From 22954a0e408afde1d8686dffb3a3dcab107a2cd3 Mon Sep 17 00:00:00 2001 From: Manuel Freiberger Date: Sun, 22 Dec 2019 10:01:35 -0800 Subject: [PATCH] Add integer bit-shift operations to the standard dialect. Rename the 'shlis' operation in the standard dialect to 'shift_left'. Add tests for this operation (these have been missing so far) and add a lowering to the 'shl' operation in the LLVM dialect. Add also 'shift_right_signed' (lowered to LLVM's 'ashr') and 'shift_right_unsigned' (lowered to 'lshr'). The original plan was to name these operations 'shift.left', 'shift.right.signed' and 'shift.right.unsigned'. This works if the operations are prefixed with 'std.' in MLIR assembly. Unfortunately during import the short form is ambigous with operations from a hypothetical 'shift' dialect. The best solution seems to omit dots in standard operations for now. Closes tensorflow/mlir#226 PiperOrigin-RevId: 286803388 --- mlir/bindings/python/test/test_py2and3.py | 2 +- mlir/include/mlir/Dialect/StandardOps/Ops.td | 49 +++++++- .../AffineToStandard/AffineToStandard.cpp | 6 +- mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp | 2 +- .../StandardToLLVM/ConvertStandardToLLVM.cpp | 35 ++++-- .../StandardToSPIRV/ConvertStandardToSPIRV.cpp | 4 +- mlir/lib/Dialect/StandardOps/Ops.cpp | 20 ++-- mlir/lib/EDSC/Builders.cpp | 4 +- mlir/lib/Transforms/Utils/LoopUtils.cpp | 12 +- mlir/test/Conversion/LoopsToGPU/imperfect_3D.mlir | 12 +- mlir/test/Conversion/LoopsToGPU/imperfect_4D.mlir | 12 +- mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir | 4 +- mlir/test/Conversion/LoopsToGPU/step_positive.mlir | 4 +- .../StandardToLLVM/convert-to-llvmir.mlir | 28 +++-- .../Conversion/StandardToSPIRV/std-to-spirv.mlir | 4 +- mlir/test/EDSC/builder-api-test.cpp | 4 +- mlir/test/IR/core-ops.mlir | 126 ++++++++++++++------- mlir/test/Target/llvmir.mlir | 6 + mlir/test/Transforms/canonicalize.mlir | 12 +- mlir/test/Transforms/constant-fold.mlir | 68 +++++------ mlir/test/Transforms/loop-coalescing.mlir | 32 +++--- mlir/test/Transforms/lower-affine.mlir | 6 +- mlir/test/Transforms/parametric-tiling.mlir | 16 +-- 23 files changed, 294 insertions(+), 174 deletions(-) diff --git a/mlir/bindings/python/test/test_py2and3.py b/mlir/bindings/python/test/test_py2and3.py index 02f8f62..2c1158f 100644 --- a/mlir/bindings/python/test/test_py2and3.py +++ b/mlir/bindings/python/test/test_py2and3.py @@ -274,7 +274,7 @@ class EdscTest: printWithCurrentFunctionName(str(self.module)) # CHECK-LABEL: testDivisions # CHECK: floordiv 42 - # CHECK: divis %{{.*}}, %{{.*}} : i32 + # CHECK: divi_signed %{{.*}}, %{{.*}} : i32 def testFunctionArgs(self): self.setUp() diff --git a/mlir/include/mlir/Dialect/StandardOps/Ops.td b/mlir/include/mlir/Dialect/StandardOps/Ops.td index 76c2ba5..c26baf6 100644 --- a/mlir/include/mlir/Dialect/StandardOps/Ops.td +++ b/mlir/include/mlir/Dialect/StandardOps/Ops.td @@ -698,12 +698,12 @@ def DivFOp : FloatArithmeticOp<"divf"> { let summary = "floating point division operation"; } -def DivISOp : IntArithmeticOp<"divis"> { +def SignedDivIOp : IntArithmeticOp<"divi_signed"> { let summary = "signed integer division operation"; let hasFolder = 1; } -def DivIUOp : IntArithmeticOp<"diviu"> { +def UnsignedDivIOp : IntArithmeticOp<"divi_unsigned"> { let summary = "unsigned integer division operation"; let hasFolder = 1; } @@ -1002,12 +1002,12 @@ def RemFOp : FloatArithmeticOp<"remf"> { let summary = "floating point division remainder operation"; } -def RemISOp : IntArithmeticOp<"remis"> { +def SignedRemIOp : IntArithmeticOp<"remi_signed"> { let summary = "signed integer division remainder operation"; let hasFolder = 1; } -def RemIUOp : IntArithmeticOp<"remiu"> { +def UnsignedRemIOp : IntArithmeticOp<"remi_unsigned"> { let summary = "unsigned integer division remainder operation"; let hasFolder = 1; } @@ -1102,8 +1102,45 @@ def SignExtendIOp : Std_Op<"sexti", }]; } -def ShlISOp : IntArithmeticOp<"shlis"> { - let summary = "signed integer shift left"; +def ShiftLeftOp : IntArithmeticOp<"shift_left"> { + let summary = "integer left-shift"; + let description = [{ + The shift_left operation shifts an integer value to the left by a variable + amount. The low order bits are filled with zeros. + + %1 = constant 5 : i8 // %1 is 0b00000101 + %2 = constant 3 : i8 + %3 = shift_left %1, %2 : (i8, i8) -> i8 // %3 is 0b00101000 + }]; +} + +def SignedShiftRightOp : IntArithmeticOp<"shift_right_signed"> { + let summary = "signed integer right-shift"; + let description = [{ + The shift_right_signed operation shifts an integer value to the right by + a variable amount. The integer is interpreted as signed. The high order + bits in the output are filled with copies of the most-significant bit + of the shifted value (which means that the sign of the value is preserved). + + %1 = constant 160 : i8 // %1 is 0b10100000 + %2 = constant 3 : i8 + %3 = shift_right_signed %1, %2 : (i8, i8) -> i8 // %3 is 0b11110100 + %4 = constant 96 : i8 // %4 is 0b01100000 + %5 = shift_right_signed %4, %2 : (i8, i8) -> i8 // %5 is 0b00001100 + }]; +} + +def UnsignedShiftRightOp : IntArithmeticOp<"shift_right_unsigned"> { + let summary = "unsigned integer right-shift"; + let description = [{ + The shift_right_unsigned operation shifts an integer value to the right by + a variable amount. The integer is interpreted as unsigned. The high order + bits are always filled with zeros. + + %1 = constant 160 : i8 // %1 is 0b10100000 + %2 = constant 3 : i8 + %3 = shift_right_unsigned %1, %2 : (i8, i8) -> i8 // %3 is 0b00010100 + }]; } def SIToFPOp : CastOp<"sitofp">, Arguments<(ins AnyType:$in)> { diff --git a/mlir/lib/Conversion/AffineToStandard/AffineToStandard.cpp b/mlir/lib/Conversion/AffineToStandard/AffineToStandard.cpp index 9208ce8..3f613c6 100644 --- a/mlir/lib/Conversion/AffineToStandard/AffineToStandard.cpp +++ b/mlir/lib/Conversion/AffineToStandard/AffineToStandard.cpp @@ -94,7 +94,7 @@ public: auto rhs = visit(expr.getRHS()); assert(lhs && rhs && "unexpected affine expr lowering failure"); - Value *remainder = builder.create(loc, lhs, rhs); + Value *remainder = builder.create(loc, lhs, rhs); Value *zeroCst = builder.create(loc, 0); Value *isRemainderNegative = builder.create(loc, CmpIPredicate::slt, remainder, zeroCst); @@ -138,7 +138,7 @@ public: Value *negatedDecremented = builder.create(loc, noneCst, lhs); Value *dividend = builder.create(loc, negative, negatedDecremented, lhs); - Value *quotient = builder.create(loc, dividend, rhs); + Value *quotient = builder.create(loc, dividend, rhs); Value *correctedQuotient = builder.create(loc, noneCst, quotient); Value *result = builder.create(loc, negative, correctedQuotient, quotient); @@ -178,7 +178,7 @@ public: Value *decremented = builder.create(loc, lhs, oneCst); Value *dividend = builder.create(loc, nonPositive, negated, decremented); - Value *quotient = builder.create(loc, dividend, rhs); + Value *quotient = builder.create(loc, dividend, rhs); Value *negatedQuotient = builder.create(loc, zeroCst, quotient); Value *incrementedQuotient = builder.create(loc, quotient, oneCst); Value *result = builder.create(loc, nonPositive, negatedQuotient, diff --git a/mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp b/mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp index c269dc5..d663ae1 100644 --- a/mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp +++ b/mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp @@ -254,7 +254,7 @@ Optional LoopToGpuConverter::collectBounds(OpTy forOp, builder.create(currentLoop.getLoc(), upperBound, lowerBound); Value *step = getOrCreateStep(currentLoop, builder); if (!isConstantOne(step)) - range = builder.create(currentLoop.getLoc(), range, step); + range = builder.create(currentLoop.getLoc(), range, step); dims.push_back(range); lbs.push_back(lowerBound); diff --git a/mlir/lib/Conversion/StandardToLLVM/ConvertStandardToLLVM.cpp b/mlir/lib/Conversion/StandardToLLVM/ConvertStandardToLLVM.cpp index ea8501b..fdc9085 100644 --- a/mlir/lib/Conversion/StandardToLLVM/ConvertStandardToLLVM.cpp +++ b/mlir/lib/Conversion/StandardToLLVM/ConvertStandardToLLVM.cpp @@ -814,16 +814,20 @@ struct SubIOpLowering : public BinaryOpLLVMOpLowering { struct MulIOpLowering : public BinaryOpLLVMOpLowering { using Super::Super; }; -struct DivISOpLowering : public BinaryOpLLVMOpLowering { +struct SignedDivIOpLowering + : public BinaryOpLLVMOpLowering { using Super::Super; }; -struct DivIUOpLowering : public BinaryOpLLVMOpLowering { +struct UnsignedDivIOpLowering + : public BinaryOpLLVMOpLowering { using Super::Super; }; -struct RemISOpLowering : public BinaryOpLLVMOpLowering { +struct SignedRemIOpLowering + : public BinaryOpLLVMOpLowering { using Super::Super; }; -struct RemIUOpLowering : public BinaryOpLLVMOpLowering { +struct UnsignedRemIOpLowering + : public BinaryOpLLVMOpLowering { using Super::Super; }; struct AndOpLowering : public BinaryOpLLVMOpLowering { @@ -862,6 +866,18 @@ struct ConstLLVMOpLowering : public OneToOneLLVMOpLowering { using Super::Super; }; +struct ShiftLeftOpLowering + : public OneToOneLLVMOpLowering { + using Super::Super; +}; +struct SignedShiftRightOpLowering + : public OneToOneLLVMOpLowering { + using Super::Super; +}; +struct UnsignedShiftRightOpLowering + : public OneToOneLLVMOpLowering { + using Super::Super; +}; // Check if the MemRefType `type` is supported by the lowering. We currently // only support memrefs with identity maps. @@ -2082,8 +2098,6 @@ void mlir::populateStdToLLVMNonMemoryConversionPatterns( CosOpLowering, ConstLLVMOpLowering, DivFOpLowering, - DivISOpLowering, - DivIUOpLowering, ExpOpLowering, LogOpLowering, Log10OpLowering, @@ -2097,18 +2111,23 @@ void mlir::populateStdToLLVMNonMemoryConversionPatterns( OrOpLowering, PrefetchOpLowering, RemFOpLowering, - RemISOpLowering, - RemIUOpLowering, ReturnOpLowering, SIToFPLowering, SelectOpLowering, + ShiftLeftOpLowering, SignExtendIOpLowering, + SignedDivIOpLowering, + SignedRemIOpLowering, + SignedShiftRightOpLowering, SplatOpLowering, SplatNdOpLowering, SubFOpLowering, SubIOpLowering, TanhOpLowering, TruncateIOpLowering, + UnsignedDivIOpLowering, + UnsignedRemIOpLowering, + UnsignedShiftRightOpLowering, XOrOpLowering, ZeroExtendIOpLowering>(*converter.getDialect(), converter); // clang-format on diff --git a/mlir/lib/Conversion/StandardToSPIRV/ConvertStandardToSPIRV.cpp b/mlir/lib/Conversion/StandardToSPIRV/ConvertStandardToSPIRV.cpp index e87bd4e..a14271e 100644 --- a/mlir/lib/Conversion/StandardToSPIRV/ConvertStandardToSPIRV.cpp +++ b/mlir/lib/Conversion/StandardToSPIRV/ConvertStandardToSPIRV.cpp @@ -316,8 +316,8 @@ void populateStandardToSPIRVPatterns(MLIRContext *context, patterns.insert, IntegerOpConversion, - IntegerOpConversion, - IntegerOpConversion, + IntegerOpConversion, + IntegerOpConversion, IntegerOpConversion, LoadOpConversion, ReturnOpConversion, SelectOpConversion, StoreOpConversion>( context, typeConverter); diff --git a/mlir/lib/Dialect/StandardOps/Ops.cpp b/mlir/lib/Dialect/StandardOps/Ops.cpp index d0fd185..4116f6f 100644 --- a/mlir/lib/Dialect/StandardOps/Ops.cpp +++ b/mlir/lib/Dialect/StandardOps/Ops.cpp @@ -1320,10 +1320,10 @@ OpFoldResult DimOp::fold(ArrayRef operands) { } //===----------------------------------------------------------------------===// -// DivISOp +// SignedDivIOp //===----------------------------------------------------------------------===// -OpFoldResult DivISOp::fold(ArrayRef operands) { +OpFoldResult SignedDivIOp::fold(ArrayRef operands) { assert(operands.size() == 2 && "binary operation takes two operands"); // Don't fold if it would overflow or if it requires a division by zero. @@ -1339,10 +1339,10 @@ OpFoldResult DivISOp::fold(ArrayRef operands) { } //===----------------------------------------------------------------------===// -// DivIUOp +// UnsignedDivIOp //===----------------------------------------------------------------------===// -OpFoldResult DivIUOp::fold(ArrayRef operands) { +OpFoldResult UnsignedDivIOp::fold(ArrayRef operands) { assert(operands.size() == 2 && "binary operation takes two operands"); // Don't fold if it would require a division by zero. @@ -1885,11 +1885,11 @@ OpFoldResult RankOp::fold(ArrayRef operands) { } //===----------------------------------------------------------------------===// -// RemISOp +// SignedRemIOp //===----------------------------------------------------------------------===// -OpFoldResult RemISOp::fold(ArrayRef operands) { - assert(operands.size() == 2 && "remis takes two operands"); +OpFoldResult SignedRemIOp::fold(ArrayRef operands) { + assert(operands.size() == 2 && "remi_signed takes two operands"); auto rhs = operands.back().dyn_cast_or_null(); if (!rhs) @@ -1911,11 +1911,11 @@ OpFoldResult RemISOp::fold(ArrayRef operands) { } //===----------------------------------------------------------------------===// -// RemIUOp +// UnsignedRemIOp //===----------------------------------------------------------------------===// -OpFoldResult RemIUOp::fold(ArrayRef operands) { - assert(operands.size() == 2 && "remiu takes two operands"); +OpFoldResult UnsignedRemIOp::fold(ArrayRef operands) { + assert(operands.size() == 2 && "remi_unsigned takes two operands"); auto rhs = operands.back().dyn_cast_or_null(); if (!rhs) diff --git a/mlir/lib/EDSC/Builders.cpp b/mlir/lib/EDSC/Builders.cpp index 2956066..47e2dfe 100644 --- a/mlir/lib/EDSC/Builders.cpp +++ b/mlir/lib/EDSC/Builders.cpp @@ -390,14 +390,14 @@ ValueHandle mlir::edsc::op::operator*(ValueHandle lhs, ValueHandle rhs) { } ValueHandle mlir::edsc::op::operator/(ValueHandle lhs, ValueHandle rhs) { - return createBinaryHandle( + return createBinaryHandle( lhs, rhs, [](AffineExpr d0, AffineExpr d1) -> AffineExpr { llvm_unreachable("only exprs of non-index type support operator/"); }); } ValueHandle mlir::edsc::op::operator%(ValueHandle lhs, ValueHandle rhs) { - return createBinaryHandle( + return createBinaryHandle( lhs, rhs, [](AffineExpr d0, AffineExpr d1) { return d0 % d1; }); } diff --git a/mlir/lib/Transforms/Utils/LoopUtils.cpp b/mlir/lib/Transforms/Utils/LoopUtils.cpp index 419df8d..3691aee 100644 --- a/mlir/lib/Transforms/Utils/LoopUtils.cpp +++ b/mlir/lib/Transforms/Utils/LoopUtils.cpp @@ -850,7 +850,7 @@ static Value *ceilDivPositive(OpBuilder &builder, Location loc, Value *dividend, Value *divisorMinusOneCst = builder.create(loc, divisor - 1); Value *divisorCst = builder.create(loc, divisor); Value *sum = builder.create(loc, dividend, divisorMinusOneCst); - return builder.create(loc, sum, divisorCst); + return builder.create(loc, sum, divisorCst); } // Build the IR that performs ceil division of a positive value by another @@ -864,7 +864,7 @@ static Value *ceilDivPositive(OpBuilder &builder, Location loc, Value *dividend, Value *cstOne = builder.create(loc, 1); Value *divisorMinusOne = builder.create(loc, divisor, cstOne); Value *sum = builder.create(loc, dividend, divisorMinusOne); - return builder.create(loc, sum, divisor); + return builder.create(loc, sum, divisor); } // Hoist the ops within `outer` that appear before `inner`. @@ -1084,12 +1084,12 @@ void mlir::coalesceLoops(MutableArrayRef loops) { for (unsigned i = 0, e = loops.size(); i < e; ++i) { unsigned idx = loops.size() - i - 1; if (i != 0) - previous = - builder.create(loc, previous, loops[idx + 1].upperBound()); + previous = builder.create(loc, previous, + loops[idx + 1].upperBound()); Value *iv = (i == e - 1) ? previous - : builder.create(loc, previous, - loops[idx].upperBound()); + : builder.create( + loc, previous, loops[idx].upperBound()); replaceAllUsesInRegionWith(loops[idx].getInductionVar(), iv, loops.back().region()); } diff --git a/mlir/test/Conversion/LoopsToGPU/imperfect_3D.mlir b/mlir/test/Conversion/LoopsToGPU/imperfect_3D.mlir index 4741c38..73f0ab7 100644 --- a/mlir/test/Conversion/LoopsToGPU/imperfect_3D.mlir +++ b/mlir/test/Conversion/LoopsToGPU/imperfect_3D.mlir @@ -49,11 +49,11 @@ module { %8 = load %arg1[%iv4, %iv6, %iv5] : memref %9 = addf %7, %8 : f32 %10 = subi %iv4, %iv1 : index - %11 = divis %10, %step1 : index + %11 = divi_signed %10, %step1 : index %12 = subi %iv5, %iv2 : index - %13 = divis %12, %step2 : index + %13 = divi_signed %12, %step2 : index %14 = subi %iv6, %iv3 : index - %15 = divis %14, %step3 : index + %15 = divi_signed %14, %step3 : index store %9, %6[%11, %13, %15] : memref } } @@ -62,11 +62,11 @@ module { loop.for %iv8 = %iv2 to %ub2 step %step2 { loop.for %iv9 = %iv3 to %ub3 step %step3 { %16 = subi %iv7, %iv1 : index - %17 = divis %16, %step1 : index + %17 = divi_signed %16, %step1 : index %18 = subi %iv8, %iv2 : index - %19 = divis %18, %step2 : index + %19 = divi_signed %18, %step2 : index %20 = subi %iv9, %iv3 : index - %21 = divis %20, %step3 : index + %21 = divi_signed %20, %step3 : index %22 = load %6[%17, %19, %21] : memref %23 = load %arg2[%iv9, %iv8, %iv7] : memref %24 = mulf %22, %23 : f32 diff --git a/mlir/test/Conversion/LoopsToGPU/imperfect_4D.mlir b/mlir/test/Conversion/LoopsToGPU/imperfect_4D.mlir index 2753cd2..2c5dd5c 100644 --- a/mlir/test/Conversion/LoopsToGPU/imperfect_4D.mlir +++ b/mlir/test/Conversion/LoopsToGPU/imperfect_4D.mlir @@ -49,11 +49,11 @@ module { %8 = load %arg1[%iv5, %iv6, %iv7, %iv8] : memref %9 = addf %7, %8 : f32 %10 = subi %iv5, %iv1 : index - %11 = divis %10, %step1 : index + %11 = divi_signed %10, %step1 : index %12 = subi %iv6, %iv2 : index - %13 = divis %12, %step2 : index + %13 = divi_signed %12, %step2 : index %14 = subi %iv7, %iv3 : index - %15 = divis %14, %step3 : index + %15 = divi_signed %14, %step3 : index store %9, %6[%11, %13, %15, %iv8] : memref } } @@ -64,11 +64,11 @@ module { loop.for %iv11 = %iv3 to %ub3 step %step3 { loop.for %iv12 = %c0 to %3 step %step4 { %18 = subi %iv9, %iv1 : index - %19 = divis %18, %step1 : index + %19 = divi_signed %18, %step1 : index %20 = subi %iv10, %iv2 : index - %21 = divis %20, %step2 : index + %21 = divi_signed %20, %step2 : index %22 = subi %iv11, %iv3 : index - %23 = divis %22, %step3 : index + %23 = divi_signed %22, %step3 : index %26 = load %6[%19, %21, %23, %iv12] : memref %27 = load %arg2[%iv9, %iv10, %iv12, %iv11] : memref %28 = mulf %26, %27 : f32 diff --git a/mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir b/mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir index a8a2d2d..f4567fb 100644 --- a/mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir +++ b/mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir @@ -6,10 +6,10 @@ func @foo(%arg0: memref, %arg1 : index) { %c42 = constant 42 : index %c3 = constant 3 : index // CHECK: subi %{{.*}}, %{{.*}} : index - // CHECK-NEXT: %[[range_i:.*]] = divis {{.*}}, %{{.*}} : index + // CHECK-NEXT: %[[range_i:.*]] = divi_signed {{.*}}, %{{.*}} : index loop.for %i0 = %c0 to %c42 step %c3 { // CHECK: subi %{{.*}}, %{{.*}} : index - // CHECK-NEXT: %[[range_j:.*]] = divis {{.*}}, %{{.*}} : index + // CHECK-NEXT: %[[range_j:.*]] = divi_signed {{.*}}, %{{.*}} : index loop.for %i1 = %c3 to %c42 step %arg1 { // CHECK: gpu.launch // CHECK-SAME: blocks diff --git a/mlir/test/Conversion/LoopsToGPU/step_positive.mlir b/mlir/test/Conversion/LoopsToGPU/step_positive.mlir index dd22f99..6bedc92 100644 --- a/mlir/test/Conversion/LoopsToGPU/step_positive.mlir +++ b/mlir/test/Conversion/LoopsToGPU/step_positive.mlir @@ -3,8 +3,8 @@ // CHECK-LABEL: @step_var func @step_var(%A : memref, %B : memref) { // Check that we divide by step. - // CHECK: %[[range_i:.*]] = divis {{.*}}, %{{.*}} - // CHECK: %[[range_j:.*]] = divis {{.*}}, %{{.*}} + // CHECK: %[[range_i:.*]] = divi_signed {{.*}}, %{{.*}} + // CHECK: %[[range_j:.*]] = divi_signed {{.*}}, %{{.*}} // CHECK: gpu.launch // CHECK-SAME: blocks(%{{[^)]*}}, %{{[^)]*}}, %{{[^)]*}}) in (%{{[^)]*}} = %[[range_i]], %{{[^)]*}} = %{{[^)]*}}, %{{[^)]*}} = %{{[^)]*}}) diff --git a/mlir/test/Conversion/StandardToLLVM/convert-to-llvmir.mlir b/mlir/test/Conversion/StandardToLLVM/convert-to-llvmir.mlir index 14e46aa..25054bb 100644 --- a/mlir/test/Conversion/StandardToLLVM/convert-to-llvmir.mlir +++ b/mlir/test/Conversion/StandardToLLVM/convert-to-llvmir.mlir @@ -371,13 +371,13 @@ func @vector_ops(%arg0: vector<4xf32>, %arg1: vector<4xi1>, %arg2: vector<4xi64> // CHECK-NEXT: %1 = llvm.fadd %arg0, %0 : !llvm<"<4 x float>"> %1 = addf %arg0, %0 : vector<4xf32> // CHECK-NEXT: %2 = llvm.sdiv %arg2, %arg2 : !llvm<"<4 x i64>"> - %3 = divis %arg2, %arg2 : vector<4xi64> + %3 = divi_signed %arg2, %arg2 : vector<4xi64> // CHECK-NEXT: %3 = llvm.udiv %arg2, %arg2 : !llvm<"<4 x i64>"> - %4 = diviu %arg2, %arg2 : vector<4xi64> + %4 = divi_unsigned %arg2, %arg2 : vector<4xi64> // CHECK-NEXT: %4 = llvm.srem %arg2, %arg2 : !llvm<"<4 x i64>"> - %5 = remis %arg2, %arg2 : vector<4xi64> + %5 = remi_signed %arg2, %arg2 : vector<4xi64> // CHECK-NEXT: %5 = llvm.urem %arg2, %arg2 : !llvm<"<4 x i64>"> - %6 = remiu %arg2, %arg2 : vector<4xi64> + %6 = remi_unsigned %arg2, %arg2 : vector<4xi64> // CHECK-NEXT: %6 = llvm.fdiv %arg0, %0 : !llvm<"<4 x float>"> %7 = divf %arg0, %0 : vector<4xf32> // CHECK-NEXT: %7 = llvm.frem %arg0, %0 : !llvm<"<4 x float>"> @@ -388,6 +388,12 @@ func @vector_ops(%arg0: vector<4xf32>, %arg1: vector<4xi1>, %arg2: vector<4xi64> %10 = or %arg2, %arg3 : vector<4xi64> // CHECK-NEXT: %10 = llvm.xor %arg2, %arg3 : !llvm<"<4 x i64>"> %11 = xor %arg2, %arg3 : vector<4xi64> +// CHECK-NEXT: %11 = llvm.shl %arg2, %arg2 : !llvm<"<4 x i64>"> + %12 = shift_left %arg2, %arg2 : vector<4xi64> +// CHECK-NEXT: %12 = llvm.ashr %arg2, %arg2 : !llvm<"<4 x i64>"> + %13 = shift_right_signed %arg2, %arg2 : vector<4xi64> +// CHECK-NEXT: %13 = llvm.lshr %arg2, %arg2 : !llvm<"<4 x i64>"> + %14 = shift_right_unsigned %arg2, %arg2 : vector<4xi64> return %1 : vector<4xf32> } @@ -401,13 +407,13 @@ func @ops(f32, f32, i32, i32) -> (f32, i32) { // CHECK-NEXT: %2 = llvm.icmp "slt" %arg2, %1 : !llvm.i32 %2 = cmpi "slt", %arg2, %1 : i32 // CHECK-NEXT: %3 = llvm.sdiv %arg2, %arg3 : !llvm.i32 - %4 = divis %arg2, %arg3 : i32 + %4 = divi_signed %arg2, %arg3 : i32 // CHECK-NEXT: %4 = llvm.udiv %arg2, %arg3 : !llvm.i32 - %5 = diviu %arg2, %arg3 : i32 + %5 = divi_unsigned %arg2, %arg3 : i32 // CHECK-NEXT: %5 = llvm.srem %arg2, %arg3 : !llvm.i32 - %6 = remis %arg2, %arg3 : i32 + %6 = remi_signed %arg2, %arg3 : i32 // CHECK-NEXT: %6 = llvm.urem %arg2, %arg3 : !llvm.i32 - %7 = remiu %arg2, %arg3 : i32 + %7 = remi_unsigned %arg2, %arg3 : i32 // CHECK-NEXT: %7 = llvm.select %2, %arg2, %arg3 : !llvm.i1, !llvm.i32 %8 = select %2, %arg2, %arg3 : i32 // CHECK-NEXT: %8 = llvm.fdiv %arg0, %arg1 : !llvm.float @@ -428,6 +434,12 @@ func @ops(f32, f32, i32, i32) -> (f32, i32) { %16 = constant 7.9e-01 : f64 // CHECK-NEXT: %16 = llvm.call @tanh(%15) : (!llvm.double) -> !llvm.double %17 = std.tanh %16 : f64 +// CHECK-NEXT: %17 = llvm.shl %arg2, %arg3 : !llvm.i32 + %18 = shift_left %arg2, %arg3 : i32 +// CHECK-NEXT: %18 = llvm.ashr %arg2, %arg3 : !llvm.i32 + %19 = shift_right_signed %arg2, %arg3 : i32 +// CHECK-NEXT: %19 = llvm.lshr %arg2, %arg3 : !llvm.i32 + %20 = shift_right_unsigned %arg2, %arg3 : i32 return %0, %4 : f32, i32 } diff --git a/mlir/test/Conversion/StandardToSPIRV/std-to-spirv.mlir b/mlir/test/Conversion/StandardToSPIRV/std-to-spirv.mlir index c113493..8876c51 100644 --- a/mlir/test/Conversion/StandardToSPIRV/std-to-spirv.mlir +++ b/mlir/test/Conversion/StandardToSPIRV/std-to-spirv.mlir @@ -88,9 +88,9 @@ func @fsub_scalar(%arg: f32) -> f32 { // CHECK-LABEL: @div_rem func @div_rem(%arg0 : i32, %arg1 : i32) { // CHECK: spv.SDiv - %0 = divis %arg0, %arg1 : i32 + %0 = divi_signed %arg0, %arg1 : i32 // CHECK: spv.SMod - %1 = remis %arg0, %arg1 : i32 + %1 = remi_signed %arg0, %arg1 : i32 return } diff --git a/mlir/test/EDSC/builder-api-test.cpp b/mlir/test/EDSC/builder-api-test.cpp index 81bb0b94..0b105ea 100644 --- a/mlir/test/EDSC/builder-api-test.cpp +++ b/mlir/test/EDSC/builder-api-test.cpp @@ -101,8 +101,8 @@ TEST_FUNC(builder_dynamic_for_func_args) { // CHECK-DAG: [[rf4:%[0-9]+]] = mulf {{.*}}, {{.*}} : f32 // CHECK: {{.*}} = subf [[rf3]], [[rf4]] : f32 // CHECK-DAG: [[ri1:%[0-9]+]] = addi {{.*}}, {{.*}} : i32 - // CHECK-DAG: [[ri2:%[0-9]+]] = divis [[ri1]], {{.*}} : i32 - // CHECK-DAG: [[ri3:%[0-9]+]] = remis [[ri2]], {{.*}} : i32 + // CHECK-DAG: [[ri2:%[0-9]+]] = divi_signed [[ri1]], {{.*}} : i32 + // CHECK-DAG: [[ri3:%[0-9]+]] = remi_signed [[ri2]], {{.*}} : i32 // CHECK-DAG: [[ri4:%[0-9]+]] = muli {{.*}}, {{.*}} : i32 // CHECK: {{.*}} = subi [[ri3]], [[ri4]] : i32 // clang-format on diff --git a/mlir/test/IR/core-ops.mlir b/mlir/test/IR/core-ops.mlir index c3a9ee4..aac0dfc 100644 --- a/mlir/test/IR/core-ops.mlir +++ b/mlir/test/IR/core-ops.mlir @@ -153,65 +153,65 @@ func @standard_instrs(tensor<4x4x?xf32>, f32, i32, index, i64, f16) { // CHECK: %{{[0-9]+}} = select %{{[0-9]+}}, %cst_4, %cst_4 : tensor<42xi32> %25 = "std.select"(%19, %tci32, %tci32) : (tensor<42 x i1>, tensor<42 x i32>, tensor<42 x i32>) -> tensor<42 x i32> - // CHECK: %{{[0-9]+}} = divis %arg2, %arg2 : i32 - %26 = divis %i, %i : i32 + // CHECK: %{{[0-9]+}} = divi_signed %arg2, %arg2 : i32 + %26 = divi_signed %i, %i : i32 - // CHECK: %{{[0-9]+}} = divis %arg3, %arg3 : index - %27 = divis %idx, %idx : index + // CHECK: %{{[0-9]+}} = divi_signed %arg3, %arg3 : index + %27 = divi_signed %idx, %idx : index - // CHECK: %{{[0-9]+}} = divis %cst_5, %cst_5 : vector<42xi32> - %28 = divis %vci32, %vci32 : vector<42 x i32> + // CHECK: %{{[0-9]+}} = divi_signed %cst_5, %cst_5 : vector<42xi32> + %28 = divi_signed %vci32, %vci32 : vector<42 x i32> - // CHECK: %{{[0-9]+}} = divis %cst_4, %cst_4 : tensor<42xi32> - %29 = divis %tci32, %tci32 : tensor<42 x i32> + // CHECK: %{{[0-9]+}} = divi_signed %cst_4, %cst_4 : tensor<42xi32> + %29 = divi_signed %tci32, %tci32 : tensor<42 x i32> - // CHECK: %{{[0-9]+}} = divis %arg2, %arg2 : i32 - %30 = "std.divis"(%i, %i) : (i32, i32) -> i32 + // CHECK: %{{[0-9]+}} = divi_signed %arg2, %arg2 : i32 + %30 = "std.divi_signed"(%i, %i) : (i32, i32) -> i32 - // CHECK: %{{[0-9]+}} = diviu %arg2, %arg2 : i32 - %31 = diviu %i, %i : i32 + // CHECK: %{{[0-9]+}} = divi_unsigned %arg2, %arg2 : i32 + %31 = divi_unsigned %i, %i : i32 - // CHECK: %{{[0-9]+}} = diviu %arg3, %arg3 : index - %32 = diviu %idx, %idx : index + // CHECK: %{{[0-9]+}} = divi_unsigned %arg3, %arg3 : index + %32 = divi_unsigned %idx, %idx : index - // CHECK: %{{[0-9]+}} = diviu %cst_5, %cst_5 : vector<42xi32> - %33 = diviu %vci32, %vci32 : vector<42 x i32> + // CHECK: %{{[0-9]+}} = divi_unsigned %cst_5, %cst_5 : vector<42xi32> + %33 = divi_unsigned %vci32, %vci32 : vector<42 x i32> - // CHECK: %{{[0-9]+}} = diviu %cst_4, %cst_4 : tensor<42xi32> - %34 = diviu %tci32, %tci32 : tensor<42 x i32> + // CHECK: %{{[0-9]+}} = divi_unsigned %cst_4, %cst_4 : tensor<42xi32> + %34 = divi_unsigned %tci32, %tci32 : tensor<42 x i32> - // CHECK: %{{[0-9]+}} = diviu %arg2, %arg2 : i32 - %35 = "std.diviu"(%i, %i) : (i32, i32) -> i32 + // CHECK: %{{[0-9]+}} = divi_unsigned %arg2, %arg2 : i32 + %35 = "std.divi_unsigned"(%i, %i) : (i32, i32) -> i32 - // CHECK: %{{[0-9]+}} = remis %arg2, %arg2 : i32 - %36 = remis %i, %i : i32 + // CHECK: %{{[0-9]+}} = remi_signed %arg2, %arg2 : i32 + %36 = remi_signed %i, %i : i32 - // CHECK: %{{[0-9]+}} = remis %arg3, %arg3 : index - %37 = remis %idx, %idx : index + // CHECK: %{{[0-9]+}} = remi_signed %arg3, %arg3 : index + %37 = remi_signed %idx, %idx : index - // CHECK: %{{[0-9]+}} = remis %cst_5, %cst_5 : vector<42xi32> - %38 = remis %vci32, %vci32 : vector<42 x i32> + // CHECK: %{{[0-9]+}} = remi_signed %cst_5, %cst_5 : vector<42xi32> + %38 = remi_signed %vci32, %vci32 : vector<42 x i32> - // CHECK: %{{[0-9]+}} = remis %cst_4, %cst_4 : tensor<42xi32> - %39 = remis %tci32, %tci32 : tensor<42 x i32> + // CHECK: %{{[0-9]+}} = remi_signed %cst_4, %cst_4 : tensor<42xi32> + %39 = remi_signed %tci32, %tci32 : tensor<42 x i32> - // CHECK: %{{[0-9]+}} = remis %arg2, %arg2 : i32 - %40 = "std.remis"(%i, %i) : (i32, i32) -> i32 + // CHECK: %{{[0-9]+}} = remi_signed %arg2, %arg2 : i32 + %40 = "std.remi_signed"(%i, %i) : (i32, i32) -> i32 - // CHECK: %{{[0-9]+}} = remiu %arg2, %arg2 : i32 - %41 = remiu %i, %i : i32 + // CHECK: %{{[0-9]+}} = remi_unsigned %arg2, %arg2 : i32 + %41 = remi_unsigned %i, %i : i32 - // CHECK: %{{[0-9]+}} = remiu %arg3, %arg3 : index - %42 = remiu %idx, %idx : index + // CHECK: %{{[0-9]+}} = remi_unsigned %arg3, %arg3 : index + %42 = remi_unsigned %idx, %idx : index - // CHECK: %{{[0-9]+}} = remiu %cst_5, %cst_5 : vector<42xi32> - %43 = remiu %vci32, %vci32 : vector<42 x i32> + // CHECK: %{{[0-9]+}} = remi_unsigned %cst_5, %cst_5 : vector<42xi32> + %43 = remi_unsigned %vci32, %vci32 : vector<42 x i32> - // CHECK: %{{[0-9]+}} = remiu %cst_4, %cst_4 : tensor<42xi32> - %44 = remiu %tci32, %tci32 : tensor<42 x i32> + // CHECK: %{{[0-9]+}} = remi_unsigned %cst_4, %cst_4 : tensor<42xi32> + %44 = remi_unsigned %tci32, %tci32 : tensor<42 x i32> - // CHECK: %{{[0-9]+}} = remiu %arg2, %arg2 : i32 - %45 = "std.remiu"(%i, %i) : (i32, i32) -> i32 + // CHECK: %{{[0-9]+}} = remi_unsigned %arg2, %arg2 : i32 + %45 = "std.remi_unsigned"(%i, %i) : (i32, i32) -> i32 // CHECK: %{{[0-9]+}} = divf %arg1, %arg1 : f32 %46 = "std.divf"(%f, %f) : (f32,f32) -> f32 @@ -448,6 +448,52 @@ func @standard_instrs(tensor<4x4x?xf32>, f32, i32, index, i64, f16) { // CHECK: %{{[0-9]+}} = tanh %arg0 : tensor<4x4x?xf32> %123 = tanh %t : tensor<4x4x?xf32> + + // CHECK: %{{[0-9]+}} = shift_left %arg2, %arg2 : i32 + %124 = "std.shift_left"(%i, %i) : (i32, i32) -> i32 + + // CHECK:%{{[0-9]+}} = shift_left %4, %4 : i32 + %125 = shift_left %i2, %i2 : i32 + + // CHECK: %{{[0-9]+}} = shift_left %arg3, %arg3 : index + %126 = shift_left %idx, %idx : index + + // CHECK: %{{[0-9]+}} = shift_left %cst_5, %cst_5 : vector<42xi32> + %127 = shift_left %vci32, %vci32 : vector<42 x i32> + + // CHECK: %{{[0-9]+}} = shift_left %cst_4, %cst_4 : tensor<42xi32> + %128 = shift_left %tci32, %tci32 : tensor<42 x i32> + + // CHECK: %{{[0-9]+}} = shift_right_signed %arg2, %arg2 : i32 + %129 = "std.shift_right_signed"(%i, %i) : (i32, i32) -> i32 + + // CHECK:%{{[0-9]+}} = shift_right_signed %4, %4 : i32 + %130 = shift_right_signed %i2, %i2 : i32 + + // CHECK: %{{[0-9]+}} = shift_right_signed %arg3, %arg3 : index + %131 = shift_right_signed %idx, %idx : index + + // CHECK: %{{[0-9]+}} = shift_right_signed %cst_5, %cst_5 : vector<42xi32> + %132 = shift_right_signed %vci32, %vci32 : vector<42 x i32> + + // CHECK: %{{[0-9]+}} = shift_right_signed %cst_4, %cst_4 : tensor<42xi32> + %133 = shift_right_signed %tci32, %tci32 : tensor<42 x i32> + + // CHECK: %{{[0-9]+}} = shift_right_unsigned %arg2, %arg2 : i32 + %134 = "std.shift_right_unsigned"(%i, %i) : (i32, i32) -> i32 + + // CHECK:%{{[0-9]+}} = shift_right_unsigned %4, %4 : i32 + %135 = shift_right_unsigned %i2, %i2 : i32 + + // CHECK: %{{[0-9]+}} = shift_right_unsigned %arg3, %arg3 : index + %136 = shift_right_unsigned %idx, %idx : index + + // CHECK: %{{[0-9]+}} = shift_right_unsigned %cst_5, %cst_5 : vector<42xi32> + %137 = shift_right_unsigned %vci32, %vci32 : vector<42 x i32> + + // CHECK: %{{[0-9]+}} = shift_right_unsigned %cst_4, %cst_4 : tensor<42xi32> + %138 = shift_right_unsigned %tci32, %tci32 : tensor<42 x i32> + return } diff --git a/mlir/test/Target/llvmir.mlir b/mlir/test/Target/llvmir.mlir index 0ad0073..9d0ee38 100644 --- a/mlir/test/Target/llvmir.mlir +++ b/mlir/test/Target/llvmir.mlir @@ -794,6 +794,12 @@ llvm.func @vector_ops(%arg0: !llvm<"<4 x float>">, %arg1: !llvm<"<4 x i1>">, %ar %10 = llvm.or %arg2, %arg2 : !llvm<"<4 x i64>"> // CHECK-NEXT: %14 = xor <4 x i64> %2, %2 %11 = llvm.xor %arg2, %arg2 : !llvm<"<4 x i64>"> +// CHECK-NEXT: %15 = shl <4 x i64> %2, %2 + %12 = llvm.shl %arg2, %arg2 : !llvm<"<4 x i64>"> +// CHECK-NEXT: %16 = lshr <4 x i64> %2, %2 + %13 = llvm.lshr %arg2, %arg2 : !llvm<"<4 x i64>"> +// CHECK-NEXT: %17 = ashr <4 x i64> %2, %2 + %14 = llvm.ashr %arg2, %arg2 : !llvm<"<4 x i64>"> // CHECK-NEXT: ret <4 x float> %4 llvm.return %1 : !llvm<"<4 x float>"> } diff --git a/mlir/test/Transforms/canonicalize.mlir b/mlir/test/Transforms/canonicalize.mlir index 07a7e7c..09db088 100644 --- a/mlir/test/Transforms/canonicalize.mlir +++ b/mlir/test/Transforms/canonicalize.mlir @@ -576,7 +576,7 @@ func @lowered_affine_mod() -> (index, index) { // CHECK-NEXT: {{.*}} = constant 41 : index %c-43 = constant -43 : index %c42 = constant 42 : index - %0 = remis %c-43, %c42 : index + %0 = remi_signed %c-43, %c42 : index %c0 = constant 0 : index %1 = cmpi "slt", %0, %c0 : index %2 = addi %0, %c42 : index @@ -584,7 +584,7 @@ func @lowered_affine_mod() -> (index, index) { // CHECK-NEXT: {{.*}} = constant 1 : index %c43 = constant 43 : index %c42_0 = constant 42 : index - %4 = remis %c43, %c42_0 : index + %4 = remi_signed %c43, %c42_0 : index %c0_1 = constant 0 : index %5 = cmpi "slt", %4, %c0_1 : index %6 = addi %4, %c42_0 : index @@ -607,7 +607,7 @@ func @lowered_affine_floordiv() -> (index, index) { %0 = cmpi "slt", %c-43, %c0 : index %1 = subi %c-1, %c-43 : index %2 = select %0, %1, %c-43 : index - %3 = divis %2, %c42 : index + %3 = divi_signed %2, %c42 : index %4 = subi %c-1, %3 : index %5 = select %0, %4, %3 : index // CHECK-NEXT: %c1 = constant 1 : index @@ -618,7 +618,7 @@ func @lowered_affine_floordiv() -> (index, index) { %6 = cmpi "slt", %c43, %c0_1 : index %7 = subi %c-1_2, %c43 : index %8 = select %6, %7, %c43 : index - %9 = divis %8, %c42_0 : index + %9 = divi_signed %8, %c42_0 : index %10 = subi %c-1_2, %9 : index %11 = select %6, %10, %9 : index return %5, %11 : index, index @@ -640,7 +640,7 @@ func @lowered_affine_ceildiv() -> (index, index) { %1 = subi %c0, %c-43 : index %2 = subi %c-43, %c1 : index %3 = select %0, %1, %2 : index - %4 = divis %3, %c42 : index + %4 = divi_signed %3, %c42 : index %5 = subi %c0, %4 : index %6 = addi %4, %c1 : index %7 = select %0, %5, %6 : index @@ -653,7 +653,7 @@ func @lowered_affine_ceildiv() -> (index, index) { %9 = subi %c0_1, %c43 : index %10 = subi %c43, %c1_2 : index %11 = select %8, %9, %10 : index - %12 = divis %11, %c42_0 : index + %12 = divi_signed %11, %c42_0 : index %13 = subi %c0_1, %12 : index %14 = addi %12, %c1_2 : index %15 = select %8, %13, %14 : index diff --git a/mlir/test/Transforms/constant-fold.mlir b/mlir/test/Transforms/constant-fold.mlir index b45fec6..a24aad2 100644 --- a/mlir/test/Transforms/constant-fold.mlir +++ b/mlir/test/Transforms/constant-fold.mlir @@ -212,8 +212,8 @@ func @mulf_splat_tensor() -> tensor<4xf32> { // ----- -// CHECK-LABEL: func @simple_divis -func @simple_divis() -> (i32, i32, i32) { +// CHECK-LABEL: func @simple_divi_signed +func @simple_divi_signed() -> (i32, i32, i32) { // CHECK-DAG: [[C0:%.+]] = constant 0 %z = constant 0 : i32 // CHECK-DAG: [[C6:%.+]] = constant 6 @@ -221,15 +221,15 @@ func @simple_divis() -> (i32, i32, i32) { %1 = constant 2 : i32 // CHECK-NEXT: [[C3:%.+]] = constant 3 : i32 - %2 = divis %0, %1 : i32 + %2 = divi_signed %0, %1 : i32 %3 = constant -2 : i32 // CHECK-NEXT: [[CM3:%.+]] = constant -3 : i32 - %4 = divis %0, %3 : i32 + %4 = divi_signed %0, %3 : i32 - // CHECK-NEXT: [[XZ:%.+]] = divis [[C6]], [[C0]] - %5 = divis %0, %z : i32 + // CHECK-NEXT: [[XZ:%.+]] = divi_signed [[C6]], [[C0]] + %5 = divi_signed %0, %z : i32 // CHECK-NEXT: return [[C3]], [[CM3]], [[XZ]] return %2, %4, %5 : i32, i32, i32 @@ -237,8 +237,8 @@ func @simple_divis() -> (i32, i32, i32) { // ----- -// CHECK-LABEL: func @divis_splat_tensor -func @divis_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { +// CHECK-LABEL: func @divi_signed_splat_tensor +func @divi_signed_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { // CHECK-DAG: [[C0:%.+]] = constant dense<0> %z = constant dense<0> : tensor<4xi32> // CHECK-DAG: [[C6:%.+]] = constant dense<6> @@ -246,15 +246,15 @@ func @divis_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { %1 = constant dense<2> : tensor<4xi32> // CHECK-NEXT: [[C3:%.+]] = constant dense<3> : tensor<4xi32> - %2 = divis %0, %1 : tensor<4xi32> + %2 = divi_signed %0, %1 : tensor<4xi32> %3 = constant dense<-2> : tensor<4xi32> // CHECK-NEXT: [[CM3:%.+]] = constant dense<-3> : tensor<4xi32> - %4 = divis %0, %3 : tensor<4xi32> + %4 = divi_signed %0, %3 : tensor<4xi32> - // CHECK-NEXT: [[XZ:%.+]] = divis [[C6]], [[C0]] - %5 = divis %0, %z : tensor<4xi32> + // CHECK-NEXT: [[XZ:%.+]] = divi_signed [[C6]], [[C0]] + %5 = divi_signed %0, %z : tensor<4xi32> // CHECK-NEXT: return [[C3]], [[CM3]], [[XZ]] return %2, %4, %5 : tensor<4xi32>, tensor<4xi32>, tensor<4xi32> @@ -262,24 +262,24 @@ func @divis_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { // ----- -// CHECK-LABEL: func @simple_diviu -func @simple_diviu() -> (i32, i32, i32) { +// CHECK-LABEL: func @simple_divi_unsigned +func @simple_divi_unsigned() -> (i32, i32, i32) { %z = constant 0 : i32 // CHECK-DAG: [[C6:%.+]] = constant 6 %0 = constant 6 : i32 %1 = constant 2 : i32 // CHECK-DAG: [[C3:%.+]] = constant 3 : i32 - %2 = diviu %0, %1 : i32 + %2 = divi_unsigned %0, %1 : i32 %3 = constant -2 : i32 // Unsigned division interprets -2 as 2^32-2, so the result is 0. // CHECK-DAG: [[C0:%.+]] = constant 0 : i32 - %4 = diviu %0, %3 : i32 + %4 = divi_unsigned %0, %3 : i32 - // CHECK-NEXT: [[XZ:%.+]] = diviu [[C6]], [[C0]] - %5 = diviu %0, %z : i32 + // CHECK-NEXT: [[XZ:%.+]] = divi_unsigned [[C6]], [[C0]] + %5 = divi_unsigned %0, %z : i32 // CHECK-NEXT: return [[C3]], [[C0]], [[XZ]] return %2, %4, %5 : i32, i32, i32 @@ -288,24 +288,24 @@ func @simple_diviu() -> (i32, i32, i32) { // ----- -// CHECK-LABEL: func @diviu_splat_tensor -func @diviu_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { +// CHECK-LABEL: func @divi_unsigned_splat_tensor +func @divi_unsigned_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { %z = constant dense<0> : tensor<4xi32> // CHECK-DAG: [[C6:%.+]] = constant dense<6> %0 = constant dense<6> : tensor<4xi32> %1 = constant dense<2> : tensor<4xi32> // CHECK-DAG: [[C3:%.+]] = constant dense<3> : tensor<4xi32> - %2 = diviu %0, %1 : tensor<4xi32> + %2 = divi_unsigned %0, %1 : tensor<4xi32> %3 = constant dense<-2> : tensor<4xi32> // Unsigned division interprets -2 as 2^32-2, so the result is 0. // CHECK-DAG: [[C0:%.+]] = constant dense<0> : tensor<4xi32> - %4 = diviu %0, %3 : tensor<4xi32> + %4 = divi_unsigned %0, %3 : tensor<4xi32> - // CHECK-NEXT: [[XZ:%.+]] = diviu [[C6]], [[C0]] - %5 = diviu %0, %z : tensor<4xi32> + // CHECK-NEXT: [[XZ:%.+]] = divi_unsigned [[C6]], [[C0]] + %5 = divi_unsigned %0, %z : tensor<4xi32> // CHECK-NEXT: return [[C3]], [[C0]], [[XZ]] return %2, %4, %5 : tensor<4xi32>, tensor<4xi32>, tensor<4xi32> @@ -313,18 +313,18 @@ func @diviu_splat_tensor() -> (tensor<4xi32>, tensor<4xi32>, tensor<4xi32>) { // ----- -// CHECK-LABEL: func @simple_remis -func @simple_remis(%a : i32) -> (i32, i32, i32) { +// CHECK-LABEL: func @simple_remi_signed +func @simple_remi_signed(%a : i32) -> (i32, i32, i32) { %0 = constant 5 : i32 %1 = constant 2 : i32 %2 = constant 1 : i32 %3 = constant -2 : i32 // CHECK-NEXT:[[C1:%.+]] = constant 1 : i32 - %4 = remis %0, %1 : i32 - %5 = remis %0, %3 : i32 + %4 = remi_signed %0, %1 : i32 + %5 = remi_signed %0, %3 : i32 // CHECK-NEXT:[[C0:%.+]] = constant 0 : i32 - %6 = remis %a, %2 : i32 + %6 = remi_signed %a, %2 : i32 // CHECK-NEXT: return [[C1]], [[C1]], [[C0]] : i32, i32, i32 return %4, %5, %6 : i32, i32, i32 @@ -332,19 +332,19 @@ func @simple_remis(%a : i32) -> (i32, i32, i32) { // ----- -// CHECK-LABEL: func @simple_remiu -func @simple_remiu(%a : i32) -> (i32, i32, i32) { +// CHECK-LABEL: func @simple_remi_unsigned +func @simple_remi_unsigned(%a : i32) -> (i32, i32, i32) { %0 = constant 5 : i32 %1 = constant 2 : i32 %2 = constant 1 : i32 %3 = constant -2 : i32 // CHECK-DAG:[[C1:%.+]] = constant 1 : i32 - %4 = remiu %0, %1 : i32 + %4 = remi_unsigned %0, %1 : i32 // CHECK-DAG:[[C5:%.+]] = constant 5 : i32 - %5 = remiu %0, %3 : i32 + %5 = remi_unsigned %0, %3 : i32 // CHECK-DAG:[[C0:%.+]] = constant 0 : i32 - %6 = remiu %a, %2 : i32 + %6 = remi_unsigned %a, %2 : i32 // CHECK-NEXT: return [[C1]], [[C5]], [[C0]] : i32, i32, i32 return %4, %5, %6 : i32, i32, i32 diff --git a/mlir/test/Transforms/loop-coalescing.mlir b/mlir/test/Transforms/loop-coalescing.mlir index 45e2b5d..d10cf19 100644 --- a/mlir/test/Transforms/loop-coalescing.mlir +++ b/mlir/test/Transforms/loop-coalescing.mlir @@ -26,10 +26,10 @@ func @one_3d_nest() { // CHECK-NOT: loop.for // Reconstruct original IVs from the linearized one. - // CHECK: %[[orig_k:.*]] = remis %[[i]], %[[orig_ub_k]] - // CHECK: %[[div:.*]] = divis %[[i]], %[[orig_ub_k]] - // CHECK: %[[orig_j:.*]] = remis %[[div]], %[[orig_ub_j]] - // CHECK: %[[orig_i:.*]] = divis %[[div]], %[[orig_ub_j]] + // CHECK: %[[orig_k:.*]] = remi_signed %[[i]], %[[orig_ub_k]] + // CHECK: %[[div:.*]] = divi_signed %[[i]], %[[orig_ub_k]] + // CHECK: %[[orig_j:.*]] = remi_signed %[[div]], %[[orig_ub_j]] + // CHECK: %[[orig_i:.*]] = divi_signed %[[div]], %[[orig_ub_j]] loop.for %j = %c0 to %c56 step %c1 { loop.for %k = %c0 to %c3 step %c1 { // CHECK: "use"(%[[orig_i]], %[[orig_j]], %[[orig_k]]) @@ -52,10 +52,10 @@ func @multi_use() { loop.for %i = %c1 to %c10 step %c1 { loop.for %j = %c1 to %c10 step %c1 { loop.for %k = %c1 to %c10 step %c1 { - // CHECK: %[[k_unshifted:.*]] = remis %[[iv]], %[[k_extent:.*]] - // CHECK: %[[ij:.*]] = divis %[[iv]], %[[k_extent]] - // CHECK: %[[j_unshifted:.*]] = remis %[[ij]], %[[j_extent:.*]] - // CHECK: %[[i_unshifted:.*]] = divis %[[ij]], %[[j_extent]] + // CHECK: %[[k_unshifted:.*]] = remi_signed %[[iv]], %[[k_extent:.*]] + // CHECK: %[[ij:.*]] = divi_signed %[[iv]], %[[k_extent]] + // CHECK: %[[j_unshifted:.*]] = remi_signed %[[ij]], %[[j_extent:.*]] + // CHECK: %[[i_unshifted:.*]] = divi_signed %[[ij]], %[[j_extent]] // CHECK: %[[k:.*]] = addi %[[k_unshifted]] // CHECK: %[[j:.*]] = addi %[[j_unshifted]] // CHECK: %[[i:.*]] = addi %[[i_unshifted]] @@ -91,7 +91,7 @@ func @unnormalized_loops() { // CHECK: %[[c1:.*]] = constant 1 // CHECK: %[[step_minus_c1:.*]] = subi %[[orig_step_i]], %[[c1]] // CHECK: %[[dividend:.*]] = addi %[[diff_i]], %[[step_minus_c1]] - // CHECK: %[[numiter_i:.*]] = divis %[[dividend]], %[[orig_step_i]] + // CHECK: %[[numiter_i:.*]] = divi_signed %[[dividend]], %[[orig_step_i]] // Normalized lower bound and step for the outer loop. // CHECK: %[[lb_i:.*]] = constant 0 @@ -99,7 +99,7 @@ func @unnormalized_loops() { // Number of iterations in the inner loop, the pattern is the same as above, // only capture the final result. - // CHECK: %[[numiter_j:.*]] = divis {{.*}}, %[[orig_step_j]] + // CHECK: %[[numiter_j:.*]] = divi_signed {{.*}}, %[[orig_step_j]] // New bounds of the outer loop. // CHECK: %[[range:.*]] = muli %[[numiter_i]], %[[numiter_j]] @@ -109,8 +109,8 @@ func @unnormalized_loops() { // CHECK-NOT: loop.for loop.for %j = %c7 to %c17 step %c3 { // The IVs are rewritten. - // CHECK: %[[normalized_j:.*]] = remis %[[i]], %[[numiter_j]] - // CHECK: %[[normalized_i:.*]] = divis %[[i]], %[[numiter_j]] + // CHECK: %[[normalized_j:.*]] = remi_signed %[[i]], %[[numiter_j]] + // CHECK: %[[normalized_i:.*]] = divi_signed %[[i]], %[[numiter_j]] // CHECK: %[[scaled_j:.*]] = muli %[[normalized_j]], %[[orig_step_j]] // CHECK: %[[orig_j:.*]] = addi %[[scaled_j]], %[[orig_lb_j]] // CHECK: %[[scaled_i:.*]] = muli %[[normalized_i]], %[[orig_step_i]] @@ -137,11 +137,11 @@ func @parametric(%lb1 : index, %ub1 : index, %step1 : index, // CHECK: %[[range1:.*]] = subi %[[orig_ub1]], %[[orig_lb1]] // CHECK: %[[orig_step1_minus_1:.*]] = subi %[[orig_step1]], %c1 // CHECK: %[[dividend1:.*]] = addi %[[range1]], %[[orig_step1_minus_1]] - // CHECK: %[[numiter1:.*]] = divis %[[dividend1]], %[[orig_step1]] + // CHECK: %[[numiter1:.*]] = divi_signed %[[dividend1]], %[[orig_step1]] // CHECK: %[[range2:.*]] = subi %[[orig_ub2]], %[[orig_lb2]] // CHECK: %[[orig_step2_minus_1:.*]] = subi %arg5, %c1 // CHECK: %[[dividend2:.*]] = addi %[[range2]], %[[orig_step2_minus_1]] - // CHECK: %[[numiter2:.*]] = divis %[[dividend2]], %[[orig_step2]] + // CHECK: %[[numiter2:.*]] = divi_signed %[[dividend2]], %[[orig_step2]] // CHECK: %[[range:.*]] = muli %[[numiter1]], %[[numiter2]] : index // Check that the outer loop is updated. @@ -151,8 +151,8 @@ func @parametric(%lb1 : index, %ub1 : index, %step1 : index, // CHECK-NOT: loop.for loop.for %j = %lb2 to %ub2 step %step2 { // Remapping of the induction variables. - // CHECK: %[[normalized_j:.*]] = remis %[[i]], %[[numiter2]] : index - // CHECK: %[[normalized_i:.*]] = divis %[[i]], %[[numiter2]] : index + // CHECK: %[[normalized_j:.*]] = remi_signed %[[i]], %[[numiter2]] : index + // CHECK: %[[normalized_i:.*]] = divi_signed %[[i]], %[[numiter2]] : index // CHECK: %[[scaled_j:.*]] = muli %[[normalized_j]], %[[orig_step2]] // CHECK: %[[orig_j:.*]] = addi %[[scaled_j]], %[[orig_lb2]] // CHECK: %[[scaled_i:.*]] = muli %[[normalized_i]], %[[orig_step1]] diff --git a/mlir/test/Transforms/lower-affine.mlir b/mlir/test/Transforms/lower-affine.mlir index dac3557..1c3de88 100644 --- a/mlir/test/Transforms/lower-affine.mlir +++ b/mlir/test/Transforms/lower-affine.mlir @@ -452,7 +452,7 @@ func @args_ret_affine_apply(index, index) -> (index, index) { // CHECK-LABEL: func @affine_apply_mod func @affine_apply_mod(%arg0 : index) -> (index) { // CHECK-NEXT: %[[c42:.*]] = constant 42 : index -// CHECK-NEXT: %[[v0:.*]] = remis %{{.*}}, %[[c42]] : index +// CHECK-NEXT: %[[v0:.*]] = remi_signed %{{.*}}, %[[c42]] : index // CHECK-NEXT: %[[c0:.*]] = constant 0 : index // CHECK-NEXT: %[[v1:.*]] = cmpi "slt", %[[v0]], %[[c0]] : index // CHECK-NEXT: %[[v2:.*]] = addi %[[v0]], %[[c42]] : index @@ -476,7 +476,7 @@ func @affine_apply_floordiv(%arg0 : index) -> (index) { // CHECK-NEXT: %[[v0:.*]] = cmpi "slt", %{{.*}}, %[[c0]] : index // CHECK-NEXT: %[[v1:.*]] = subi %[[cm1]], %{{.*}} : index // CHECK-NEXT: %[[v2:.*]] = select %[[v0]], %[[v1]], %{{.*}} : index -// CHECK-NEXT: %[[v3:.*]] = divis %[[v2]], %[[c42]] : index +// CHECK-NEXT: %[[v3:.*]] = divi_signed %[[v2]], %[[c42]] : index // CHECK-NEXT: %[[v4:.*]] = subi %[[cm1]], %[[v3]] : index // CHECK-NEXT: %[[v5:.*]] = select %[[v0]], %[[v4]], %[[v3]] : index %0 = affine.apply #mapfloordiv (%arg0) @@ -499,7 +499,7 @@ func @affine_apply_ceildiv(%arg0 : index) -> (index) { // CHECK-NEXT: %[[v1:.*]] = subi %[[c0]], %{{.*}} : index // CHECK-NEXT: %[[v2:.*]] = subi %{{.*}}, %[[c1]] : index // CHECK-NEXT: %[[v3:.*]] = select %[[v0]], %[[v1]], %[[v2]] : index -// CHECK-NEXT: %[[v4:.*]] = divis %[[v3]], %[[c42]] : index +// CHECK-NEXT: %[[v4:.*]] = divi_signed %[[v3]], %[[c42]] : index // CHECK-NEXT: %[[v5:.*]] = subi %[[c0]], %[[v4]] : index // CHECK-NEXT: %[[v6:.*]] = addi %[[v4]], %[[c1]] : index // CHECK-NEXT: %[[v7:.*]] = select %[[v0]], %[[v5]], %[[v6]] : index diff --git a/mlir/test/Transforms/parametric-tiling.mlir b/mlir/test/Transforms/parametric-tiling.mlir index 2e715fd..afa33cb 100644 --- a/mlir/test/Transforms/parametric-tiling.mlir +++ b/mlir/test/Transforms/parametric-tiling.mlir @@ -12,11 +12,11 @@ func @rectangular(%arg0: memref) { // COMMON: %[[diff:.*]] = subi %c44, %c2 // COMMON: %[[adjustment:.*]] = subi %c1, %c1_{{.*}} // COMMON-NEXT: %[[diff_adj:.*]] = addi %[[diff]], %[[adjustment]] - // COMMON-NEXT: %[[range:.*]] = divis %[[diff_adj]], %c1 + // COMMON-NEXT: %[[range:.*]] = divi_signed %[[diff_adj]], %c1 // Ceildiv to get the parametric tile size. // COMMON: %[[sum:.*]] = addi %[[range]], %c6 - // COMMON-NEXT: %[[size:.*]] = divis %[[sum]], %c7 + // COMMON-NEXT: %[[size:.*]] = divi_signed %[[sum]], %c7 // New outer step (original is %c1). // COMMON-NEXT: %[[step:.*]] = muli %c1, %[[size]] @@ -26,11 +26,11 @@ func @rectangular(%arg0: memref) { // TILE_74: %[[diff2:.*]] = subi %c44, %c1 // TILE_74: %[[adjustment2:.*]] = subi %c2, %c1_{{.*}} // TILE_74-NEXT: %[[diff2_adj:.*]] = addi %[[diff2]], %[[adjustment2]] - // TILE_74-NEXT: %[[range2:.*]] = divis %[[diff2_adj]], %c2 + // TILE_74-NEXT: %[[range2:.*]] = divi_signed %[[diff2_adj]], %c2 // Ceildiv to get the parametric tile size for the second original loop. // TILE_74: %[[sum2:.*]] = addi %[[range2]], %c3 - // TILE_74-NEXT: %[[size2:.*]] = divis %[[sum2]], %c4 + // TILE_74-NEXT: %[[size2:.*]] = divi_signed %[[sum2]], %c4 // New inner step (original is %c2). // TILE_74-NEXT: %[[step2:.*]] = muli %c2, %[[size2]] @@ -76,11 +76,11 @@ func @triangular(%arg0: memref) { // COMMON: %[[diff:.*]] = subi %c44, %c2 // COMMON: %[[adjustment:.*]] = subi %c1, %c1_{{.*}} // COMMON-NEXT: %[[diff_adj:.*]] = addi %[[diff]], %[[adjustment]] - // COMMON-NEXT: %[[range:.*]] = divis %[[diff_adj]], %c1 + // COMMON-NEXT: %[[range:.*]] = divi_signed %[[diff_adj]], %c1 // Ceildiv to get the parametric tile size. // COMMON: %[[sum:.*]] = addi %[[range]], %c6 - // COMMON-NEXT: %[[size:.*]] = divis %[[sum]], %c7 + // COMMON-NEXT: %[[size:.*]] = divi_signed %[[sum]], %c7 // New outer step (original is %c1). // COMMON-NEXT: %[[step:.*]] = muli %c1, %[[size]] @@ -95,11 +95,11 @@ func @triangular(%arg0: memref) { // where step is known to be %c2. // TILE_74: %[[diff2:.*]] = subi %[[i]], %c1 // TILE_74-NEXT: %[[diff2_adj:.*]] = addi %[[diff2]], %[[adjustment2]] - // TILE_74-NEXT: %[[range2:.*]] = divis %[[diff2_adj]], %c2 + // TILE_74-NEXT: %[[range2:.*]] = divi_signed %[[diff2_adj]], %c2 // Ceildiv to get the parametric tile size for the second original loop. // TILE_74: %[[sum2:.*]] = addi %[[range2]], %c3 - // TILE_74-NEXT: %[[size2:.*]] = divis %[[sum2]], %c4 + // TILE_74-NEXT: %[[size2:.*]] = divi_signed %[[sum2]], %c4 // New inner step (original is %c2). // TILE_74-NEXT: %[[step2:.*]] = muli %c2, %[[size2]] -- 2.7.4