Skip to content

[X86] Consistently use f128 libcalls #142386

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 2 commits into from
Jun 3, 2025
Merged

[X86] Consistently use f128 libcalls #142386

merged 2 commits into from
Jun 3, 2025

Conversation

nikic
Copy link
Contributor

@nikic nikic commented Jun 2, 2025

On x86, the *l libcalls are for 80-bit extended precision. fp128 needs to use the *f128 libcalls instead.

Add a few missing ones, esp. for FP min/max.

Also use the f128 libcalls on x86-32. I believe the situation there is the same as on x86-64.

@nikic
[X86] Consistently use f128 libcalls
79064db 
On x86, the `*l` libcalls are for 80-bit extended precision.
`fp128` needs to use the `*f128` libcalls instead.

Add a few missing ones, esp. for FP min/max.

Also use the `f128` libcalls on x86-32.
@nikic nikic requested review from RKSimon and phoebewang June 2, 2025 13:07
@llvmbot
Copy link
Member

llvmbot commented Jun 2, 2025 •
edited
Loading

@llvm/pr-subscribers-backend-x86

@llvm/pr-subscribers-llvm-ir

Author: Nikita Popov (nikic)

Changes

On x86, the *l libcalls are for 80-bit extended precision. fp128 needs to use the *f128 libcalls instead.

Add a few missing ones, esp. for FP min/max.

Also use the f128 libcalls on x86-32. I believe the situation there is the same as on x86-64.

Patch is 21.87 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/142386.diff

4 Files Affected:

  • (modified) llvm/lib/IR/RuntimeLibcalls.cpp (+7-2)
  • (modified) llvm/test/CodeGen/X86/fminimum-fmaximum-i686.ll (+4-4)
  • (modified) llvm/test/CodeGen/X86/fp128-libcalls-strict.ll (+32-32)
  • (modified) llvm/test/CodeGen/X86/fp128-libcalls.ll (+22-22)
diff --git a/llvm/lib/IR/RuntimeLibcalls.cpp b/llvm/lib/IR/RuntimeLibcalls.cpp
index 90c3bf0db0236..324636531b8fa 100644
--- a/llvm/lib/IR/RuntimeLibcalls.cpp
+++ b/llvm/lib/IR/RuntimeLibcalls.cpp
@@ -25,8 +25,8 @@ void RuntimeLibcallsInfo::initLibcalls(const Triple &TT) {
   for (int LC = 0; LC < RTLIB::UNKNOWN_LIBCALL; ++LC)
     setLibcallCallingConv((RTLIB::Libcall)LC, CallingConv::C);
 
-  // Use the f128 variants of math functions on x86_64
-  if (TT.getArch() == Triple::ArchType::x86_64 && TT.isGNUEnvironment()) {
+  // Use the f128 variants of math functions on x86
+  if (TT.isX86() && TT.isGNUEnvironment()) {
     setLibcallName(RTLIB::REM_F128, "fmodf128");
     setLibcallName(RTLIB::FMA_F128, "fmaf128");
     setLibcallName(RTLIB::SQRT_F128, "sqrtf128");
@@ -65,12 +65,17 @@ void RuntimeLibcallsInfo::initLibcalls(const Triple &TT) {
     setLibcallName(RTLIB::COPYSIGN_F128, "copysignf128");
     setLibcallName(RTLIB::FMIN_F128, "fminf128");
     setLibcallName(RTLIB::FMAX_F128, "fmaxf128");
+    setLibcallName(RTLIB::FMINIMUM_F128, "fminimumf128");
+    setLibcallName(RTLIB::FMAXIMUM_F128, "fmaximumf128");
+    setLibcallName(RTLIB::FMINIMUMNUM_F128, "fminimum_numf128");
+    setLibcallName(RTLIB::FMAXIMUMNUM_F128, "fmaximum_numf128");
     setLibcallName(RTLIB::LROUND_F128, "lroundf128");
     setLibcallName(RTLIB::LLROUND_F128, "llroundf128");
     setLibcallName(RTLIB::LRINT_F128, "lrintf128");
     setLibcallName(RTLIB::LLRINT_F128, "llrintf128");
     setLibcallName(RTLIB::LDEXP_F128, "ldexpf128");
     setLibcallName(RTLIB::FREXP_F128, "frexpf128");
+    setLibcallName(RTLIB::MODF_F128, "modff128");
   }
 
   // For IEEE quad-precision libcall names, PPC uses "kf" instead of "tf".
diff --git a/llvm/test/CodeGen/X86/fminimum-fmaximum-i686.ll b/llvm/test/CodeGen/X86/fminimum-fmaximum-i686.ll
index 4f930d84c9595..d3f85ce51edea 100644
--- a/llvm/test/CodeGen/X86/fminimum-fmaximum-i686.ll
+++ b/llvm/test/CodeGen/X86/fminimum-fmaximum-i686.ll
@@ -239,7 +239,7 @@ define fp128 @maximum_fp128(fp128 %x, fp128 %y) nounwind {
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl %eax
-; CHECK-NEXT:    calll fmaximuml
+; CHECK-NEXT:    calll fmaximumf128
 ; CHECK-NEXT:    addl $44, %esp
 ; CHECK-NEXT:    movl (%esp), %eax
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
@@ -496,7 +496,7 @@ define fp128 @minimum_fp128(fp128 %x, fp128 %y) nounwind {
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl %eax
-; CHECK-NEXT:    calll fminimuml
+; CHECK-NEXT:    calll fminimumf128
 ; CHECK-NEXT:    addl $44, %esp
 ; CHECK-NEXT:    movl (%esp), %eax
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
@@ -533,7 +533,7 @@ define fp128 @maximumnum_fp128(fp128 %x, fp128 %y) nounwind {
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl %eax
-; CHECK-NEXT:    calll fmaximum_numl
+; CHECK-NEXT:    calll fmaximum_numf128
 ; CHECK-NEXT:    addl $44, %esp
 ; CHECK-NEXT:    movl (%esp), %eax
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
@@ -570,7 +570,7 @@ define fp128 @minimumnum_fp128(fp128 %x, fp128 %y) nounwind {
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl {{[0-9]+}}(%esp)
 ; CHECK-NEXT:    pushl %eax
-; CHECK-NEXT:    calll fminimum_numl
+; CHECK-NEXT:    calll fminimum_numf128
 ; CHECK-NEXT:    addl $44, %esp
 ; CHECK-NEXT:    movl (%esp), %eax
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
diff --git a/llvm/test/CodeGen/X86/fp128-libcalls-strict.ll b/llvm/test/CodeGen/X86/fp128-libcalls-strict.ll
index a85b53ea62ac7..3ac4415d075c9 100644
--- a/llvm/test/CodeGen/X86/fp128-libcalls-strict.ll
+++ b/llvm/test/CodeGen/X86/fp128-libcalls-strict.ll
@@ -452,7 +452,7 @@ define fp128 @fma(fp128 %x, fp128 %y, fp128 %z) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmal
+; X86-NEXT:    calll fmaf128
 ; X86-NEXT:    addl $60, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -552,7 +552,7 @@ define fp128 @frem(fp128 %x, fp128 %y) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmodl
+; X86-NEXT:    calll fmodf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -641,7 +641,7 @@ define fp128 @ceil(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll ceill
+; X86-NEXT:    calll ceilf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -723,7 +723,7 @@ define fp128 @acos(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll acosl
+; X86-NEXT:    calll acosf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -805,7 +805,7 @@ define fp128 @cos(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll cosl
+; X86-NEXT:    calll cosf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -887,7 +887,7 @@ define fp128 @cosh(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll coshl
+; X86-NEXT:    calll coshf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -969,7 +969,7 @@ define fp128 @exp(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll expl
+; X86-NEXT:    calll expf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1051,7 +1051,7 @@ define fp128 @exp2(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll exp2l
+; X86-NEXT:    calll exp2f128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1133,7 +1133,7 @@ define fp128 @floor(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll floorl
+; X86-NEXT:    calll floorf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1215,7 +1215,7 @@ define fp128 @log(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll logl
+; X86-NEXT:    calll logf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1297,7 +1297,7 @@ define fp128 @log10(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll log10l
+; X86-NEXT:    calll log10f128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1379,7 +1379,7 @@ define fp128 @log2(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll log2l
+; X86-NEXT:    calll log2f128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1465,7 +1465,7 @@ define fp128 @maxnum(fp128 %x, fp128 %y) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmaxl
+; X86-NEXT:    calll fmaxf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1558,7 +1558,7 @@ define fp128 @minnum(fp128 %x, fp128 %y) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fminl
+; X86-NEXT:    calll fminf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1647,7 +1647,7 @@ define fp128 @nearbyint(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll nearbyintl
+; X86-NEXT:    calll nearbyintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1733,7 +1733,7 @@ define fp128 @pow(fp128 %x, fp128 %y) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll powl
+; X86-NEXT:    calll powf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1913,7 +1913,7 @@ define fp128 @rint(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll rintl
+; X86-NEXT:    calll rintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1995,7 +1995,7 @@ define fp128 @round(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll roundl
+; X86-NEXT:    calll roundf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2077,7 +2077,7 @@ define fp128 @roundeven(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll roundevenl
+; X86-NEXT:    calll roundevenf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2159,7 +2159,7 @@ define fp128 @asin(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll asinl
+; X86-NEXT:    calll asinf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2241,7 +2241,7 @@ define fp128 @sin(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll sinl
+; X86-NEXT:    calll sinf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2323,7 +2323,7 @@ define fp128 @sinh(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll sinhl
+; X86-NEXT:    calll sinhf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2405,7 +2405,7 @@ define fp128 @sqrt(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll sqrtl
+; X86-NEXT:    calll sqrtf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2487,7 +2487,7 @@ define fp128 @atan(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll atanl
+; X86-NEXT:    calll atanf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2573,7 +2573,7 @@ define fp128 @atan2(fp128 %x, fp128 %y) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll atan2l
+; X86-NEXT:    calll atan2f128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2662,7 +2662,7 @@ define fp128 @tan(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll tanl
+; X86-NEXT:    calll tanf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2744,7 +2744,7 @@ define fp128 @tanh(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll tanhl
+; X86-NEXT:    calll tanhf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2826,7 +2826,7 @@ define fp128 @trunc(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll truncl
+; X86-NEXT:    calll truncf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2903,7 +2903,7 @@ define i32 @lrint(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
-; X86-NEXT:    calll lrintl
+; X86-NEXT:    calll lrintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    retl
 ;
@@ -2953,7 +2953,7 @@ define i64 @llrint(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
-; X86-NEXT:    calll llrintl
+; X86-NEXT:    calll llrintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    retl
 ;
@@ -3003,7 +3003,7 @@ define i32 @lround(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
-; X86-NEXT:    calll lroundl
+; X86-NEXT:    calll lroundf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    retl
 ;
@@ -3053,7 +3053,7 @@ define i64 @llround(fp128 %x) nounwind strictfp {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
-; X86-NEXT:    calll llroundl
+; X86-NEXT:    calll llroundf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    retl
 ;
diff --git a/llvm/test/CodeGen/X86/fp128-libcalls.ll b/llvm/test/CodeGen/X86/fp128-libcalls.ll
index 4c716ace79a30..f727a79078627 100644
--- a/llvm/test/CodeGen/X86/fp128-libcalls.ll
+++ b/llvm/test/CodeGen/X86/fp128-libcalls.ll
@@ -841,7 +841,7 @@ define dso_local void @Test128Rem(fp128 %d1, fp128 %d2) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmodl
+; X86-NEXT:    calll fmodf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -933,7 +933,7 @@ define dso_local void @Test128_1Rem(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl vf128+4
 ; X86-NEXT:    pushl vf128
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmodl
+; X86-NEXT:    calll fmodf128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1018,7 +1018,7 @@ define dso_local void @Test128Sqrt(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll sqrtl
+; X86-NEXT:    calll sqrtf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1096,7 +1096,7 @@ define dso_local void @Test128Sin(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll sinl
+; X86-NEXT:    calll sinf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1174,7 +1174,7 @@ define dso_local void @Test128Cos(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll cosl
+; X86-NEXT:    calll cosf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1252,7 +1252,7 @@ define dso_local void @Test128Ceil(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll ceill
+; X86-NEXT:    calll ceilf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1330,7 +1330,7 @@ define dso_local void @Test128Floor(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll floorl
+; X86-NEXT:    calll floorf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1408,7 +1408,7 @@ define dso_local void @Test128Trunc(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll truncl
+; X86-NEXT:    calll truncf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1486,7 +1486,7 @@ define dso_local void @Test128Nearbyint(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll nearbyintl
+; X86-NEXT:    calll nearbyintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1564,7 +1564,7 @@ define dso_local void @Test128Rint(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll rintl
+; X86-NEXT:    calll rintf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1642,7 +1642,7 @@ define dso_local void @Test128Round(fp128 %d1) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll roundl
+; X86-NEXT:    calll roundf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, vf128
@@ -1723,7 +1723,7 @@ define fp128 @Test128FMA(fp128 %a, fp128 %b, fp128 %c) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll fmal
+; X86-NEXT:    calll fmaf128
 ; X86-NEXT:    addl $60, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1814,7 +1814,7 @@ define fp128 @Test128Acos(fp128 %a) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll acosl
+; X86-NEXT:    calll acosf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1889,7 +1889,7 @@ define fp128 @Test128Asin(fp128 %a) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll asinl
+; X86-NEXT:    calll asinf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -1964,7 +1964,7 @@ define fp128 @Test128Atan(fp128 %a) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll atanl
+; X86-NEXT:    calll atanf128
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2043,7 +2043,7 @@ define fp128 @Test128Atan2(fp128 %a, fp128 %b) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll atan2l
+; X86-NEXT:    calll atan2f128
 ; X86-NEXT:    addl $44, %esp
 ; X86-NEXT:    movaps (%esp), %xmm0
 ; X86-NEXT:    movaps %xmm0, (%esi)
@@ -2125,7 +2125,7 @@ define fp128 @Test128Cosh(fp128 %a) nounwind {
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; ...
[truncated]

RKSimon
RKSimon approved these changes Jun 2, 2025
View reviewed changes
Copy link
Collaborator

@RKSimon RKSimon left a comment

Choose a reason for hiding this comment

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

LGTM - although this probably should be mentioned in the release notes

Do we have any test-suite benchmarks that use f128?

@nikic
Copy link
Contributor Author

nikic commented Jun 2, 2025

LGTM - although this probably should be mentioned in the release notes

Which part do you have in mind? That the f128 libcalls also get used on x86-32?

Do we have any test-suite benchmarks that use f128?

Based on a grep of float128: no. There are only some fortran tests for C interop.

@RKSimon
Copy link
Collaborator

RKSimon commented Jun 2, 2025

Which part do you have in mind? That the f128 libcalls also get used on x86-32?

Yes - does this match gcc behaviour btw?

@phoebewang
Copy link
Contributor

Which part do you have in mind? That the f128 libcalls also get used on x86-32?

Yes - does this match gcc behaviour btw?

Seems libquadmath uses q suffix. Why we use f128 here?

@nikic
Copy link
Contributor Author

nikic commented Jun 3, 2025

This is using glibcs native f128 support, not libquadmath. See https://gcc.godbolt.org/z/fWeh3a98Y (for both x86-64 and x86-32). If you use the *l libcalls instead you get extra tfxf conversions.

@phoebewang
Copy link
Contributor

This is using glibcs native f128 support, not libquadmath. See https://gcc.godbolt.org/z/fWeh3a98Y (for both x86-64 and x86-32). If you use the *l libcalls instead you get extra tfxf conversions.

Thanks! The symbol works for both. https://gcc.godbolt.org/z/6xhfxe4ha

phoebewang
phoebewang approved these changes Jun 3, 2025
View reviewed changes
Copy link
Contributor

@phoebewang phoebewang left a comment

Choose a reason for hiding this comment

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

LGTM.

RKSimon
RKSimon approved these changes Jun 3, 2025
View reviewed changes
Copy link
Collaborator

@RKSimon RKSimon left a comment

Choose a reason for hiding this comment

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

LGTM - cheers

@nikic nikic merged commit 7547ff5 into llvm:main Jun 3, 2025
10 of 12 checks passed
@nikic nikic deleted the x86-f128-libcalls branch June 3, 2025 10:28
rorth pushed a commit to rorth/llvm-project that referenced this pull request Jun 11, 2025
@nikic @rorth
[X86] Consistently use f128 libcalls (llvm#142386)
cfaf8fb 
On x86, the `*l` libcalls are for 80-bit extended precision. `fp128`
needs to use the `*f128` libcalls instead.

Add a few missing ones, esp. for FP min/max.

Also use the `f128` libcalls on x86-32. I believe the situation there is
the same as on x86-64.
DhruvSrivastavaX pushed a commit to DhruvSrivastavaX/lldb-for-aix that referenced this pull request Jun 12, 2025
@nikic @DhruvSrivastavaX
[X86] Consistently use f128 libcalls (llvm#142386)
d0465eb 
On x86, the `*l` libcalls are for 80-bit extended precision. `fp128`
needs to use the `*f128` libcalls instead.

Add a few missing ones, esp. for FP min/max.

Also use the `f128` libcalls on x86-32. I believe the situation there is
the same as on x86-64.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@RKSimon RKSimon RKSimon approved these changes

@phoebewang phoebewang phoebewang approved these changes

Assignees
No one assigned
Labels
backend:X86 llvm:ir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@nikic @llvmbot @RKSimon @phoebewang