[MC][Tablegen] Allow models to describe the retire control unit for llvm-mca.

[andreadb]

Summary

This patch adds the ability to describe properties of the hardware retire control unit.

Tablegen class RetireControlUnit has been added for this purpose (see TargetSchedule.td).

A RetireControlUnit specifies the size of the reorder buffer, as well as the maximum number of opcodes that can be retired every cycle.

A zero (or negative) value for the reorder buffer size means: "the size is unknown". If the size is unknown, then llvm-mca defaults it to the value of field SchedMachineModel::MicroOpBufferSize.
A zero or negative number of opcodes retired per cycle means: "there is no restriction on the number of instructions that can be retired every cycle".

Models can optionally specify an instance of RetireControlUnit. There can only be up-to one RetireControlUnit definition per scheduling model.

Information related to the RCU (RetireControlUnit) is stored in (two new fields of) MCExtraProcessorInfo.
llvm-mca loads that information when it initializes the DispatchUnit / RetireControlUnit (see Dispatch.h/Dispatch.cpp).

This patch fixes PR36661.

Please let me know if okay to commit

-Andrea

[courbet]

tools/llvm-mca/Dispatch.cpp

262 ↗

(On Diff #140958)

Can we get rid of the MaxRetirePerCycle ctor parameter ? Is there any reason why the user would override this ?

tools/llvm-mca/Dispatch.h

194 ↗

(On Diff #140958)

Any reason for this to be in a separate function ? What about moving the ctor definition to the cpp file and inlining the code ?
Else when reading the ctor definition here it's not obvious that AvailableSlots might be overridden in initialize().

[andreadb]

tools/llvm-mca/Dispatch.cpp

262 ↗

(On Diff #140958)

The original idea was to let users override the retire-per-cycle.
However, with this patch, the max-retire-per-cycle can now be set via tablegen.

That being said, most processors don't provide a max-retire-per-cycle. This patch would only fix the BtVer2 model.
So, I am tempted to keep the flag for now (although, I don't have a strong opinion about it). Alternatively, I can commit a separate patch that removes the flag, and the rebase this patch. What do you think?

tools/llvm-mca/Dispatch.h

194 ↗

(On Diff #140958)

No reason. I will move the constructor to the cpp file.

[courbet]

Why did the tests change ? I thought that this change would not impact tests.

tools/llvm-mca/Dispatch.cpp

262 ↗

(On Diff #140958)

I'm pretty sure that I would stick to tablegen-specified models as a user. And if you're developing you own processor, I hope that you have your own tools :) So the flag does not sound that useful to me given that it makes a bit harder to understand this code.
I don't have a strong opinion though, so feel free to keep the flag if i'ts useful to you.

[andreadb]

In D45259#1058226, @courbet wrote:

Why did the tests change ? I thought that this change would not impact tests.

It does have an impact on the retire stage.

On BtVer2, the retire throughput is 2 instructions per cycle. It does affect the timeline at least.

tools/llvm-mca/Dispatch.cpp

262 ↗

(On Diff #140958)

That's a good point (especially if you have your own tools).
It could have been used to experiment/play with the RCU (with SMT processors we might emulate a different retire throughput).
That being said, the argument is a bit weak. SMT introduces other issues related to the partitioning of resources which cannot be fully tweaked via flags.

If okay for you, I am going to commit a change that removes the flag and updates the docs in preparation for this patch. Then I rebase this patch.

[andreadb]

Patch updated.

Addressed review comments.

[andreadb]

tools/llvm-mca/Dispatch.cpp

262 ↗

(On Diff #140958)

I went ahead and removed that flag at r329274.

[RKSimon]

LGTM with a couple of minors

include/llvm/MC/MCSchedule.h

176 ↗

(On Diff #141142)

These comments repeat a lot of what is said in TargetSchedule.td - make the comments briefer here?

tools/llvm-mca/Dispatch.h

262 ↗

(On Diff #141142)

Are you confident that we won't need llvm::MCSubtargetInfo again anytime soon?

[andreadb]

Thanks Simon.

include/llvm/MC/MCSchedule.h

176 ↗

(On Diff #141142)

I will simplify the comment.

tools/llvm-mca/Dispatch.h

262 ↗

(On Diff #141142)

Ideally, all the information needed by the dispatch logic should be accessible through the scheduling model. I may be wrong, but I don't think that we will need it again.

[benhamilton]

llvm/trunk/utils/TableGen/SubtargetEmitter.cpp

616

Looks like this broke the build:

http://green.lab.llvm.org/green//job/clang-stage1-cmake-RA-incremental/47722/consoleFull#-172429723349ba4694-19c4-4d7e-bec5-911270d8a58c

/Users/buildslave/jenkins/workspace/clang-stage1-cmake-RA-incremental/llvm/utils/TableGen/SubtargetEmitter.cpp:616:9: error: no matching function for call to 'max'
        std::max(ReorderBufferSize, RCU->getValueAsInt("ReorderBufferSize"));
        ^~~~~~~~

P8077

[andreadb]

I am going to fix it now.
I guess I should have used int64_t ...

llvm/trunk/include/llvm/MC/MCSchedule.h

 };
 
 /// Provide extra details about the machine processor.
 ///
 /// This is a collection of "optional" processor information that is not
 /// normally used by the LLVM machine schedulers, but that can be consumed by
 /// external tools like llvm-mca to improve the quality of the peformance
 /// analysis.
-/// In future, the plan is to extend this struct with extra information (for
-/// example: maximum number of instructions retired per cycle; actual size of
-/// the reorder buffer; etc.).
 struct MCExtraProcessorInfo {
+  // Actual size of the reorder buffer in hardware.
+  unsigned ReorderBufferSize;
+  // Number of instructions retired per cycle.
+  unsigned MaxRetirePerCycle;
   const MCRegisterFileDesc *RegisterFiles;
   unsigned NumRegisterFiles;
   const MCRegisterCostEntry *RegisterCostTable;
   unsigned NumRegisterCostEntries;
 };
 
 /// Machine model for scheduling, bundling, and heuristics.
 ///

llvm/trunk/include/llvm/Target/TargetSchedule.td

 // SchedModel will usually be provided by surrounding let statement
 // and ties this SchedAlias mapping to a processor.
 class SchedAlias<SchedReadWrite match, SchedReadWrite alias> {
   SchedReadWrite MatchRW = match;
   SchedReadWrite AliasRW = alias;
   SchedMachineModel SchedModel = ?;
 }
 
-// Alow the definition of processor register files.
+// Allow the definition of processor register files.
 // Each processor register file declares the number of physical registers, as
 // well as a optional register cost information. The cost of a register R is the
 // number of physical registers used to rename R (at register renaming stage).
 // That value defaults to 1, to all the registers contained in the register
 // file. The set of target register files is inferred from the list of register
 // classes. Register costs are defined at register class granularity. An empty
 // list of register classes means that this register file contains all the
 // registers defined by the target.
 class RegisterFile<int numPhysRegs, list<RegisterClass> Classes = [],
                    list<int> Costs = []> {
   list<RegisterClass> RegClasses = Classes;
   list<int> RegCosts = Costs;
   int NumPhysRegs = numPhysRegs;
   SchedMachineModel SchedModel = ?;
 }
+
+// Describe the retire control unit.
+// A retire control unit specifies the size of the reorder buffer, as well as
+// the maximum number of opcodes that can be retired every cycle.
+// A value less-than-or-equal-to zero for field 'ReorderBufferSize' means: "the
+// size is unknown". The idea is that external tools can fall-back to using
+// field MicroOpBufferSize in SchedModel if the reorder buffer size is unknown.
+// A zero or negative value for field 'MaxRetirePerCycle' means "no
+// restrictions on the number of instructions retired per cycle".
+// Models can optionally specify up to one instance of RetireControlUnit per
+// scheduling model.
+class RetireControlUnit<int bufferSize, int retirePerCycle> {
+  int ReorderBufferSize = bufferSize;
+  int MaxRetirePerCycle = retirePerCycle;
+  SchedMachineModel SchedModel = ?;
+}
+

llvm/trunk/lib/Target/X86/X86ScheduleBtVer2.td

 // Reference: www.realworldtech.com/jaguar/4/
 def IntegerPRF : RegisterFile<64, [GR8, GR16, GR32, GR64, CCR]>;
 
 // The Jaguar FP Retire Queue renames SIMD and FP uOps onto a pool of 72 SSE
 // registers. Operations on 256-bit data types are cracked into two COPs.
 // Reference: www.realworldtech.com/jaguar/4/
 def FpuPRF: RegisterFile<72, [VR64, VR128, VR256], [1, 1, 2]>;
 
+// The retire control unit (RCU) can track up to 64 macro-ops in-flight. It can
+// retire up to two macro-ops per cycle.
+// Reference: "Software Optimization Guide for AMD Family 16h Processors"
+def RCU : RetireControlUnit<64, 2>;
+
 // Integer Pipe Scheduler
 def JALU01 : ProcResGroup<[JALU0, JALU1]> {
   let BufferSize=20;
 }
 
 // AGU Pipe Scheduler
 def JLSAGU : ProcResGroup<[JLAGU, JSAGU]> {
   let BufferSize=12;

llvm/trunk/test/tools/llvm-mca/X86/BtVer2/dot-product.s

 
 # CHECK:      Timeline view:
 # CHECK-NEXT:                 012345
 # CHECK-NEXT: Index 0123456789
 
 # CHECK:      [0,0] DeeER.    .    .  vmulps  %xmm0, %xmm1, %xmm2
 # CHECK-NEXT: [0,1] D==eeeER  .    .  vhaddps %xmm2, %xmm2, %xmm3
 # CHECK-NEXT: [0,2] .D====eeeER    .  vhaddps %xmm3, %xmm3, %xmm4
-
 # CHECK:      [1,0] .DeeE-----R    .  vmulps  %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [1,1] . D=eeeE--R    .  vhaddps %xmm2, %xmm2, %xmm3
+# CHECK-NEXT: [1,1] . D=eeeE---R   .  vhaddps %xmm2, %xmm2, %xmm3
 # CHECK-NEXT: [1,2] . D====eeeER   .  vhaddps %xmm3, %xmm3, %xmm4
-
-# CHECK:      [2,0] .  DeeE----R   .  vmulps  %xmm0, %xmm1, %xmm2
+# CHECK:      [2,0] .  DeeE-----R  .  vmulps  %xmm0, %xmm1, %xmm2
 # CHECK-NEXT: [2,1] .  D====eeeER  .  vhaddps %xmm2, %xmm2, %xmm3
 # CHECK-NEXT: [2,2] .   D======eeeER  vhaddps %xmm3, %xmm3, %xmm4
 
+
 # CHECK:      Average Wait times (based on the timeline view):
 # CHECK-NEXT: [0]: Executions
 # CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
 # CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
 # CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
 
 # CHECK:            [0]    [1]    [2]    [3]
-# CHECK-NEXT: 0.     3     1.0    1.0    3.0    vmulps  %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: 1.     3     3.3    0.7    0.7    vhaddps %xmm2, %xmm2, %xmm3
+# CHECK-NEXT: 0.     3     1.0    1.0    3.3    vmulps  %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: 1.     3     3.3    0.7    1.0    vhaddps %xmm2, %xmm2, %xmm3
 # CHECK-NEXT: 2.     3     5.7    0.0    0.0    vhaddps %xmm3, %xmm3, %xmm4

llvm/trunk/test/tools/llvm-mca/X86/BtVer2/pipes-fpu.s

 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -      -     1.00    -      -      -       vcvttps2dq  %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -      -      -     1.00    -      -      -      -      -      -      -     1.00     vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -       vaddps  %xmm0, %xmm1, %xmm2
 # CHECK-NEXT:  -      -      -      -     21.00   -     1.00    -      -      -      -      -      -      -       vsqrtps %xmm0, %xmm2
 # CHECK-NEXT:  -      -      -     2.00    -     2.00    -      -      -      -      -      -      -      -       vaddps  %ymm0, %ymm1, %ymm2
 # CHECK-NEXT:  -      -      -      -     42.00   -     2.00    -      -      -      -      -      -      -       vsqrtps %ymm0, %ymm2
 
 
-# CHECK:      Timeline view:
-# CHECK-NEXT:                 0123456789          0123456789          0123456789        
-# CHECK-NEXT: Index 0123456789          0123456789          0123456789          01234567
 
-# CHECK:      [0,0] DeeeeER   .    .    .    .    .    .    .    .    .    .    .    . .  vpmulld %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [0,1] .DeE--R   .    .    .    .    .    .    .    .    .    .    .    . .  vpand %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [0,2] . DeeeER  .    .    .    .    .    .    .    .    .    .    .    . .  vcvttps2dq  %xmm0, %xmm2
-# CHECK-NEXT: [0,3] .  DeeE-R .    .    .    .    .    .    .    .    .    .    .    . .  vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [0,4] .  DeeeER .    .    .    .    .    .    .    .    .    .    .    . .  vaddps  %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [0,5] .   DeeeeeeeeeeeeeeeeeeeeeER  .    .    .    .    .    .    .    . .  vsqrtps %xmm0, %xmm2
-# CHECK-NEXT: [0,6] .    DeeeE-----------------R  .    .    .    .    .    .    .    . .  vaddps  %ymm0, %ymm1, %ymm2
-# CHECK-NEXT: [0,7] .     D===================eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeER  vsqrtps %ymm0, %ymm2
-
-# CHECK:      [1,0] .    .DeeeeE--------------------------------------------------------R vpmulld %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [1,1] .    . DeE----------------------------------------------------------R vpand %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [1,2] .    .  DeeeE-------------------------------------------------------R vcvttps2dq  %xmm0, %xmm2
-# CHECK-NEXT: [1,3] .    .  DeeE--------------------------------------------------------R vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: [1,4] .    .   DeeeE------------------------------------------------------R vaddps  %xmm0, %xmm1, %xmm2
+# CHECK:      Timeline view:
+# CHECK-NEXT:                 0123456789          0123456789          0123456789          0
+# CHECK-NEXT: Index 0123456789          0123456789          0123456789          0123456789 
+# CHECK:      [0,0] DeeeeER   .    .    .    .    .    .    .    .    .    .    .    .    . vpmulld %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [0,1] .DeE--R   .    .    .    .    .    .    .    .    .    .    .    .    . vpand %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [0,2] . DeeeER  .    .    .    .    .    .    .    .    .    .    .    .    . vcvttps2dq  %xmm0, %xmm2
+# CHECK-NEXT: [0,3] . DeeE-R  .    .    .    .    .    .    .    .    .    .    .    .    . vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [0,4] .  DeeeER .    .    .    .    .    .    .    .    .    .    .    .    . vaddps  %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [0,5] .  DeeeeeeeeeeeeeeeeeeeeeER   .    .    .    .    .    .    .    .    . vsqrtps %xmm0, %xmm2
+# CHECK-NEXT: [0,6] .   DeeeE-----------------R   .    .    .    .    .    .    .    .    . vaddps  %ymm0, %ymm1, %ymm2
+# CHECK-NEXT: [0,7] .    D===================eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeER . vsqrtps %ymm0, %ymm2
+# CHECK:      [1,0] .    .DeeeeE--------------------------------------------------------R . vpmulld %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [1,1] .    . DeE-----------------------------------------------------------R. vpand %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [1,2] .    .  DeeeE--------------------------------------------------------R. vcvttps2dq  %xmm0, %xmm2
+# CHECK-NEXT: [1,3] .    .  DeeE----------------------------------------------------------R vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: [1,4] .    .   DeeeE--------------------------------------------------------R vaddps  %xmm0, %xmm1, %xmm2
 
 
 # CHECK:      Average Wait times (based on the timeline view):
 # CHECK-NEXT: [0]: Executions
 # CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
 # CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
 # CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
 
-# CHECK:            [0]    [1]    [2]    [3]
+# CHECK:       [0]    [1]    [2]    [3]
 # CHECK-NEXT: 0.     2     1.0    1.0    28.0   vpmulld %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: 1.     2     1.0    1.0    30.0   vpand %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: 2.     2     1.0    1.0    27.5   vcvttps2dq  %xmm0, %xmm2
-# CHECK-NEXT: 3.     2     1.0    1.0    28.5   vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
-# CHECK-NEXT: 4.     2     1.0    1.0    27.0   vaddps  %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: 1.     2     1.0    1.0    30.5   vpand %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: 2.     2     1.0    1.0    28.0   vcvttps2dq  %xmm0, %xmm2
+# CHECK-NEXT: 3.     2     1.0    1.0    29.5   vpclmulqdq  $0, %xmm0, %xmm1, %xmm2
+# CHECK-NEXT: 4.     2     1.0    1.0    28.0   vaddps  %xmm0, %xmm1, %xmm2
 # CHECK-NEXT: 5.     1     1.0    1.0    0.0    vsqrtps %xmm0, %xmm2
 # CHECK-NEXT: 6.     1     1.0    1.0    17.0   vaddps  %ymm0, %ymm1, %ymm2
 # CHECK-NEXT: 7.     1     20.0   20.0   0.0    vsqrtps %ymm0, %ymm2

llvm/trunk/test/tools/llvm-mca/X86/BtVer2/register-files-5.s

 
 
 # CHECK:      Timeline view:
 # CHECK-NEXT:                 0123456789          0123456789          0123456789          
 # CHECK-NEXT: Index 0123456789          0123456789          0123456789          0123456789
 
 # CHECK:      [0,0] DeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeER    .    .    .    .    .   .  vdivps  %ymm0, %ymm0, %ymm1
 # CHECK-NEXT: [0,1] .DeeeE----------------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm2
-# CHECK-NEXT: [0,2] . D=eeeE--------------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm3
-# CHECK-NEXT: [0,3] .  D==eeeE------------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm4
-# CHECK-NEXT: [0,4] .   D===eeeE----------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm5
-# CHECK-NEXT: [0,5] .    D====eeeE--------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm6
-# CHECK-NEXT: [0,6] .    .D=====eeeE------------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm7
-# CHECK-NEXT: [0,7] .    . D======eeeE----------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm8
-# CHECK-NEXT: [0,8] .    .  D=======eeeE--------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm9
-# CHECK-NEXT: [0,9] .    .   D========eeeE------------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm10
-# CHECK-NEXT: [0,10]  .    .    D=========eeeE----------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm11
-# CHECK-NEXT: [0,11]  .    .    .D==========eeeE--------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm12
-# CHECK-NEXT: [0,12]  .    .    . D===========eeeE------------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm13
-# CHECK-NEXT: [0,13]  .    .    .  D============eeeE----------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm14
-# CHECK-NEXT: [0,14]  .    .    .   D=============eeeE--------R    .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm15
-# CHECK-NEXT: [0,15]  .    .    .    D==============eeeE------R    .    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm0
-# CHECK-NEXT: [0,16]  .    .    .    .D================eeeE---R    .    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm3
-# CHECK-NEXT: [0,17]  .    .    .    . D=================eeeE-R    .    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm4
-# CHECK-NEXT: [0,18]  .    .    .    .  D==================eeeER   .    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm5
-# CHECK-NEXT: [0,19]  .    .    .    .   D===================eeeER .    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm6
-# CHECK-NEXT: [0,20]  .    .    .    .    D====================eeeER    .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm7
-# CHECK-NEXT: [0,21]  .    .    .    .    .D=====================eeeER  .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm8
-# CHECK-NEXT: [0,22]  .    .    .    .    . D======================eeeER.    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm9
+# CHECK-NEXT: [0,2] . D=eeeE---------------------------------R   .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm3
+# CHECK-NEXT: [0,3] .  D==eeeE-------------------------------R   .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm4
+# CHECK-NEXT: [0,4] .   D===eeeE------------------------------R  .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm5
+# CHECK-NEXT: [0,5] .    D====eeeE----------------------------R  .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm6
+# CHECK-NEXT: [0,6] .    .D=====eeeE---------------------------R .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm7
+# CHECK-NEXT: [0,7] .    . D======eeeE-------------------------R .    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm8
+# CHECK-NEXT: [0,8] .    .  D=======eeeE------------------------R.    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm9
+# CHECK-NEXT: [0,9] .    .   D========eeeE----------------------R.    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm10
+# CHECK-NEXT: [0,10]  .    .    D=========eeeE---------------------R    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm11
+# CHECK-NEXT: [0,11]  .    .    .D==========eeeE-------------------R    .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm12
+# CHECK-NEXT: [0,12]  .    .    . D===========eeeE------------------R   .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm13
+# CHECK-NEXT: [0,13]  .    .    .  D============eeeE----------------R   .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm14
+# CHECK-NEXT: [0,14]  .    .    .   D=============eeeE---------------R  .    .    .    .   .  vaddps  %ymm0, %ymm0, %ymm15
+# CHECK-NEXT: [0,15]  .    .    .    D==============eeeE-------------R  .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm0
+# CHECK-NEXT: [0,16]  .    .    .    .D================eeeE-----------R .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm3
+# CHECK-NEXT: [0,17]  .    .    .    . D=================eeeE---------R .    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm4
+# CHECK-NEXT: [0,18]  .    .    .    .  D==================eeeE--------R.    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm5
+# CHECK-NEXT: [0,19]  .    .    .    .   D===================eeeE------R.    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm6
+# CHECK-NEXT: [0,20]  .    .    .    .    D====================eeeE-----R    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm7
+# CHECK-NEXT: [0,21]  .    .    .    .    .D=====================eeeE---R    .    .    .   .  vaddps  %ymm2, %ymm0, %ymm8
+# CHECK-NEXT: [0,22]  .    .    .    .    . D======================eeeE--R   .    .    .   .  vaddps  %ymm2, %ymm0, %ymm9
 # CHECK-NEXT: [0,23]  .    .    .    .    .  D=======================eeeER   .    .    .   .  vaddps  %ymm2, %ymm0, %ymm10
 # CHECK-NEXT: [0,24]  .    .    .    .    .   D========================eeeER .    .    .   .  vaddps  %ymm2, %ymm0, %ymm11
 # CHECK-NEXT: [0,25]  .    .    .    .    .    D=========================eeeER    .    .   .  vaddps  %ymm2, %ymm0, %ymm12
 # CHECK-NEXT: [0,26]  .    .    .    .    .    .D==========================eeeER  .    .   .  vaddps  %ymm2, %ymm0, %ymm13
 # CHECK-NEXT: [0,27]  .    .    .    .    .    . D===========================eeeER.    .   .  vaddps  %ymm2, %ymm0, %ymm14
 # CHECK-NEXT: [0,28]  .    .    .    .    .    .  D============================eeeER   .   .  vaddps  %ymm2, %ymm0, %ymm15
 # CHECK-NEXT: [0,29]  .    .    .    .    .    .   D=============================eeeER .   .  vaddps  %ymm3, %ymm0, %ymm2
 # CHECK-NEXT: [0,30]  .    .    .    .    .    .    D==============================eeeER   .  vaddps  %ymm3, %ymm0, %ymm4
 # CHECK-NEXT: [0,31]  .    .    .    .    .    .    .D===============================eeeER .  vaddps  %ymm3, %ymm0, %ymm5
 # CHECK-NEXT: [0,32]  .    .    .    .    .    .    .    .    D========================eeeER  vaddps  %ymm3, %ymm0, %ymm6

llvm/trunk/tools/llvm-mca/Backend.h

Backend(const llvm::MCSubtargetInfo &Subtarget,
           const llvm::MCRegisterInfo &MRI, InstrBuilder &B, SourceMgr &Source,
           unsigned DispatchWidth = 0, unsigned RegisterFileSize = 0,
           unsigned LoadQueueSize = 0, unsigned StoreQueueSize = 0,
           bool AssumeNoAlias = false)
       : STI(Subtarget), IB(B),
         HWS(llvm::make_unique<Scheduler>(this, Subtarget.getSchedModel(),
                                          LoadQueueSize, StoreQueueSize,
                                          AssumeNoAlias)),
-        DU(llvm::make_unique<DispatchUnit>(
-            this, STI, MRI, Subtarget.getSchedModel().MicroOpBufferSize,
-            RegisterFileSize, DispatchWidth, HWS.get())),
+        DU(llvm::make_unique<DispatchUnit>(this, Subtarget.getSchedModel(), MRI,
+                                           RegisterFileSize, DispatchWidth,
+                                           HWS.get())),
         SM(Source), Cycles(0) {
     HWS->setDispatchUnit(DU.get());
   }
 
   void run() {
     while (SM.hasNext() || !DU->isRCUEmpty())
       runCycle(Cycles++);
   }

llvm/trunk/tools/llvm-mca/Dispatch.h

private:
   unsigned NextAvailableSlotIdx;
   unsigned CurrentInstructionSlotIdx;
   unsigned AvailableSlots;
   unsigned MaxRetirePerCycle; // 0 means no limit.
   std::vector<RUToken> Queue;
   DispatchUnit *Owner;
 
 public:
-  RetireControlUnit(unsigned NumSlots, DispatchUnit *DU)
-      : NextAvailableSlotIdx(0), CurrentInstructionSlotIdx(0),
-        AvailableSlots(NumSlots), MaxRetirePerCycle(0), Owner(DU) {
-    assert(NumSlots && "Expected at least one slot!");
-    Queue.resize(NumSlots);
-  }
+  RetireControlUnit(const llvm::MCSchedModel &SM, DispatchUnit *DU);
 
   bool isFull() const { return !AvailableSlots; }
   bool isEmpty() const { return AvailableSlots == Queue.size(); }
   bool isAvailable(unsigned Quantity = 1) const {
     // Some instructions may declare a number of uOps which exceedes the size
     // of the reorder buffer. To avoid problems, cap the amount of slots to
     // the size of the reorder buffer.
     Quantity = std::min(Quantity, static_cast<unsigned>(Queue.size()));
class DispatchUnit {
   bool checkRCU(unsigned Index, const InstrDesc &Desc);
   bool checkScheduler(unsigned Index, const InstrDesc &Desc);
 
   void updateRAWDependencies(ReadState &RS, const llvm::MCSubtargetInfo &STI);
   void notifyInstructionDispatched(unsigned IID,
                                    llvm::ArrayRef<unsigned> UsedPhysRegs);
 
 public:
-  DispatchUnit(Backend *B, const llvm::MCSubtargetInfo &STI,
-               const llvm::MCRegisterInfo &MRI, unsigned MicroOpBufferSize,
-               unsigned RegisterFileSize, unsigned MaxDispatchWidth,
-               Scheduler *Sched)
+  DispatchUnit(Backend *B, const llvm::MCSchedModel &SM,
+               const llvm::MCRegisterInfo &MRI, unsigned RegisterFileSize,
+               unsigned MaxDispatchWidth, Scheduler *Sched)
       : DispatchWidth(MaxDispatchWidth), AvailableEntries(MaxDispatchWidth),
         CarryOver(0U), SC(Sched),
-        RAT(llvm::make_unique<RegisterFile>(STI.getSchedModel(), MRI,
-                                            RegisterFileSize)),
-        RCU(llvm::make_unique<RetireControlUnit>(MicroOpBufferSize, this)),
-        Owner(B) {}
+        RAT(llvm::make_unique<RegisterFile>(SM, MRI, RegisterFileSize)),
+        RCU(llvm::make_unique<RetireControlUnit>(SM, this)), Owner(B) {}
 
   unsigned getDispatchWidth() const { return DispatchWidth; }
 
   bool isAvailable(unsigned NumEntries) const {
     return NumEntries <= AvailableEntries || AvailableEntries == DispatchWidth;
   }
 
   bool isRCUEmpty() const { return RCU->isEmpty(); }

llvm/trunk/tools/llvm-mca/Dispatch.cpp

for (unsigned I = 0, E = getNumRegisterFiles(); I < E; ++I) {
     dbgs() << "Register File #" << I;
     const RegisterMappingTracker &RMT = RegisterFiles[I];
     dbgs() << "\n  TotalMappings:        " << RMT.TotalMappings
            << "\n  NumUsedMappings:      " << RMT.NumUsedMappings << '\n';
   }
 }
 #endif
 
+RetireControlUnit::RetireControlUnit(const llvm::MCSchedModel &SM,
+                                     DispatchUnit *DU)
+    : NextAvailableSlotIdx(0), CurrentInstructionSlotIdx(0),
+      AvailableSlots(SM.MicroOpBufferSize), MaxRetirePerCycle(0), Owner(DU) {
+  // Check if the scheduling model provides extra information about the machine
+  // processor. If so, then use that information to set the reorder buffer size
+  // and the maximum number of instructions retired per cycle.
+  if (SM.hasExtraProcessorInfo()) {
+    const MCExtraProcessorInfo &EPI = SM.getExtraProcessorInfo();
+    if (EPI.ReorderBufferSize)
+      AvailableSlots = EPI.ReorderBufferSize;
+    MaxRetirePerCycle = EPI.MaxRetirePerCycle;
+  }
+
+  assert(AvailableSlots && "Invalid reorder buffer size!");
+  Queue.resize(AvailableSlots);
+}
+
 // Reserves a number of slots, and returns a new token.
 unsigned RetireControlUnit::reserveSlot(unsigned Index, unsigned NumMicroOps) {
   assert(isAvailable(NumMicroOps));
   unsigned NormalizedQuantity =
       std::min(NumMicroOps, static_cast<unsigned>(Queue.size()));
   // Zero latency instructions may have zero mOps. Artificially bump this
   // value to 1. Although zero latency instructions don't consume scheduler
   // resources, they still consume one slot in the retire queue.

llvm/trunk/utils/TableGen/CodeGenSchedule.h

struct CodeGenProcModel {
   RecVec ReadAdvanceDefs;
 
   // Per-operand machine model resources associated with this processor.
   RecVec ProcResourceDefs;
 
   // List of Register Files.
   std::vector<CodeGenRegisterFile> RegisterFiles;
 
+  // Optional Retire Control Unit definition.
+  Record *RetireControlUnit;
+
   CodeGenProcModel(unsigned Idx, std::string Name, Record *MDef,
                    Record *IDef) :
-    Index(Idx), ModelName(std::move(Name)), ModelDef(MDef), ItinsDef(IDef) {}
+    Index(Idx), ModelName(std::move(Name)), ModelDef(MDef), ItinsDef(IDef),
+    RetireControlUnit(nullptr) {}
 
   bool hasItineraries() const {
     return !ItinsDef->getValueAsListOfDefs("IID").empty();
   }
 
   bool hasInstrSchedModel() const {
     return !WriteResDefs.empty() || !ItinRWDefs.empty();
   }
 
   bool hasExtraProcessorInfo() const {
-    return !RegisterFiles.empty();
+    return RetireControlUnit || !RegisterFiles.empty();
   }
 
   unsigned getProcResourceIdx(Record *PRDef) const;
 
   bool isUnsupported(const CodeGenInstruction &Inst) const;
 
 #ifndef NDEBUG
   void dump() const;
private:
 
   void collectSchedRW();
 
   std::string genRWName(ArrayRef<unsigned> Seq, bool IsRead);
   unsigned findRWForSequence(ArrayRef<unsigned> Seq, bool IsRead);
 
   void collectSchedClasses();
 
+  void collectRetireControlUnits();
+
   void collectRegisterFiles();
 
+  void collectOptionalProcessorInfo();
+
   std::string createSchedClassName(Record *ItinClassDef,
                                    ArrayRef<unsigned> OperWrites,
                                    ArrayRef<unsigned> OperReads);
   std::string createSchedClassName(const RecVec &InstDefs);
   void createInstRWClass(Record *InstRWDef);
 
   void collectProcItins();
 

llvm/trunk/utils/TableGen/CodeGenSchedule.cpp

CodeGenSchedModels::CodeGenSchedModels(RecordKeeper &RK,
   // Infer new SchedClasses from SchedVariant.
   inferSchedClasses();
 
   // Populate each CodeGenProcModel's WriteResDefs, ReadAdvanceDefs, and
   // ProcResourceDefs.
   DEBUG(dbgs() << "\n+++ RESOURCE DEFINITIONS (collectProcResources) +++\n");
   collectProcResources();
 
+  // Collect optional processor description.
+  collectOptionalProcessorInfo();
+
+  checkCompleteness();
+}
+
+void CodeGenSchedModels::collectRetireControlUnits() {
+  RecVec Units = Records.getAllDerivedDefinitions("RetireControlUnit");
+
+  for (Record *RCU : Units) {
+    CodeGenProcModel &PM = getProcModel(RCU->getValueAsDef("SchedModel"));
+    if (PM.RetireControlUnit) {
+      PrintError(RCU->getLoc(),
+                 "Expected a single RetireControlUnit definition");
+      PrintNote(PM.RetireControlUnit->getLoc(),
+                "Previous definition of RetireControlUnit was here");
+    }
+    PM.RetireControlUnit = RCU;
+  }
+}
+
+/// Collect optional processor information.
+void CodeGenSchedModels::collectOptionalProcessorInfo() {
   // Find register file definitions for each processor.
   collectRegisterFiles();
 
-  checkCompleteness();
+  // Collect processor RetireControlUnit descriptors if available.
+  collectRetireControlUnits();
 }
 
 /// Gather all processor models.
 void CodeGenSchedModels::collectProcModels() {
   RecVec ProcRecords = Records.getAllDerivedDefinitions("Processor");
   std::sort(ProcRecords.begin(), ProcRecords.end(), LessRecordFieldName());
 
   // Reserve space because we can. Reallocation would be ok.

llvm/trunk/utils/TableGen/SubtargetEmitter.cpp

for (Record *RUDef : ResUnits) {
         OS << "  " << ProcModel.getProcResourceIdx(RU) << ", ";
       }
     }
     OS << "  // " << PRDef->getName() << "\n";
   }
   OS << "};\n";
 }
 
+static void EmitRetireControlUnitInfo(const CodeGenProcModel &ProcModel,
+                                      raw_ostream &OS) {
+  long ReorderBufferSize = 0, MaxRetirePerCycle = 0;
+  if (Record *RCU = ProcModel.RetireControlUnit) {
+    ReorderBufferSize =
+        std::max(ReorderBufferSize, RCU->getValueAsInt("ReorderBufferSize"));
+    MaxRetirePerCycle =
+        std::max(MaxRetirePerCycle, RCU->getValueAsInt("MaxRetirePerCycle"));
+  }
+
+  OS << ReorderBufferSize << ", // ReorderBufferSize\n  ";
+  OS << MaxRetirePerCycle << ", // MaxRetirePerCycle\n  ";
+}
+
 static void EmitRegisterFileInfo(const CodeGenProcModel &ProcModel,
                                  unsigned NumRegisterFiles,
                                  unsigned NumCostEntries, raw_ostream &OS) {
   if (NumRegisterFiles)
     OS << ProcModel.ModelName << "RegisterFiles,\n  " << (1 + NumRegisterFiles);
   else
     OS << "nullptr,\n  0";
 
void SubtargetEmitter::EmitExtraProcessorInfo(const CodeGenProcModel &ProcModel,
   // Generate a table of register file descriptors (one entry per each user
   // defined register file), and a table of register costs.
   unsigned NumCostEntries = EmitRegisterFileTables(ProcModel, OS);
 
   // Now generate a table for the extra processor info.
   OS << "\nstatic const llvm::MCExtraProcessorInfo " << ProcModel.ModelName
      << "ExtraInfo = {\n  ";
 
+  // Add information related to the retire control unit.
+  EmitRetireControlUnitInfo(ProcModel, OS);
+
   // Add information related to the register files (i.e. where to find register
   // file descriptors and register costs).
   EmitRegisterFileInfo(ProcModel, ProcModel.RegisterFiles.size(),
                        NumCostEntries, OS);
 
   OS << "};\n";
 }