Skip to content

[mlir][complex] Prevent underflow in complex.abs (#79786) #81092

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

Jump to bottom
Merged
merged 3 commits into from
Feb 10, 2024
Merged

[mlir][complex] Prevent underflow in complex.abs (#79786) #81092

merged 3 commits into from
Feb 10, 2024

Conversation

Lewuathe
Copy link
Member

@Lewuathe Lewuathe commented Feb 8, 2024 •
edited
Loading

Fix the issue found in the previous change. 70fb96a It is necessary to use the absolute value in the case of either real or imag part is zero as shown in the original algorithm.

See: #79786

@llvmbot llvmbot added mlir mlir:complex MLIR complex dialect labels Feb 8, 2024
@llvmbot
Copy link
Member

llvmbot commented Feb 8, 2024 •
edited
Loading

@llvm/pr-subscribers-mlir-complex

@llvm/pr-subscribers-mlir

Author: Kai Sasaki (Lewuathe)

Changes

Fix the issue found in the previous change. 70fb96a It is necessary to use the absolute value in the case of either real or imag part is zero.

See: #79786

Full diff: https://github.com/llvm/llvm-project/pull/81092.diff

4 Files Affected:

  • (modified) mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp (+44-14)
  • (modified) mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir (+103-22)
  • (modified) mlir/test/Conversion/ComplexToStandard/full-conversion.mlir (+23-4)
  • (modified) mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir (+52)
diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 4c9dad9e2c1731..cc315110f9be20 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -26,29 +26,59 @@ namespace mlir {
 using namespace mlir;
 
 namespace {
+// The algorithm is listed in https://dl.acm.org/doi/pdf/10.1145/363717.363780.
 struct AbsOpConversion : public OpConversionPattern<complex::AbsOp> {
   using OpConversionPattern<complex::AbsOp>::OpConversionPattern;
 
   LogicalResult
   matchAndRewrite(complex::AbsOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    auto loc = op.getLoc();
-    auto type = op.getType();
+    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
     arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
 
-    Value real =
-        rewriter.create<complex::ReOp>(loc, type, adaptor.getComplex());
-    Value imag =
-        rewriter.create<complex::ImOp>(loc, type, adaptor.getComplex());
-    Value realSqr =
-        rewriter.create<arith::MulFOp>(loc, real, real, fmf.getValue());
-    Value imagSqr =
-        rewriter.create<arith::MulFOp>(loc, imag, imag, fmf.getValue());
-    Value sqNorm =
-        rewriter.create<arith::AddFOp>(loc, realSqr, imagSqr, fmf.getValue());
-
-    rewriter.replaceOpWithNewOp<math::SqrtOp>(op, sqNorm);
+    Type elementType = op.getType();
+    Value arg = adaptor.getComplex();
+
+    Value zero =
+        b.create<arith::ConstantOp>(elementType, b.getZeroAttr(elementType));
+    Value one = b.create<arith::ConstantOp>(elementType,
+                                            b.getFloatAttr(elementType, 1.0));
+
+    Value real = b.create<complex::ReOp>(elementType, arg);
+    Value imag = b.create<complex::ImOp>(elementType, arg);
+
+    Value realIsZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero);
+    Value imagIsZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero);
+
+    // Real > Imag
+    Value imagDivReal = b.create<arith::DivFOp>(imag, real, fmf.getValue());
+    Value imagSq =
+        b.create<arith::MulFOp>(imagDivReal, imagDivReal, fmf.getValue());
+    Value imagSqPlusOne = b.create<arith::AddFOp>(imagSq, one, fmf.getValue());
+    Value imagSqrt = b.create<math::SqrtOp>(imagSqPlusOne, fmf.getValue());
+    Value realAbs = b.create<math::AbsFOp>(real, fmf.getValue());
+    Value absImag = b.create<arith::MulFOp>(imagSqrt, realAbs, fmf.getValue());
+
+    // Real <= Imag
+    Value realDivImag = b.create<arith::DivFOp>(real, imag, fmf.getValue());
+    Value realSq =
+        b.create<arith::MulFOp>(realDivImag, realDivImag, fmf.getValue());
+    Value realSqPlusOne = b.create<arith::AddFOp>(realSq, one, fmf.getValue());
+    Value realSqrt = b.create<math::SqrtOp>(realSqPlusOne, fmf.getValue());
+    Value imagAbs = b.create<math::AbsFOp>(imag, fmf.getValue());
+    Value absReal = b.create<arith::MulFOp>(realSqrt, imagAbs, fmf.getValue());
+
+    rewriter.replaceOpWithNewOp<arith::SelectOp>(
+        op, realIsZero, imagAbs,
+        b.create<arith::SelectOp>(
+            imagIsZero, realAbs,
+            b.create<arith::SelectOp>(
+                b.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, real, imag),
+                absImag, absReal)));
+
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 8fa29ea43854a4..1fe843b1447ab3 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -7,13 +7,30 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg: complex<f32>
   return %abs : f32
 }
+
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK-DAG: %[[REAL_SQ:.*]] = arith.mulf %[[REAL]], %[[REAL]] : f32
-// CHECK-DAG: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] : f32
-// CHECK: %[[SQ_NORM:.*]] = arith.addf %[[REAL_SQ]], %[[IMAG_SQ]] : f32
-// CHECK: %[[NORM:.*]] = math.sqrt %[[SQ_NORM]] : f32
-// CHECK: return %[[NORM]] : f32
+// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
+// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] : f32
+// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
+// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
+// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
+// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
+// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
+// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
+// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
+// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
+// CHECK: %[[ABS3:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
+// CHECK: return %[[ABS3]] : f32
 
 // -----
 
@@ -241,12 +258,28 @@ func.func @complex_log(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg: complex<f32>
   return %log : complex<f32>
 }
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[SQR_REAL:.*]] = arith.mulf %[[REAL]], %[[REAL]] : f32
-// CHECK: %[[SQR_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] : f32
-// CHECK: %[[SQ_NORM:.*]] = arith.addf %[[SQR_REAL]], %[[SQR_IMAG]] : f32
-// CHECK: %[[NORM:.*]] = math.sqrt %[[SQ_NORM]] : f32
+// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
+// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] : f32
+// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
+// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
+// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
+// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
+// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
+// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
+// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
+// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
+// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
 // CHECK: %[[RESULT_REAL:.*]] = math.log %[[NORM]] : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
@@ -469,12 +502,28 @@ func.func @complex_sign(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[REAL_IS_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
 // CHECK: %[[IMAG_IS_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
 // CHECK: %[[IS_ZERO:.*]] = arith.andi %[[REAL_IS_ZERO]], %[[IMAG_IS_ZERO]] : i1
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[SQR_REAL:.*]] = arith.mulf %[[REAL2]], %[[REAL2]] : f32
-// CHECK: %[[SQR_IMAG:.*]] = arith.mulf %[[IMAG2]], %[[IMAG2]] : f32
-// CHECK: %[[SQ_NORM:.*]] = arith.addf %[[SQR_REAL]], %[[SQR_IMAG]] : f32
-// CHECK: %[[NORM:.*]] = math.sqrt %[[SQ_NORM]] : f32
+// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL2]], %[[ZERO]] : f32
+// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG2]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG2]], %[[REAL2]] : f32
+// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
+// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
+// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL2]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
+// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL2]], %[[IMAG2]] : f32
+// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
+// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
+// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG2]] : f32
+// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
+// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL2]], %[[IMAG2]] : f32
+// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
+// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
+// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
 // CHECK: %[[REAL_SIGN:.*]] = arith.divf %[[REAL]], %[[NORM]] : f32
 // CHECK: %[[IMAG_SIGN:.*]] = arith.divf %[[IMAG]], %[[NORM]] : f32
 // CHECK: %[[SIGN:.*]] = complex.create %[[REAL_SIGN]], %[[IMAG_SIGN]] : complex<f32>
@@ -716,13 +765,29 @@ func.func @complex_abs_with_fmf(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg fastmath<nnan,contract> : complex<f32>
   return %abs : f32
 }
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK-DAG: %[[REAL_SQ:.*]] = arith.mulf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK-DAG: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_NORM:.*]] = arith.addf %[[REAL_SQ]], %[[IMAG_SQ]] fastmath<nnan,contract> : f32
-// CHECK: %[[NORM:.*]] = math.sqrt %[[SQ_NORM]] : f32
-// CHECK: return %[[NORM]] : f32
+// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
+// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
+// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
+// CHECK: %[[ABS3:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
+// CHECK: return %[[ABS3]] : f32
 
 // -----
 
@@ -807,12 +872,28 @@ func.func @complex_log_with_fmf(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg fastmath<nnan,contract> : complex<f32>
   return %log : complex<f32>
 }
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[SQR_REAL:.*]] = arith.mulf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQR_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_NORM:.*]] = arith.addf %[[SQR_REAL]], %[[SQR_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[NORM:.*]] = math.sqrt %[[SQ_NORM]] : f32
+// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
+// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
+// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
+// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
 // CHECK: %[[RESULT_REAL:.*]] = math.log %[[NORM]] fastmath<nnan,contract> : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
diff --git a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
index 9983dd46f09433..0f23e20167f491 100644
--- a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
@@ -6,12 +6,31 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg: complex<f32>
   return %abs : f32
 }
+// CHECK: %[[ZERO:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
+// CHECK: %[[ONE:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
 // CHECK: %[[REAL:.*]] = llvm.extractvalue %[[ARG]][0] : ![[C_TY]]
 // CHECK: %[[IMAG:.*]] = llvm.extractvalue %[[ARG]][1] : ![[C_TY]]
-// CHECK-DAG: %[[REAL_SQ:.*]] = llvm.fmul %[[REAL]], %[[REAL]]  : f32
-// CHECK-DAG: %[[IMAG_SQ:.*]] = llvm.fmul %[[IMAG]], %[[IMAG]]  : f32
-// CHECK: %[[SQ_NORM:.*]] = llvm.fadd %[[REAL_SQ]], %[[IMAG_SQ]]  : f32
-// CHECK: %[[NORM:.*]] = llvm.intr.sqrt(%[[SQ_NORM]]) : (f32) -> f32
+// CHECK: %[[REAL_IS_ZERO:.*]] = llvm.fcmp "oeq" %[[REAL]], %[[ZERO]] : f32
+// CHECK: %[[IMAG_IS_ZERO:.*]] = llvm.fcmp "oeq" %[[IMAG]], %[[ZERO]] : f32
+
+// CHECK: %[[IMAG_DIV_REAL:.*]] = llvm.fdiv %[[IMAG]], %[[REAL]] : f32
+// CHECK: %[[IMAG_SQ:.*]] = llvm.fmul %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]]  : f32
+// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = llvm.fadd %[[IMAG_SQ]], %[[ONE]] : f32
+// CHECK: %[[IMAG_SQRT:.*]] = llvm.intr.sqrt(%[[IMAG_SQ_PLUS_ONE]]) : (f32) -> f32
+// CHECK: %[[REAL_ABS:.*]] = llvm.intr.fabs(%[[REAL]]) : (f32) -> f32
+// CHECK: %[[ABS_IMAG:.*]] = llvm.fmul %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
+
+// CHECK: %[[REAL_DIV_IMAG:.*]] = llvm.fdiv %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[REAL_SQ:.*]] = llvm.fmul %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
+// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = llvm.fadd %[[REAL_SQ]], %[[ONE]]  : f32
+// CHECK: %[[REAL_SQRT:.*]] = llvm.intr.sqrt(%[[REAL_SQ_PLUS_ONE]])  : (f32) -> f32
+// CHECK: %[[IMAG_ABS:.*]] = llvm.intr.fabs(%[[IMAG]]) : (f32) -> f32
+// CHECK: %[[ABS_REAL:.*]] = llvm.fmul %[[REAL_SQRT]], %[[IMAG_ABS]]  : f32
+
+// CHECK: %[[REAL_GT_IMAG:.*]] = llvm.fcmp "ogt" %[[REAL]], %[[IMAG]] : f32
+// CHECK: %[[ABS1:.*]] = llvm.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : i1, f32
+// CHECK: %[[ABS2:.*]] = llvm.select %[[IMAG_IS_ZERO]], %[[REAL_ABS]], %[[ABS1]] : i1, f32
+// CHECK: %[[NORM:.*]] = llvm.select %[[REAL_IS_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : i1, f32
 // CHECK: llvm.return %[[NORM]] : f32
 
 // CHECK-LABEL: llvm.func @complex_eq
diff --git a/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir b/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
index 349b92a7aefa2e..2d6a4815380e84 100644
--- a/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
+++ b/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
@@ -106,6 +106,27 @@ func.func @angle(%arg: complex<f32>) -> f32 {
   func.return %angle : f32
 }
 
+func.func @test_element_f64(%input: tensor<?xcomplex<f64>>,
+                      %func: (complex<f64>) -> f64) {
+  %c0 = arith.constant 0 : index
+  %c1 = arith.constant 1 : index
+  %size = tensor.dim %input, %c0: tensor<?xcomplex<f64>>
+
+  scf.for %i = %c0 to %size step %c1 {
+    %elem = tensor.extract %input[%i]: tensor<?xcomplex<f64>>
+
+    %val = func.call_indirect %func(%elem) : (complex<f64>) -> f64
+    vector.print %val : f64
+    scf.yield
+  }
+  func.return
+}
+
+func.func @abs(%arg: complex<f64>) -> f64 {
+  %abs = complex.abs %arg : complex<f64>
+  func.return %abs : f64
+}
+
 func.func @entry() {
   // complex.sqrt test
   %sqrt_test = arith.constant dense<[
@@ -300,5 +321,36 @@ func.func @entry() {
   call @test_element(%angle_test_cast, %angle_func)
     : (tensor<?xcomplex<f32>>, (complex<f32>) -> f32) -> ()
 
+  // complex.abs test
+  %abs_test = arith.constant dense<[
+    (1.0, 1.0),
+    // CHECK:  1.414
+    (1.0e300, 1.0e300),
+    // CHECK-NEXT:  1.41421e+300
+    (1.0e-300, 1.0e-300),
+    // CHECK-NEXT:  1.41421e-300
+    (5.0, 0.0),
+    // CHECK-NEXT:  5
+    (0.0, 6.0),
+    // CHECK-NEXT:  6
+    (7.0, 8.0),
+    // CHECK-NEXT:  10.6301
+    (-1.0, -1.0),
+    // CHECK-NEXT: 1.414
+    (-1.0e300, -1.0e300),
+    // CHECK-NEXT:  1.41421e+300
+    (-1.0, 0.0),
+    // CHECK-NEXT:  1
+    (0.0, -1.0)
+    // CHECK-NEXT:  1
+  ]> : tensor<10xcomplex<f64>>
+  %abs_test_cast = tensor.cast %abs_test
+    :  tensor<10xcomplex<f64>> to tensor<?xcomplex<f64>>
+
+  %abs_func = func.constant @abs : (complex<f64>) -> f64
+
+  call @test_element_f64(%abs_test_cast, %abs_func)
+    : (tensor<?xcomplex<f64>>, (complex<f64>) -> f64) -> ()
+
   func.return
 }

Lewuathe
Lewuathe commented Feb 8, 2024
View reviewed changes
mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
Value absReal = b.create<arith::MulFOp>(realSqrt, imagAbs, fmf.getValue());

rewriter.replaceOpWithNewOp<arith::SelectOp>(
op, realIsZero, imagAbs,
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Use imagAbs if real is zero.

mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
rewriter.replaceOpWithNewOp<arith::SelectOp>(
op, realIsZero, imagAbs,
b.create<arith::SelectOp>(
imagIsZero, realAbs,
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Use realAbs if imag is zero.

mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
(-1.0e300, -1.0e300),
// CHECK-NEXT: 1.41421e+300
(-1.0, 0.0),
// CHECK-NEXT: 1
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add test case the imag part is zero.

mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
(-1.0, 0.0),
// CHECK-NEXT: 1
(0.0, -1.0)
// CHECK-NEXT: 1
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add test case where the real is zero.

@Lewuathe
[mlir][complex] Prevent underflow in complex.abs (llvm#79786)
f4e9fa6 
The previous PR was not enough about the way to handle the negative value.
It is necessary to take the absolute value of the given real (or
imaginary) part to be multiplied with the sqrt part in the case of
either is zero.

See: llvm#76316
@Lewuathe Lewuathe force-pushed the prevent-underflow-in-complex-abs branch from 1d3d9a1 to f4e9fa6 Compare February 9, 2024 03:45
Lewuathe
Lewuathe commented Feb 10, 2024
View reviewed changes
Copy link
Member Author

@Lewuathe Lewuathe left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@d0k @matthias-springer @joker-eph Sorry for asking you to review the same change couple of times. I have detected and fixed the cause of the previous negative value case error. Could you review this when you get a chance?

@Lewuathe Lewuathe merged commit b17348c into llvm:main Feb 10, 2024
@Lewuathe
Copy link
Member Author

@d0k Thank you so much for reviewing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@d0k d0k d0k approved these changes

@matthias-springer matthias-springer Awaiting requested review from matthias-springer

@joker-eph joker-eph Awaiting requested review from joker-eph

Assignees
No one assigned
Labels
mlir:complex MLIR complex dialect mlir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@Lewuathe @llvmbot @d0k