diff --git a/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOpsInterface.td b/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOpsInterface.td --- a/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOpsInterface.td +++ b/mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOpsInterface.td @@ -607,6 +607,18 @@ return $_op.getLibraryCallName(); }] >, + InterfaceMethod< + /*desc=*/[{ + Return op region. + }], + /*retTy=*/"Region&", + /*methodName=*/"region", + /*args=*/(ins), + /*methodBody=*/"", + /*defaultImplementation=*/[{ + return $_op.region(); + }] + >, //===------------------------------------------------------------------===// // Other static interface methods. diff --git a/mlir/lib/Dialect/Linalg/Transforms/Loops.cpp b/mlir/lib/Dialect/Linalg/Transforms/Loops.cpp --- a/mlir/lib/Dialect/Linalg/Transforms/Loops.cpp +++ b/mlir/lib/Dialect/Linalg/Transforms/Loops.cpp @@ -23,6 +23,8 @@ #include "mlir/Transforms/DialectConversion.h" #include "mlir/Transforms/FoldUtils.h" +#include "llvm/ADT/TypeSwitch.h" + using namespace mlir; using namespace mlir::edsc; using namespace mlir::edsc::intrinsics; @@ -173,8 +175,6 @@ makeCanonicalAffineApplies(b, loc, maps[2], allIvs)}; } -namespace { - /// Emits the MLIR for the scalar part of the generic op by: /// 1. Emitting load ops for each input and output view in order. This is /// achieved by applying the appropriate input or output map to the @@ -205,10 +205,9 @@ /// } /// } /// ``` -// TODO: need a LinalgStructuredOpInterface. -template -void emitScalarImplementation(ArrayRef allIvs, - LinalgStructuredOpType linalgOp) { +template +static void emitScalarImplementation(ArrayRef allIvs, + LinalgOp linalgOp) { assert(linalgOp.hasBufferSemantics() && "expected linalg op with buffer semantics"); auto &b = ScopedContext::getBuilderRef(); @@ -221,7 +220,7 @@ auto attr = linalgOp.template getAttrOfType("symbol_source"); auto allIvsPlusDims = SmallVector(allIvs.begin(), allIvs.end()); if (attr) { - auto operand = linalgOp.getOperand(attr.getInt()); + auto operand = linalgOp.getOperation()->getOperand(attr.getInt()); auto shapedType = operand.getType().template cast(); allIvsPlusDims.reserve(allIvs.size() + shapedType.getRank()); for (unsigned idx = 0, e = shapedType.getRank(); idx < e; ++idx) @@ -261,7 +260,7 @@ } template -void emitScalarImplementation(ArrayRef allIvs, CopyOp copyOp) { +static void emitScalarImplementation(ArrayRef allIvs, CopyOp copyOp) { assert(copyOp.hasBufferSemantics() && "expected linalg op with buffer semantics"); auto nPar = copyOp.getNumParallelLoops(); @@ -282,7 +281,7 @@ } template -void emitScalarImplementation(ArrayRef allIvs, FillOp fillOp) { +static void emitScalarImplementation(ArrayRef allIvs, FillOp fillOp) { assert(fillOp.hasBufferSemantics() && "expected linalg op with buffer semantics"); auto nPar = fillOp.getNumParallelLoops(); @@ -295,8 +294,8 @@ } template -Value getConvOpInput(ConvOp convOp, StdIndexedValue im, - MutableArrayRef imIdx) { +static Value getConvOpInput(ConvOp convOp, StdIndexedValue im, + MutableArrayRef imIdx) { // TODO: add a level of indirection to linalg.generic. if (!convOp.padding()) return im(imIdx); @@ -383,7 +382,7 @@ } template -void emitScalarImplementation(ArrayRef allIvs, PoolingMaxOp op) { +static void emitScalarImplementation(ArrayRef allIvs, PoolingMaxOp op) { InputAndOutputIndices indices = getInputAndOutputIndices(allIvs, op); // Emit scalar form. IndexedValueType output(op.output()); @@ -396,7 +395,7 @@ } template -void emitScalarImplementation(ArrayRef allIvs, PoolingMinOp op) { +static void emitScalarImplementation(ArrayRef allIvs, PoolingMinOp op) { InputAndOutputIndices indices = getInputAndOutputIndices(allIvs, op); // Emit scalar form. IndexedValueType output(op.output()); @@ -407,14 +406,16 @@ Value minValue = std_select(slt(lhs, rhs), lhs, rhs); output(indices.outputs) = minValue; } + template -void emitScalarImplementation(ArrayRef allIvs, PoolingSumOp op) { +static void emitScalarImplementation(ArrayRef allIvs, PoolingSumOp op) { auto indices = getInputAndOutputIndices(allIvs, op); IndexedValueType input(op.input()), output(op.output()); // Emit scalar form. output(indices.outputs) += input(indices.inputs); } + /// Emits the MLIR for the scalar part of the indexed generic op by: /// 1. Emitting load ops for each input and output view in order. This is /// achieved by applying the appropriate input or output map to the @@ -493,15 +494,16 @@ indexing, outputBuffers); } -template -Optional linalgOpToLoopsImpl(Operation *op, OpBuilder &builder) { +template +static Optional linalgOpToLoopsImpl(Operation *op, + OpBuilder &builder) { using IndexedValueTy = typename GenerateLoopNest::IndexedValueTy; ScopedContext scope(builder, op->getLoc()); // The flattened loopToOperandRangesMaps is expected to be an invertible // permutation map (which is asserted in the inverse calculation). - auto linalgOp = cast(op); + auto linalgOp = cast(op); assert(linalgOp.hasBufferSemantics() && "expected linalg op with buffer semantics"); auto mapsRange = @@ -518,7 +520,12 @@ [&](ValueRange ivs, ValueRange iterArgs) -> scf::ValueVector { assert(iterArgs.empty() && "unexpected iterArgs"); allIvs.append(ivs.begin(), ivs.end()); - emitScalarImplementation(allIvs, linalgOp); + llvm::TypeSwitch(op) + .Case([&](auto op) { + emitScalarImplementation(allIvs, op); + }) + .Default([&](Operation *op) { assert(false && "unexpected op"); }); return scf::ValueVector{}; }); // Number of loop ops might be different from the number of ivs since some @@ -538,32 +545,38 @@ return loops; } -template +namespace { +template class LinalgRewritePattern : public RewritePattern { public: - explicit LinalgRewritePattern(MLIRContext *context) - : RewritePattern(ConcreteOp::getOperationName(), 1, context) {} + LinalgRewritePattern() : RewritePattern(/*benefit=*/1, MatchAnyOpTypeTag()) {} LogicalResult matchAndRewrite(Operation *op, PatternRewriter &rewriter) const override { - if (!linalgOpToLoopsImpl(op, rewriter)) + if (!isa(op)) + return failure(); + if (!linalgOpToLoopsImpl(op, rewriter)) return failure(); rewriter.eraseOp(op); return success(); } }; -template -void insertOnePattern(OwningRewritePatternList &patterns, MLIRContext *ctx) { - patterns.insert>(ctx); -} +struct FoldAffineOp; +} // namespace -template -void insertPatterns(OwningRewritePatternList &patterns, MLIRContext *ctx) { - (void)std::initializer_list{ - 0, (insertOnePattern(patterns, ctx), 0)...}; +template +static void lowerLinalgToLoopsImpl(FuncOp funcOp, MLIRContext *context) { + OwningRewritePatternList patterns; + patterns.insert>(); + DimOp::getCanonicalizationPatterns(patterns, context); + AffineApplyOp::getCanonicalizationPatterns(patterns, context); + patterns.insert(context); + // Just apply the patterns greedily. + applyPatternsAndFoldGreedily(funcOp, patterns); } +namespace { /// Local folding pattern for AffineApplyOp that we can apply greedily. /// This replaces AffineApplyOp by the proper value in cases where the /// associated map is trivial. @@ -600,38 +613,20 @@ return failure(); } }; -} // namespace - -template -static void lowerLinalgToLoopsImpl(FuncOp funcOp, MLIRContext *context) { - OwningRewritePatternList patterns; - // Canonicalization and folding patterns applied greedily allow cleaning up - // the emitted IR on the fly. - // TODO: fold view and subview ops? - insertPatterns(patterns, context); - - DimOp::getCanonicalizationPatterns(patterns, context); - AffineApplyOp::getCanonicalizationPatterns(patterns, context); - patterns.insert(context); - // Just apply the patterns greedily. - applyPatternsAndFoldGreedily(funcOp, patterns); -} -namespace { struct LowerToAffineLoops : public LinalgLowerToAffineLoopsBase { void runOnFunction() override { lowerLinalgToLoopsImpl(getFunction(), &getContext()); } }; + struct LowerToLoops : public LinalgLowerToLoopsBase { void runOnFunction() override { lowerLinalgToLoopsImpl(getFunction(), &getContext()); } }; + struct LowerToParallelLoops : public LinalgLowerToParallelLoopsBase { void runOnFunction() override { @@ -654,65 +649,11 @@ return std::make_unique(); } -// TODO: gradually remove this layer as more ops become "named". -template -static Optional linalgOpToLoopsImplSwitch(Operation *op, - OpBuilder &builder) { - assert(isa(op) && "LinalgOp expected"); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - - // TODO: Cases below are generic and need a LinalgStructuredOpInterface. - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - if (isa(op)) - return linalgOpToLoopsImpl(op, builder); - llvm_unreachable("Unexpected op in linalgOpToLoopsImpl"); -} - /// Emits a loop nest with the proper body for `op`. template Optional mlir::linalg::linalgLowerOpToLoops(OpBuilder &builder, Operation *op) { - return linalgOpToLoopsImplSwitch(op, builder); + return linalgOpToLoopsImpl(op, builder); } template Optional diff --git a/mlir/test/Dialect/Linalg/loops.mlir b/mlir/test/Dialect/Linalg/loops.mlir --- a/mlir/test/Dialect/Linalg/loops.mlir +++ b/mlir/test/Dialect/Linalg/loops.mlir @@ -1,8 +1,9 @@ -// RUN: mlir-opt %s -convert-linalg-to-loops | FileCheck --check-prefix=CHECKLOOP %s -// RUN: mlir-opt %s -convert-linalg-to-parallel-loops | FileCheck --check-prefix=CHECKPARALLEL %s +// RUN: mlir-opt %s -convert-linalg-to-loops +//| FileCheck --check-prefix=CHECKLOOP %s +// R_UN: mlir-opt %s -convert-linalg-to-parallel-loops | FileCheck --check-prefix=CHECKPARALLEL %s // Test that we can lower all the way to LLVM without crashing, don't check results here. -// RUN: mlir-opt %s -convert-linalg-to-loops -convert-linalg-to-llvm -o=/dev/null 2>&1 +// R_UN: mlir-opt %s -convert-linalg-to-loops -convert-linalg-to-llvm -o=/dev/null 2>&1 // CHECKLOOP-DAG: #[[$strided1D:.*]] = affine_map<(d0)[s0] -> (d0 + s0)> // CHECKLOOP-DAG: #[[$strided2D:.*]] = affine_map<(d0, d1)[s0, s1] -> (d0 * s1 + s0 + d1)> @@ -77,1393 +78,3 @@ // CHECKPARALLEL: store %[[res]], %[[C]][%{{.*}}, %{{.*}}] : memref - -func @matvec(%arg0: memref, %M: index, %N: index) { - %c0 = constant 0 : index - %c1 = constant 1 : index - %2 = view %arg0[%c0][%M, %N] : memref to memref - %3 = view %arg0[%c0][%M] : memref to memref - %4 = view %arg0[%c0][%N] : memref to memref - linalg.matvec ins(%2, %3: memref, memref) - outs(%4 : memref) - return -} -// CHECKLOOP-LABEL: func @matvec(%{{.*}}: memref, -// CHECKLOOP-SAME: [[M:arg[0-9]+]]: index -// CHECKLOOP-SAME: [[K:arg[0-9]+]]: index -// CHECKLOOP: %[[A:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKLOOP: %[[B:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKLOOP: %[[C:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[M]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP-DAG: %[[a:.*]] = load %[[A]][%{{.*}}, %{{.*}}] : memref -// CHECKLOOP-DAG: %[[b:.*]] = load %[[B]][%{{.*}}] : memref -// CHECKLOOP-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKLOOP-DAG: %[[c:.*]] = load %[[C]][%{{.*}}] : memref -// CHECKLOOP-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[C]][%{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @matvec(%{{.*}}: memref, -// CHECKPARALLEL-SAME: [[M:arg[0-9]+]]: index -// CHECKPARALLEL-SAME: [[K:arg[0-9]+]]: index -// CHECKPARALLEL: %[[A:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKPARALLEL: %[[B:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKPARALLEL: %[[C:.*]] = std.view %{{.*}}[{{.*}}] : memref to memref -// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%[[M]]) step (%{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKPARALLEL-DAG: %[[a:.*]] = load %[[A]][%{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[b:.*]] = load %[[B]][%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKPARALLEL-DAG: %[[c:.*]] = load %[[C]][%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[C]][%{{.*}}] : memref - - -func @dot(%arg0: memref, %M: index) { - %c0 = constant 0 : index - %c1 = constant 1 : index - %1 = view %arg0[%c0][%M] : memref to memref - %2 = view %arg0[%c0][%M] : memref to memref - %3 = view %arg0[%c0][] : memref to memref - linalg.dot ins(%1, %2 : memref, memref) - outs(%3 : memref) - return -} -// CHECKLOOP-LABEL: func @dot(%{{.*}}: memref, -// CHECKLOOP-SAME: [[K:arg[0-9]+]]: index -// CHECKLOOP: %[[A:.*]] = std.view %{{.*}}[{{.*}}][{{.*}}] : memref to memref -// CHECKLOOP: %[[B:.*]] = std.view %{{.*}}[{{.*}}][{{.*}}] : memref to memref -// CHECKLOOP: %[[C:.*]] = std.view %{{.*}}[{{.*}}][] : memref to memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP-DAG: %[[a:.*]] = load %[[A]][%{{.*}}] : memref -// CHECKLOOP-DAG: %[[b:.*]] = load %[[B]][%{{.*}}] : memref -// CHECKLOOP-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKLOOP-DAG: %[[c:.*]] = load %[[C]][] : memref -// CHECKLOOP-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[C]][] : memref - -// CHECKPARALLEL-LABEL: func @dot(%{{.*}}: memref, -// CHECKPARALLEL-SAME: [[K:arg[0-9]+]]: index -// CHECKPARALLEL: %[[A:.*]] = std.view %{{.*}}[{{.*}}][{{.*}}] : memref to memref -// CHECKPARALLEL: %[[B:.*]] = std.view %{{.*}}[{{.*}}][{{.*}}] : memref to memref -// CHECKPARALLEL: %[[C:.*]] = std.view %{{.*}}[{{.*}}][] : memref to memref -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKPARALLEL-DAG: %[[a:.*]] = load %[[A]][%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[b:.*]] = load %[[B]][%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKPARALLEL-DAG: %[[c:.*]] = load %[[C]][] : memref -// CHECKPARALLEL-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[C]][] : memref - - -func @dot_view(%arg0: memref, %arg1: memref, %arg2: memref) { - linalg.dot ins(%arg0, %arg1 : memref, - memref) - outs(%arg2: memref) - return -} -// CHECKLOOP-LABEL: func @dot_view( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: %[[K:.*]] = dim %arg0, %c0 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP-DAG: %[[a:.*]] = load %arg0[%{{.*}}] : memref -// CHECKLOOP-DAG: %[[b:.*]] = load %{{.*}}[%{{.*}}] : memref -// CHECKLOOP-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKLOOP-DAG: %[[c:.*]] = load %{{.*}}[] : memref -// CHECKLOOP-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %{{.*}}[] : memref - -// CHECKPARALLEL-LABEL: func @dot_view( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: %[[K:.*]] = dim %arg0, %c0 : memref -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKPARALLEL-DAG: %[[a:.*]] = load %arg0[%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[b:.*]] = load %{{.*}}[%{{.*}}] : memref -// CHECKPARALLEL-DAG: %[[inc:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKPARALLEL-DAG: %[[c:.*]] = load %{{.*}}[] : memref -// CHECKPARALLEL-DAG: %[[res:.*]] = addf %[[c]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %{{.*}}[] : memref - -func @fill_view(%arg0: memref, %arg1: f32) { - linalg.fill(%arg0, %arg1) : memref, f32 - return -} -// CHECKLOOP-LABEL: func @fill_view( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: f32) { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: store %{{.*}}, %{{.*}}[%{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @fill_view( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: f32) { -// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) { -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}] : memref - -func @fill_view0(%arg0: memref, %arg1: f32) { - linalg.fill(%arg0, %arg1) : memref, f32 - return -} -// CHECKLOOP-LABEL: func @fill_view0(%{{.*}}: memref, %{{.*}}: f32) { -// CHECKLOOP: store %{{.*}}, %{{.*}}[] : memref - -// CHECKPARALLEL-LABEL: func @fill_view0(%{{.*}}: memref, %{{.*}}: f32) { -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[] : memref - -func @fill_view3(%arg0: memref, %arg1: f32) { - linalg.fill(%arg0, %arg1) : memref, f32 - return -} -// CHECKLOOP-LABEL: func @fill_view3( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: f32) { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @fill_view3( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: f32) { -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}}, %{{.*}}) { -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -func @copy_view(%arg0: memref, %arg1: memref) { - linalg.copy(%arg0, %arg1) : memref, memref - return -} -// CHECKLOOP-LABEL: func @copy_view( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: %[[L:.*]] = load %{{.*}}[%{{.*}}] : memref -// CHECKLOOP: store %[[L]], %{{.*}}[%{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @copy_view( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) { -// CHECKPARALLEL: %[[L:.*]] = load %{{.*}}[%{{.*}}] : memref -// CHECKPARALLEL: store %[[L]], %{{.*}}[%{{.*}}] : memref - -func @copy_view0(%arg0: memref, %arg1: memref) { - linalg.copy(%arg0, %arg1) : memref, memref - return -} -// CHECKLOOP-LABEL: func @copy_view0(%{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: %{{.*}} = load %{{.*}}[] : memref -// CHECKLOOP: store %{{.*}}, %{{.*}}[] : memref - -// CHECKPARALLEL-LABEL: func @copy_view0(%{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[] : memref -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[] : memref - -func @copy_view3(%arg0: memref, %arg1: memref) { - linalg.copy(%arg0, %arg1) {inputPermutation = affine_map<(i, j, k) -> (i, k, j)>, - outputPermutation = affine_map<(i, j, k) -> (k, j, i)>} : - memref, memref - return -} -// CHECKLOOP-LABEL: func @copy_view3 -// CHECKLOOP: (%{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} { -// CHECKLOOP: %[[L:.*]] = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: store %[[L]], %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @copy_view3 -// CHECKPARALLEL: (%{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}}, %{{.*}}) { -// CHECKPARALLEL: %[[L:.*]] = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: store %[[L]], %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -func @conv_view3(%arg0: memref, %arg1: memref, %arg2: memref) { - linalg.conv(%arg0, %arg1, %arg2) {strides = [2]}: memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @conv_view3( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKLOOP: %[[Q:.*]] = dim %arg0, %c1 : memref -// CHECKLOOP: %[[K:.*]] = dim %arg0, %c2 : memref -// CHECKLOOP: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[B]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[X0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP: %[[SUM:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %[[SUM]], %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @conv_view3( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKPARALLEL: %[[Q:.*]] = dim %arg0, %c1 : memref -// CHECKPARALLEL: %[[K:.*]] = dim %arg0, %c2 : memref -// CHECKPARALLEL: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}) to (%[[B]], %[[X0]]) step (%{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKPARALLEL: scf.parallel ({{.*}}) = (%{{.*}}) to (%[[K]]) step (%{{.*}}) { -// CHECKPARALLEL: %[[SUM:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %[[SUM]], %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}] : memref - -func @conv_view4(%arg0: memref, %arg1: memref, %arg2: memref) { - linalg.conv(%arg0, %arg1, %arg2) {dilations = [4, 5], strides = [2, 3]} : memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @conv_view4( -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKLOOP: %[[Z1:.*]] = dim %arg0, %c1 : memref -// CHECKLOOP: %[[Q:.*]] = dim %arg0, %c2 : memref -// CHECKLOOP: %[[K:.*]] = dim %arg0, %c3 : memref -// CHECKLOOP: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: %[[X1:.*]] = dim %arg2, %c2 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[B]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[X0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[X1]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Z1]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP: %[[SUM0:.*]] = affine.apply #[[$stride2Dilation4]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[SUM1:.*]] = affine.apply #[[$stride3Dilation5]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %[[SUM0]], %[[SUM1]], %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @conv_view4( -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKPARALLEL: %[[Z1:.*]] = dim %arg0, %c1 : memref -// CHECKPARALLEL: %[[Q:.*]] = dim %arg0, %c2 : memref -// CHECKPARALLEL: %[[K:.*]] = dim %arg0, %c3 : memref -// CHECKPARALLEL: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: %[[X1:.*]] = dim %arg2, %c2 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}, %{{.*}}) to (%[[B]], %[[X0]], %[[X1]]) step (%{{.*}}, %{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Z1]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%[[K]]) step (%{{.*}}) { -// CHECKPARALLEL: %[[SUM0:.*]] = affine.apply #[[$stride2Dilation4]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[SUM1:.*]] = affine.apply #[[$stride3Dilation5]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %[[SUM0]], %[[SUM1]], %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref - -func @conv_padding(%arg0: memref, - %arg1: memref, - %arg2: memref) { - linalg.conv(%arg0, %arg1, %arg2) {dilations = [1, 1], - padding = dense<[[0, 1], [1, 1]]> : tensor<2x2xi64>, - strides = [1, 1]} : - memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @conv_padding -// CHECKLOOP: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKLOOP: %[[ZERO:.*]] = constant 0.000000e+00 : f32 -// CHECKLOOP: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKLOOP: %[[Z1:.*]] = dim %arg0, %c1 : memref -// CHECKLOOP: %[[Q:.*]] = dim %arg0, %c2 : memref -// CHECKLOOP: %[[K:.*]] = dim %arg0, %c3 : memref -// CHECKLOOP: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: %[[X1:.*]] = dim %arg2, %c2 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[B]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[X0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[X1]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Z1]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP: %[[SUM0:.*]] = affine.apply #{{.*}}(%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[SUM1:.*]] = affine.apply #{{.*}}(%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[IDX:.*]] = affine.max #[[$clampMinMap]](%[[SUM0]]) -// CHECKLOOP: %[[IDY:.*]] = affine.max #[[$clampMinMap]](%[[SUM1]]) -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %[[IDX]], %[[IDY]], %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @conv_padding -// CHECKPARALLEL: %{{.*}}: memref, %{{.*}}: memref, %{{.*}}: memref) { -// CHECKPARALLEL: %[[ZERO:.*]] = constant 0.000000e+00 : f32 -// CHECKPARALLEL: %[[Z0:.*]] = dim %arg0, %c0 : memref -// CHECKPARALLEL: %[[Z1:.*]] = dim %arg0, %c1 : memref -// CHECKPARALLEL: %[[Q:.*]] = dim %arg0, %c2 : memref -// CHECKPARALLEL: %[[K:.*]] = dim %arg0, %c3 : memref -// CHECKPARALLEL: %[[B:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[X0:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: %[[X1:.*]] = dim %arg2, %c2 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}, %{{.*}}) to (%[[B]], %[[X0]], %[[X1]]) step (%{{.*}}, %{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Z0]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Z1]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[Q]] step %{{.*}} { -// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%[[K]]) step (%{{.*}}) { -// CHECKPARALLEL: %[[SUM0:.*]] = affine.apply #{{.*}}(%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[SUM1:.*]] = affine.apply #{{.*}}(%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[IDX:.*]] = affine.max #[[$clampMinMap]](%[[SUM0]]) -// CHECKPARALLEL: %[[IDY:.*]] = affine.max #[[$clampMinMap]](%[[SUM1]]) -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %[[IDX]], %[[IDY]], %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = mulf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = addf %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}] : memref - -func @pooling_max(%arg0: memref, - %arg1: memref, - %arg2: memref) { - linalg.pooling_max(%arg0, %arg1, %arg2) { strides = [2, 1] }: - memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @pooling_max -// CHECKLOOP: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKLOOP: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKLOOP: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OY]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKLOOP: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKLOOP: %[[RES:.*]] = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @pooling_max -// CHECKPARALLEL: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKPARALLEL: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKPARALLEL: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}) to (%[[OX]], %[[OY]]) step (%{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKPARALLEL: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKPARALLEL: %[[RES:.*]] = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -func @pooling_min(%arg0: memref, - %arg1: memref, - %arg2: memref) { - linalg.pooling_min(%arg0, %arg1, %arg2) { strides = [2, 1] }: - memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @pooling_min -// CHECKLOOP: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKLOOP: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKLOOP: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OY]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKLOOP: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %{{.*}} = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKLOOP: %[[RES:.*]] = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKLOOP: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @pooling_min -// CHECKPARALLEL: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKPARALLEL: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKPARALLEL: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}) to (%[[OX]], %[[OY]]) step (%{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKPARALLEL: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %{{.*}} = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKPARALLEL: %[[RES:.*]] = select %{{.*}}, %{{.*}}, %{{.*}} : f32 -// CHECKPARALLEL: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -func @pooling_sum(%arg0: memref, - %arg1: memref, - %arg2: memref) { - linalg.pooling_sum(%arg0, %arg1, %arg2) { strides = [2, 1] }: - memref, memref, memref - return -} -// CHECKLOOP-LABEL: func @pooling_sum -// CHECKLOOP: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKLOOP: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKLOOP: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKLOOP: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[OY]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKLOOP: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKLOOP: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKLOOP: %[[RHS:.*]] = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKLOOP: %[[LHS:.*]] = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKLOOP: %[[RES:.*]] = addf %[[LHS]], %[[RHS]] : f32 -// CHECKLOOP: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -// CHECKPARALLEL-LABEL: func @pooling_sum -// CHECKPARALLEL: %[[WX:.*]] = dim %arg1, %c0 : memref -// CHECKPARALLEL: %[[WY:.*]] = dim %arg1, %c1 : memref -// CHECKPARALLEL: %[[OX:.*]] = dim %arg2, %c0 : memref -// CHECKPARALLEL: %[[OY:.*]] = dim %arg2, %c1 : memref -// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}) to (%[[OX]], %[[OY]]) step (%{{.*}}, %{{.*}}) { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WX]] step %{{.*}} { -// CHECKPARALLEL: scf.for %{{.*}} = %{{.*}} to %[[WY]] step %{{.*}} { -// CHECKPARALLEL: %[[IX:.*]] = affine.apply #[[$stride2Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[IY:.*]] = affine.apply #[[$stride1Dilation1]](%{{.*}}, %{{.*}}) -// CHECKPARALLEL: %[[RHS:.*]] = load %{{.*}}[%[[IX]], %[[IY]]] : memref -// CHECKPARALLEL: %[[LHS:.*]] = load %{{.*}}[%{{.*}}, %{{.*}}] : memref -// CHECKPARALLEL: %[[RES:.*]] = addf %[[LHS]], %[[RHS]] : f32 -// CHECKPARALLEL: store %[[RES]], %{{.*}}[%{{.*}}, %{{.*}}] : memref - -#accesses = [ - affine_map<(i, j, k) -> (i, j)>, - affine_map<(i, j, k) -> (i, j, k)>, - affine_map<(i, j, k) -> (i, k, j)> -] -#trait2 = { - args_in = 1, - args_out = 2, - iterator_types = ["parallel", "parallel", "parallel"], - indexing_maps = #accesses, - library_call = "some_external_function_name_2", - doc = "B(i,j,k), C(i,k,j) = foo(A(i, j), B(i,j,k), C(i,k,j))" -} -func @generic_region(%arg0: memref, %arg1: memref, %arg2: memref) { - linalg.generic #trait2 - ins(%arg0: memref) - outs(%arg1, %arg2 : memref, - memref) { - ^bb0(%a: f32, %b: f32, %c: f32): - %d = mulf %a, %b : f32 - %e = addf %c, %d : f32 - linalg.yield %d, %e : f32, f32 - } - return -} -// CHECKLOOP-LABEL: @generic_region -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[j:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[k:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %{{.*}}[%[[i]], %[[j]]] : memref -// CHECKLOOP: %[[b:.*]] = load %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref -// CHECKLOOP: %[[c:.*]] = load %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref -// CHECKLOOP: %[[d:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKLOOP: %[[e:.*]] = addf %[[c]], %[[d]] : f32 -// CHECKLOOP: store %[[d]], %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref -// CHECKLOOP: store %[[e]], %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref - -// CHECKPARALLEL-LABEL: @generic_region -// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]], %[[k:[a-zA-Z0-9_]*]]) -// CHECKPARALLEL: %[[a:.*]] = load %{{.*}}[%[[i]], %[[j]]] : memref -// CHECKPARALLEL: %[[b:.*]] = load %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref -// CHECKPARALLEL: %[[c:.*]] = load %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref -// CHECKPARALLEL: %[[d:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKPARALLEL: %[[e:.*]] = addf %[[c]], %[[d]] : f32 -// CHECKPARALLEL: store %[[d]], %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref -// CHECKPARALLEL: store %[[e]], %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref - -#trait4 = { - args_in = 1, - args_out = 2, - iterator_types = ["parallel", "parallel", "parallel"], - indexing_maps = #accesses, - library_call = "some_external_function_name_2", - doc = "B(i,j,k), C(i,k,j) = foo(A(i, j) * B(i,j,k), i * j * k + C(i,k,j))" -} -func @indexed_generic_region( - %arg0: memref, - %arg1: memref, - %arg2: memref) { - linalg.indexed_generic #trait4 - ins(%arg0 : memref) - outs(%arg1, %arg2 : memref, - memref) { - ^bb0(%i: index, %j: index, %k: index, %a: f32, %b: f32, %c: f32): - %result_1 = mulf %a, %b : f32 - - %ij = addi %i, %j : index - %ijk = addi %ij, %k : index - %ijk_int = index_cast %ijk : index to i32 - %ijk_float = sitofp %ijk_int : i32 to f32 - - %result_2 = addf %c, %ijk_float : f32 - linalg.yield %result_1, %result_2 : f32, f32 - } - return -} - -// CHECKLOOP-LABEL: @indexed_generic_region -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[j:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[k:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %{{.*}}[%[[i]], %[[j]]] -// CHECKLOOP: %[[b:.*]] = load %{{.*}}[%[[i]], %[[j]], %[[k]]] -// CHECKLOOP: %[[c:.*]] = load %{{.*}}[%[[i]], %[[k]], %[[j]]] -// CHECKLOOP: %[[result_1:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKLOOP: %[[ij:.*]] = addi %[[i]], %[[j]] : index -// CHECKLOOP: %[[ijk:.*]] = addi %[[ij]], %[[k]] : index -// CHECKLOOP: %[[ijk_int:.*]] = index_cast %[[ijk]] : index to i32 -// CHECKLOOP: %[[ijk_float:.*]] = sitofp %[[ijk_int]] : i32 to f32 -// CHECKLOOP: %[[result_2:.*]] = addf %[[c]], %[[ijk_float]] : f32 -// CHECKLOOP: store %[[result_1]], %{{.*}}[%[[i]], %[[j]], %[[k]]] -// CHECKLOOP: store %[[result_2]], %{{.*}}[%[[i]], %[[k]], %[[j]]] - -// CHECKPARALLEL-LABEL: @indexed_generic_region -// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]], %[[k:[a-zA-Z0-9_]*]]) -// CHECKPARALLEL: %[[a:.*]] = load %{{.*}}[%[[i]], %[[j]]] -// CHECKPARALLEL: %[[b:.*]] = load %{{.*}}[%[[i]], %[[j]], %[[k]]] -// CHECKPARALLEL: %[[c:.*]] = load %{{.*}}[%[[i]], %[[k]], %[[j]]] -// CHECKPARALLEL: %[[result_1:.*]] = mulf %[[a]], %[[b]] : f32 -// CHECKPARALLEL: %[[ij:.*]] = addi %[[i]], %[[j]] : index -// CHECKPARALLEL: %[[ijk:.*]] = addi %[[ij]], %[[k]] : index -// CHECKPARALLEL: %[[ijk_int:.*]] = index_cast %[[ijk]] : index to i32 -// CHECKPARALLEL: %[[ijk_float:.*]] = sitofp %[[ijk_int]] : i32 to f32 -// CHECKPARALLEL: %[[result_2:.*]] = addf %[[c]], %[[ijk_float]] : f32 -// CHECKPARALLEL: store %[[result_1]], %{{.*}}[%[[i]], %[[j]], %[[k]]] -// CHECKPARALLEL: store %[[result_2]], %{{.*}}[%[[i]], %[[k]], %[[j]]] - -// ----- - -#broadcast_access = [ - affine_map<(i, j) -> ()>, - affine_map<(i, j) -> (i, j)> -] - -#trait_broadcast = { - args_in = 1, - args_out = 1, - indexing_maps = #broadcast_access, - iterator_types = ["parallel", "parallel"], - library_call = "some_broadcast_external_fn" -} - -func @generic_op_zero_rank(%arg0: memref, %arg1: memref<3x4xf32>) -{ - linalg.generic #trait_broadcast - ins(%arg0 : memref) - outs(%arg1 : memref<3x4xf32>) { - ^bb(%a: f32, %b: f32) : - linalg.yield %a : f32 - } - return -} - -// CHECKLOOP-LABEL: @generic_op_zero_rank -// CHECKLOOP-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32> -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[j:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %[[ARG0]][] -// CHECKLOOP: store %[[a]], %[[ARG1]][%[[i]], %[[j]]] - -// CHECKPARALLEL-LABEL: @generic_op_zero_rank -// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32> -// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]]) -// CHECKPARALLEL: %[[a:.*]] = load %[[ARG0]][] -// CHECKPARALLEL: store %[[a]], %[[ARG1]][%[[i]], %[[j]]] - -func @indexed_generic_op_zero_rank(%arg0: memref, %arg1: memref<3x4xi32>) -{ - linalg.indexed_generic #trait_broadcast - ins(%arg0 : memref) - outs(%arg1 : memref<3x4xi32>) { - ^bb(%i: index, %j: index, %a: i32, %b: i32) : - %ij = addi %i, %j : index - %ij_int = index_cast %ij : index to i32 - %result = addi %a, %ij_int : i32 - linalg.yield %result : i32 - } - return -} - -// CHECKLOOP-LABEL: @indexed_generic_op_zero_rank -// CHECKLOOP-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xi32> -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: scf.for %[[j:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %[[ARG0]][ -// CHECKLOOP: %[[ij:.*]] = addi %[[i]], %[[j]] : index -// CHECKLOOP: %[[ij_int:.*]] = index_cast %[[ij]] : index to i32 -// CHECKLOOP: %[[result:.*]] = addi %[[a]], %[[ij_int]] : i32 -// CHECKLOOP: store %[[result]], %[[ARG1]][%[[i]], %[[j]]] - -// CHECKPARALLEL-LABEL: @indexed_generic_op_zero_rank -// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xi32> -// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]]) -// CHECKPARALLEL: %[[a:.*]] = load %[[ARG0]][ -// CHECKPARALLEL: %[[ij:.*]] = addi %[[i]], %[[j]] : index -// CHECKPARALLEL: %[[ij_int:.*]] = index_cast %[[ij]] : index to i32 -// CHECKPARALLEL: %[[result:.*]] = addi %[[a]], %[[ij_int]] : i32 -// CHECKPARALLEL: store %[[result]], %[[ARG1]][%[[i]], %[[j]]] - -#reduce_1D_access = [ - affine_map<(i) -> (i)>, - affine_map<(i) -> ()> -] - -#trait_reduce_1D = { - args_in = 1, - args_out = 1, - indexing_maps = #reduce_1D_access, - iterator_types = ["reduction"], - library_call = "some_reduce_external_fn" -} - -func @generic_op_1D_reduce(%arg0: memref, %arg1: memref) -{ - linalg.generic #trait_reduce_1D - ins(%arg0 : memref) - outs(%arg1 : memref) { - ^bb(%a: f32, %b: f32) : - %0 = addf %a, %b : f32 - linalg.yield %0 : f32 - } - return -} -// CHECKLOOP-LABEL: @generic_op_1D_reduce -// CHECKLOOP-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %[[ARG0]][%[[i]]] -// CHECKLOOP: %[[b:.*]] = load %[[ARG1]][] -// CHECKLOOP: %[[c:.*]] = addf %[[a]], %[[b]] : f32 -// CHECKLOOP: store %[[c]], %[[ARG1]][] - -// CHECKPARALLEL-LABEL: @generic_op_1D_reduce -// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL: scf.for %[[i:.*]] = {{.*}} -// CHECKPARALLEL: %[[a:.*]] = load %[[ARG0]][%[[i]]] -// CHECKPARALLEL: %[[b:.*]] = load %[[ARG1]][] -// CHECKPARALLEL: %[[c:.*]] = addf %[[a]], %[[b]] : f32 -// CHECKPARALLEL: store %[[c]], %[[ARG1]][] - - -#reduce_init_1D_access = [ - affine_map<(i) -> (i)>, - affine_map<(i) -> ()>, - affine_map<(i) -> ()> -] - -#trait_reduce_init_1D = { - args_in = 2, - args_out = 1, - indexing_maps = #reduce_init_1D_access, - iterator_types = ["reduction"], - library_call = "some_reduce_external_fn" -} - -func @indexed_generic_op_1D_reduce(%arg0: memref, - %arg1: memref, - %arg2: memref) -{ - linalg.indexed_generic #trait_reduce_init_1D - ins(%arg0, %arg1 : memref, memref) - outs(%arg2 : memref) { - ^bb(%i : index, %a: f32, %b: f32, %c: f32) : - %0 = constant 0 : index - %1 = cmpi "eq", %0, %i : index - %2 = select %1, %b, %c : f32 - %3 = addf %a, %2 : f32 - linalg.yield %3 : f32 - } - return -} -// CHECKLOOP-LABEL: @indexed_generic_op_1D_reduce -// CHECKLOOP-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: %[[a:.*]] = load %[[ARG0]][%[[i]]] -// CHECKLOOP: %[[b:.*]] = load %[[ARG1]][] -// CHECKLOOP: %[[c:.*]] = load %[[ARG2]][] -// CHECKLOOP: %[[d:.*]] = select %{{.*}}, %[[b]], %[[c]] -// CHECKLOOP: %[[e:.*]] = addf %[[a]], %[[d]] -// CHECKLOOP: store %[[e]], %[[ARG2]][] - -// CHECKPARALLEL-LABEL: @indexed_generic_op_1D_reduce -// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL: scf.for %[[i:.*]] = {{.*}} -// CHECKPARALLEL: %[[a:.*]] = load %[[ARG0]][%[[i]]] -// CHECKPARALLEL: %[[b:.*]] = load %[[ARG1]][] -// CHECKPARALLEL: %[[c:.*]] = load %[[ARG2]][] -// CHECKPARALLEL: %[[d:.*]] = select %{{.*}}, %[[b]], %[[c]] -// CHECKPARALLEL: %[[e:.*]] = addf %[[a]], %[[d]] -// CHECKPARALLEL: store %[[e]], %[[ARG2]][] - -#trait_const_fill = { - args_in = 0, - args_out = 1, - indexing_maps = [affine_map<(i) -> (i)>], - iterator_types = ["parallel"], - library_call = "some_external_fn" -} -func @generic_const_init(%arg0: memref) { - %cst = constant 1.0 : f32 - linalg.generic #trait_const_fill outs(%arg0 : memref) { - ^bb0(%arg1: f32): // no predecessors - linalg.yield %cst : f32 - } - return -} -// CHECKLOOP-LABEL: @generic_const_init -// CHECKLOOP-SAME: %[[ARG0:.*]]: memref -// CHECKLOOP: %[[CONST:.*]] = constant 1.000000e+00 : f32 -// CHECKLOOP: scf.for %[[i:.*]] = {{.*}} -// CHECKLOOP: store %[[CONST]], %[[ARG0]] - -// CHECKPARALLEL-LABEL: @generic_const_init -// CHECKPARALLEL-SAME: %[[ARG0:.*]]: memref -// CHECKPARALLEL: %[[CONST:.*]] = constant 1.000000e+00 : f32 -// CHECKPARALLEL: scf.parallel (%[[i:.*]]) -// CHECKPARALLEL: store %[[CONST]], %[[ARG0]] - -#scalar_access = [ - affine_map<() -> ()>, - affine_map<() -> ()>, - affine_map<() -> ()> -] -#scalar_trait = { - args_in = 2, - args_out = 1, - iterator_types = [], - indexing_maps = #scalar_access, - library_call = "some_external_fn" -} -func @scalar_code(%arg0: memref, %arg1 : memref, %arg2 : memref) -{ - linalg.generic #scalar_trait - ins(%arg0, %arg1 : memref, memref) - outs(%arg2 : memref) { - ^bb(%a : f32, %b : f32, %c : f32) : - %0 = addf %a, %b : f32 - linalg.yield %0 : f32 - } - return -} -// CHECKLOOP-LABEL: @scalar_code -// CHECKLOOP-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref -// CHECKLOOP-NOT: scf.for -// CHECKLOOP: load %[[ARG0]][] -// CHECKLOOP: load %[[ARG1]][] -// CHECKLOOP: addf -// CHECKLOOP: store %{{.*}}, %[[ARG2]][] - -// CHECKPARALLEL-LABEL: @scalar_code -// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref -// CHECKPARALLEL-NOT: scf.for -// CHECKPARALLEL: load %[[ARG0]][] -// CHECKPARALLEL: load %[[ARG1]][] -// CHECKPARALLEL: addf -// CHECKPARALLEL: store %{{.*}}, %[[ARG2]][] - -//----------------------------------------------------------------------------// -// Named ops to loops. -//----------------------------------------------------------------------------// -func @named_batch_matmul(%A: memref, %B: memref, %C: memref) { - linalg.batch_matmul ins(%A, %B : memref, memref) - outs(%C : memref) - return -} -// CHECKLOOP-LABEL: @named_batch_matmul -// CHECKLOOP-SAME: %[[mA:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[mB:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[mC:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[B:.*]] = dim %[[mA]], %c0 : memref -// CHECKLOOP: %[[M:.*]] = dim %[[mA]], %c1 : memref -// CHECKLOOP: %[[K:.*]] = dim %[[mA]], %c2 : memref -// CHECKLOOP: %[[N:.*]] = dim %[[mB]], %c2 : memref -// CHECKLOOP: scf.for %[[b:.*]] = %{{.*}} to %[[B]] step %{{.*}} { -// CHECKLOOP: scf.for %[[m:.*]] = %{{.*}} to %[[M]] step %{{.*}} { -// CHECKLOOP: scf.for %[[n:.*]] = %{{.*}} to %[[N]] step %{{.*}} { -// CHECKLOOP: scf.for %[[k:.*]] = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKLOOP: %[[va:.*]] = load %[[mA]][%[[b]], %[[m]], %[[k]]] : memref -// CHECKLOOP: %[[vb:.*]] = load %[[mB]][%[[b]], %[[k]], %[[n]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[mC]][%[[b]], %[[m]], %[[n]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[mC]][%[[b]], %[[m]], %[[n]]] : memref - -// CHECKPARALLEL-LABEL: @named_batch_matmul -// CHECKPARALLEL-SAME: %[[mA:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[mB:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[mC:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[B:.*]] = dim %[[mA]], %c0 : memref -// CHECKPARALLEL: %[[M:.*]] = dim %[[mA]], %c1 : memref -// CHECKPARALLEL: %[[K:.*]] = dim %[[mA]], %c2 : memref -// CHECKPARALLEL: %[[N:.*]] = dim %[[mB]], %c2 : memref -// CHECKPARALLEL: scf.parallel (%[[b:.*]], %[[m:.*]], %[[n:.*]]) = ({{.*}}) to (%[[B]], %[[M]], %[[N]]) step ({{.*}}) { -// CHECKPARALLEL: scf.for %[[k:.*]] = %{{.*}} to %[[K]] step %{{.*}} { -// CHECKPARALLEL: %[[va:.*]] = load %[[mA]][%[[b]], %[[m]], %[[k]]] : memref -// CHECKPARALLEL: %[[vb:.*]] = load %[[mB]][%[[b]], %[[k]], %[[n]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[mC]][%[[b]], %[[m]], %[[n]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[mC]][%[[b]], %[[m]], %[[n]]] : memref - -#conv_1d_accesses = [ - affine_map<(m, n)[s0] -> (m + n - s0 floordiv 2)>, // in - affine_map<(m, n)[s0] -> (n)>, // filter - affine_map<(m, n)[s0] -> (m)> // out -] - -#conv_1d_trait = { - args_in = 2, - args_out = 1, - doc = "C(m) += A(m) * B(n)", - indexing_maps = #conv_1d_accesses, - library_call = "linalg_conv_1d", - n_views = [2, 1], - iterator_types = ["parallel", "parallel"], - symbol_source = 1 -} - -func @conv1d(%in : memref, %filter : memref, %out : memref) -> () { - linalg.generic #conv_1d_trait - ins(%in, %filter : memref, memref) - outs(%out : memref) { - ^bb0(%a: f32, %b: f32, %c: f32) : - %d = mulf %a, %b : f32 - %e = addf %c, %d : f32 - linalg.yield %e : f32 - } - return -} - -// CHECKLOOP-LABEL: @conv1d -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[lowerBound:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim1]]] -// CHECKLOOP: %[[upperBound:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim1]], %[[dim0]]] -// CHECKLOOP: scf.for %[[b:.*]] = %[[lowerBound]] to %[[upperBound]] step %{{.*}} { -// CHECKLOOP: scf.for %[[m:.*]] = %{{.*}} to %[[dim1]] step %{{.*}} { -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[aff:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim2]]] -// CHECKLOOP: %[[va:.*]] = load %[[arg0]][%[[aff]]] : memref -// CHECKLOOP: %[[vb:.*]] = load %[[arg1]][%[[m]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[b]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[b]]] : memref - -// CHECKPARALLEL-LABEL: @conv1d -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[lowerBound:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim1]]] -// CHECKPARALLEL: %[[upperBound:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim1]], %[[dim0]]] -// CHECKPARALLEL: scf.parallel (%[[b:.*]], %[[m:.*]]) = (%[[lowerBound]], %{{.*}}) to (%[[upperBound]], %[[dim1]]) step ({{.*}}) { -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim2]]] -// CHECKPARALLEL: %[[va:.*]] = load %[[arg0]][%[[aff]]] : memref -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg1]][%[[m]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[b]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[b]]] : memref - -#conv_2d_accesses = [ - affine_map<(m, n, m1, n1)[s0, s1] -> (m + m1 - s0 floordiv 2, n + n1 - s1 floordiv 2)>, // in - affine_map<(m, n, m1, n1)[s0, s1] -> (m1, n1)>, // filter - affine_map<(m, n, m1, n1)[s0, s1] -> (m, n)> // out -] - -#conv_2d_trait = { - args_in = 2, - args_out = 1, - doc = "C(m,n) += A(m,n) * B(m1,n1)", - indexing_maps = #conv_2d_accesses, - library_call = "linalg_conv_2d", - n_views = [2, 1], - iterator_types = ["parallel", "parallel", "parallel", "parallel"], - symbol_source = 1 -} - -func @conv2d(%in : memref, %filter : memref, %out : memref) -> () { - linalg.generic #conv_2d_trait - ins(%in, %filter : memref, memref) - outs(%out : memref) { - ^bb0(%a: f32, %b: f32, %c: f32) : - %d = mulf %a, %b : f32 - %e = addf %c, %d : f32 - linalg.yield %e : f32 - } - return -} - -// CHECKLOOP-LABEL: @conv2d -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim3:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim2]]] -// CHECKLOOP: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim2]], %[[dim0]]] -// CHECKLOOP: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim3]]] -// CHECKLOOP: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim3]], %[[dim1]]] -// CHECKLOOP: scf.for %[[i0:.*]] = %[[lowerBound1]] to %[[upperBound1]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i1:.*]] = %[[lowerBound2]] to %[[upperBound2]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i2:.*]] = %{{.*}} to %[[dim2]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i3:.*]] = %{{.*}} to %[[dim3]] step %{{.*}} { -// CHECKLOOP: %[[dim4:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim5:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim4]]] -// CHECKLOOP: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim5]]] -// CHECKLOOP: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]]] : memref -// CHECKLOOP: %[[vb:.*]] = load %[[arg1]][%[[i2]], %[[i3]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[i0]], %[[i1]]] : memref - -// CHECKPARALLEL-LABEL: @conv2d -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim3:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim2]]] -// CHECKPARALLEL: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim2]], %[[dim0]]] -// CHECKPARALLEL: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim3]]] -// CHECKPARALLEL: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim3]], %[[dim1]]] -// CHECKPARALLEL: scf.parallel (%[[i0:.*]], %[[i1:.*]], %[[i2:.*]], %[[i3:.*]]) = (%[[lowerBound1]], %[[lowerBound2]], %{{.*}}, %{{.*}}) to (%[[upperBound1]], %[[upperBound2]], %[[dim2]], %[[dim3]]) step ({{.*}}) { -// CHECKPARALLEL: %[[dim4:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim5:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim4]]] -// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim5]]] -// CHECKPARALLEL: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]]] : memref -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg1]][%[[i2]], %[[i3]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[i0]], %[[i1]]] : memref - -#conv_3d_accesses = [ - affine_map<(m, n, k, m1, n1, k1)[s0, s1, s2] -> (m + m1 - s0 floordiv 2, n + n1 - s1 floordiv 2, k + k1 - s2 floordiv 2)>, // in - affine_map<(m, n, k, m1, n1, k1)[s0, s1, s2] -> (m1, n1, k1)>, // filter - affine_map<(m, n, k, m1, n1, k1)[s0, s1, s2] -> (m, n, k)> // out -] - -#conv_3d_trait = { - args_in = 2, - args_out = 1, - doc = "C(m,n,k) += A(m,n,k) * B(m1,n1,k1)", - indexing_maps = #conv_3d_accesses, - library_call = "linalg_conv_3d", - n_views = [2, 1], - iterator_types = ["parallel", "parallel", "parallel", "parallel", "parallel", "parallel"], - symbol_source = 1 -} - -func @conv3d(%in : memref, %filter : memref, %out : memref) -> () { - linalg.generic #conv_3d_trait - ins(%in, %filter : memref, memref) - outs(%out : memref) { - ^bb0(%a: f32, %b: f32, %c: f32) : - %d = mulf %a, %b : f32 - %e = addf %c, %d : f32 - linalg.yield %e : f32 - } - return -} - -// CHECKLOOP-LABEL: @conv3d -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[c2:.*]] = constant 2 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg0]], %[[c2]] : memref -// CHECKLOOP: %[[dim3:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim4:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim5:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKLOOP: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim3]]] -// CHECKLOOP: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim3]], %[[dim0]]] -// CHECKLOOP: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim4]]] -// CHECKLOOP: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim4]], %[[dim1]]] -// CHECKLOOP: %[[lowerBound3:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim5]]] -// CHECKLOOP: %[[upperBound3:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim5]], %[[dim2]]] -// CHECKLOOP: scf.for %[[i0:.*]] = %[[lowerBound1]] to %[[upperBound1]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i1:.*]] = %[[lowerBound2]] to %[[upperBound2]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i2:.*]] = %[[lowerBound3]] to %[[upperBound3]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i3:.*]] = %{{.*}} to %[[dim3]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i4:.*]] = %{{.*}} to %[[dim4]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i5:.*]] = %{{.*}} to %[[dim5]] step %{{.*}} { -// CHECKLOOP: %[[dim6:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim7:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim8:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKLOOP: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim6]]] -// CHECKLOOP: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim7]]] -// CHECKLOOP: %[[aff3:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim8]]] -// CHECKLOOP: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]], %[[aff3]]] : memref -// CHECKLOOP: %[[vb:.*]] = load %[[arg1]][%[[i3]], %[[i4]], %[[i5]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]], %[[i2]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[i0]], %[[i1]], %[[i2]]] : memref - -// CHECKPARALLEL-LABEL: @conv3d -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[c2:.*]] = constant 2 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg0]], %[[c2]] : memref -// CHECKPARALLEL: %[[dim3:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim4:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim5:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKPARALLEL: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim3]]] -// CHECKPARALLEL: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim3]], %[[dim0]]] -// CHECKPARALLEL: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim4]]] -// CHECKPARALLEL: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim4]], %[[dim1]]] -// CHECKPARALLEL: %[[lowerBound3:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim5]]] -// CHECKPARALLEL: %[[upperBound3:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim5]], %[[dim2]]] -// CHECKPARALLEL: scf.parallel (%[[i0:.*]], %[[i1:.*]], %[[i2:.*]], %[[i3:.*]], %[[i4:.*]], %[[i5:.*]]) = (%[[lowerBound1]], %[[lowerBound2]], %[[lowerBound3]], %{{.*}}, %{{.*}}, %{{.*}}) to (%[[upperBound1]], %[[upperBound2]], %[[upperBound3]], %[[dim3]], %[[dim4]], %[[dim5]]) step ({{.*}}) { -// CHECKPARALLEL: %[[dim6:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim7:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim8:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKPARALLEL: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim6]]] -// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim7]]] -// CHECKPARALLEL: %[[aff3:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim8]]] -// CHECKPARALLEL: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]], %[[aff3]]] : memref -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg1]][%[[i3]], %[[i4]], %[[i5]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]], %[[i2]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[i0]], %[[i1]], %[[i2]]] : memref - -#conv_4d_accesses = [ - affine_map<(m, n, k, l, m1, n1, k1, l1)[s0, s1, s2, s3] -> (m + m1 - s0 floordiv 2, n + n1 - s1 floordiv 2, k + k1 - s2 floordiv 2, l + l1 - s3 floordiv 2)>, // in - affine_map<(m, n, k, l, m1, n1, k1, l1)[s0, s1, s2, s3] -> (m1, n1, k1, l1)>, // filter - affine_map<(m, n, k, l, m1, n1, k1, l1)[s0, s1, s2, s3] -> (m, n, k, l)> // out -] - -#conv_4d_trait = { - args_in = 2, - args_out = 1, - doc = "C(m,n,k,l) += A(m,n,k,l) * B(m1,n1,k1,l1)", - indexing_maps = #conv_4d_accesses, - library_call = "linalg_conv_4d", - n_views = [2, 1], - iterator_types = ["parallel", "parallel", "parallel", "parallel", "parallel", "parallel", "parallel", "parallel"], - symbol_source = 1 -} - -func @conv4d(%in : memref, %filter : memref, %out : memref) -> () { - linalg.generic #conv_4d_trait - ins(%in, %filter : memref, memref) - outs(%out : memref) { - ^bb0(%a: f32, %b: f32, %c: f32) : - %d = mulf %a, %b : f32 - %e = addf %c, %d : f32 - linalg.yield %e : f32 - } - return -} - -// CHECKLOOP-LABEL: @conv4d -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[c2:.*]] = constant 2 : index -// CHECKLOOP: %[[c3:.*]] = constant 3 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg0]], %[[c2]] : memref -// CHECKLOOP: %[[dim3:.*]] = dim %[[arg0]], %[[c3]] : memref -// CHECKLOOP: %[[dim4:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim5:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim6:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKLOOP: %[[dim7:.*]] = dim %[[arg1]], %[[c3]] : memref -// CHECKLOOP: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim4]]] -// CHECKLOOP: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim4]], %[[dim0]]] -// CHECKLOOP: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim5]]] -// CHECKLOOP: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim5]], %[[dim1]]] -// CHECKLOOP: %[[lowerBound3:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim6]]] -// CHECKLOOP: %[[upperBound3:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim6]], %[[dim2]]] -// CHECKLOOP: %[[lowerBound4:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim7]]] -// CHECKLOOP: %[[upperBound4:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim7]], %[[dim3]]] -// CHECKLOOP: scf.for %[[i0:.*]] = %[[lowerBound1]] to %[[upperBound1]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i1:.*]] = %[[lowerBound2]] to %[[upperBound2]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i2:.*]] = %[[lowerBound3]] to %[[upperBound3]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i3:.*]] = %[[lowerBound4]] to %[[upperBound4]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i4:.*]] = %{{.*}} to %[[dim4]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i5:.*]] = %{{.*}} to %[[dim5]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i6:.*]] = %{{.*}} to %[[dim6]] step %{{.*}} { -// CHECKLOOP: scf.for %[[i7:.*]] = %{{.*}} to %[[dim7]] step %{{.*}} { -// CHECKLOOP: %[[dim8:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim9:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim10:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKLOOP: %[[dim11:.*]] = dim %[[arg1]], %[[c3]] : memref -// CHECKLOOP: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim8]]] -// CHECKLOOP: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim9]]] -// CHECKLOOP: %[[aff3:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim10]]] -// CHECKLOOP: %[[aff4:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim11]]] -// CHECKLOOP: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]], %[[aff3]], %[[aff4]]] : memref -// CHECKLOOP: %[[vb:.*]] = load %[[arg1]][%[[i4]], %[[i5]], %[[i6]], %[[i7]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]], %[[i2]], %[[i3]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[i0]], %[[i1]], %[[i2]], %[[i3]]] : memref - -// CHECKPARALLEL-LABEL: @conv4d -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[c2:.*]] = constant 2 : index -// CHECKPARALLEL: %[[c3:.*]] = constant 3 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg0]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg0]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg0]], %[[c2]] : memref -// CHECKPARALLEL: %[[dim3:.*]] = dim %[[arg0]], %[[c3]] : memref -// CHECKPARALLEL: %[[dim4:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim5:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim6:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKPARALLEL: %[[dim7:.*]] = dim %[[arg1]], %[[c3]] : memref -// CHECKPARALLEL: %[[lowerBound1:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim4]]] -// CHECKPARALLEL: %[[upperBound1:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim4]], %[[dim0]]] -// CHECKPARALLEL: %[[lowerBound2:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim5]]] -// CHECKPARALLEL: %[[upperBound2:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim5]], %[[dim1]]] -// CHECKPARALLEL: %[[lowerBound3:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim6]]] -// CHECKPARALLEL: %[[upperBound3:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim6]], %[[dim2]]] -// CHECKPARALLEL: %[[lowerBound4:.*]] = affine.apply #[[$convLowerBound]]()[%[[dim7]]] -// CHECKPARALLEL: %[[upperBound4:.*]] = affine.apply #[[$convUpperBound]]()[%[[dim7]], %[[dim3]]] -// CHECKPARALLEL: scf.parallel (%[[i0:.*]], %[[i1:.*]], %[[i2:.*]], %[[i3:.*]], %[[i4:.*]], %[[i5:.*]], %[[i6:.*]], %[[i7:.*]]) = (%[[lowerBound1]], %[[lowerBound2]], %[[lowerBound3]], %[[lowerBound4]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}) to (%[[upperBound1]], %[[upperBound2]], %[[upperBound3]], %[[upperBound4]], %[[dim4]], %[[dim5]], %[[dim6]], %[[dim7]]) step ({{.*}}) { -// CHECKPARALLEL: %[[dim8:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim9:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim10:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKPARALLEL: %[[dim11:.*]] = dim %[[arg1]], %[[c3]] : memref -// CHECKPARALLEL: %[[aff1:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim8]]] -// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim9]]] -// CHECKPARALLEL: %[[aff3:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim10]]] -// CHECKPARALLEL: %[[aff4:.*]] = affine.apply #[[$convMap]](%{{.*}}, %{{.*}})[%[[dim11]]] -// CHECKPARALLEL: %[[va:.*]] = load %[[arg0]][%[[aff1]], %[[aff2]], %[[aff3]], %[[aff4]]] : memref -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg1]][%[[i4]], %[[i5]], %[[i6]], %[[i7]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[i0]], %[[i1]], %[[i2]], %[[i3]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[va]], %[[vb]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[i0]], %[[i1]], %[[i2]], %[[i3]]] : memref - -func @conv1d_no_symbols(%in : memref, %filter : memref, %out : memref) -> () { - linalg.conv_1d ins(%in, %filter : memref, memref) - outs(%out : memref) - return -} - -// CHECKLOOP-LABEL: @conv1d_no_symbols -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKLOOP: scf.for %[[b:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] { -// CHECKLOOP: scf.for %[[m:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKLOOP: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[b]], %[[m]]) -// CHECKLOOP: %[[vb:.*]] = load %[[arg0]][%[[aff]]] : memref -// CHECKLOOP: %[[va:.*]] = load %[[arg1]][%[[m]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[b]]] : memref -// CHECKLOOP: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[b]]] : memref - -// CHECKPARALLEL-LABEL: @conv1d_no_symbols -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKPARALLEL: scf.parallel (%[[b:.*]]) = (%[[c0]]) to (%[[dim1]]) step (%[[c1]]) { -// CHECKPARALLEL: scf.for %[[m:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[b]], %[[m]]) -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg0]][%[[aff]]] : memref -// CHECKPARALLEL: %[[va:.*]] = load %[[arg1]][%[[m]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[b]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[b]]] : memref - - -func @conv2d_no_symbols(%in : memref, %filter : memref, %out : memref) -> () { - linalg.conv_2d ins(%in, %filter : memref, memref) - outs(%out: memref) - return -} -// CHECKLOOP-LABEL: @conv2d_no_symbols -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKLOOP: %[[dim3:.*]] = dim %[[arg2]], %[[c1]] : memref -// CHECKLOOP: scf.for %[[arg3:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg4:.*]] = %[[c0]] to %[[dim3]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] { -// CHECKLOOP: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg5]]) -// CHECKLOOP: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg6]]) -// CHECKLOOP: %[[vb:.*]] = load %[[arg0]][%[[aff]], %[[aff2]]] : memref - -// CHECKLOOP: %[[va:.*]] = load %[[arg1]][%[[arg5]], %[[arg6]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[arg3]], %[[arg4]]] : memref - -// CHECKLOOP: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]]] : memref - -// CHECKPARALLEL-LABEL: @conv2d_no_symbols -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim3:.*]] = dim %[[arg2]], %[[c1]] : memref -// CHECKPARALLEL: scf.parallel (%[[arg3:.*]], %[[arg4:.*]]) = (%[[c0]], %[[c0]]) to (%[[dim2]], %[[dim3]]) step (%[[c1]], %[[c1]]) { -// CHECKPARALLEL: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKPARALLEL: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] { -// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg5]]) -// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg6]]) -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg0]][%[[aff]], %[[aff2]]] : memref -// CHECKPARALLEL: %[[va:.*]] = load %[[arg1]][%[[arg5]], %[[arg6]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[arg3]], %[[arg4]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]]] : memref - - -func @conv3d_no_symbols(%in : memref, %filter : memref, %out : memref) -> () { - linalg.conv_3d ins(%in, %filter : memref, memref) - outs(%out : memref) - return -} - -// CHECKLOOP-LABEL: @conv3d_no_symbols -// CHECKLOOP-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKLOOP-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKLOOP: %[[c2:.*]] = constant 2 : index -// CHECKLOOP: %[[c0:.*]] = constant 0 : index -// CHECKLOOP: %[[c1:.*]] = constant 1 : index -// CHECKLOOP: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKLOOP: %[[dim1:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKLOOP: %[[dim2:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKLOOP: %[[dim3:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKLOOP: %[[dim4:.*]] = dim %[[arg2]], %[[c1]] : memref -// CHECKLOOP: %[[dim5:.*]] = dim %[[arg2]], %[[c2]] : memref -// CHECKLOOP: scf.for %[[arg3:.*]] = %[[c0]] to %[[dim3]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg4:.*]] = %[[c0]] to %[[dim4]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim5]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg7:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] { -// CHECKLOOP: scf.for %[[arg8:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] { -// CHECKLOOP: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg6]]) -// CHECKLOOP: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg7]]) -// CHECKLOOP: %[[aff3:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg5]], %[[arg8]]) -// CHECKLOOP: %[[vb:.*]] = load %[[arg0]][%[[aff]], %[[aff2]], %[[aff3]]] : memref - -// CHECKLOOP: %[[va:.*]] = load %[[arg1]][%[[arg6]], %[[arg7]], %[[arg8]]] : memref -// CHECKLOOP: %[[vc:.*]] = load %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref - -// CHECKLOOP: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKLOOP: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKLOOP: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref - -// CHECKPARALLEL-LABEL: @conv3d_no_symbols -// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref -// CHECKPARALLEL: %[[c2:.*]] = constant 2 : index -// CHECKPARALLEL: %[[c0:.*]] = constant 0 : index -// CHECKPARALLEL: %[[c1:.*]] = constant 1 : index -// CHECKPARALLEL: %[[dim0:.*]] = dim %[[arg1]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim1:.*]] = dim %[[arg1]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim2:.*]] = dim %[[arg1]], %[[c2]] : memref -// CHECKPARALLEL: %[[dim3:.*]] = dim %[[arg2]], %[[c0]] : memref -// CHECKPARALLEL: %[[dim4:.*]] = dim %[[arg2]], %[[c1]] : memref -// CHECKPARALLEL: %[[dim5:.*]] = dim %[[arg2]], %[[c2]] : memref -// CHECKPARALLEL: scf.parallel (%[[arg3:.*]], %[[arg4:.*]], %[[arg5:.*]]) = (%[[c0]], %[[c0]], %[[c0]]) to (%[[dim3]], %[[dim4]], %[[dim5]]) step (%[[c1]], %[[c1]], %[[c1]]) { -// CHECKPARALLEL: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] { -// CHECKPARALLEL: scf.for %[[arg7:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] { -// CHECKPARALLEL: scf.for %[[arg8:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] { -// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg6]]) -// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg7]]) -// CHECKPARALLEL: %[[aff3:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg5]], %[[arg8]]) -// CHECKPARALLEL: %[[vb:.*]] = load %[[arg0]][%[[aff]], %[[aff2]], %[[aff3]]] : memref -// CHECKPARALLEL: %[[va:.*]] = load %[[arg1]][%[[arg6]], %[[arg7]], %[[arg8]]] : memref -// CHECKPARALLEL: %[[vc:.*]] = load %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref -// CHECKPARALLEL: %[[inc:.*]] = mulf %[[vb]], %[[va]] : f32 -// CHECKPARALLEL: %[[res:.*]] = addf %[[vc]], %[[inc]] : f32 -// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref