[libcxx-commits] [clang] [libc] [libunwind] [clang-tools-extra] [llvm] [lld] [libcxxabi] [lldb] [libcxx] [compiler-rt] [flang] Fix a bug in Smith's algorithm used in complex div/mul. (PR #78330)
Zahira Ammarguellat via libcxx-commits
libcxx-commits at lists.llvm.org
Tue Jan 16 14:38:07 PST 2024
https://github.com/zahiraam updated https://github.com/llvm/llvm-project/pull/78330
>From 8f8917528e30d2ba67f669cfd1a893bc85c21121 Mon Sep 17 00:00:00 2001
From: Ammarguellat <zahira.ammarguellat at intel.com>
Date: Tue, 16 Jan 2024 11:24:03 -0800
Subject: [PATCH] Fixed a bug in Smith's algorithm and made sure last option in
command line rules.
---
clang/lib/CodeGen/CGExprComplex.cpp | 8 ++++++--
clang/lib/Driver/ToolChains/Clang.cpp | 19 ++++++++++---------
clang/test/CodeGen/cx-complex-range.c | 16 ++++++++++++++++
clang/test/Driver/range.c | 7 +++++++
4 files changed, 39 insertions(+), 11 deletions(-)
diff --git a/clang/lib/CodeGen/CGExprComplex.cpp b/clang/lib/CodeGen/CGExprComplex.cpp
index e532794b71bdb4..6fbd8f19eeb50a 100644
--- a/clang/lib/CodeGen/CGExprComplex.cpp
+++ b/clang/lib/CodeGen/CGExprComplex.cpp
@@ -936,7 +936,7 @@ ComplexPairTy ComplexExprEmitter::EmitRangeReductionDiv(llvm::Value *LHSr,
llvm::Value *RC = Builder.CreateFMul(CdD, RHSr); // rc
llvm::Value *DpRC = Builder.CreateFAdd(RHSi, RC); // tmp=d+rc
- llvm::Value *T7 = Builder.CreateFMul(LHSr, RC); // ar
+ llvm::Value *T7 = Builder.CreateFMul(LHSr, CdD); // ar
llvm::Value *T8 = Builder.CreateFAdd(T7, LHSi); // ar+b
llvm::Value *DSTFr = Builder.CreateFDiv(T8, DpRC); // (ar+b)/tmp
@@ -978,7 +978,11 @@ ComplexPairTy ComplexExprEmitter::EmitBinDiv(const BinOpInfo &Op) {
return EmitRangeReductionDiv(LHSr, LHSi, RHSr, RHSi);
else if (Op.FPFeatures.getComplexRange() == LangOptions::CX_Limited)
return EmitAlgebraicDiv(LHSr, LHSi, RHSr, RHSi);
- else if (!CGF.getLangOpts().FastMath) {
+ else if (!CGF.getLangOpts().FastMath ||
+ // '-ffast-math' is used in the command line but followed by an
+ // '-fno-cx-limited-range'.
+ (CGF.getLangOpts().FastMath &&
+ Op.FPFeatures.getComplexRange() == LangOptions::CX_Full)) {
LHSi = OrigLHSi;
// If we have a complex operand on the RHS and FastMath is not allowed, we
// delegate to a libcall to handle all of the complexities and minimize
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 9edae3fec91a87..0c05d1db396071 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -2753,6 +2753,7 @@ static void RenderFloatingPointOptions(const ToolChain &TC, const Driver &D,
StringRef Float16ExcessPrecision = "";
StringRef BFloat16ExcessPrecision = "";
LangOptions::ComplexRangeKind Range = LangOptions::ComplexRangeKind::CX_Full;
+ std::string ComplexRangeStr = "";
if (const Arg *A = Args.getLastArg(options::OPT_flimited_precision_EQ)) {
CmdArgs.push_back("-mlimit-float-precision");
@@ -2768,24 +2769,24 @@ static void RenderFloatingPointOptions(const ToolChain &TC, const Driver &D,
case options::OPT_fcx_limited_range: {
EmitComplexRangeDiag(D, Range, LangOptions::ComplexRangeKind::CX_Limited);
Range = LangOptions::ComplexRangeKind::CX_Limited;
- std::string ComplexRangeStr = RenderComplexRangeOption("limited");
- if (!ComplexRangeStr.empty())
- CmdArgs.push_back(Args.MakeArgString(ComplexRangeStr));
+ ComplexRangeStr = RenderComplexRangeOption("limited");
break;
}
case options::OPT_fno_cx_limited_range:
+ EmitComplexRangeDiag(D, Range, LangOptions::ComplexRangeKind::CX_Full);
Range = LangOptions::ComplexRangeKind::CX_Full;
+ ComplexRangeStr = RenderComplexRangeOption("full");
break;
case options::OPT_fcx_fortran_rules: {
EmitComplexRangeDiag(D, Range, LangOptions::ComplexRangeKind::CX_Fortran);
Range = LangOptions::ComplexRangeKind::CX_Fortran;
- std::string ComplexRangeStr = RenderComplexRangeOption("fortran");
- if (!ComplexRangeStr.empty())
- CmdArgs.push_back(Args.MakeArgString(ComplexRangeStr));
+ ComplexRangeStr = RenderComplexRangeOption("fortran");
break;
}
case options::OPT_fno_cx_fortran_rules:
+ EmitComplexRangeDiag(D, Range, LangOptions::ComplexRangeKind::CX_Full);
Range = LangOptions::ComplexRangeKind::CX_Full;
+ ComplexRangeStr = RenderComplexRangeOption("full");
break;
case options::OPT_ffp_model_EQ: {
// If -ffp-model= is seen, reset to fno-fast-math
@@ -3056,9 +3057,7 @@ static void RenderFloatingPointOptions(const ToolChain &TC, const Driver &D,
SeenUnsafeMathModeOption = true;
// ffast-math enables fortran rules for complex multiplication and
// division.
- std::string ComplexRangeStr = RenderComplexRangeOption("limited");
- if (!ComplexRangeStr.empty())
- CmdArgs.push_back(Args.MakeArgString(ComplexRangeStr));
+ ComplexRangeStr = RenderComplexRangeOption("limited");
break;
}
case options::OPT_fno_fast_math:
@@ -3215,6 +3214,8 @@ static void RenderFloatingPointOptions(const ToolChain &TC, const Driver &D,
options::OPT_fstrict_float_cast_overflow, false))
CmdArgs.push_back("-fno-strict-float-cast-overflow");
+ if (!ComplexRangeStr.empty())
+ CmdArgs.push_back(Args.MakeArgString(ComplexRangeStr));
if (Args.hasArg(options::OPT_fcx_limited_range))
CmdArgs.push_back("-fcx-limited-range");
if (Args.hasArg(options::OPT_fcx_fortran_rules))
diff --git a/clang/test/CodeGen/cx-complex-range.c b/clang/test/CodeGen/cx-complex-range.c
index 8368fa611335cc..2d8507c710f202 100644
--- a/clang/test/CodeGen/cx-complex-range.c
+++ b/clang/test/CodeGen/cx-complex-range.c
@@ -15,9 +15,25 @@
// RUN: -ffast-math -complex-range=limited -emit-llvm -o - %s \
// RUN: | FileCheck %s --check-prefix=LMTD-FAST
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
+// RUN: -ffast-math -complex-range=full -emit-llvm -o - %s \
+// RUN: | FileCheck %s --check-prefix=FULL
+
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
// RUN: -fno-cx-fortran-rules -o - | FileCheck %s --check-prefix=FULL
+// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
+// RUN: -fcx-limited-range -fno-cx-limited-range -o - \
+// RUN: | FileCheck %s --check-prefix=FULL
+
+// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
+// RUN: -fno-cx-limited-range -fcx-limited-range -o - \
+// RUN: | FileCheck %s --check-prefix=FULL
+
+// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
+// RUN: -fno-cx-fortran-rules -fcx-fortran-rules -o - \
+// RUN: | FileCheck %s --check-prefix=FULL
+
_Complex float div(_Complex float a, _Complex float b) {
// LABEL: define {{.*}} @div(
// FULL: call {{.*}} @__divsc3
diff --git a/clang/test/Driver/range.c b/clang/test/Driver/range.c
index 8d456a997d6967..045d9c7d3d802a 100644
--- a/clang/test/Driver/range.c
+++ b/clang/test/Driver/range.c
@@ -6,6 +6,9 @@
// RUN: %clang -### -target x86_64 -fno-cx-limited-range -c %s 2>&1 \
// RUN: | FileCheck %s
+// RUN: %clang -### -target x86_64 -fcx-limited-range -fno-cx-limited-range \
+// RUN: -c %s 2>&1 | FileCheck --check-prefix=FULL %s
+
// RUN: %clang -### -target x86_64 -fcx-fortran-rules -c %s 2>&1 \
// RUN: | FileCheck --check-prefix=FRTRN %s
@@ -29,7 +32,11 @@
// RUN: %clang -### -target x86_64 -fcx-limited-range -ffast-math -c %s 2>&1 \
// RUN: | FileCheck --check-prefix=LMTD %s
+// RUN: %clang -### -target x86_64 -ffast-math -fno-cx-limited-range -c %s 2>&1 \
+// RUN: | FileCheck --check-prefix=FULL %s
+
// LMTD: -complex-range=limited
+// FULL: -complex-range=full
// LMTD-NOT: -complex-range=fortran
// CHECK-NOT: -complex-range=limited
// FRTRN: -complex-range=fortran
More information about the libcxx-commits
mailing list