[mlir][math] powi with negative exponent should invert at the end #135735
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Previously, an FPowI operation would invert the base *before* performing a sequence of multiplications, but this led to discrepancies between LLVM pow intrinsic folding and that coming from the math dialect. See compiler-rt's version, which does the inverse at the end of the calculation: compiler-rt/lib/builtins/powidf2.c
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-mlir-math Author: Asher Mancinelli (ashermancinelli) ChangesPreviously, an FPowI operation would invert the base before performing a sequence of multiplications, but this led to discrepancies between LLVM pow intrinsic folding and that coming from the math dialect. See compiler-rt's version, which does the inverse at the end of the calculation: compiler-rt/lib/builtins/powidf2.c Full diff: https://github.com/llvm/llvm-project/pull/135735.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Math/Transforms/AlgebraicSimplification.cpp b/mlir/lib/Dialect/Math/Transforms/AlgebraicSimplification.cpp
index dcace489673f0..13e2a4b5541b2 100644
--- a/mlir/lib/Dialect/Math/Transforms/AlgebraicSimplification.cpp
+++ b/mlir/lib/Dialect/Math/Transforms/AlgebraicSimplification.cpp
@@ -197,11 +197,6 @@ PowIStrengthReduction<PowIOpTy, DivOpTy, MulOpTy>::matchAndRewrite(
if (exponentValue > exponentThreshold)
return failure();
- // Inverse the base for negative exponent, i.e. for
- // `[fi]powi(x, negative_exponent)` set `x` to `1 / x`.
- if (exponentIsNegative)
- base = rewriter.create<DivOpTy>(loc, bcast(one), base);
-
Value result = base;
// Transform to naive sequence of multiplications:
// * For positive exponent case replace:
@@ -215,6 +210,11 @@ PowIStrengthReduction<PowIOpTy, DivOpTy, MulOpTy>::matchAndRewrite(
for (unsigned i = 1; i < exponentValue; ++i)
result = rewriter.create<MulOpTy>(loc, result, base);
+ // Inverse the base for negative exponent, i.e. for
+ // `[fi]powi(x, negative_exponent)` set `x` to `1 / x`.
+ if (exponentIsNegative)
+ result = rewriter.create<DivOpTy>(loc, bcast(one), result);
+
rewriter.replaceOp(op, result);
return success();
}
diff --git a/mlir/test/Dialect/Math/algebraic-simplification.mlir b/mlir/test/Dialect/Math/algebraic-simplification.mlir
index a97ecc52a17e9..e0e2b9853a2a1 100644
--- a/mlir/test/Dialect/Math/algebraic-simplification.mlir
+++ b/mlir/test/Dialect/Math/algebraic-simplification.mlir
@@ -135,11 +135,11 @@ func.func @ipowi_exp_two(%arg0: i32, %arg1: vector<4xi32>) -> (i32, vector<4xi32
// CHECK-DAG: %[[CST_V:.*]] = arith.constant dense<1> : vector<4xi32>
// CHECK: %[[SCALAR0:.*]] = arith.muli %[[ARG0]], %[[ARG0]]
// CHECK: %[[VECTOR0:.*]] = arith.muli %[[ARG1]], %[[ARG1]]
- // CHECK: %[[SCALAR1:.*]] = arith.divsi %[[CST_S]], %[[ARG0]]
- // CHECK: %[[SMUL:.*]] = arith.muli %[[SCALAR1]], %[[SCALAR1]]
- // CHECK: %[[VECTOR1:.*]] = arith.divsi %[[CST_V]], %[[ARG1]]
- // CHECK: %[[VMUL:.*]] = arith.muli %[[VECTOR1]], %[[VECTOR1]]
- // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SMUL]], %[[VMUL]]
+ // CHECK: %[[SMUL:.*]] = arith.muli %[[ARG0]], %[[ARG0]]
+ // CHECK: %[[SCALAR1:.*]] = arith.divsi %[[CST_S]], %[[SMUL]]
+ // CHECK: %[[VMUL:.*]] = arith.muli %[[ARG1]], %[[ARG1]]
+ // CHECK: %[[VECTOR1:.*]] = arith.divsi %[[CST_V]], %[[VMUL]]
+ // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SCALAR1]], %[[VECTOR1]]
%c1 = arith.constant 2 : i32
%v1 = arith.constant dense <2> : vector<4xi32>
%0 = math.ipowi %arg0, %c1 : i32
@@ -162,13 +162,13 @@ func.func @ipowi_exp_three(%arg0: i32, %arg1: vector<4xi32>) -> (i32, vector<4xi
// CHECK: %[[SCALAR0:.*]] = arith.muli %[[SMUL0]], %[[ARG0]]
// CHECK: %[[VMUL0:.*]] = arith.muli %[[ARG1]], %[[ARG1]]
// CHECK: %[[VECTOR0:.*]] = arith.muli %[[VMUL0]], %[[ARG1]]
- // CHECK: %[[SCALAR1:.*]] = arith.divsi %[[CST_S]], %[[ARG0]]
- // CHECK: %[[SMUL1:.*]] = arith.muli %[[SCALAR1]], %[[SCALAR1]]
- // CHECK: %[[SMUL2:.*]] = arith.muli %[[SMUL1]], %[[SCALAR1]]
- // CHECK: %[[VECTOR1:.*]] = arith.divsi %[[CST_V]], %[[ARG1]]
- // CHECK: %[[VMUL1:.*]] = arith.muli %[[VECTOR1]], %[[VECTOR1]]
- // CHECK: %[[VMUL2:.*]] = arith.muli %[[VMUL1]], %[[VECTOR1]]
- // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SMUL2]], %[[VMUL2]]
+ // CHECK: %[[SMUL1:.*]] = arith.muli %[[ARG0]], %[[ARG0]]
+ // CHECK: %[[SMUL2:.*]] = arith.muli %[[SMUL1]], %[[ARG0]]
+ // CHECK: %[[SCALAR1:.*]] = arith.divsi %[[CST_S]], %[[SMUL2]]
+ // CHECK: %[[VMUL1:.*]] = arith.muli %[[ARG1]], %[[ARG1]]
+ // CHECK: %[[VMUL2:.*]] = arith.muli %[[VMUL1]], %[[ARG1]]
+ // CHECK: %[[VECTOR1:.*]] = arith.divsi %[[CST_V]], %[[VMUL2]]
+ // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SCALAR1]], %[[VECTOR1]]
%c1 = arith.constant 3 : i32
%v1 = arith.constant dense <3> : vector<4xi32>
%0 = math.ipowi %arg0, %c1 : i32
@@ -225,11 +225,11 @@ func.func @fpowi_exp_two(%arg0: f32, %arg1: vector<4xf32>) -> (f32, vector<4xf32
// CHECK-DAG: %[[CST_V:.*]] = arith.constant dense<1.000000e+00> : vector<4xf32>
// CHECK: %[[SCALAR0:.*]] = arith.mulf %[[ARG0]], %[[ARG0]]
// CHECK: %[[VECTOR0:.*]] = arith.mulf %[[ARG1]], %[[ARG1]]
- // CHECK: %[[SCALAR1:.*]] = arith.divf %[[CST_S]], %[[ARG0]]
- // CHECK: %[[SMUL:.*]] = arith.mulf %[[SCALAR1]], %[[SCALAR1]]
- // CHECK: %[[VECTOR1:.*]] = arith.divf %[[CST_V]], %[[ARG1]]
- // CHECK: %[[VMUL:.*]] = arith.mulf %[[VECTOR1]], %[[VECTOR1]]
- // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SMUL]], %[[VMUL]]
+ // CHECK: %[[SMUL:.*]] = arith.mulf %[[ARG0]], %[[ARG0]]
+ // CHECK: %[[SCALAR1:.*]] = arith.divf %[[CST_S]], %[[SMUL]]
+ // CHECK: %[[VMUL:.*]] = arith.mulf %[[ARG1]], %[[ARG1]]
+ // CHECK: %[[VECTOR1:.*]] = arith.divf %[[CST_V]], %[[VMUL]]
+ // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SCALAR1]], %[[VECTOR1]]
%c1 = arith.constant 2 : i32
%v1 = arith.constant dense <2> : vector<4xi32>
%0 = math.fpowi %arg0, %c1 : f32, i32
@@ -252,13 +252,13 @@ func.func @fpowi_exp_three(%arg0: f32, %arg1: vector<4xf32>) -> (f32, vector<4xf
// CHECK: %[[SCALAR0:.*]] = arith.mulf %[[SMUL0]], %[[ARG0]]
// CHECK: %[[VMUL0:.*]] = arith.mulf %[[ARG1]], %[[ARG1]]
// CHECK: %[[VECTOR0:.*]] = arith.mulf %[[VMUL0]], %[[ARG1]]
- // CHECK: %[[SCALAR1:.*]] = arith.divf %[[CST_S]], %[[ARG0]]
- // CHECK: %[[SMUL1:.*]] = arith.mulf %[[SCALAR1]], %[[SCALAR1]]
- // CHECK: %[[SMUL2:.*]] = arith.mulf %[[SMUL1]], %[[SCALAR1]]
- // CHECK: %[[VECTOR1:.*]] = arith.divf %[[CST_V]], %[[ARG1]]
- // CHECK: %[[VMUL1:.*]] = arith.mulf %[[VECTOR1]], %[[VECTOR1]]
- // CHECK: %[[VMUL2:.*]] = arith.mulf %[[VMUL1]], %[[VECTOR1]]
- // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SMUL2]], %[[VMUL2]]
+ // CHECK: %[[SMUL1:.*]] = arith.mulf %[[ARG0]], %[[ARG0]]
+ // CHECK: %[[SMUL2:.*]] = arith.mulf %[[SMUL1]], %[[ARG0]]
+ // CHECK: %[[SCALAR1:.*]] = arith.divf %[[CST_S]], %[[SMUL2]]
+ // CHECK: %[[VMUL1:.*]] = arith.mulf %[[ARG1]], %[[ARG1]]
+ // CHECK: %[[VMUL2:.*]] = arith.mulf %[[VMUL1]], %[[ARG1]]
+ // CHECK: %[[VECTOR1:.*]] = arith.divf %[[CST_V]], %[[VMUL2]]
+ // CHECK: return %[[SCALAR0]], %[[VECTOR0]], %[[SCALAR1]], %[[VECTOR1]]
%c1 = arith.constant 3 : i32
%v1 = arith.constant dense <3> : vector<4xi32>
%0 = math.fpowi %arg0, %c1 : f32, i32
|
var-const
pushed a commit
to ldionne/llvm-project
that referenced
this pull request
Apr 17, 2025
…vm#135735) Previously, an FPowI operation would invert the base *before* performing a sequence of multiplications, but this led to discrepancies between LLVM pow intrinsic folding and that coming from the math dialect. See compiler-rt's version, which does the inverse at the end of the calculation: compiler-rt/lib/builtins/powidf2.c
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Previously, an FPowI operation would invert the base before performing a sequence of multiplications, but this led to discrepancies between LLVM pow intrinsic folding and that coming from the math dialect.
See compiler-rt's version, which does the inverse at the end of the calculation: compiler-rt/lib/builtins/powidf2.c