[Mlir-commits] [mlir] 12ffc26 - [mlir][spirv] Define spv.ImageDrefGather operation
Weiwei Li
llvmlistbot at llvm.org
Thu Apr 8 17:20:45 PDT 2021
Author: Weiwei Li
Date: 2021-04-08T20:15:54-04:00
New Revision: 12ffc260678dd2a4ce0fbfba562bfd1af9fe8efe
URL: https://github.com/llvm/llvm-project/commit/12ffc260678dd2a4ce0fbfba562bfd1af9fe8efe
DIFF: https://github.com/llvm/llvm-project/commit/12ffc260678dd2a4ce0fbfba562bfd1af9fe8efe.diff
LOG: [mlir][spirv] Define spv.ImageDrefGather operation
This patch doesn't support the optional operands of ImageDrefGather. The support of optional operands will be implemented later.
co-authered-by: Alan Liu <alanliu.yf at gmail.com>
Differential Revision: https://reviews.llvm.org/D100128
Added:
Modified:
mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
mlir/include/mlir/Dialect/SPIRV/IR/SPIRVImageOps.td
mlir/lib/Dialect/SPIRV/IR/SPIRVOps.cpp
mlir/test/Dialect/SPIRV/IR/image-ops.mlir
mlir/test/Target/SPIRV/image-ops.mlir
Removed:
################################################################################
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
index d293a6a88afd4..579981f2c1010 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
@@ -3048,6 +3048,7 @@ def SPV_Void : TypeAlias<NoneType, "void">;
def SPV_Bool : TypeAlias<I1, "bool">;
def SPV_Integer : AnyIntOfWidths<[8, 16, 32, 64]>;
def SPV_Int32 : TypeAlias<I32, "Int32">;
+def SPV_Float32 : TypeAlias<F32, "Float32">;
def SPV_Float : FloatOfWidths<[16, 32, 64]>;
def SPV_Float16or32 : FloatOfWidths<[16, 32]>;
def SPV_Vector : VectorOfLengthAndType<[2, 3, 4, 8, 16],
@@ -3202,6 +3203,7 @@ def SPV_OC_OpCompositeConstruct : I32EnumAttrCase<"OpCompositeConstruct",
def SPV_OC_OpCompositeExtract : I32EnumAttrCase<"OpCompositeExtract", 81>;
def SPV_OC_OpCompositeInsert : I32EnumAttrCase<"OpCompositeInsert", 82>;
def SPV_OC_OpTranspose : I32EnumAttrCase<"OpTranspose", 84>;
+def SPV_OC_OpImageDrefGather : I32EnumAttrCase<"OpImageDrefGather", 97>;
def SPV_OC_OpImage : I32EnumAttrCase<"OpImage", 100>;
def SPV_OC_OpConvertFToU : I32EnumAttrCase<"OpConvertFToU", 109>;
def SPV_OC_OpConvertFToS : I32EnumAttrCase<"OpConvertFToS", 110>;
@@ -3341,37 +3343,38 @@ def SPV_OpcodeAttr :
SPV_OC_OpMemberDecorate, SPV_OC_OpVectorExtractDynamic,
SPV_OC_OpVectorInsertDynamic, SPV_OC_OpVectorShuffle,
SPV_OC_OpCompositeConstruct, SPV_OC_OpCompositeExtract,
- SPV_OC_OpCompositeInsert, SPV_OC_OpTranspose, SPV_OC_OpImage,
- SPV_OC_OpConvertFToU, SPV_OC_OpConvertFToS, SPV_OC_OpConvertSToF,
- SPV_OC_OpConvertUToF, SPV_OC_OpUConvert, SPV_OC_OpSConvert, SPV_OC_OpFConvert,
- SPV_OC_OpBitcast, SPV_OC_OpSNegate, SPV_OC_OpFNegate, SPV_OC_OpIAdd,
- SPV_OC_OpFAdd, SPV_OC_OpISub, SPV_OC_OpFSub, SPV_OC_OpIMul, SPV_OC_OpFMul,
- SPV_OC_OpUDiv, SPV_OC_OpSDiv, SPV_OC_OpFDiv, SPV_OC_OpUMod, SPV_OC_OpSRem,
- SPV_OC_OpSMod, SPV_OC_OpFRem, SPV_OC_OpFMod, SPV_OC_OpMatrixTimesScalar,
- SPV_OC_OpMatrixTimesMatrix, SPV_OC_OpIsNan, SPV_OC_OpIsInf, SPV_OC_OpOrdered,
- SPV_OC_OpUnordered, SPV_OC_OpLogicalEqual, SPV_OC_OpLogicalNotEqual,
- SPV_OC_OpLogicalOr, SPV_OC_OpLogicalAnd, SPV_OC_OpLogicalNot, SPV_OC_OpSelect,
- SPV_OC_OpIEqual, SPV_OC_OpINotEqual, SPV_OC_OpUGreaterThan,
- SPV_OC_OpSGreaterThan, SPV_OC_OpUGreaterThanEqual, SPV_OC_OpSGreaterThanEqual,
- SPV_OC_OpULessThan, SPV_OC_OpSLessThan, SPV_OC_OpULessThanEqual,
- SPV_OC_OpSLessThanEqual, SPV_OC_OpFOrdEqual, SPV_OC_OpFUnordEqual,
- SPV_OC_OpFOrdNotEqual, SPV_OC_OpFUnordNotEqual, SPV_OC_OpFOrdLessThan,
- SPV_OC_OpFUnordLessThan, SPV_OC_OpFOrdGreaterThan, SPV_OC_OpFUnordGreaterThan,
- SPV_OC_OpFOrdLessThanEqual, SPV_OC_OpFUnordLessThanEqual,
- SPV_OC_OpFOrdGreaterThanEqual, SPV_OC_OpFUnordGreaterThanEqual,
- SPV_OC_OpShiftRightLogical, SPV_OC_OpShiftRightArithmetic,
- SPV_OC_OpShiftLeftLogical, SPV_OC_OpBitwiseOr, SPV_OC_OpBitwiseXor,
- SPV_OC_OpBitwiseAnd, SPV_OC_OpNot, SPV_OC_OpBitFieldInsert,
- SPV_OC_OpBitFieldSExtract, SPV_OC_OpBitFieldUExtract, SPV_OC_OpBitReverse,
- SPV_OC_OpBitCount, SPV_OC_OpControlBarrier, SPV_OC_OpMemoryBarrier,
- SPV_OC_OpAtomicCompareExchangeWeak, SPV_OC_OpAtomicIIncrement,
- SPV_OC_OpAtomicIDecrement, SPV_OC_OpAtomicIAdd, SPV_OC_OpAtomicISub,
- SPV_OC_OpAtomicSMin, SPV_OC_OpAtomicUMin, SPV_OC_OpAtomicSMax,
- SPV_OC_OpAtomicUMax, SPV_OC_OpAtomicAnd, SPV_OC_OpAtomicOr, SPV_OC_OpAtomicXor,
- SPV_OC_OpPhi, SPV_OC_OpLoopMerge, SPV_OC_OpSelectionMerge, SPV_OC_OpLabel,
- SPV_OC_OpBranch, SPV_OC_OpBranchConditional, SPV_OC_OpReturn,
- SPV_OC_OpReturnValue, SPV_OC_OpUnreachable, SPV_OC_OpGroupBroadcast,
- SPV_OC_OpNoLine, SPV_OC_OpModuleProcessed, SPV_OC_OpGroupNonUniformElect,
+ SPV_OC_OpCompositeInsert, SPV_OC_OpTranspose, SPV_OC_OpImageDrefGather,
+ SPV_OC_OpImage, SPV_OC_OpConvertFToU, SPV_OC_OpConvertFToS,
+ SPV_OC_OpConvertSToF, SPV_OC_OpConvertUToF, SPV_OC_OpUConvert,
+ SPV_OC_OpSConvert, SPV_OC_OpFConvert, SPV_OC_OpBitcast, SPV_OC_OpSNegate,
+ SPV_OC_OpFNegate, SPV_OC_OpIAdd, SPV_OC_OpFAdd, SPV_OC_OpISub, SPV_OC_OpFSub,
+ SPV_OC_OpIMul, SPV_OC_OpFMul, SPV_OC_OpUDiv, SPV_OC_OpSDiv, SPV_OC_OpFDiv,
+ SPV_OC_OpUMod, SPV_OC_OpSRem, SPV_OC_OpSMod, SPV_OC_OpFRem, SPV_OC_OpFMod,
+ SPV_OC_OpMatrixTimesScalar, SPV_OC_OpMatrixTimesMatrix, SPV_OC_OpIsNan,
+ SPV_OC_OpIsInf, SPV_OC_OpOrdered, SPV_OC_OpUnordered, SPV_OC_OpLogicalEqual,
+ SPV_OC_OpLogicalNotEqual, SPV_OC_OpLogicalOr, SPV_OC_OpLogicalAnd,
+ SPV_OC_OpLogicalNot, SPV_OC_OpSelect, SPV_OC_OpIEqual, SPV_OC_OpINotEqual,
+ SPV_OC_OpUGreaterThan, SPV_OC_OpSGreaterThan, SPV_OC_OpUGreaterThanEqual,
+ SPV_OC_OpSGreaterThanEqual, SPV_OC_OpULessThan, SPV_OC_OpSLessThan,
+ SPV_OC_OpULessThanEqual, SPV_OC_OpSLessThanEqual, SPV_OC_OpFOrdEqual,
+ SPV_OC_OpFUnordEqual, SPV_OC_OpFOrdNotEqual, SPV_OC_OpFUnordNotEqual,
+ SPV_OC_OpFOrdLessThan, SPV_OC_OpFUnordLessThan, SPV_OC_OpFOrdGreaterThan,
+ SPV_OC_OpFUnordGreaterThan, SPV_OC_OpFOrdLessThanEqual,
+ SPV_OC_OpFUnordLessThanEqual, SPV_OC_OpFOrdGreaterThanEqual,
+ SPV_OC_OpFUnordGreaterThanEqual, SPV_OC_OpShiftRightLogical,
+ SPV_OC_OpShiftRightArithmetic, SPV_OC_OpShiftLeftLogical, SPV_OC_OpBitwiseOr,
+ SPV_OC_OpBitwiseXor, SPV_OC_OpBitwiseAnd, SPV_OC_OpNot,
+ SPV_OC_OpBitFieldInsert, SPV_OC_OpBitFieldSExtract, SPV_OC_OpBitFieldUExtract,
+ SPV_OC_OpBitReverse, SPV_OC_OpBitCount, SPV_OC_OpControlBarrier,
+ SPV_OC_OpMemoryBarrier, SPV_OC_OpAtomicCompareExchangeWeak,
+ SPV_OC_OpAtomicIIncrement, SPV_OC_OpAtomicIDecrement, SPV_OC_OpAtomicIAdd,
+ SPV_OC_OpAtomicISub, SPV_OC_OpAtomicSMin, SPV_OC_OpAtomicUMin,
+ SPV_OC_OpAtomicSMax, SPV_OC_OpAtomicUMax, SPV_OC_OpAtomicAnd,
+ SPV_OC_OpAtomicOr, SPV_OC_OpAtomicXor, SPV_OC_OpPhi, SPV_OC_OpLoopMerge,
+ SPV_OC_OpSelectionMerge, SPV_OC_OpLabel, SPV_OC_OpBranch,
+ SPV_OC_OpBranchConditional, SPV_OC_OpReturn, SPV_OC_OpReturnValue,
+ SPV_OC_OpUnreachable, SPV_OC_OpGroupBroadcast, SPV_OC_OpNoLine,
+ SPV_OC_OpModuleProcessed, SPV_OC_OpGroupNonUniformElect,
SPV_OC_OpGroupNonUniformBroadcast, SPV_OC_OpGroupNonUniformBallot,
SPV_OC_OpGroupNonUniformIAdd, SPV_OC_OpGroupNonUniformFAdd,
SPV_OC_OpGroupNonUniformIMul, SPV_OC_OpGroupNonUniformFMul,
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVImageOps.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVImageOps.td
index 2c50b0c9bcfa9..0ceab0be9f253 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVImageOps.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVImageOps.td
@@ -17,6 +17,63 @@
include "mlir/Dialect/SPIRV/IR/SPIRVBase.td"
include "mlir/Interfaces/SideEffectInterfaces.td"
+// -----
+
+def SPV_ImageDrefGatherOp : SPV_Op<"ImageDrefGather", [NoSideEffect]> {
+ let summary = "Gathers the requested depth-comparison from four texels.";
+
+ let description = [{
+ Result Type must be a vector of four components of floating-point type
+ or integer type. Its components must be the same as Sampled Type of the
+ underlying OpTypeImage (unless that underlying Sampled Type is
+ OpTypeVoid). It has one component per gathered texel.
+
+ Sampled Image must be an object whose type is OpTypeSampledImage. Its
+ OpTypeImage must have a Dim of 2D, Cube, or Rect. The MS operand of the
+ underlying OpTypeImage must be 0.
+
+ Coordinate must be a scalar or vector of floating-point type. It
+ contains (u[, v] … [, array layer]) as needed by the definition of
+ Sampled Image.
+
+ Dref is the depth-comparison reference value. It must be a 32-bit
+ floating-point type scalar.
+
+ Image Operands encodes what operands follow, as per Image Operands.
+
+ <!-- End of AutoGen section -->
+
+ #### Example:
+
+ ```mlir
+ %0 = spv.ImageDrefGather %1 : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %2 : vector<4xf32>, %3 : f32 -> vector<4xi32>
+ ```
+ }];
+
+ let availability = [
+ MinVersion<SPV_V_1_0>,
+ MaxVersion<SPV_V_1_5>,
+ Extension<[]>,
+ Capability<[SPV_C_Shader]>
+ ];
+
+ let arguments = (ins
+ SPV_AnySampledImage:$sampledimage,
+ SPV_ScalarOrVectorOf<SPV_Float>:$coordinate,
+ SPV_Float:$dref
+ );
+
+ let results = (outs
+ SPV_Vector:$result
+ );
+
+ let assemblyFormat = "attr-dict $sampledimage `:` type($sampledimage) `,` $coordinate `:` type($coordinate) `,` $dref `:` type($dref) `->` type($result)";
+
+ let verifier = [{ return ::verify(*this); }];
+
+}
+
+
// -----
def SPV_ImageOp : SPV_Op<"Image",
diff --git a/mlir/lib/Dialect/SPIRV/IR/SPIRVOps.cpp b/mlir/lib/Dialect/SPIRV/IR/SPIRVOps.cpp
index 21bcfe458ab7f..31a92c3515992 100644
--- a/mlir/lib/Dialect/SPIRV/IR/SPIRVOps.cpp
+++ b/mlir/lib/Dialect/SPIRV/IR/SPIRVOps.cpp
@@ -3608,6 +3608,46 @@ static LogicalResult verify(spirv::GLSLLdexpOp ldexpOp) {
return success();
}
+//===----------------------------------------------------------------------===//
+// spv.ImageDrefGather
+//===----------------------------------------------------------------------===//
+
+static LogicalResult verify(spirv::ImageDrefGatherOp imageDrefGatherOp) {
+ // TODO: Support optional operands.
+ VectorType resultType =
+ imageDrefGatherOp.result().getType().cast<VectorType>();
+ auto sampledImageType = imageDrefGatherOp.sampledimage()
+ .getType()
+ .cast<spirv::SampledImageType>();
+ auto imageType = sampledImageType.getImageType().cast<spirv::ImageType>();
+
+ if (resultType.getNumElements() != 4)
+ return imageDrefGatherOp.emitOpError(
+ "result type must be a vector of four components");
+
+ Type elementType = resultType.getElementType();
+ Type sampledElementType = imageType.getElementType();
+ if (!sampledElementType.isa<NoneType>() && elementType != sampledElementType)
+ return imageDrefGatherOp.emitOpError(
+ "the component type of result must be the same as sampled type of the "
+ "underlying image type");
+
+ spirv::Dim imageDim = imageType.getDim();
+ spirv::ImageSamplingInfo imageMS = imageType.getSamplingInfo();
+
+ if (imageDim != spirv::Dim::Dim2D && imageDim != spirv::Dim::Cube &&
+ imageDim != spirv::Dim::Rect)
+ return imageDrefGatherOp.emitOpError(
+ "the Dim operand of the underlying image type must be 2D, Cube, or "
+ "Rect");
+
+ if (imageMS != spirv::ImageSamplingInfo::SingleSampled)
+ return imageDrefGatherOp.emitOpError(
+ "the MS operand of the underlying image type must be 0");
+
+ return success();
+}
+
namespace mlir {
namespace spirv {
diff --git a/mlir/test/Dialect/SPIRV/IR/image-ops.mlir b/mlir/test/Dialect/SPIRV/IR/image-ops.mlir
index ef5d2cf94ef2e..d83339f5f92de 100644
--- a/mlir/test/Dialect/SPIRV/IR/image-ops.mlir
+++ b/mlir/test/Dialect/SPIRV/IR/image-ops.mlir
@@ -1,5 +1,50 @@
// RUN: mlir-opt -split-input-file -verify-diagnostics %s | FileCheck %s
+
+//===----------------------------------------------------------------------===//
+// spv.ImageDrefGather
+//===----------------------------------------------------------------------===//
+
+func @image_dref_gather(%arg0 : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) -> () {
+ // CHECK: spv.ImageDrefGather {{.*}} : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, {{.*}} : vector<4xf32>, {{.*}} : f32 -> vector<4xi32>
+ %0 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<4xi32>
+ spv.Return
+}
+
+// -----
+
+func @image_dref_gather_error_result_type(%arg0 : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) -> () {
+ // expected-error @+1 {{result type must be a vector of four components}}
+ %0 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<i32, Dim2D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<3xi32>
+ spv.Return
+}
+
+// -----
+
+func @image_dref_gather_error_same_type(%arg0 : !spv.sampled_image<!spv.image<i32, Rect, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) -> () {
+ // expected-error @+1 {{the component type of result must be the same as sampled type of the underlying image type}}
+ %0 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<i32, Rect, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<4xf32>
+ spv.Return
+}
+
+// -----
+
+func @image_dref_gather_error_dim(%arg0 : !spv.sampled_image<!spv.image<i32, Dim1D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) -> () {
+ // expected-error @+1 {{the Dim operand of the underlying image type must be 2D, Cube, or Rect}}
+ %0 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<i32, Dim1D, NoDepth, NonArrayed, SingleSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<4xi32>
+ spv.Return
+}
+
+// -----
+
+func @image_dref_gather_error_ms(%arg0 : !spv.sampled_image<!spv.image<i32, Cube, NoDepth, NonArrayed, MultiSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) -> () {
+ // expected-error @+1 {{the MS operand of the underlying image type must be 0}}
+ %0 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<i32, Cube, NoDepth, NonArrayed, MultiSampled, NoSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<4xi32>
+ spv.Return
+}
+
+// -----
+
//===----------------------------------------------------------------------===//
// spv.Image
//===----------------------------------------------------------------------===//
diff --git a/mlir/test/Target/SPIRV/image-ops.mlir b/mlir/test/Target/SPIRV/image-ops.mlir
index 56c880227e676..4e00bf6523432 100644
--- a/mlir/test/Target/SPIRV/image-ops.mlir
+++ b/mlir/test/Target/SPIRV/image-ops.mlir
@@ -1,9 +1,11 @@
// RUN: mlir-translate -test-spirv-roundtrip %s | FileCheck %s
spv.module Logical GLSL450 requires #spv.vce<v1.0, [Shader], []> {
- spv.func @image(%arg0 : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>) "None" {
+ spv.func @image(%arg0 : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32) "None" {
// CHECK: {{%.*}} = spv.Image {{%.*}} : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>
%0 = spv.Image %arg0 : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>
+ // CHECK: {{%.*}} = spv.ImageDrefGather {{%.*}} : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>, {{%.*}} : vector<4xf32>, {{%.*}} : f32 -> vector<4xf32>
+ %1 = spv.ImageDrefGather %arg0 : !spv.sampled_image<!spv.image<f32, Dim2D, NoDepth, NonArrayed, SingleSampled, NeedSampler, Unknown>>, %arg1 : vector<4xf32>, %arg2 : f32 -> vector<4xf32>
spv.Return
}
}
More information about the Mlir-commits
mailing list