[AMDGPU] Add parameterization for optimized shared memory variables (#82508)

erman-gurses · web-flow · commit 87c0260f45e5 · 2024-02-27T23:28:12.000-05:00
- This PR adds parameterization for shared memory variables that are
used for optimization: `sharedMemoryLineSizeBytes` and
`defaultVectorSizeBits.`
- The default values are set to 128 for both variables since it gives
zero bank conflicts.
diff --git a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td b/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td
@@ -13,8 +13,8 @@ include "mlir/Dialect/Transform/IR/TransformAttrs.td"
 include "mlir/Dialect/Transform/IR/TransformDialect.td"
 include "mlir/Dialect/Transform/IR/TransformInterfaces.td"
 include "mlir/Dialect/Transform/IR/TransformTypes.td"
-include "mlir/Interfaces/SideEffectInterfaces.td"
 
+include "mlir/Interfaces/SideEffectInterfaces.td"
 //===----------------------------------------------------------------------===//
 // ApplyOptimizeSharedMemoryReadsAndWritesOp
 //===----------------------------------------------------------------------===//
@@ -28,7 +28,9 @@ def ApplyOptimizeSharedMemoryReadsAndWritesOp :
     reads/writes with the goal of avoiding bank conflicts.
   }];
 
-  let arguments = (ins TransformHandleTypeInterface:$target);
+  let arguments = (ins TransformHandleTypeInterface:$target,
+                    DefaultValuedOptionalAttr<I64Attr, "128">:$sharedMemoryLineSizeBytes,
+                    DefaultValuedOptionalAttr<I64Attr, "128">:$defaultVectorSizeBits);
   let results = (outs);
 
   let assemblyFormat = "$target attr-dict `:` functional-type(operands, results)";
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td
@@ -37,10 +37,17 @@ def OptimizeSharedMemory : Pass<"amdgpu-optimize-shared-memory"> {
     attempts to optimize reads/writes from a memref representing GPU shared
     memory in order to avoid bank conflicts.
   }];
-
   let dependentDialects = [
     "memref::MemRefDialect", "vector::VectorDialect"
   ];
+  let options = [
+    Option<"sharedMemoryLineSizeBytes", "shared-memory-line-size-bytes", "int64_t",
+           /*default=*/"128",
+           "Shared memory line size in bytes">,
+    Option<"defaultVectorSizeBits", "default-vector-size-bits", "int64_t",
+           /*default=*/"128",
+           "Default vector size in bits">,
+  ];
 }
 
 #endif // MLIR_DIALECT_AMDGPU_TRANSFORMS_PASSES_TD_
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h
@@ -45,11 +45,15 @@ namespace amdgpu {
 /// function that depends on the row Index. The permutation function is chosen
 /// to ensure that sequential distributed+vectorized reads/writes down a single
 /// dimension of the memref have minimal conflicts.
-LogicalResult optimizeSharedMemoryReadsAndWrites(Operation *parentOp,
-                                                 Value memrefValue);
+LogicalResult
+optimizeSharedMemoryReadsAndWrites(Operation *parentOp, Value memrefValue,
+                                   int64_t sharedMemoryLineSizeBytes,
+                                   int64_t defaultVectorSizeBits);
 
 std::optional<LogicalResult>
-optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp);
+optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp,
+                                     int64_t sharedMemoryLineSizeBytes,
+                                     int64_t defaultVectorSizeBits);
 
 } // namespace amdgpu
 } // namespace mlir
diff --git a/mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp b/mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp
@@ -27,7 +27,8 @@ DiagnosedSilenceableFailure
 ApplyOptimizeSharedMemoryReadsAndWritesOp::applyToOne(
     TransformRewriter &rewriter, FuncOp funcOp, ApplyToEachResultList &results,
     TransformState &state) {
-  optimizeSharedMemoryReadsAndWritesOp(funcOp);
+  optimizeSharedMemoryReadsAndWritesOp(funcOp, getSharedMemoryLineSizeBytes(),
+                                       getDefaultVectorSizeBits());
   return DiagnosedSilenceableFailure::success();
 }
 
diff --git a/mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp b/mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp
@@ -35,31 +35,26 @@ namespace amdgpu {
 using namespace mlir;
 using namespace mlir::amdgpu;
 
-/// The size of a shared memory line according to AMD documentation.
-/// https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/instinct-mi200-cdna2-instruction-set-architecture.pdf
-constexpr int64_t kSharedMemoryLineSizeBytes = 64;
-/// We optimize for 64bit accesses, but this can be made an argument in the
-/// future.
-constexpr int64_t kDefaultVectorSizeBits = 64;
-
 /// Uses `srcIndexValue` to permute `tgtIndexValue` via
 /// `result = xor(floordiv(srcIdxVal,permuteEveryN),
 ///               floordiv(tgtIdxVal,vectorSize)))
 ///            + tgtIdxVal % vectorSize`
 /// This is done using an optimized sequence of `arith` operations.
 static Value permuteVectorOffset(OpBuilder &b, Location loc,
                                  ArrayRef<Value> indices, MemRefType memrefTy,
-                                 int64_t srcDim, int64_t tgtDim) {
+                                 int64_t srcDim, int64_t tgtDim,
+                                 int64_t sharedMemoryLineSizeBytes,
+                                 int64_t defaultVectorSizeBits) {
   // Adjust the src index to change how often the permutation changes
   // if necessary.
   Value src = indices[srcDim];
 
   // We only want to permute every N iterations of the target dim where N is
   // ceil(sharedMemoryLineSizeBytes / dimSizeBytes(tgtDim)).
   const int64_t permuteEveryN = std::max<int64_t>(
-      1, kSharedMemoryLineSizeBytes / ((memrefTy.getDimSize(tgtDim) *
-                                        memrefTy.getElementTypeBitWidth()) /
-                                       8));
+      1, sharedMemoryLineSizeBytes / ((memrefTy.getDimSize(tgtDim) *
+                                       memrefTy.getElementTypeBitWidth()) /
+                                      8));
 
   // clang-format off
   // Index bit representation (b0 = least significant bit) for dim(1)
@@ -71,7 +66,7 @@ static Value permuteVectorOffset(OpBuilder &b, Location loc,
   // bits[N:M] = vector index
   // clang-format on
   int64_t n =
-      llvm::Log2_64(kDefaultVectorSizeBits / memrefTy.getElementTypeBitWidth());
+      llvm::Log2_64(defaultVectorSizeBits / memrefTy.getElementTypeBitWidth());
   int64_t m = llvm::Log2_64(memrefTy.getDimSize(tgtDim));
 
   // Capture bits[0:(M-N)] of src by first creating a (M-N) mask.
@@ -105,9 +100,11 @@ static Value permuteVectorOffset(OpBuilder &b, Location loc,
 static void transformIndices(OpBuilder &builder, Location loc,
                              SmallVector<Value, 4> &indices,
                              MemRefType memrefTy, int64_t srcDim,
-                             int64_t tgtDim) {
+                             int64_t tgtDim, int64_t sharedMemoryLineSizeBytes,
+                             int64_t defaultVectorSizeBits) {
   indices[tgtDim] =
-      permuteVectorOffset(builder, loc, indices, memrefTy, srcDim, tgtDim);
+      permuteVectorOffset(builder, loc, indices, memrefTy, srcDim, tgtDim,
+                          sharedMemoryLineSizeBytes, defaultVectorSizeBits);
 }
 
 // Return all operations within `parentOp` that read from or write to
@@ -149,8 +146,9 @@ getShmReadAndWriteOps(Operation *parentOp, Value shmMemRef,
   return success();
 }
 
-LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(Operation *parentOp,
-                                                         Value memrefValue) {
+LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(
+    Operation *parentOp, Value memrefValue, int64_t sharedMemoryLineSizeBytes,
+    int64_t defaultVectorSizeBits) {
   auto memRefType = dyn_cast<MemRefType>(memrefValue.getType());
   if (!memRefType ||
       !amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(memRefType))
@@ -167,10 +165,10 @@ LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(Operation *parentOp,
   // If dim[rank-1] is small enough to fit 8 rows in a 128B line.
   const int64_t rowSize = memRefType.getDimSize(memRefType.getRank() - 1);
   const int64_t rowsPerLine =
-      (8 * kSharedMemoryLineSizeBytes / memRefType.getElementTypeBitWidth()) /
+      (8 * sharedMemoryLineSizeBytes / memRefType.getElementTypeBitWidth()) /
       rowSize;
   const int64_t threadGroupSize =
-      1LL << (7 - llvm::Log2_64(kDefaultVectorSizeBits / 8));
+      1LL << (7 - llvm::Log2_64(defaultVectorSizeBits / 8));
   if (rowsPerLine >= threadGroupSize)
     return failure();
 
@@ -198,7 +196,8 @@ LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(Operation *parentOp,
     auto indices = amdgpu::getIndices(shmWriteOp);
     SmallVector<Value, 4> transformedIndices(indices->begin(), indices->end());
     transformIndices(builder, shmWriteOp->getLoc(), transformedIndices,
-                     memRefType, srcDim, tgtDim);
+                     memRefType, srcDim, tgtDim, sharedMemoryLineSizeBytes,
+                     defaultVectorSizeBits);
     amdgpu::setIndices(shmWriteOp, transformedIndices);
   }
 
@@ -210,24 +209,28 @@ LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(Operation *parentOp,
     auto indices = amdgpu::getIndices(shmReadOp);
     SmallVector<Value, 4> transformedIndices(indices->begin(), indices->end());
     transformIndices(builder, shmReadOp->getLoc(), transformedIndices,
-                     memRefType, srcDim, tgtDim);
+                     memRefType, srcDim, tgtDim, sharedMemoryLineSizeBytes,
+                     defaultVectorSizeBits);
     amdgpu::setIndices(shmReadOp, transformedIndices);
   }
 
   return success();
 }
 
 std::optional<LogicalResult>
-amdgpu::optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp) {
+amdgpu::optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp,
+                                             int64_t sharedMemoryLineSizeBytes,
+                                             int64_t defaultVectorSizeBits) {
   SmallVector<memref::AllocOp> shmAllocOps;
   funcOp.walk([&](memref::AllocOp allocOp) {
     if (!amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(allocOp.getType()))
       return;
     shmAllocOps.push_back(allocOp);
   });
   for (auto allocOp : shmAllocOps) {
-    if (failed(amdgpu::optimizeSharedMemoryReadsAndWrites(funcOp,
-                                                          allocOp.getMemref())))
+    if (failed(amdgpu::optimizeSharedMemoryReadsAndWrites(
+            funcOp, allocOp.getMemref(), sharedMemoryLineSizeBytes,
+            defaultVectorSizeBits)))
       return failure();
   }
   return success();
@@ -237,7 +240,8 @@ struct OptimizeSharedMemoryPass
     : public amdgpu::impl::OptimizeSharedMemoryBase<OptimizeSharedMemoryPass> {
 public:
   OptimizeSharedMemoryPass() = default;
-
+  OptimizeSharedMemoryPass(const OptimizeSharedMemoryOptions &options)
+      : OptimizeSharedMemoryBase(options) {}
   void runOnOperation() override {
     Operation *op = getOperation();
     SmallVector<memref::AllocOp> shmAllocOps;
@@ -248,8 +252,9 @@ struct OptimizeSharedMemoryPass
       shmAllocOps.push_back(allocOp);
     });
     for (auto allocOp : shmAllocOps) {
-      if (failed(optimizeSharedMemoryReadsAndWrites(getOperation(),
-                                                    allocOp.getMemref())))
+      if (failed(optimizeSharedMemoryReadsAndWrites(op, allocOp.getMemref(),
+                                                    sharedMemoryLineSizeBytes,
+                                                    defaultVectorSizeBits)))
         return;
     }
   }
diff --git a/mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir b/mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir
@@ -1,56 +1,50 @@
-// RUN: mlir-opt  %s --pass-pipeline='builtin.module(func.func(amdgpu-optimize-shared-memory))' | FileCheck %s
+// RUN: mlir-opt %s --pass-pipeline='builtin.module(func.func(amdgpu-optimize-shared-memory))' | FileCheck %s
   
   // CHECK: @optimize_shmem([[arg0:%.+]]: memref<{{.*}}>, [[readRow:%.+]]: index, [[readCol:%.+]]: index, [[writeRow:%.+]]: index, [[writeCol:%.+]]: index, [[fragRow:%.+]]: index, [[fragCol:%.+]]: index, [[fragColPerm:%.+]]: index, [[stRow:%.+]]: index, [[stCol:%.+]]: index)
-  func.func @optimize_shmem(%arg0: memref<4096x4096xf16>, 
+  func.func @optimize_shmem(%arg0: memref<4096x4096xf16>,
                     %readRow: index, %readCol: index,
                     %writeRow: index, %writeCol: index,
-                    %fragRow: index, %fragCol: index, 
+                    %fragRow: index, %fragCol: index,
                     %fragColPerm: index,
                     %stRow: index, %stCol: index) {
-    // CHECK:    %[[cst:.+]] = arith.constant 0.000000e+00 : f16                  
+    // CHECK:    %[[cst:.+]] = arith.constant 0.000000e+00 : f16
     %cst = arith.constant 0.000000e+00 : f16
 
     // CHECK: [[shmA:%.+]] = memref.alloc
     // CHECK: [[shmB:%.+]] = memref.alloc
     %shmA = memref.alloc() {alignment = 64 : i64} : memref<128x32xf16, 3>
     %shmB = memref.alloc() {alignment = 64 : i64} : memref<256x32xf16, 3>
 
-    // CHECK: %[[D0:.+]] = vector.transfer_read [[arg0:%.+]][[[readRow:%.+]], [[readCol:%.+]]], [[cst:.+]] {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
     %0 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c7:%.+]] = arith.constant 7 : index                  
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c7]]       
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index                 
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]     
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]  
-    // CHECK: vector.transfer_write %[[D0:.+]], [[shmB]][[[writeRow:%.+]], [[writeCol:%.+]]] {in_bounds = [true, true]} : vector<1x8xf16>, memref<256x32xf16, 3>
+    // CHECK: [[c6:%.+]] = arith.constant 6 : index
+    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
+    // CHECK: [[c2:%.+]] = arith.constant 2 : index
+    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
+    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
     vector.transfer_write %0, %shmB[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<256x32xf16, 3>
     gpu.barrier
     gpu.barrier
-    // CHECK: [[c7:%.+]] = arith.constant 7 : index                     
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c7]]     
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index                 
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]       
+    // CHECK: [[c6:%.+]] = arith.constant 6 : index
+    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
+    // CHECK: [[c2:%.+]] = arith.constant 2 : index
+    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
     // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]] 
-    // CHECK:  vector.load [[shmB:%.+]][[[fragRow:%.+]], [[fragColPerm]]] : memref<256x32xf16, 3>, vector<8xf16>
     %1 = vector.load %shmB[%fragRow, %fragColPerm] : memref<256x32xf16, 3>, vector<8xf16>
 
-    // CHECK: %[[D2:.+]] = vector.transfer_read [[arg0:%.+]][[[readRow:%.+]], [[readCol:%.+]]], [[cst:.+]] {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
     %2 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c7:%.+]] = arith.constant 7 : index                  
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c7]]       
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index                 
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]     
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]  
-    // CHECK: vector.transfer_write %[[D2:.+]], [[shmA:%.+]][[[writeRow:%.+]], [[writeCol:%.+]]] {in_bounds = [true, true]} : vector<1x8xf16>, memref<128x32xf16, 3>
+    // CHECK: [[c6:%.+]] = arith.constant 6 : index
+    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
+    // CHECK: [[c2:%.+]] = arith.constant 2 : index
+    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
+    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
     vector.transfer_write %2, %shmA[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<128x32xf16, 3>
     gpu.barrier
     gpu.barrier
-    // CHECK: [[c7:%.+]] = arith.constant 7 : index                     
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c7]]          
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index                     
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]] 
+    // CHECK: [[c6:%.+]] = arith.constant 6 : index
+    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
+    // CHECK: [[c2:%.+]] = arith.constant 2 : index
+    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
     // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]]
-    // CHECK:  vector.load [[shmA:%.+]][[[fragRow:%.+]], [[fragColPerm]]] : memref<128x32xf16, 3>, vector<8xf16>
     %3 = vector.load %shmA[%fragRow, %fragColPerm] : memref<128x32xf16, 3>, vector<8xf16>
     return
   }
diff --git a/mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir b/mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir

Original file line number	Diff line number	Diff line change
`@@ -27,7 +27,8 @@ DiagnosedSilenceableFailure`
`27`	`27`	`ApplyOptimizeSharedMemoryReadsAndWritesOp::applyToOne(`
`28`	`28`	`TransformRewriter &rewriter, FuncOp funcOp, ApplyToEachResultList &results,`
`29`	`29`	`TransformState &state) {`
`30`		`- optimizeSharedMemoryReadsAndWritesOp(funcOp);`
	`30`	`+ optimizeSharedMemoryReadsAndWritesOp(funcOp, getSharedMemoryLineSizeBytes(),`
	`31`	`+ getDefaultVectorSizeBits());`
`31`	`32`	`return DiagnosedSilenceableFailure::success();`
`32`	`33`	`}`
`33`	`34`