multiple, 4 reg, D-form
+// ASIMD load, 1 element, multiple, 4 reg, Q-form
+def : InstRW<[THX3T110Write_6Cyc_LS01],
+            (instregex "^LD1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_6Cyc_LS01, WriteAdr],
+            (instregex "^LD1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 1 element, one lane, B/H/S
+// ASIMD load, 1 element, one lane, D
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123],
+            (instregex "^LD1i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD1i(8|16|32|64)_POST$")>;
+
+// ASIMD load, 1 element, all lanes, D-form, B/H/S
+// ASIMD load, 1 element, all lanes, D-form, D
+// ASIMD load, 1 element, all lanes, Q-form
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123],
+            (instregex "^LD1Rv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD1Rv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 2 element, multiple, D-form, B/H/S
+// ASIMD load, 2 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123],
+            (instregex "^LD2Twov(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD2Twov(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 2 element, one lane, B/H
+// ASIMD load, 2 element, one lane, S
+// ASIMD load, 2 element, one lane, D
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123],
+            (instregex "^LD2i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD2i(8|16|32|64)_POST$")>;
+
+// ASIMD load, 2 element, all lanes, D-form, B/H/S
+// ASIMD load, 2 element, all lanes, D-form, D
+// ASIMD load, 2 element, all lanes, Q-form
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123],
+            (instregex "^LD2Rv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_5Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD2Rv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 3 element, multiple, D-form, B/H/S
+// ASIMD load, 3 element, multiple, Q-form, B/H/S
+// ASIMD load, 3 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_8Cyc_LS01_F0123],
+            (instregex "^LD3Threev(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_8Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD3Threev(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 3 element, one lone, B/H
+// ASIMD load, 3 element, one lane, S
+// ASIMD load, 3 element, one lane, D
+def : InstRW<[THX3T110Write_7Cyc_LS01_F0123],
+            (instregex "^LD3i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_7Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD3i(8|16|32|64)_POST$")>;
+
+// ASIMD load, 3 element, all lanes, D-form, B/H/S
+// ASIMD load, 3 element, all lanes, D-form, D
+// ASIMD load, 3 element, all lanes, Q-form, B/H/S
+// ASIMD load, 3 element, all lanes, Q-form, D
+def : InstRW<[THX3T110Write_7Cyc_LS01_F0123],
+            (instregex "^LD3Rv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_7Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD3Rv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 4 element, multiple, D-form, B/H/S
+// ASIMD load, 4 element, multiple, Q-form, B/H/S
+// ASIMD load, 4 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_8Cyc_LS01_F0123],
+            (instregex "^LD4Fourv(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_8Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD4Fourv(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD load, 4 element, one lane, B/H
+// ASIMD load, 4 element, one lane, S
+// ASIMD load, 4 element, one lane, D
+def : InstRW<[THX3T110Write_6Cyc_LS01_F0123],
+            (instregex "^LD4i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD4i(8|16|32|64)_POST$")>;
+
+// ASIMD load, 4 element, all lanes, D-form, B/H/S
+// ASIMD load, 4 element, all lanes, D-form, D
+// ASIMD load, 4 element, all lanes, Q-form, B/H/S
+// ASIMD load, 4 element, all lanes, Q-form, D
+def : InstRW<[THX3T110Write_6Cyc_LS01_F0123],
+            (instregex "^LD4Rv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_6Cyc_LS01_F0123, WriteAdr],
+            (instregex "^LD4Rv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+//--
+// 3.16 ASIMD Store Instructions
+//--
+
+// ASIMD store, 1 element, multiple, 1 reg, D-form
+// ASIMD store, 1 element, multiple, 1 reg, Q-form
+def : InstRW<[THX3T110Write_1Cyc_LS01],
+            (instregex "^ST1Onev(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01, WriteAdr],
+            (instregex "^ST1Onev(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 1 element, multiple, 2 reg, D-form
+// ASIMD store, 1 element, multiple, 2 reg, Q-form
+def : InstRW<[THX3T110Write_1Cyc_LS01],
+            (instregex "^ST1Twov(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01, WriteAdr],
+            (instregex "^ST1Twov(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 1 element, multiple, 3 reg, D-form
+// ASIMD store, 1 element, multiple, 3 reg, Q-form
+def : InstRW<[THX3T110Write_1Cyc_LS01],
+            (instregex "^ST1Threev(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01, WriteAdr],
+            (instregex "^ST1Threev(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 1 element, multiple, 4 reg, D-form
+// ASIMD store, 1 element, multiple, 4 reg, Q-form
+def : InstRW<[THX3T110Write_1Cyc_LS01],
+            (instregex "^ST1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01, WriteAdr],
+            (instregex "^ST1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 1 element, one lane, B/H/S
+// ASIMD store, 1 element, one lane, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST1i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST1i(8|16|32|64)_POST$")>;
+
+// ASIMD store, 2 element, multiple, D-form, B/H/S
+// ASIMD store, 2 element, multiple, Q-form, B/H/S
+// ASIMD store, 2 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST2Twov(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST2Twov(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 2 element, one lane, B/H/S
+// ASIMD store, 2 element, one lane, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST2i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST2i(8|16|32|64)_POST$")>;
+
+// ASIMD store, 3 element, multiple, D-form, B/H/S
+// ASIMD store, 3 element, multiple, Q-form, B/H/S
+// ASIMD store, 3 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST3Threev(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST3Threev(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 3 element, one lane, B/H
+// ASIMD store, 3 element, one lane, S
+// ASIMD store, 3 element, one lane, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST3i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST3i(8|16|32|64)_POST$")>;
+
+// ASIMD store, 4 element, multiple, D-form, B/H/S
+// ASIMD store, 4 element, multiple, Q-form, B/H/S
+// ASIMD store, 4 element, multiple, Q-form, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST4Fourv(8b|4h|2s|16b|8h|4s|2d)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST4Fourv(8b|4h|2s|16b|8h|4s|2d)_POST$")>;
+
+// ASIMD store, 4 element, one lane, B/H
+// ASIMD store, 4 element, one lane, S
+// ASIMD store, 4 element, one lane, D
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123],
+            (instregex "^ST4i(8|16|32|64)$")>;
+def : InstRW<[THX3T110Write_1Cyc_LS01_F0123, WriteAdr],
+            (instregex "^ST4i(8|16|32|64)_POST$")>;
+
+// V8.1a Atomics (LSE)
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs CASB, CASH, CASW, CASX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs CASAB, CASAH, CASAW, CASAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs CASLB, CASLH, CASLW, CASLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs CASALB, CASALH, CASALW, CASALX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDLARB, LDLARH, LDLARW, LDLARX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDADDB, LDADDH, LDADDW, LDADDX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDADDAB, LDADDAH, LDADDAW, LDADDAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDADDLB, LDADDLH, LDADDLW, LDADDLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs LDADDALB, LDADDALH, LDADDALW, LDADDALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDCLRB, LDCLRH, LDCLRW, LDCLRX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDCLRAB, LDCLRAH, LDCLRAW, LDCLRAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDCLRLB, LDCLRLH, LDCLRLW, LDCLRLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs LDCLRALB, LDCLRALH, LDCLRALW, LDCLRALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDEORB, LDEORH, LDEORW, LDEORX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDEORAB, LDEORAH, LDEORAW, LDEORAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDEORLB, LDEORLH, LDEORLW, LDEORLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs LDEORALB, LDEORALH, LDEORALW, LDEORALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDSETB, LDSETH, LDSETW, LDSETX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDSETAB, LDSETAH, LDSETAW, LDSETAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs LDSETLB, LDSETLH, LDSETLW, LDSETLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs LDSETALB, LDSETALH, LDSETALW, LDSETALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDSMAXB, LDSMAXH, LDSMAXW, LDSMAXX,
+             LDSMAXAB, LDSMAXAH, LDSMAXAW, LDSMAXAX,
+             LDSMAXLB, LDSMAXLH, LDSMAXLW, LDSMAXLX,
+             LDSMAXALB, LDSMAXALH, LDSMAXALW, LDSMAXALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDSMINB, LDSMINH, LDSMINW, LDSMINX,
+             LDSMINAB, LDSMINAH, LDSMINAW, LDSMINAX,
+             LDSMINLB, LDSMINLH, LDSMINLW, LDSMINLX,
+             LDSMINALB, LDSMINALH, LDSMINALW, LDSMINALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDUMAXB, LDUMAXH, LDUMAXW, LDUMAXX,
+             LDUMAXAB, LDUMAXAH, LDUMAXAW, LDUMAXAX,
+             LDUMAXLB, LDUMAXLH, LDUMAXLW, LDUMAXLX,
+             LDUMAXALB, LDUMAXALH, LDUMAXALW, LDUMAXALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs LDUMINB, LDUMINH, LDUMINW, LDUMINX,
+             LDUMINAB, LDUMINAH, LDUMINAW, LDUMINAX,
+             LDUMINLB, LDUMINLH, LDUMINLW, LDUMINLX,
+             LDUMINALB, LDUMINALH, LDUMINALW, LDUMINALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs SWPB, SWPH, SWPW, SWPX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs SWPAB, SWPAH, SWPAW, SWPAX)>;
+
+def : InstRW<[THX3T110Write_6Cyc_I0123, WriteAtomic],
+            (instrs SWPLB, SWPLH, SWPLW, SWPLX)>;
+
+def : InstRW<[THX3T110Write_8Cyc_I0123, WriteAtomic],
+            (instrs SWPALB, SWPALH, SWPALW, SWPALX)>;
+
+def : InstRW<[THX3T110Write_4Cyc_I0123, WriteAtomic],
+            (instrs STLLRB, STLLRH, STLLRW, STLLRX)>;
+
+// V8.3a PAC
+def : InstRW<[THX3T110Write_11Cyc_LS01_I1], (instregex "^LDRAA", "^LDRAB")>;
+def : InstRW<[THX3T110Write_8Cyc_I123],
+            (instrs BLRAA, BLRAAZ, BLRAB, BLRABZ,
+                    BRAA, BRAAZ, BRAB, BRABZ)>;
+def : InstRW<[THX3T110Write_8Cyc_I123], (instrs RETAA, RETAB)>;
+
+} // SchedModel = ThunderX3T110Model

llvm/lib/Target/AArch64/AArch64Subtarget.h

enum ARMProcFamilyEnum : uint8_t {
     NeoverseE1,
     NeoverseN1,
     Saphira,
     ThunderX2T99,
     ThunderX,
     ThunderXT81,
     ThunderXT83,
     ThunderXT88,
-    TSV110
+    TSV110,
+    ThunderX3T110
   };
 
 protected:
   /// ARMProcFamily - ARM processor family: Cortex-A53, Cortex-A57, and others.
   ARMProcFamilyEnum ARMProcFamily = Others;
 
   bool HasV8_1aOps = false;
   bool HasV8_2aOps = false;

llvm/lib/Target/AArch64/AArch64Subtarget.cpp

case ThunderXT83:
     // FIXME: remove this to enable 64-bit SLP if performance looks good.
     MinVectorRegisterBitWidth = 128;
     break;
   case TSV110:
     CacheLineSize = 64;
     PrefFunctionLogAlignment = 4;
     PrefLoopLogAlignment = 2;
     break;
+  case ThunderX3T110:
+    CacheLineSize = 64;
+    PrefFunctionLogAlignment = 4;
+    PrefLoopLogAlignment = 2;
+    MaxInterleaveFactor = 4;
+    PrefetchDistance = 128;
+    MinPrefetchStride = 1024;
+    MaxPrefetchIterationsAhead = 4;
+    // FIXME: remove this to enable 64-bit SLP if performance looks good.
+    MinVectorRegisterBitWidth = 128;
+    break;
   }
 }
 
 AArch64Subtarget::AArch64Subtarget(const Triple &TT, const std::string &CPU,
                                    const std::string &FS,
                                    const TargetMachine &TM, bool LittleEndian)
     : AArch64GenSubtargetInfo(TT, CPU, FS),
       ReserveXRegister(AArch64::GPR64commonRegClass.getNumRegs()),

llvm/test/CodeGen/AArch64/aarch64-combine-fmul-fsub.mir

 # RUN: llc -run-pass=machine-combiner -o - -mtriple=aarch64-unknown-linux -mcpu=cortex-a57 -enable-unsafe-fp-math -machine-combiner-verify-pattern-order=true %s | FileCheck --check-prefixes=UNPROFITABLE,ALL %s
 # RUN: llc -run-pass=machine-combiner -o - -mtriple=aarch64-unknown-linux -mcpu=falkor -enable-unsafe-fp-math %s -machine-combiner-verify-pattern-order=true | FileCheck --check-prefixes=PROFITABLE,ALL %s
 # RUN: llc -run-pass=machine-combiner -o - -mtriple=aarch64-unknown-linux -mcpu=exynos-m3 -enable-unsafe-fp-math -machine-combiner-verify-pattern-order=true %s | FileCheck --check-prefixes=PROFITABLE,ALL %s
 # RUN: llc -run-pass=machine-combiner -o - -mtriple=aarch64-unknown-linux -mcpu=thunderx2t99 -enable-unsafe-fp-math -machine-combiner-verify-pattern-order=true %s | FileCheck --check-prefixes=PROFITABLE,ALL %s
+# RUN: llc -run-pass=machine-combiner -o - -mtriple=aarch64-unknown-linux -mcpu=thunderx3t110 -enable-unsafe-fp-math -machine-combiner-verify-pattern-order=true %s | FileCheck --check-prefixes=PROFITABLE,ALL %s
 #
 name:            f1_2s
 registers:
   - { id: 0, class: fpr64 }
   - { id: 1, class: fpr64 }
   - { id: 2, class: fpr64 }
   - { id: 3, class: fpr64 }
   - { id: 4, class: fpr64 }

llvm/test/CodeGen/AArch64/cpus.ll

 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=neoverse-n1 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=exynos-m3 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=exynos-m4 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=exynos-m5 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=falkor 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=saphira 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=kryo 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=thunderx2t99 2>&1 | FileCheck %s
+; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=thunderx3t110 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=tsv110 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=apple-latest 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=a64fx 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=invalidcpu 2>&1 | FileCheck %s --check-prefix=INVALID
 
 ; CHECK-NOT: {{.*}}  is not a recognized processor for this target
 ; INVALID: {{.*}}  is not a recognized processor for this target
 
 define i32 @f(i64 %z) {
   ret i32 0
 }

llvm/test/CodeGen/AArch64/machine-combiner-madd.ll

 ; Test all AArch64 subarches with scheduling models.
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=cortex-a57 < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=cortex-a72 < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=cortex-a73 < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=cyclone    < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=exynos-m3  < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=kryo       < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=thunderx2t99 < %s | FileCheck %s
+; RUN: llc -mtriple=aarch64-linux-gnu -mcpu=thunderx3t110 < %s | FileCheck %s
 
 ; Make sure that inst-combine fuses the multiply add in the addressing mode of
 ; the load.
 
 ; CHECK-LABEL: fun:
 ; CHECK-NOT: mul
 ; CHECK:     madd
 ; CHECK-NOT: mul

llvm/test/CodeGen/AArch64/preferred-function-alignment.ll

 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=kryo < %s | FileCheck --check-prefixes=ALIGN2,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=neoverse-e1 < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=neoverse-n1 < %s | FileCheck --check-prefixes=ALIGN4,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderx < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderxt81 < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderxt83 < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderxt88 < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderx2t99 < %s | FileCheck --check-prefixes=ALIGN3,CHECK %s
+; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=thunderx3t110 < %s | FileCheck --check-prefixes=ALIGN4,CHECK %s
 ; RUN: llc -mtriple=aarch64-unknown-linux -mcpu=exynos-m3 < %s | FileCheck --check-prefixes=ALIGN5,CHECK %s
 
 define void @test() {
   ret void
 }
 
 ; CHECK-LABEL: test
 ; ALIGN2: .p2align 2

llvm/test/CodeGen/AArch64/remat.ll

 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=exynos-m4 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=exynos-m5 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=falkor -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=saphira -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=kryo -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=thunderx2t99 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=tsv110 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mattr=+custom-cheap-as-move -o - %s | FileCheck %s
+; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=thunderx3t110 -o - %s | FileCheck %s
 
 %X = type { i64, i64, i64 }
 declare void @f(%X*)
 define void @t() {
 entry:
   %tmp = alloca %X
   call void @f(%X* %tmp)
 ; CHECK: add x0, sp, #8
 ; CHECK-NOT: mov
 ; CHECK-NEXT: bl f
   call void @f(%X* %tmp)
 ; CHECK: add x0, sp, #8
 ; CHECK-NOT: mov
 ; CHECK-NEXT: bl f
   ret void
 }

llvm/unittests/Support/TargetParserTest.cpp

"neoverse-n1", "armv8.2-a", "crypto-neon-fp-armv8",
       AArch64::AEK_PROFILE | AArch64::AEK_RAS | AArch64::AEK_RCPC |
       AArch64::AEK_RDM | AArch64::AEK_SIMD | AArch64::AEK_SSBS,
      "8.2-A"));
   EXPECT_TRUE(testAArch64CPU(
       "thunderx2t99", "armv8.1-a", "crypto-neon-fp-armv8",
       AArch64::AEK_CRC | AArch64::AEK_CRYPTO | AArch64::AEK_LSE |
       AArch64::AEK_RDM | AArch64::AEK_FP | AArch64::AEK_SIMD, "8.1-A"));
   EXPECT_TRUE(testAArch64CPU(
+      "thunderx3t110", "armv8.3-a", "crypto-neon-fp-armv8",
+      AArch64::AEK_CRC | AArch64::AEK_CRYPTO | AArch64::AEK_LSE |
+          AArch64::AEK_RDM | AArch64::AEK_FP | AArch64::AEK_SIMD |
+          AArch64::AEK_PROFILE | AArch64::AEK_RAS | AArch64::AEK_RAND |
+          AArch64::AEK_RCPC,
+      "8.3-A"));
+  EXPECT_TRUE(testAArch64CPU(
       "thunderx", "armv8-a", "crypto-neon-fp-armv8",
       AArch64::AEK_CRC | AArch64::AEK_CRYPTO | AArch64::AEK_SIMD |
       AArch64::AEK_FP | AArch64::AEK_PROFILE,
       "8-A"));
   EXPECT_TRUE(testAArch64CPU(
       "thunderxt81", "armv8-a", "crypto-neon-fp-armv8",
       AArch64::AEK_CRC | AArch64::AEK_CRYPTO | AArch64::AEK_SIMD |
       AArch64::AEK_FP | AArch64::AEK_PROFILE,
TEST(TargetParserTest, testAArch64CPU) {
   EXPECT_TRUE(testAArch64CPU(
       "carmel", "armv8.2-a", "crypto-neon-fp-armv8",
       AArch64::AEK_CRC | AArch64::AEK_CRYPTO | AArch64::AEK_FP |
       AArch64::AEK_SIMD | AArch64::AEK_FP16 | AArch64::AEK_RAS |
       AArch64::AEK_LSE | AArch64::AEK_RDM,
       "8.2-A"));
 }
 
-static constexpr unsigned NumAArch64CPUArchs = 38;
+static constexpr unsigned NumAArch64CPUArchs = 39;
 
 TEST(TargetParserTest, testAArch64CPUArchList) {
   SmallVector<StringRef, NumAArch64CPUArchs> List;
   AArch64::fillValidCPUArchList(List);
 
   // No list exists for these in this test suite, so ensure all are
   // valid, and match the expected 'magic' count.
   EXPECT_EQ(List.size(), NumAArch64CPUArchs);