Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Codegen][DT] Remove tensor.pad logics entirely from materialization. #18130

Merged
merged 2 commits into from
Aug 7, 2024
Merged

[Codegen][DT] Remove tensor.pad logics entirely from materialization. #18130

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
884936a
[Codegen][DT] Remove tensor.pad logics entirely from materialization.
hanhanW Aug 6, 2024
00f9a6b
fix lit test
hanhanW Aug 7, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 5 additions & 28 deletions compiler/src/iree/compiler/Codegen/Common/MaterializeEncodingIntoPackUnPack.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,27 +188,9 @@ static void transposeInPlace(MaterializeEncodingInfo &info) {
// to `pack` and `unpack` operations respectively.
//===---------------------------------------------------------------------===//

/// Utility method to get the optional padding value to use with pack operation
/// if source is defined using a `tensor.pad` operation. Note `source` is
/// passed by reference. It is updated to use the source of the pad operation.
static std::optional<Value> getPaddingValue(Value &source) {
auto padOp = source.getDefiningOp<tensor::PadOp>();
if (!padOp || padOp.getNofold() || !padOp.hasZeroLowPad()) {
return std::nullopt;
}

Value constantPaddingValue = padOp.getConstantPaddingValue();
if (!constantPaddingValue) {
return std::nullopt;
}

source = padOp.getSource();
return constantPaddingValue;
}

/// Utility method to convert from `set_encoding` op to `pack` operation.
/// For now this takes a `paddingValue` as input. The source is also taken
/// as input so that these could be used with `OpConversionPatterns`.
/// Utility method to convert from `set_encoding` op to `pack` operation with
/// zero padding values. The source is also taken as input so that these could
/// be used with `OpConversionPatterns`.
static FailureOr<tensor::PackOp> lowerSetEncodingOpToPackOp(
RewriterBase &rewriter, IREE::Encoding::SetEncodingOp encodingOp,
Value source, MaterializeEncodingFn materializeEncodingFn,
Expand All @@ -235,13 +217,8 @@ static FailureOr<tensor::PackOp> lowerSetEncodingOpToPackOp(
return rewriter.notifyMatchFailure(
encodingOp, "failed to generate runtime tile size query");
}
std::optional<Value> paddingValue;
if (encoding.getRoundDimsToArray().empty()) {
paddingValue = getPaddingValue(source);
} else {
paddingValue = rewriter.create<arith::ConstantOp>(
loc, rewriter.getZeroAttr(resultType.getElementType()));
}
Value paddingValue = rewriter.create<arith::ConstantOp>(
loc, rewriter.getZeroAttr(resultType.getElementType()));
SmallVector<OpFoldResult> sourceDims =
tensor::getMixedSizes(rewriter, loc, source);
SmallVector<OpFoldResult> resultDims = tensor::PackOp::getResultShape(
Expand Down
46 changes: 2 additions & 44 deletions compiler/src/iree/compiler/Codegen/Common/test/materialize_encoding_into_nop.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,48 +14,6 @@ func.func @pack_unpack_gemm_lhs(%arg0 : tensor<?x?xf32>) -> tensor<?x?xf32> {

// -----

#map = affine_map<(d0, d1, d2) -> (d0, d2)>
#map1 = affine_map<(d0, d1, d2) -> (d2, d1)>
#map2 = affine_map<(d0, d1, d2) -> (d0, d1)>
func.func @pad_gemm(%arg0 : tensor<100x250xf32>, %arg1 : tensor<250x500xf32>, %arg2 : tensor<100x500xf32>) -> tensor<100x500xf32> {
%pad_value = arith.constant 0.0 : f32
%pad_lhs = tensor.pad %arg0 low[0, 0] high[4, 2] {
^bb0(%b0: index, %b1 : index):
tensor.yield %pad_value : f32
} : tensor<100x250xf32> to tensor<104x252xf32>
%lhs = iree_encoding.set_encoding %pad_lhs : tensor<104x252xf32> -> tensor<104x252xf32, #iree_encoding.encoding<operand_index = 0, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>
%pad_rhs = tensor.pad %arg1 low[0, 0] high[2, 4] {
^bb0(%b0: index, %b1 : index):
tensor.yield %pad_value : f32
} : tensor<250x500xf32> to tensor<252x504xf32>
%rhs = iree_encoding.set_encoding %pad_rhs : tensor<252x504xf32> -> tensor<252x504xf32, #iree_encoding.encoding<operand_index = 1, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>
%pad_output = tensor.pad %arg2 low[0, 0] high[4, 4] {
^bb0(%b0: index, %b1 : index):
tensor.yield %pad_value : f32
} : tensor<100x500xf32> to tensor<104x504xf32>
%output = iree_encoding.set_encoding %pad_output : tensor<104x504xf32> -> tensor<104x504xf32, #iree_encoding.encoding<operand_index = 2, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>
%gemm_packed = linalg.matmul ins(%lhs, %rhs : tensor<104x252xf32, #iree_encoding.encoding<operand_index = 0, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>, tensor<252x504xf32, #iree_encoding.encoding<operand_index = 1, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>)
outs(%output : tensor<104x504xf32, #iree_encoding.encoding<operand_index = 2, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>) -> tensor<104x504xf32, #iree_encoding.encoding<operand_index = 2, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>>
%gemm = iree_encoding.unset_encoding %gemm_packed : tensor<104x504xf32, #iree_encoding.encoding<operand_index = 2, op_type = matmul, element_types = [f32, f32, f32], user_indexing_maps = [#map, #map1, #map2]>> -> tensor<104x504xf32>
%result = tensor.extract_slice %gemm[0, 0] [100, 500] [1, 1] : tensor<104x504xf32> to tensor<100x500xf32>
return %result : tensor<100x500xf32>
}
// CHECK: func @pad_gemm(
// CHECK-SAME: %[[ARG0:.+]]: tensor<100x250xf32>
// CHECK-SAME: %[[ARG1:.+]]: tensor<250x500xf32>
// CHECK-SAME: %[[ARG2:.+]]: tensor<100x500xf32>
// CHECK: %[[CST:.+]] = arith.constant 0.0
// CHECK-DAG: %[[LHS:.+]] = tensor.pad %[[ARG0]]
// CHECK-DAG: %[[RHS:.+]] = tensor.pad %[[ARG1]]
// CHECK-DAG: %[[DEST:.+]] = tensor.pad %[[ARG2]]
// CHECK: %[[GEMM:.+]] = linalg.matmul
// CHECK-SAME: ins(%[[LHS]], %[[RHS]] :
// CHECK-SAME: outs(%[[DEST]] :
// CHECK: %[[RES:.+]] = tensor.extract_slice %[[GEMM]]
// CHECK: return %[[RES]]

// -----

#map = affine_map<(d0, d1, d2) -> (d0, d2)>
#map1 = affine_map<(d0, d1, d2) -> (d2, d1)>
#map2 = affine_map<(d0, d1, d2) -> (d0, d1)>
Expand Down Expand Up @@ -103,8 +61,8 @@ func.func @gemm_fill_dynamic(%arg0 : tensor<?x?xf32>, %arg1 : tensor<?x?xf32>) -
// CHECK-SAME: %[[RHS:[a-zA-Z0-9]+]]: tensor<?x?xf32>
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
// CHECK-DAG: %[[D1:.+]] = tensor.dim %[[ARG1]], %[[C1]]
// CHECK-DAG: %[[D0:.+]] = tensor.dim %[[LHS]], %[[C0]]
// CHECK-DAG: %[[D1:.+]] = tensor.dim %[[RHS]], %[[C1]]
// CHECK-DAG: %[[EMPTY:.+]] = tensor.empty(%[[D0]], %[[D1]]) : tensor<?x?xf32>
// CHECK: %[[FILL:.+]] = linalg.fill
// CHECK-SAME: outs(%[[EMPTY]] :
Expand Down
Loading