[Mlir-commits] [mlir] [mlir][spirv] Expand support for TOSA Extended Instruction Set (00100… (PR #176908)
Davide Grohmann
llvmlistbot at llvm.org
Tue Jan 20 04:05:45 PST 2026
https://github.com/davidegrohmann created https://github.com/llvm/llvm-project/pull/176908
…0.1)
This patch expands support for the TOSA Extended Instruction Set (001000.1) to the SPIR-V dialect in MLIR. The TOSA extended instruction set provides a standardized set of machine learning operations designed to be used within `spirv.ARM.Graph` operations (corresponding to OpGraphARM in SPV_ARM_graph) and typed with `!spirv.arm.tensor<...>` (corresponding to OpTypeTensorARM in SPV_ARM_tensor).
The change introduces:
* Extending dialect plumbing for import, serialization, and deserialization of the TOSA extended instruction set.
* The `spirv.Tosa.*Conv*` convolution operation from TOSA extended instruction, each lowering to the corresponding `OpExtInst`.
* Verification enforcing that new convolution operations appears only within `spirv.ARM.Graph` regions, operates on `!spirv.arm.tensor<...>` types, and is well-formed according to the TOSA 001000.1 specification.
All convolution operations from TOSA 001000.1 extended instructions are introduced: [parser, printer, verifier, and round-trip tests using MLIR’s SPIR-V serialization/deserialization infrastructure are included.
This work aligns with Khronos SPIR-V TOSA specifications.
Specification:
https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html
Change-Id: I32ca642362dbad0cfb172f5738f8ff62b6745b85
>From 4d5a6fb202288a5075a6941a9e36a5a89025bb5b Mon Sep 17 00:00:00 2001
From: Davide Grohmann <davide.grohmann at arm.com>
Date: Tue, 20 Jan 2026 10:44:53 +0100
Subject: [PATCH] [mlir][spirv] Expand support for TOSA Extended Instruction
Set (001000.1)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This patch expands support for the TOSA Extended Instruction Set
(001000.1) to the SPIR-V dialect in MLIR. The TOSA extended
instruction set provides a standardized set of machine learning
operations designed to be used within `spirv.ARM.Graph` operations
(corresponding to OpGraphARM in SPV_ARM_graph) and typed with
`!spirv.arm.tensor<...>` (corresponding to OpTypeTensorARM in
SPV_ARM_tensor).
The change introduces:
* Extending dialect plumbing for import, serialization, and
deserialization of the TOSA extended instruction set.
* The `spirv.Tosa.*Conv*` convolution operation from TOSA extended
instruction, each lowering to the corresponding `OpExtInst`.
* Verification enforcing that new convolution operations appears only
within `spirv.ARM.Graph` regions, operates on
`!spirv.arm.tensor<...>` types, and is well-formed according to the
TOSA 001000.1 specification.
All convolution operations from TOSA 001000.1 extended instructions
are introduced: [parser, printer, verifier, and round-trip tests using
MLIR’s SPIR-V serialization/deserialization infrastructure are
included.
This work aligns with Khronos SPIR-V TOSA specifications.
Specification:
https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html
Signed-off-by: Davide Grohmann <davide.grohmann at arm.com>
Change-Id: I32ca642362dbad0cfb172f5738f8ff62b6745b85
---
.../mlir/Dialect/SPIRV/IR/SPIRVBase.td | 11 +
.../mlir/Dialect/SPIRV/IR/SPIRVTosaOps.td | 250 +++++++++++++
.../mlir/Dialect/SPIRV/IR/SPIRVTosaTypes.td | 29 ++
mlir/lib/Dialect/SPIRV/IR/SPIRVTosaOps.cpp | 90 +++++
.../SPIRV/IR/tosa-ops-verification.mlir | 337 ++++++++++++++++++
mlir/test/Dialect/SPIRV/IR/tosa-ops.mlir | 104 ++++++
mlir/test/Target/SPIRV/tosa-ops.mlir | 184 ++++++++++
mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp | 13 +-
8 files changed, 1016 insertions(+), 2 deletions(-)
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
index 21010d91dc47c..4ea6d784dd88f 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVBase.td
@@ -4915,6 +4915,17 @@ def SPIRV_FPFastMathModeAttr :
// SPIR-V TOSA enum definitions.
//===----------------------------------------------------------------------===//
+// NOTE: This is an attribute in the SPIR-V *dialect* but a constant (<id>) in
+// SPIR-V proper.
+def SPIRV_TosaExtAccTypeAttr : SPIRV_I32EnumAttr<
+ "TosaExtAccType", "Tosa Ext Acculumator Type", "tosa_ext_acc_type",
+ [
+ I32EnumAttrCase<"INT32", 1>,
+ I32EnumAttrCase<"FP16", 2>,
+ I32EnumAttrCase<"FP32", 3>,
+ I32EnumAttrCase<"INT48", 4>,
+ ]>;
+
// NOTE: This is an attribute in the SPIR-V *dialect* but a constant (<id>) in
// SPIR-V proper.
def SPIRV_TosaExtNaNPropagationModeAttr : SPIRV_I32EnumAttr<
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaOps.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaOps.td
index 6c6a318db4827..6fd368af6ec7e 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaOps.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaOps.td
@@ -83,4 +83,254 @@ def SPIRV_TosaArgMaxOp : SPIRV_TosaOp<"ArgMax", 0, [Pure]> {
}];
}
+
+def SPIRV_TosaConv2DOp : SPIRV_TosaOp<"Conv2D", 2, [Pure]> {
+ let summary = "2D Convolution operator.";
+
+ let description = [{
+ Performs a 2D convolution over the given tensor input, using the weight
+ tensor. Implementations may choose to skip calculation of multiplies in
+ the padding area.
+
+ References:
+ * https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html#_conv2d
+ * https://www.mlplatform.org/tosa/tosa_spec_1_0_1.html#_conv2d
+
+ #### Example:
+ ```mlir
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ ```
+ }];
+
+ let arguments = (ins
+ SPIRV_Int32_1DTensorArmOfLength4Attr: $pad,
+ SPIRV_Int32_1DTensorArmOfLength2Attr: $stride,
+ SPIRV_Int32_1DTensorArmOfLength2Attr: $dilation,
+ SPIRV_TosaExtAccTypeAttr: $acc_type,
+ SPIRV_BoolConstAttr: $local_bound,
+ SPIRV_TosaNumerical_TensorArm4D: $input,
+ SPIRV_TosaNumerical_TensorArm4D: $weight,
+ SPIRV_TosaNumerical_TensorArm1D: $bias,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $input_zp,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $weight_zp
+ );
+
+ let results = (outs
+ SPIRV_TosaNumerical_TensorArm4D: $output
+ );
+
+ let hasVerifier = 1;
+
+ let assemblyFormat = [{
+ `pad` `=` $pad `,` `stride` `=` $stride `,`
+ `dilation` `=` $dilation `,` `acc_type` `=` $acc_type `,`
+ `local_bound` `=` $local_bound `,`
+ $input `,` $weight `,` $bias `,` $input_zp `,` $weight_zp
+ attr-dict `:` type(operands) `->` type(results)
+ }];
+
+ let extraClassDeclaration = [{
+ ::mlir::spirv::TensorArmType getInputType() {
+ return cast<::mlir::spirv::TensorArmType>(getInput().getType());
+ }
+ ::mlir::spirv::TensorArmType getWeightType() {
+ return cast<::mlir::spirv::TensorArmType>(getWeight().getType());
+ }
+ ::mlir::spirv::TensorArmType getBiasType() {
+ return cast<::mlir::spirv::TensorArmType>(getBias().getType());
+ }
+ ::mlir::spirv::TensorArmType getResultType() {
+ return cast<::mlir::spirv::TensorArmType>(getType());
+ }
+ }];
+}
+
+
+def SPIRV_TosaConv3DOp : SPIRV_TosaOp<"Conv3D", 3, [Pure]> {
+ let summary = "3D Convolution operator.";
+
+ let description = [{
+ Performs a 3D convolution over the given input tensor. Implementations
+ may choose to skip calculation of multiplies in the padding area.
+
+ References:
+ * https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html#_conv3d
+ * https://www.mlplatform.org/tosa/tosa_spec_1_0_1.html#_conv3d
+
+ #### Example:
+ ```mlir
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x9x21x14x1xi8>, !spirv.arm.tensor<2x1x2x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x9x20x14x2xi32>
+ %7 = spirv.Tosa.Conv3D pad = dense<[0, 1, 1, 0, 0, 1]> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<[1, 1, 7]> : !spirv.arm.tensor<3xi32>, acc_type = <FP32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x2x65539x1x2xf32>, !spirv.arm.tensor<1x1x1x1x2xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x3x65540x2x1xf32>
+ ```
+ }];
+
+ let arguments = (ins
+ SPIRV_Int32_1DTensorArmOfLength6Attr: $pad,
+ SPIRV_Int32_1DTensorArmOfLength3Attr: $stride,
+ SPIRV_Int32_1DTensorArmOfLength3Attr: $dilation,
+ SPIRV_TosaExtAccTypeAttr: $acc_type,
+ SPIRV_BoolConstAttr: $local_bound,
+ SPIRV_TosaNumerical_TensorArm5D: $input,
+ SPIRV_TosaNumerical_TensorArm5D: $weight,
+ SPIRV_TosaNumerical_TensorArm1D: $bias,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $input_zp,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $weight_zp
+ );
+
+ let results = (outs
+ SPIRV_TosaNumerical_TensorArm5D: $output
+ );
+
+ let hasVerifier = 1;
+
+ let assemblyFormat = [{
+ `pad` `=` $pad `,` `stride` `=` $stride `,`
+ `dilation` `=` $dilation `,` `acc_type` `=` $acc_type `,`
+ `local_bound` `=` $local_bound `,`
+ $input `,` $weight `,` $bias `,` $input_zp `,` $weight_zp
+ attr-dict `:` type(operands) `->` type(results)
+ }];
+
+ let extraClassDeclaration = [{
+ ::mlir::spirv::TensorArmType getInputType() {
+ return cast<::mlir::spirv::TensorArmType>(getInput().getType());
+ }
+ ::mlir::spirv::TensorArmType getWeightType() {
+ return cast<::mlir::spirv::TensorArmType>(getWeight().getType());
+ }
+ ::mlir::spirv::TensorArmType getBiasType() {
+ return cast<::mlir::spirv::TensorArmType>(getBias().getType());
+ }
+ ::mlir::spirv::TensorArmType getResultType() {
+ return cast<::mlir::spirv::TensorArmType>(getType());
+ }
+ }];
+}
+
+
+def SPIRV_TosaDepthwiseConv2DOp : SPIRV_TosaOp<"DepthwiseConv2D", 4, [Pure]> {
+ let summary = "Depthwise 2D Convolution operator.";
+
+ let description = [{
+ Performs 2D convolutions separately over each channel of the given tensor
+ input, using the weight tensor. Implementations may choose to skip
+ calculation of multiplies in the padding area.
+
+ References:
+ * https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html#_depthwise_conv2d
+ * https://www.mlplatform.org/tosa/tosa_spec_1_0_1.html#_depthwise_conv2d
+
+ #### Example:
+ ```mlir
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<7> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x4x65537x1xi8>, !spirv.arm.tensor<1x3x1x4xi8>, !spirv.arm.tensor<4xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x4x32762x4xi32>
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[0, 1, 1, 1]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[1, 7]> : !spirv.arm.tensor<2xi32>, acc_type = <FP32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65540x1x3xf32>, !spirv.arm.tensor<1x1x3x1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x65541x2x3xf32>
+ ```
+ }];
+
+ let arguments = (ins
+ SPIRV_Int32_1DTensorArmOfLength4Attr: $pad,
+ SPIRV_Int32_1DTensorArmOfLength2Attr: $stride,
+ SPIRV_Int32_1DTensorArmOfLength2Attr: $dilation,
+ SPIRV_TosaExtAccTypeAttr: $acc_type,
+ SPIRV_BoolConstAttr: $local_bound,
+ SPIRV_TosaNumerical_TensorArm4D: $input,
+ SPIRV_TosaNumerical_TensorArm4D: $weight,
+ SPIRV_TosaNumerical_TensorArm1D: $bias,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $input_zp,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $weight_zp
+ );
+
+ let results = (outs
+ SPIRV_TosaNumerical_TensorArm4D: $output
+ );
+
+ let hasVerifier = 1;
+
+ let assemblyFormat = [{
+ `pad` `=` $pad `,` `stride` `=` $stride `,`
+ `dilation` `=` $dilation `,` `acc_type` `=` $acc_type `,`
+ `local_bound` `=` $local_bound `,`
+ $input `,` $weight `,` $bias `,` $input_zp `,` $weight_zp
+ attr-dict `:` type(operands) `->` type(results)
+ }];
+
+ let extraClassDeclaration = [{
+ ::mlir::spirv::TensorArmType getInputType() {
+ return cast<::mlir::spirv::TensorArmType>(getInput().getType());
+ }
+ ::mlir::spirv::TensorArmType getWeightType() {
+ return cast<::mlir::spirv::TensorArmType>(getWeight().getType());
+ }
+ ::mlir::spirv::TensorArmType getBiasType() {
+ return cast<::mlir::spirv::TensorArmType>(getBias().getType());
+ }
+ ::mlir::spirv::TensorArmType getResultType() {
+ return cast<::mlir::spirv::TensorArmType>(getType());
+ }
+ }];
+}
+
+
+def SPIRV_TosaTransposeConv2DOp : SPIRV_TosaOp<"TransposeConv2D", 9, [Pure]> {
+ let summary = "Transpose 2D Convolution operator.";
+
+ let description = [{
+ Performs a 2D transposed convolution over the given tensor input, using the
+ weights tensor. Implementations may choose to skip calculation of multiplies
+ by zero at fractional input positions.
+
+ References:
+ * https://github.khronos.org/SPIRV-Registry/extended/TOSA.001000.1.html#_transpose_conv2d
+ * https://www.mlplatform.org/tosa/tosa_spec_1_0_1.html#_transpose_conv2d
+
+ #### Example:
+ ```mlir
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<1x13x33x3xi16>, !spirv.arm.tensor<11x1x3x3xi8>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x13x35x11xi64>
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<[0, 1, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 8]> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<10x24x9x13xf16>, !spirv.arm.tensor<14x1x1x13xf16>, !spirv.arm.tensor<14xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<10x25x65x14xf16>
+ ```
+ }];
+
+ let arguments = (ins
+ SPIRV_Int32_1DTensorArmOfLength4Attr: $out_pad,
+ SPIRV_Int32_1DTensorArmOfLength2Attr: $stride,
+ SPIRV_TosaExtAccTypeAttr: $acc_type,
+ SPIRV_BoolConstAttr: $local_bound,
+ SPIRV_TosaNumerical_TensorArm4D: $input,
+ SPIRV_TosaNumerical_TensorArm4D: $weight,
+ SPIRV_TosaNumerical_TensorArm1D: $bias,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $input_zp,
+ SPIRV_TosaNumerical_1DTensorArmOfLength1: $weight_zp
+ );
+
+ let results = (outs
+ SPIRV_TosaNumerical_TensorArm4D: $output
+ );
+
+ let hasVerifier = 1;
+
+ let assemblyFormat = [{
+ `out_pad` `=` $out_pad `,` `stride` `=` $stride `,`
+ `acc_type` `=` $acc_type `,` `local_bound` `=` $local_bound `,`
+ $input `,` $weight `,` $bias `,` $input_zp `,` $weight_zp
+ attr-dict `:` type(operands) `->` type(results)
+ }];
+
+ let extraClassDeclaration = [{
+ ::mlir::spirv::TensorArmType getInputType() {
+ return cast<::mlir::spirv::TensorArmType>(getInput().getType());
+ }
+ ::mlir::spirv::TensorArmType getWeightType() {
+ return cast<::mlir::spirv::TensorArmType>(getWeight().getType());
+ }
+ ::mlir::spirv::TensorArmType getBiasType() {
+ return cast<::mlir::spirv::TensorArmType>(getBias().getType());
+ }
+ ::mlir::spirv::TensorArmType getResultType() {
+ return cast<::mlir::spirv::TensorArmType>(getType());
+ }
+ }];
+}
+
+
#endif // MLIR_DIALECT_SPIRV_IR_TOSA_OPS
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaTypes.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaTypes.td
index e731388182eb4..7e2c37f74b437 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaTypes.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVTosaTypes.td
@@ -13,6 +13,7 @@
#ifndef MLIR_DIALECT_SPIRV_IR_TOSA_TYPES
#define MLIR_DIALECT_SPIRV_IR_TOSA_TYPES
+include "mlir/IR/CommonAttrConstraints.td"
include "mlir/Dialect/SPIRV/IR/SPIRVBase.td"
def SPIRV_TosaInteger : AnyIntOfWidths<[8, 16, 32, 64]>;
@@ -21,6 +22,7 @@ def SPIRV_TosaNumerical : AnyTypeOf<[SPIRV_TosaInteger, SPIRV_TosaFloat]>;
def SPIRV_TosaAny : AnyTypeOf<[SPIRV_TosaNumerical, SPIRV_Bool]>;
def SPIRV_TensorArmAxisAttr : ConfinedAttr<I32Attr, [IntNonNegative, IntMaxValue<5>]>;
+def SPIRV_BoolConstAttr : ConfinedAttr<BoolAttr, []>;
// TensorARM Types
@@ -35,7 +37,34 @@ class TensorArmRankOf<list<Type> allowedTypes, list<int> ranks>
[HasAnyRankOfPred<ranks>],
!interleave(!foreach(rank, ranks, rank # "D"), "/") # " tensorArm">;
+def SPIRV_TosaNumerical_TensorArm1D : TensorArmRankOf<[SPIRV_TosaNumerical], [1]>;
+def SPIRV_TosaNumerical_TensorArm4D : TensorArmRankOf<[SPIRV_TosaNumerical], [4]>;
+def SPIRV_TosaNumerical_TensorArm5D : TensorArmRankOf<[SPIRV_TosaNumerical], [5]>;
+
def SPIRV_TosaNumerical_TensorArm : TensorArmRankOf<[SPIRV_TosaNumerical], [1, 2, 3, 4, 5, 6]>;
def SPIRV_Int32_TensorArmUpTo5D : TensorArmRankOf<[SPIRV_Int32], [1, 2, 3, 4, 5]>;
+class Is1DTensorArmOfLength<list<int> allowedLengths> :
+ And<[HasAnyRankOfPred<[1]>,
+ Or<!foreach(allowedlength, allowedLengths,
+ CPred<[{::llvm::cast<::mlir::spirv::TensorArmType>($_self).getShape()[0] == }]
+ # allowedlength>)>]>;
+
+class SPIRV_1DTensorArmOfLengthAndType<list<int> allowedLengths, list<Type> allowedTypes> :
+ ContainerType<AnyTypeOf<allowedTypes>, Is1DTensorArmOfLength<allowedLengths>,
+ "::llvm::cast<::mlir::spirv::TensorArmType>($_self).getElementType()",
+ "rank 1 tensorArm of length " # !interleave(allowedLengths, "/"),
+ "::mlir::spirv::TensorArmType">;
+
+def SPIRV_DenseElementAttrsWithTensorArmType : AttrConstraint<
+ CPred<"::llvm::isa<::mlir::spirv::TensorArmType>(::llvm::cast<::mlir::DenseElementsAttr>($_self).getType())">,
+ "Attr with type = spirv::TensorArmType">;
+
+def SPIRV_Int32_1DTensorArmOfLength2Attr : ConfinedAttr<RankedI32ElementsAttr<[2]>, [SPIRV_DenseElementAttrsWithTensorArmType]>;
+def SPIRV_Int32_1DTensorArmOfLength3Attr : ConfinedAttr<RankedI32ElementsAttr<[3]>, [SPIRV_DenseElementAttrsWithTensorArmType]>;
+def SPIRV_Int32_1DTensorArmOfLength4Attr : ConfinedAttr<RankedI32ElementsAttr<[4]>, [SPIRV_DenseElementAttrsWithTensorArmType]>;
+def SPIRV_Int32_1DTensorArmOfLength6Attr : ConfinedAttr<RankedI32ElementsAttr<[6]>, [SPIRV_DenseElementAttrsWithTensorArmType]>;
+
+def SPIRV_TosaNumerical_1DTensorArmOfLength1 : SPIRV_1DTensorArmOfLengthAndType<[1], [SPIRV_TosaNumerical]>;
+
#endif // MLIR_DIALECT_SPIRV_IR_TOSA_TYPES
diff --git a/mlir/lib/Dialect/SPIRV/IR/SPIRVTosaOps.cpp b/mlir/lib/Dialect/SPIRV/IR/SPIRVTosaOps.cpp
index 4f3c91d4a1c12..0a755bf7c1b00 100644
--- a/mlir/lib/Dialect/SPIRV/IR/SPIRVTosaOps.cpp
+++ b/mlir/lib/Dialect/SPIRV/IR/SPIRVTosaOps.cpp
@@ -20,6 +20,72 @@ namespace mlir::spirv {
// TOSA Operator Verifiers.
//===----------------------------------------------------------------------===//
+namespace {
+
+template <typename T>
+static LogicalResult verifyConvOp(T op) {
+ ShapedType inputTy = op.getInputType();
+ ShapedType biasTy = op.getBiasType();
+ ShapedType resultTy = op.getResultType();
+
+ Type inputETy = inputTy.getElementType();
+ Type biasETy = biasTy.getElementType();
+ Type resultETy = resultTy.getElementType();
+
+ if (inputETy.isInteger() && !inputETy.isInteger(8) &&
+ !inputETy.isInteger(16)) {
+ return op.emitOpError(
+ "input element type can only be of width 8 or 16 when integer type");
+ }
+
+ if (inputETy.isInteger(8) && !resultETy.isInteger(32)) {
+ return op.emitOpError("expect result type to be i32, got ") << resultETy;
+ }
+
+ if (inputETy.isInteger(16) && !resultETy.isInteger(64)) {
+ return op.emitOpError("expect result type to be i64, got ") << resultETy;
+ }
+
+ if (inputETy.isF16() && !resultETy.isF16()) {
+ return op.emitOpError("expect result type to be f16, got ") << resultETy;
+ }
+
+ if (inputETy.isF32() && !resultETy.isF32()) {
+ return op.emitOpError("expect result type to be f32, got ") << resultETy;
+ }
+
+ if (biasETy != resultETy) {
+ return op.emitOpError("element types of bias and result must be the same");
+ }
+
+ TosaExtAccType accType = op.getAccType();
+ if (inputETy.isInteger(8) && accType != TosaExtAccType::INT32) {
+ return op.emitOpError("accumulator type for i8 tensorARM is not i32");
+ }
+
+ if (inputETy.isInteger(16) && accType != TosaExtAccType::INT48) {
+ return op.emitOpError("accumulator type for i16 tensorARM is not i48");
+ }
+
+ if (inputETy.isF16() &&
+ !(accType == TosaExtAccType::FP16 || accType == TosaExtAccType::FP32)) {
+ return op.emitOpError(
+ "accumulator type for f16 tensorARM is not f16 or f32");
+ }
+
+ if (inputETy.isBF16() && accType != TosaExtAccType::FP32) {
+ return op.emitOpError("accumulator type for bf16 tensorARM is not f32");
+ }
+
+ if (inputETy.isF32() && accType != TosaExtAccType::FP32) {
+ return op.emitOpError("accumulator type for f32 tensorARM is not f32");
+ }
+
+ return success();
+}
+
+} // namespace
+
//===----------------------------------------------------------------------===//
// spirv.TosaArgmaxOp
//===----------------------------------------------------------------------===//
@@ -46,4 +112,28 @@ LogicalResult TosaArgMaxOp::verify() {
return success();
}
+//===----------------------------------------------------------------------===//
+// spirv.TosaConv2DOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult TosaConv2DOp::verify() { return verifyConvOp(*this); }
+
+//===----------------------------------------------------------------------===//
+// spirv.TosaConv3DOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult TosaConv3DOp::verify() { return verifyConvOp(*this); }
+
+//===----------------------------------------------------------------------===//
+// SPIRV Tosa DepthwiseConv2D Ops:
+//===----------------------------------------------------------------------===//
+
+LogicalResult TosaDepthwiseConv2DOp::verify() { return verifyConvOp(*this); }
+
+//===----------------------------------------------------------------------===//
+// SPIRV Tosa TransposeConv2D Ops:
+//===----------------------------------------------------------------------===//
+
+LogicalResult TosaTransposeConv2DOp::verify() { return verifyConvOp(*this); }
+
} // namespace mlir::spirv
diff --git a/mlir/test/Dialect/SPIRV/IR/tosa-ops-verification.mlir b/mlir/test/Dialect/SPIRV/IR/tosa-ops-verification.mlir
index a6496316f9881..cb7863ba9c1a0 100644
--- a/mlir/test/Dialect/SPIRV/IR/tosa-ops-verification.mlir
+++ b/mlir/test/Dialect/SPIRV/IR/tosa-ops-verification.mlir
@@ -21,3 +21,340 @@ spirv.ARM.Graph @argmax_axis_value_not_in_input_rank_range(%arg0: !spirv.arm.ten
%2 = spirv.Tosa.ArgMax axis = 4, nan_mode = <Propagate>, %arg0 : !spirv.arm.tensor<3x28x17x17xi8> -> !spirv.arm.tensor<3x28x17xi32>
spirv.ARM.GraphOutputs %2 : !spirv.arm.tensor<3x28x17xi32>
}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv2D
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv2d_wrong_input_integer_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi32>, %arg1: !spirv.arm.tensor<7x1x1x1xi32>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi64>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi64>
+ // expected-error @+1 {{op input element type can only be of width 8 or 16 when integer type}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi32>, !spirv.arm.tensor<7x1x1x1xi32>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @conv2d_mismatch_result_element_type_i8_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi16>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i32, got 'i16'}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi16>
+}
+
+spirv.ARM.Graph @conv2d_mismatch_result_element_type_i16_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i64, got 'i32'}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @conv2d_mismatch_result_element_type_f16_input(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op expect result type to be f16, got 'f32'}}
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
+
+spirv.ARM.Graph @conv2d_mismatch_result_element_type_f32_input(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op expect result type to be f32, got 'f16'}}
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @conv2d_bias_element_type_must_be_same_as_result_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op element types of bias and result must be the same}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @conv2d_accumulator_must_be_INT32_for_i8_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op accumulator type for i8 tensorARM is not i32}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @conv2d_accumulator_must_be_INT48_for_i16_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi16>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi16>
+ // expected-error @+1 {{op accumulator type for i16 tensorARM is not i48}}
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi16> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @conv2d_accumulator_must_be_either_FP16_or_FP32_for_f16_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op accumulator type for f16 tensorARM is not f16 or f32}}
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @conv2d_accumulator_must_be_either_FP32_for_f32_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op accumulator type for f32 tensorARM is not f32}}
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
+
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv3D
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv3d_wrong_input_integer_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi32>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi32>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7x1xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi64>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi64>
+ // expected-error @+1 {{op input element type can only be of width 8 or 16 when integer type}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi32>, !spirv.arm.tensor<7x1x1x1x1xi32>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64> -> !spirv.arm.tensor<1x65536x2x7x1xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi64>
+}
+
+spirv.ARM.Graph @conv3d_mismatch_result_element_type_i8_input(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7x1xi16>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i32, got 'i16'}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi8>, !spirv.arm.tensor<7x1x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7x1xi16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi16>
+}
+
+spirv.ARM.Graph @conv3d_mismatch_result_element_type_i16_input(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7x1xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i64, got 'i32'}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi16>, !spirv.arm.tensor<7x1x1x1x1xi16>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7x1xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi32>
+}
+
+spirv.ARM.Graph @conv3d_mismatch_result_element_type_f16_input(%arg0: !spirv.arm.tensor<1x34x18x27x1xf16>, %arg1: !spirv.arm.tensor<11x1x1x27x1xf16>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11x1xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op expect result type to be f16, got 'f32'}}
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27x1xf16>, !spirv.arm.tensor<11x1x1x27x1xf16>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11x1xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11x1xf32>
+}
+
+spirv.ARM.Graph @conv3d_mismatch_result_element_type_f32_input(%arg0: !spirv.arm.tensor<1x34x18x27x1xf32>, %arg1: !spirv.arm.tensor<11x1x1x27x1xf32>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11x1xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op expect result type to be f32, got 'f16'}}
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27x1xf32>, !spirv.arm.tensor<11x1x1x27x1xf32>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11x1xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11x1xf16>
+}
+
+spirv.ARM.Graph @conv3d_bias_element_type_must_be_same_as_result_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7x1xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op element types of bias and result must be the same}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi8>, !spirv.arm.tensor<7x1x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7x1xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi32>
+}
+
+spirv.ARM.Graph @conv3d_accumulator_must_be_INT32_for_i8_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7x1xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op accumulator type for i8 tensorARM is not i32}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi8>, !spirv.arm.tensor<7x1x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7x1xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi32>
+}
+
+spirv.ARM.Graph @conv3d_accumulator_must_be_INT48_for_i16_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7x1xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi16>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi16>
+ // expected-error @+1 {{op accumulator type for i16 tensorARM is not i48}}
+ %7 = spirv.Tosa.Conv3D pad = dense<[1, 0, 0, 0, 0, 0]> : !spirv.arm.tensor<6xi32>, stride = dense<[1, 2, 3]> : !spirv.arm.tensor<3xi32>, dilation = dense<[7, 1, 1]> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1x1xi16>, !spirv.arm.tensor<7x1x1x1x1xi16>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi16> -> !spirv.arm.tensor<1x65536x2x7x1xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7x1xi64>
+}
+
+spirv.ARM.Graph @conv3d_accumulator_must_be_either_FP16_or_FP32_for_f16_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27x1xf16>, %arg1: !spirv.arm.tensor<11x1x1x27x1xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11x1xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op accumulator type for f16 tensorARM is not f16 or f32}}
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27x1xf16>, !spirv.arm.tensor<11x1x1x27x1xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11x1xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11x1xf16>
+}
+
+spirv.ARM.Graph @conv3d_accumulator_must_be_either_FP32_for_f32_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27x1xf32>, %arg1: !spirv.arm.tensor<11x1x1x27x1xf32>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11x1xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op accumulator type for f32 tensorARM is not f32}}
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27x1xf32>, !spirv.arm.tensor<11x1x1x27x1xf32>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11x1xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11x1xf32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.DepthwiseConv2D
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @depthwise_conv2d_wrong_input_integer_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi32>, %arg1: !spirv.arm.tensor<7x1x1x1xi32>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi64>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi64>
+ // expected-error @+1 {{op input element type can only be of width 8 or 16 when integer type}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi32>, !spirv.arm.tensor<7x1x1x1xi32>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_mismatch_result_element_type_i8_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi16>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i32, got 'i16'}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi16>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_mismatch_result_element_type_i16_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i64, got 'i32'}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_mismatch_result_element_type_f16_input(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op expect result type to be f16, got 'f32'}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_mismatch_result_element_type_f32_input(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op expect result type to be f32, got 'f16'}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_bias_element_type_must_be_same_as_result_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op element types of bias and result must be the same}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_accumulator_must_be_INT32_for_i8_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op accumulator type for i8 tensorARM is not i32}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_accumulator_must_be_INT48_for_i16_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi16>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi16>
+ // expected-error @+1 {{op accumulator type for i16 tensorARM is not i48}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi16> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_accumulator_must_be_either_FP16_or_FP32_for_f16_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op accumulator type for f16 tensorARM is not f16 or f32}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @depthwise_conv2d_accumulator_must_be_either_FP32_for_f32_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op accumulator type for f32 tensorARM is not f32}}
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.TransposeConv2D
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @transpose_conv2d_wrong_input_integer_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi32>, %arg1: !spirv.arm.tensor<7x1x1x1xi32>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi64>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi64>
+ // expected-error @+1 {{op input element type can only be of width 8 or 16 when integer type}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi32>, !spirv.arm.tensor<7x1x1x1xi32>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi64> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @transpose_conv2d_mismatch_result_element_type_i8_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi16>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i32, got 'i16'}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi16>
+}
+
+spirv.ARM.Graph @transpose_conv2d_mismatch_result_element_type_i16_input(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op expect result type to be i64, got 'i32'}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @transpose_conv2d_mismatch_result_element_type_f16_input(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op expect result type to be f16, got 'f32'}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
+
+spirv.ARM.Graph @transpose_conv2d_mismatch_result_element_type_f32_input(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op expect result type to be f32, got 'f16'}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @transpose_conv2d_bias_element_type_must_be_same_as_result_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi16>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op element types of bias and result must be the same}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @transpose_conv2d_accumulator_must_be_INT32_for_i8_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // expected-error @+1 {{op accumulator type for i8 tensorARM is not i32}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+spirv.ARM.Graph @transpose_conv2d_accumulator_must_be_INT48_for_i16_input_element_type(%arg0: !spirv.arm.tensor<1x65535x3x1xi16>, %arg1: !spirv.arm.tensor<7x1x1x1xi16>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x65536x2x7xi64>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi16>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi16>
+ // expected-error @+1 {{op accumulator type for i16 tensorARM is not i48}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi16>, !spirv.arm.tensor<7x1x1x1xi16>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi16> -> !spirv.arm.tensor<1x65536x2x7xi64>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi64>
+}
+
+spirv.ARM.Graph @transpose_conv2d_accumulator_must_be_either_FP16_or_FP32_for_f16_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // expected-error @+1 {{op accumulator type for f16 tensorARM is not f16 or f32}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+spirv.ARM.Graph @transpose_conv2d_accumulator_must_be_either_FP32_for_f32_input_element_type(%arg0: !spirv.arm.tensor<1x34x18x27xf32>, %arg1: !spirv.arm.tensor<11x1x1x27xf32>, %arg2: !spirv.arm.tensor<11xf32>) -> (!spirv.arm.tensor<1x34x18x11xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // expected-error @+1 {{op accumulator type for f32 tensorARM is not f32}}
+ %7 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf32>, !spirv.arm.tensor<11x1x1x27xf32>, !spirv.arm.tensor<11xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x34x18x11xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf32>
+}
diff --git a/mlir/test/Dialect/SPIRV/IR/tosa-ops.mlir b/mlir/test/Dialect/SPIRV/IR/tosa-ops.mlir
index c9832b903b79e..3d6cb8a29eb38 100644
--- a/mlir/test/Dialect/SPIRV/IR/tosa-ops.mlir
+++ b/mlir/test/Dialect/SPIRV/IR/tosa-ops.mlir
@@ -21,3 +21,107 @@ spirv.ARM.Graph @argmax_fp(%arg0: !spirv.arm.tensor<2x2x7x14xf32>) -> (!spirv.ar
// CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<2x2x14xi32>
spirv.ARM.GraphOutputs %2 : !spirv.arm.tensor<2x2x14xi32>
}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv2d_int(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv2d_fp(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv3D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv3d_int(%arg0: !spirv.arm.tensor<1x9x21x14x1xi8>, %arg1: !spirv.arm.tensor<2x1x2x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x9x20x14x2xi32>) {
+ %5 = spirv.Constant dense<123> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<121> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x9x21x14x1xi8>, !spirv.arm.tensor<2x1x2x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x9x20x14x2xi32>
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x9x21x14x1xi8>, !spirv.arm.tensor<2x1x2x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x9x20x14x2xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x9x20x14x2xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x9x20x14x2xi32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv3D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @conv3d_fp(%arg0: !spirv.arm.tensor<1x2x65539x1x2xf32>, %arg1: !spirv.arm.tensor<1x1x1x1x2xf32>, %arg2: !spirv.arm.tensor<1xf32>) -> (!spirv.arm.tensor<1x3x65540x2x1xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv3D pad = dense<[0, 1, 1, 0, 0, 1]> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<[1, 1, 7]> : !spirv.arm.tensor<3xi32>, acc_type = <FP32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x2x65539x1x2xf32>, !spirv.arm.tensor<1x1x1x1x2xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x3x65540x2x1xf32>
+ %7 = spirv.Tosa.Conv3D pad = dense<[0, 1, 1, 0, 0, 1]> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<[1, 1, 7]> : !spirv.arm.tensor<3xi32>, acc_type = <FP32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x2x65539x1x2xf32>, !spirv.arm.tensor<1x1x1x1x2xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x3x65540x2x1xf32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x3x65540x2x1xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x3x65540x2x1xf32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.DepthwiseConv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @depthwiseconv2d_int(%arg0: !spirv.arm.tensor<1x4x65537x1xi8>, %arg1: !spirv.arm.tensor<1x3x1x4xi8>, %arg2: !spirv.arm.tensor<4xi32>) -> (!spirv.arm.tensor<1x4x32762x4xi32>) {
+ %5 = spirv.Constant dense<58> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<-106> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<7> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x4x65537x1xi8>, !spirv.arm.tensor<1x3x1x4xi8>, !spirv.arm.tensor<4xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x4x32762x4xi32>
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<7> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x4x65537x1xi8>, !spirv.arm.tensor<1x3x1x4xi8>, !spirv.arm.tensor<4xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x4x32762x4xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x4x32762x4xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x4x32762x4xi32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.DepthwiseConv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @depthwiseconv2d_fp(%arg0: !spirv.arm.tensor<1x65540x1x3xf32>, %arg1: !spirv.arm.tensor<1x1x3x1xf32>, %arg2: !spirv.arm.tensor<1xf32>) -> (!spirv.arm.tensor<1x65541x2x3xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // CHECK: {{%.*}} = spirv.Tosa.DepthwiseConv2D pad = dense<[0, 1, 1, 1]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[1, 7]> : !spirv.arm.tensor<2xi32>, acc_type = <FP32>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x65540x1x3xf32>, !spirv.arm.tensor<1x1x3x1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x65541x2x3xf32>
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[0, 1, 1, 1]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[1, 7]> : !spirv.arm.tensor<2xi32>, acc_type = <FP32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65540x1x3xf32>, !spirv.arm.tensor<1x1x3x1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x65541x2x3xf32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x65541x2x3xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65541x2x3xf32>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.TransposeConv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @transposeconv2d_int(%arg0: !spirv.arm.tensor<1x13x33x3xi16>, %arg1: !spirv.arm.tensor<11x1x3x3xi8>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x13x35x11xi64>) {
+ %4 = spirv.Constant dense<0> : !spirv.arm.tensor<1xi16>
+ %5 = spirv.Constant dense<88> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x13x33x3xi16>, !spirv.arm.tensor<11x1x3x3xi8>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x13x35x11xi64>
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<1x13x33x3xi16>, !spirv.arm.tensor<11x1x3x3xi8>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x13x35x11xi64>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x13x35x11xi64>
+ spirv.ARM.GraphOutputs %6 : !spirv.arm.tensor<1x13x35x11xi64>
+}
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.TransposeConv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+spirv.ARM.Graph @transposeconv2d_fp(%arg0: !spirv.arm.tensor<10x24x9x13xf16>, %arg1: !spirv.arm.tensor<14x1x1x13xf16>, %arg2: !spirv.arm.tensor<14xf16>) -> (!spirv.arm.tensor<10x25x65x14xf16>) {
+ %4 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // CHECK: {{%.*}} = spirv.Tosa.TransposeConv2D out_pad = dense<[0, 1, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 8]> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<10x24x9x13xf16>, !spirv.arm.tensor<14x1x1x13xf16>, !spirv.arm.tensor<14xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<10x25x65x14xf16>
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<[0, 1, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 8]> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<10x24x9x13xf16>, !spirv.arm.tensor<14x1x1x13xf16>, !spirv.arm.tensor<14xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<10x25x65x14xf16>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<10x25x65x14xf16>
+ spirv.ARM.GraphOutputs %6 : !spirv.arm.tensor<10x25x65x14xf16>
+}
diff --git a/mlir/test/Target/SPIRV/tosa-ops.mlir b/mlir/test/Target/SPIRV/tosa-ops.mlir
index 8c0429bca68e4..38ff0535b3b11 100644
--- a/mlir/test/Target/SPIRV/tosa-ops.mlir
+++ b/mlir/test/Target/SPIRV/tosa-ops.mlir
@@ -39,3 +39,187 @@ spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader
spirv.ARM.GraphOutputs %2 : !spirv.arm.tensor<2x2x14xi32>
}
}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @conv2d_int_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x65535x3x1xi8>, UniformConstant>
+ spirv.GlobalVariable @conv2d_int_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<7x1x1x1xi8>, UniformConstant>
+ spirv.GlobalVariable @conv2d_int_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<1xi32>, UniformConstant>
+ spirv.GlobalVariable @conv2d_int_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x65536x2x7xi32>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @conv2d_int, @conv2d_int_arg_0, @conv2d_int_arg_1, @conv2d_int_arg_2, @conv2d_int_res_0
+ spirv.ARM.Graph @conv2d_int(%arg0: !spirv.arm.tensor<1x65535x3x1xi8>, %arg1: !spirv.arm.tensor<7x1x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x65536x2x7xi32>) {
+ %5 = spirv.Constant dense<35> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<57> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ %7 = spirv.Tosa.Conv2D pad = dense<[1, 0, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[7, 1]> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65535x3x1xi8>, !spirv.arm.tensor<7x1x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x65536x2x7xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x65536x2x7xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65536x2x7xi32>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @conv2d_fp_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x34x18x27xf16>, UniformConstant>
+ spirv.GlobalVariable @conv2d_fp_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<11x1x1x27xf16>, UniformConstant>
+ spirv.GlobalVariable @conv2d_fp_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<11xf16>, UniformConstant>
+ spirv.GlobalVariable @conv2d_fp_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x34x18x11xf16>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @conv2d_fp, @conv2d_fp_arg_0, @conv2d_fp_arg_1, @conv2d_fp_arg_2, @conv2d_fp_res_0
+ spirv.ARM.Graph @conv2d_fp(%arg0: !spirv.arm.tensor<1x34x18x27xf16>, %arg1: !spirv.arm.tensor<11x1x1x27xf16>, %arg2: !spirv.arm.tensor<11xf16>) -> (!spirv.arm.tensor<1x34x18x11xf16>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ %7 = spirv.Tosa.Conv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, dilation = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x34x18x27xf16>, !spirv.arm.tensor<11x1x1x27xf16>, !spirv.arm.tensor<11xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<1x34x18x11xf16>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x34x18x11xf16>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x34x18x11xf16>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv3D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @conv3d_int_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x9x21x14x1xi8>, UniformConstant>
+ spirv.GlobalVariable @conv3d_int_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<2x1x2x1x1xi8>, UniformConstant>
+ spirv.GlobalVariable @conv3d_int_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<1xi32>, UniformConstant>
+ spirv.GlobalVariable @conv3d_int_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x9x20x14x2xi32>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @conv3d_int, @conv3d_int_arg_0, @conv3d_int_arg_1, @conv3d_int_arg_2, @conv3d_int_res_0
+ spirv.ARM.Graph @conv3d_int(%arg0: !spirv.arm.tensor<1x9x21x14x1xi8>, %arg1: !spirv.arm.tensor<2x1x2x1x1xi8>, %arg2: !spirv.arm.tensor<1xi32>) -> (!spirv.arm.tensor<1x9x20x14x2xi32>) {
+ %5 = spirv.Constant dense<123> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<121> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x9x21x14x1xi8>, !spirv.arm.tensor<2x1x2x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x9x20x14x2xi32>
+ %7 = spirv.Tosa.Conv3D pad = dense<0> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<1> : !spirv.arm.tensor<3xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x9x21x14x1xi8>, !spirv.arm.tensor<2x1x2x1x1xi8>, !spirv.arm.tensor<1xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x9x20x14x2xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x9x20x14x2xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x9x20x14x2xi32>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.Conv3D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @conv3d_fp_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x2x65539x1x2xf32>, UniformConstant>
+ spirv.GlobalVariable @conv3d_fp_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<1x1x1x1x2xf32>, UniformConstant>
+ spirv.GlobalVariable @conv3d_fp_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<1xf32>, UniformConstant>
+ spirv.GlobalVariable @conv3d_fp_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x3x65540x2x1xf32>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @conv3d_fp, @conv3d_fp_arg_0, @conv3d_fp_arg_1, @conv3d_fp_arg_2, @conv3d_fp_res_0
+ spirv.ARM.Graph @conv3d_fp(%arg0: !spirv.arm.tensor<1x2x65539x1x2xf32>, %arg1: !spirv.arm.tensor<1x1x1x1x2xf32>, %arg2: !spirv.arm.tensor<1xf32>) -> (!spirv.arm.tensor<1x3x65540x2x1xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // CHECK: {{%.*}} = spirv.Tosa.Conv3D pad = dense<[0, 1, 1, 0, 0, 1]> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<[1, 1, 7]> : !spirv.arm.tensor<3xi32>, acc_type = <FP32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x2x65539x1x2xf32>, !spirv.arm.tensor<1x1x1x1x2xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x3x65540x2x1xf32>
+ %7 = spirv.Tosa.Conv3D pad = dense<[0, 1, 1, 0, 0, 1]> : !spirv.arm.tensor<6xi32>, stride = dense<1> : !spirv.arm.tensor<3xi32>, dilation = dense<[1, 1, 7]> : !spirv.arm.tensor<3xi32>, acc_type = <FP32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x2x65539x1x2xf32>, !spirv.arm.tensor<1x1x1x1x2xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x3x65540x2x1xf32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x3x65540x2x1xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x3x65540x2x1xf32>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.DepthwiseConv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @depthwiseconv2d_int_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x4x65537x1xi8>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_int_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<1x3x1x4xi8>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_int_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<4xi32>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_int_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x4x32762x4xi32>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @depthwiseconv2d_int, @depthwiseconv2d_int_arg_0, @depthwiseconv2d_int_arg_1, @depthwiseconv2d_int_arg_2, @depthwiseconv2d_int_res_0
+ spirv.ARM.Graph @depthwiseconv2d_int(%arg0: !spirv.arm.tensor<1x4x65537x1xi8>, %arg1: !spirv.arm.tensor<1x3x1x4xi8>, %arg2: !spirv.arm.tensor<4xi32>) -> (!spirv.arm.tensor<1x4x32762x4xi32>) {
+ %5 = spirv.Constant dense<58> : !spirv.arm.tensor<1xi8>
+ %6 = spirv.Constant dense<-106> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<7> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x4x65537x1xi8>, !spirv.arm.tensor<1x3x1x4xi8>, !spirv.arm.tensor<4xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x4x32762x4xi32>
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<7> : !spirv.arm.tensor<2xi32>, acc_type = <INT32>, local_bound = false, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x4x65537x1xi8>, !spirv.arm.tensor<1x3x1x4xi8>, !spirv.arm.tensor<4xi32>, !spirv.arm.tensor<1xi8>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x4x32762x4xi32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x4x32762x4xi32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x4x32762x4xi32>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.DepthwiseConv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @depthwiseconv2d_fp_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x65540x1x3xf32>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_fp_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<1x1x3x1xf32>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_fp_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<1xf32>, UniformConstant>
+ spirv.GlobalVariable @depthwiseconv2d_fp_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x65541x2x3xf32>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @depthwiseconv2d_fp, @depthwiseconv2d_fp_arg_0, @depthwiseconv2d_fp_arg_1, @depthwiseconv2d_fp_arg_2, @depthwiseconv2d_fp_res_0
+ spirv.ARM.Graph @depthwiseconv2d_fp(%arg0: !spirv.arm.tensor<1x65540x1x3xf32>, %arg1: !spirv.arm.tensor<1x1x3x1xf32>, %arg2: !spirv.arm.tensor<1xf32>) -> (!spirv.arm.tensor<1x65541x2x3xf32>) {
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ %6 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf32>
+ // CHECK: {{%.*}} = spirv.Tosa.DepthwiseConv2D pad = dense<[0, 1, 1, 1]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[1, 7]> : !spirv.arm.tensor<2xi32>, acc_type = <FP32>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x65540x1x3xf32>, !spirv.arm.tensor<1x1x3x1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x65541x2x3xf32>
+ %7 = spirv.Tosa.DepthwiseConv2D pad = dense<[0, 1, 1, 1]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 2]> : !spirv.arm.tensor<2xi32>, dilation = dense<[1, 7]> : !spirv.arm.tensor<2xi32>, acc_type = <FP32>, local_bound = true, %arg0, %arg1, %arg2, %5, %6 : !spirv.arm.tensor<1x65540x1x3xf32>, !spirv.arm.tensor<1x1x3x1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32>, !spirv.arm.tensor<1xf32> -> !spirv.arm.tensor<1x65541x2x3xf32>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x65541x2x3xf32>
+ spirv.ARM.GraphOutputs %7 : !spirv.arm.tensor<1x65541x2x3xf32>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.TransposeConv2D - PRO-INT
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @transposeconv2d_int_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<1x13x33x3xi16>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_int_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<11x1x3x3xi8>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_int_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<1xi64>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_int_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<1x13x35x11xi64>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @transposeconv2d_int, @transposeconv2d_int_arg_0, @transposeconv2d_int_arg_1, @transposeconv2d_int_arg_2, @transposeconv2d_int_res_0
+ spirv.ARM.Graph @transposeconv2d_int(%arg0: !spirv.arm.tensor<1x13x33x3xi16>, %arg1: !spirv.arm.tensor<11x1x3x3xi8>, %arg2: !spirv.arm.tensor<1xi64>) -> (!spirv.arm.tensor<1x13x35x11xi64>) {
+ %4 = spirv.Constant dense<0> : !spirv.arm.tensor<1xi16>
+ %5 = spirv.Constant dense<88> : !spirv.arm.tensor<1xi8>
+ // CHECK: {{%.*}} = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<1x13x33x3xi16>, !spirv.arm.tensor<11x1x3x3xi8>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x13x35x11xi64>
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<0> : !spirv.arm.tensor<4xi32>, stride = dense<1> : !spirv.arm.tensor<2xi32>, acc_type = <INT48>, local_bound = false, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<1x13x33x3xi16>, !spirv.arm.tensor<11x1x3x3xi8>, !spirv.arm.tensor<1xi64>, !spirv.arm.tensor<1xi16>, !spirv.arm.tensor<1xi8> -> !spirv.arm.tensor<1x13x35x11xi64>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<1x13x35x11xi64>
+ spirv.ARM.GraphOutputs %6 : !spirv.arm.tensor<1x13x35x11xi64>
+ }
+}
+
+// -----
+
+//===----------------------------------------------------------------------===//
+// spirv.TOSA.TransposeConv2D - PRO-FP
+//===----------------------------------------------------------------------===//
+
+// CHECK: spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]>
+spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int8, Int16, Int64, Float16, TensorsARM, GraphARM], [SPV_ARM_tensors, SPV_ARM_graph, SPV_KHR_vulkan_memory_model]> {
+ spirv.GlobalVariable @transposeconv2d_fp_arg_0 bind(0, 0) : !spirv.ptr<!spirv.arm.tensor<10x24x9x13xf16>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_fp_arg_1 bind(0, 1) : !spirv.ptr<!spirv.arm.tensor<14x1x1x13xf16>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_fp_arg_2 bind(0, 2) : !spirv.ptr<!spirv.arm.tensor<14xf16>, UniformConstant>
+ spirv.GlobalVariable @transposeconv2d_fp_res_0 bind(1, 0) : !spirv.ptr<!spirv.arm.tensor<10x25x65x14xf16>, UniformConstant>
+ spirv.ARM.GraphEntryPoint @transposeconv2d_fp, @transposeconv2d_fp_arg_0, @transposeconv2d_fp_arg_1, @transposeconv2d_fp_arg_2, @transposeconv2d_fp_res_0
+ spirv.ARM.Graph @transposeconv2d_fp(%arg0: !spirv.arm.tensor<10x24x9x13xf16>, %arg1: !spirv.arm.tensor<14x1x1x13xf16>, %arg2: !spirv.arm.tensor<14xf16>) -> (!spirv.arm.tensor<10x25x65x14xf16>) {
+ %4 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ %5 = spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<1xf16>
+ // CHECK: {{%.*}} = spirv.Tosa.TransposeConv2D out_pad = dense<[0, 1, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 8]> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, {{%.*}}, {{%.*}} : !spirv.arm.tensor<10x24x9x13xf16>, !spirv.arm.tensor<14x1x1x13xf16>, !spirv.arm.tensor<14xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<10x25x65x14xf16>
+ %6 = spirv.Tosa.TransposeConv2D out_pad = dense<[0, 1, 0, 0]> : !spirv.arm.tensor<4xi32>, stride = dense<[1, 8]> : !spirv.arm.tensor<2xi32>, acc_type = <FP16>, local_bound = true, %arg0, %arg1, %arg2, %4, %5 : !spirv.arm.tensor<10x24x9x13xf16>, !spirv.arm.tensor<14x1x1x13xf16>, !spirv.arm.tensor<14xf16>, !spirv.arm.tensor<1xf16>, !spirv.arm.tensor<1xf16> -> !spirv.arm.tensor<10x25x65x14xf16>
+ // CHECK: spirv.ARM.GraphOutputs {{%.*}} : !spirv.arm.tensor<10x25x65x14xf16>
+ spirv.ARM.GraphOutputs %6 : !spirv.arm.tensor<10x25x65x14xf16>
+ }
+}
diff --git a/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp b/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
index f3327e31aae04..71992d3fff663 100644
--- a/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
+++ b/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
@@ -501,6 +501,7 @@ constexpr llvm::StringLiteral constantIdEnumAttrs[] = {
"SPIRV_KHR_CooperativeMatrixLayoutAttr",
"SPIRV_MemorySemanticsAttr",
"SPIRV_MatrixLayoutAttr",
+ "SPIRV_TosaExtAccTypeAttr",
"SPIRV_TosaExtNaNPropagationModeAttr",
};
@@ -556,11 +557,17 @@ static void emitAttributeSerialization(const Attribute &attr,
os << tabs << " return failure();\n";
os << tabs << " }\n";
os << tabs << formatv(" {0}.push_back(attrTypeID);\n", operandList);
- } else if (attr.getAttrDefName() == "SPIRV_TensorArmAxisAttr") {
+ } else if (attr.getAttrDefName() == "SPIRV_TensorArmAxisAttr" ||
+ attr.getAttrDefName() == "SPIRV_BoolConstAttr") {
os << tabs
<< formatv(
" {0}.push_back(prepareConstantScalar({1}.getLoc(), attr));\n",
operandList, opVar);
+ } else if (attr.getAttrDefName().contains("TensorArm")) {
+ os << tabs
+ << formatv(" {0}.push_back(prepareConstant({1}.getLoc(), "
+ "llvm::cast<DenseElementsAttr>(attr).getType(), attr));\n",
+ operandList, opVar);
} else {
PrintFatalError(
loc,
@@ -855,7 +862,9 @@ static void emitAttributeDeserialization(const Attribute &attr,
<< formatv("{0}.push_back(opBuilder.getNamedAttr(\"{1}\", "
"TypeAttr::get(getType({2}[{3}++]))));\n",
attrList, attrName, words, wordIndex);
- } else if (attr.getAttrDefName() == "SPIRV_TensorArmAxisAttr") {
+ } else if (attr.getAttrDefName() == "SPIRV_TensorArmAxisAttr" ||
+ attr.getAttrDefName() == "SPIRV_BoolConstAttr" ||
+ attr.getAttrDefName().contains("TensorArm")) {
os << tabs
<< formatv("std::optional<std::pair<Attribute, Type>> c = "
"getConstant({0}[{1}++]);\n",
More information about the Mlir-commits
mailing list