[SPIRV] Addition of matrix multiply accumulate operands #138665
Conversation
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-backend-spir-v Author: Aadesh Premkumar (aadeshps-mcw) Changes--Added Matrix multiply accumulate operands for the extension SPV_INTEL_subgroup_matrix_multiply_accumulate Full diff: https://github.com/llvm/llvm-project/pull/138665.diff 3 Files Affected:
diff --git a/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVBaseInfo.h b/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVBaseInfo.h
index d009244a92259..3435e319fc816 100644
--- a/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVBaseInfo.h
+++ b/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVBaseInfo.h
@@ -57,6 +57,11 @@ namespace MemoryModel {
#include "SPIRVGenTables.inc"
} // namespace MemoryModel
+namespace MatrixMultiplyAccumulate {
+#define GET_MatrixMultiplyAccumulate_DECL
+#include "SPIRVGenTables.inc"
+} // namespace MatrixMultiplyAccumulate
+
namespace ExecutionMode {
#define GET_ExecutionMode_DECL
#include "SPIRVGenTables.inc"
diff --git a/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVInstPrinter.cpp b/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVInstPrinter.cpp
index b486132077e3b..4688b8dc7dcd0 100644
--- a/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVInstPrinter.cpp
+++ b/llvm/lib/Target/SPIRV/MCTargetDesc/SPIRVInstPrinter.cpp
@@ -242,6 +242,37 @@ void SPIRVInstPrinter::printInst(const MCInst *MI, uint64_t Address,
}
break;
}
+ case SPIRV::OpSubgroupMatrixMultiplyAccumulateINTEL: {
+ const unsigned NumOps = MI->getNumOperands();
+ if (NumFixedOps == NumOps)
+ break; // No extra operands, so no flags to process
+ OS << ' ';
+ // Extract the last operand only if it exists
+ if (NumOps > NumFixedOps) {
+ const unsigned Flags = MI->getOperand(NumOps - 1).getImm();
+ if (Flags == 0) {
+ printSymbolicOperand<
+ OperandCategory::MatrixMultiplyAccumulateOperandsOperand>(
+ MI, NumOps - 1, OS);
+ } else {
+ std::string Buffer;
+ for (unsigned Mask = 0x1;
+ Mask != SPIRV::MatrixMultiplyAccumulate::
+ MatrixBPackedBFloat16INTEL; // Replace with
+ // actual last flag
+ Mask <<= 1) {
+ if (Flags & Mask) {
+ if (!Buffer.empty())
+ Buffer += '|';
+ Buffer += getSymbolicOperandMnemonic(
+ OperandCategory::MatrixMultiplyAccumulateOperandsOperand, Mask);
+ }
+ }
+ OS << Buffer;
+ }
+ }
+ break;
+ }
default:
printRemainingVariableOps(MI, NumFixedOps, OS);
break;
diff --git a/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td b/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td
index cc32200a0a261..854368247b336 100644
--- a/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td
+++ b/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td
@@ -172,6 +172,7 @@ def KernelProfilingInfoOperand : OperandCategory;
def OpcodeOperand : OperandCategory;
def CooperativeMatrixLayoutOperand : OperandCategory;
def CooperativeMatrixOperandsOperand : OperandCategory;
+def MatrixMultiplyAccumulateOperandsOperand :OperandCategory;
//===----------------------------------------------------------------------===//
// Multiclass used to define Extesions enum values and at the same time
@@ -1746,3 +1747,40 @@ defm MatrixAAndBTF32ComponentsINTEL : CooperativeMatrixOperandsOperand<0x20, [SP
defm MatrixAAndBBFloat16ComponentsINTEL : CooperativeMatrixOperandsOperand<0x40, [SPV_INTEL_joint_matrix], [CooperativeMatrixBFloat16ComponentTypeINTEL]>;
defm MatrixCBFloat16ComponentsINTEL : CooperativeMatrixOperandsOperand<0x80, [SPV_INTEL_joint_matrix], [CooperativeMatrixBFloat16ComponentTypeINTEL]>;
defm MatrixResultBFloat16ComponentsINTEL : CooperativeMatrixOperandsOperand<0x100, [SPV_INTEL_joint_matrix], [CooperativeMatrixBFloat16ComponentTypeINTEL]>;
+
+//===----------------------------------------------------------------------===//
+// Multiclass used to define Matrix Multiply Accumulate Operands enum values and at the same time
+// SymbolicOperand entries with string mnemonics and capabilities.
+//===----------------------------------------------------------------------===//
+def MatrixMultiplyAccumulate : GenericEnum, Operand<i32> {
+ let FilterClass = "MatrixMultiplyAccumulate";
+ let NameField = "Name";
+ let ValueField = "Value";
+ let PrintMethod = !strconcat("printSymbolicOperand<OperandCategory::", FilterClass, "Operand>");
+}
+
+class MatrixMultiplyAccumulate<string name, bits<32> value> {
+ string Name = name;
+ bits<32> Value = value;
+}
+
+multiclass MatrixMultiplyAccumulateOperandsOperand<bits<32> value, list<Extension> reqExtensions> {
+def : MatrixMultiplyAccumulate<NAME, value>;
+ defm : SymbolicOperandWithRequirements< MatrixMultiplyAccumulateOperandsOperand, value, NAME, 0, 0, reqExtensions, []>;
+}
+
+defm None : MatrixMultiplyAccumulateOperandsOperand<0x0, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixASignedComponentsINTEL : MatrixMultiplyAccumulateOperandsOperand<0x1, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBSignedComponentsINTEL : MatrixMultiplyAccumulateOperandsOperand<0x2, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixCBFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x4, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixResultBFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x8, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixAPackedInt8INTEL : MatrixMultiplyAccumulateOperandsOperand<0x10, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBPackedInt8INTEL : MatrixMultiplyAccumulateOperandsOperand<0x20, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixAPackedInt4INTEL : MatrixMultiplyAccumulateOperandsOperand<0x40, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBPackedInt4INTEL : MatrixMultiplyAccumulateOperandsOperand<0x80, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixATF32INTEL : MatrixMultiplyAccumulateOperandsOperand<0x100, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBTF32INTEL : MatrixMultiplyAccumulateOperandsOperand<0x200, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixAPackedFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x400, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBPackedFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x800, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixAPackedBFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x1000, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
+defm MatrixBPackedBFloat16INTEL : MatrixMultiplyAccumulateOperandsOperand<0x2000, [SPV_INTEL_subgroup_matrix_multiply_accumulate]>;
|
006813f to
4ad6d5b
Compare
|
Ping |
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
…EL_subgroup_matrix_multiply_accumulate
4ad6d5b to
f764d86
Compare
|
@aadeshps-mcw Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR. Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues. How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
--Added Matrix multiply accumulate operands for the extension SPV_INTEL_subgroup_matrix_multiply_accumulate