RuntimeLibcalls: Add sincos to the RISCV runtime libcall set by arsenm · Pull Request #215115 · llvm/llvm-project

arsenm · 2026-08-09T17:02:19Z

Inspection of the glibc sources suggests this is generically
available, with the target variance being for long double support.
The set of library functions is a large historical mess I'm attempting
to untangle. The traditional system had a large set of defaulted calls,
but sincos was a case which was explicitly enabled, and I'm assuming
riscv just never got around to adding it. It will be easier to
reorganize the library functions if synthetic architectural glibc
variance is eliminated.

Co-authored-by: Claude (Claude-Opus-4.8) noreply@anthropic.com

Inspection of the glibc sources suggests this is generically available, with the target variance being for long double support. The set of library functions is a large historical mess I'm attempting to untangle. The traditional system had a large set of defaulted calls, but sincos was a case which was explicitly enabled, and I'm assuming riscv just never got around to adding it. It will be easier to reorganize the library functions if synthetic architectural glibc variance is eliminated. Co-authored-by: Claude (Claude-Opus-4.8) <noreply@anthropic.com>

arsenm · 2026-08-09T17:02:32Z

RuntimeLibcalls: Add sincos to the RISCV runtime libcall set #215115 👈 (View in Graphite)
main

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmorg-github-actions · 2026-08-09T17:03:45Z

@llvm/pr-subscribers-llvm-ir

@llvm/pr-subscribers-backend-risc-v

Author: Matt Arsenault (arsenm)

Changes

Inspection of the glibc sources suggests this is generically
available, with the target variance being for long double support.
The set of library functions is a large historical mess I'm attempting
to untangle. The traditional system had a large set of defaulted calls,
but sincos was a case which was explicitly enabled, and I'm assuming
riscv just never got around to adding it. It will be easier to
reorganize the library functions if synthetic architectural glibc
variance is eliminated.

Co-authored-by: Claude (Claude-Opus-4.8) <noreply@anthropic.com>

Full diff: https://github.com/llvm/llvm-project/pull/215115.diff

3 Files Affected:

(modified) llvm/include/llvm/IR/RuntimeLibcalls.td (+1)
(added) llvm/test/CodeGen/RISCV/llvm.sincos.ll (+136)
(modified) llvm/test/CodeGen/RISCV/sincos-expansion.ll (+16-24)

diff --git a/llvm/include/llvm/IR/RuntimeLibcalls.td b/llvm/include/llvm/IR/RuntimeLibcalls.td
index b0d9663cca395..1b504c8de6add 100644
--- a/llvm/include/llvm/IR/RuntimeLibcalls.td
+++ b/llvm/include/llvm/IR/RuntimeLibcalls.td
@@ -3185,6 +3185,7 @@ def RISCVSystemLibrary
     : SystemRuntimeLibrary<isRISCV,
       (add DefaultRuntimeLibcallImpls,
            exp10f, exp10, exp10l_f128,
+           LibmHasSinCosF32, LibmHasSinCosF64, LibmHasSinCosF128,
            __riscv_flush_icache,
            LibcallImpls<(add Int128RTLibcalls), isRISCV64>,
            DefaultStackProtector,
diff --git a/llvm/test/CodeGen/RISCV/llvm.sincos.ll b/llvm/test/CodeGen/RISCV/llvm.sincos.ll
new file mode 100644
index 0000000000000..bb4935ba885ab
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/llvm.sincos.ll
@@ -0,0 +1,136 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=riscv64-linux-gnu -mattr=+d < %s | FileCheck %s --check-prefix=RV64
+; RUN: llc -mtriple=riscv32-linux-gnu -mattr=+d < %s | FileCheck %s --check-prefix=RV32
+
+; Test the llvm.sincos intrinsic lowering to the sincos libcalls, which glibc
+; provides (sincosf/sincos/sincosl). RISCV long double is fp128, so f128 uses
+; sincosl.
+
+define { float, float } @test_sincos_f32(float %a) nounwind {
+; RV64-LABEL: test_sincos_f32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    addi a0, sp, 4
+; RV64-NEXT:    mv a1, sp
+; RV64-NEXT:    call sincosf
+; RV64-NEXT:    flw fa0, 4(sp)
+; RV64-NEXT:    flw fa1, 0(sp)
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+;
+; RV32-LABEL: test_sincos_f32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    addi a1, sp, 4
+; RV32-NEXT:    call sincosf
+; RV32-NEXT:    flw fa0, 8(sp)
+; RV32-NEXT:    flw fa1, 4(sp)
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+  %result = call { float, float } @llvm.sincos.f32(float %a)
+  ret { float, float } %result
+}
+
+define { double, double } @test_sincos_f64(double %a) nounwind {
+; RV64-LABEL: test_sincos_f64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -32
+; RV64-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64-NEXT:    addi a0, sp, 16
+; RV64-NEXT:    addi a1, sp, 8
+; RV64-NEXT:    call sincos
+; RV64-NEXT:    fld fa0, 16(sp)
+; RV64-NEXT:    fld fa1, 8(sp)
+; RV64-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 32
+; RV64-NEXT:    ret
+;
+; RV32-LABEL: test_sincos_f64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -32
+; RV32-NEXT:    sw ra, 28(sp) # 4-byte Folded Spill
+; RV32-NEXT:    addi a0, sp, 16
+; RV32-NEXT:    addi a1, sp, 8
+; RV32-NEXT:    call sincos
+; RV32-NEXT:    fld fa0, 16(sp)
+; RV32-NEXT:    fld fa1, 8(sp)
+; RV32-NEXT:    lw ra, 28(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 32
+; RV32-NEXT:    ret
+  %result = call { double, double } @llvm.sincos.f64(double %a)
+  ret { double, double } %result
+}
+
+define { fp128, fp128 } @test_sincos_f128(fp128 %a) nounwind {
+; RV64-LABEL: test_sincos_f128:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -48
+; RV64-NEXT:    sd ra, 40(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
+; RV64-NEXT:    mv a4, a2
+; RV64-NEXT:    mv s0, a0
+; RV64-NEXT:    addi a2, sp, 16
+; RV64-NEXT:    mv a3, sp
+; RV64-NEXT:    mv a0, a1
+; RV64-NEXT:    mv a1, a4
+; RV64-NEXT:    call sincosl
+; RV64-NEXT:    ld a0, 16(sp)
+; RV64-NEXT:    ld a1, 24(sp)
+; RV64-NEXT:    ld a2, 0(sp)
+; RV64-NEXT:    ld a3, 8(sp)
+; RV64-NEXT:    sd a0, 0(s0)
+; RV64-NEXT:    sd a1, 8(s0)
+; RV64-NEXT:    sd a2, 16(s0)
+; RV64-NEXT:    sd a3, 24(s0)
+; RV64-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 48
+; RV64-NEXT:    ret
+;
+; RV32-LABEL: test_sincos_f128:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -80
+; RV32-NEXT:    sw ra, 76(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s0, 72(sp) # 4-byte Folded Spill
+; RV32-NEXT:    lw a4, 0(a1)
+; RV32-NEXT:    lw a5, 4(a1)
+; RV32-NEXT:    lw a6, 8(a1)
+; RV32-NEXT:    lw a7, 12(a1)
+; RV32-NEXT:    mv s0, a0
+; RV32-NEXT:    addi a0, sp, 24
+; RV32-NEXT:    addi a1, sp, 8
+; RV32-NEXT:    addi a2, sp, 56
+; RV32-NEXT:    addi a3, sp, 40
+; RV32-NEXT:    sw a4, 8(sp)
+; RV32-NEXT:    sw a5, 12(sp)
+; RV32-NEXT:    sw a6, 16(sp)
+; RV32-NEXT:    sw a7, 20(sp)
+; RV32-NEXT:    call sincosl
+; RV32-NEXT:    lw a0, 40(sp)
+; RV32-NEXT:    lw a1, 44(sp)
+; RV32-NEXT:    lw a2, 48(sp)
+; RV32-NEXT:    lw a3, 52(sp)
+; RV32-NEXT:    lw a4, 56(sp)
+; RV32-NEXT:    lw a5, 60(sp)
+; RV32-NEXT:    lw a6, 64(sp)
+; RV32-NEXT:    lw a7, 68(sp)
+; RV32-NEXT:    sw a0, 16(s0)
+; RV32-NEXT:    sw a1, 20(s0)
+; RV32-NEXT:    sw a2, 24(s0)
+; RV32-NEXT:    sw a3, 28(s0)
+; RV32-NEXT:    sw a4, 0(s0)
+; RV32-NEXT:    sw a5, 4(s0)
+; RV32-NEXT:    sw a6, 8(s0)
+; RV32-NEXT:    sw a7, 12(s0)
+; RV32-NEXT:    lw ra, 76(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s0, 72(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 80
+; RV32-NEXT:    ret
+  %result = call { fp128, fp128 } @llvm.sincos.f128(fp128 %a)
+  ret { fp128, fp128 } %result
+}
diff --git a/llvm/test/CodeGen/RISCV/sincos-expansion.ll b/llvm/test/CodeGen/RISCV/sincos-expansion.ll
index 6f37d67134cb3..606768a8e2f53 100644
--- a/llvm/test/CodeGen/RISCV/sincos-expansion.ll
+++ b/llvm/test/CodeGen/RISCV/sincos-expansion.ll
@@ -7,20 +7,16 @@
 define float @test_sincos_f32(float %f) nounwind {
 ; CHECK-LABEL: test_sincos_f32:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    addi sp, sp, -32
-; CHECK-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fsd fs0, 16(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fsd fs1, 8(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fmv.s fs0, fa0
-; CHECK-NEXT:    call sinf
-; CHECK-NEXT:    fmv.s fs1, fa0
-; CHECK-NEXT:    fmv.s fa0, fs0
-; CHECK-NEXT:    call cosf
-; CHECK-NEXT:    fadd.s fa0, fs1, fa0
-; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
-; CHECK-NEXT:    fld fs0, 16(sp) # 8-byte Folded Reload
-; CHECK-NEXT:    fld fs1, 8(sp) # 8-byte Folded Reload
-; CHECK-NEXT:    addi sp, sp, 32
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    addi a0, sp, 4
+; CHECK-NEXT:    mv a1, sp
+; CHECK-NEXT:    call sincosf
+; CHECK-NEXT:    flw fa5, 0(sp)
+; CHECK-NEXT:    flw fa4, 4(sp)
+; CHECK-NEXT:    fadd.s fa0, fa4, fa5
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %sin = call float @sinf(float %f) readnone
   %cos = call float @cosf(float %f) readnone
@@ -57,17 +53,13 @@ define double @test_sincos_f64(double %f) nounwind {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    addi sp, sp, -32
 ; CHECK-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fsd fs0, 16(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fsd fs1, 8(sp) # 8-byte Folded Spill
-; CHECK-NEXT:    fmv.d fs0, fa0
-; CHECK-NEXT:    call sin
-; CHECK-NEXT:    fmv.d fs1, fa0
-; CHECK-NEXT:    fmv.d fa0, fs0
-; CHECK-NEXT:    call cos
-; CHECK-NEXT:    fadd.d fa0, fs1, fa0
+; CHECK-NEXT:    addi a0, sp, 16
+; CHECK-NEXT:    addi a1, sp, 8
+; CHECK-NEXT:    call sincos
+; CHECK-NEXT:    fld fa5, 8(sp)
+; CHECK-NEXT:    fld fa4, 16(sp)
+; CHECK-NEXT:    fadd.d fa0, fa4, fa5
 ; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
-; CHECK-NEXT:    fld fs0, 16(sp) # 8-byte Folded Reload
-; CHECK-NEXT:    fld fs1, 8(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    addi sp, sp, 32
 ; CHECK-NEXT:    ret
   %sin = call double @sin(double %f) readnone

nemanjai · 2026-08-10T13:09:45Z

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
 ; RUN: llc -mtriple=riscv64-linux-gnu -mattr=+d -verify-machineinstrs -o - %s | FileCheck %s

 ; Test adapted from AArch64. Demonstrates the cases where sin and cos libcalls


The purpose and behaviour of this test seem a bit puzzling to me. The test suggests that it is meant to demonstrate the cases where sin+cos are combined into sincos, but prior to this change, that didn't happen for any cases. And with this change, the 32/64 bit versions are combined when the readnone attribute is present, but this doesn't happen for the 128 bit version. Why?

f128 / long double are more broken than not across most targets. I'm separately working on fixing that (#214944 is part)

nemanjai · 2026-08-10T13:09:50Z

    : SystemRuntimeLibrary<isRISCV,
      (add DefaultRuntimeLibcallImpls,
           exp10f, exp10, exp10l_f128,
+           LibmHasSinCosF32, LibmHasSinCosF64, LibmHasSinCosF128,


Just out of curiosity - if these are generally available on all targets, why not just get rid of the predicates?

It's available on all targets, with glibc not any libc

Co-authored-by: Craig Topper <craig.topper@sifive.com>

lenary

LGTM.

This was proposed before in #168708 - which seemed to have stalled due to f128 problems, but you've acknowledged there's more to do there to fix those.

topperc

LGTM

arsenm added the backend:RISC-V label Aug 9, 2026 — with Graphite App

arsenm requested review from asb, ilovepi, lenary and topperc August 9, 2026 17:03

arsenm marked this pull request as ready for review August 9, 2026 17:06

llvmorg-github-actions Bot added the llvm:ir label Aug 9, 2026

nemanjai reviewed Aug 10, 2026

View reviewed changes

topperc reviewed Aug 10, 2026

View reviewed changes

Comment thread llvm/test/CodeGen/RISCV/llvm.sincos.ll Outdated

spelling

9ee4dd8

Co-authored-by: Craig Topper <craig.topper@sifive.com>

lenary approved these changes Aug 10, 2026

View reviewed changes

ilovepi approved these changes Aug 10, 2026

View reviewed changes

topperc approved these changes Aug 10, 2026

View reviewed changes

Navigation Menu

Uh oh!

RuntimeLibcalls: Add sincos to the RISCV runtime libcall set - #215115

RuntimeLibcalls: Add sincos to the RISCV runtime libcall set#215115
arsenm wants to merge 2 commits into
mainfrom
users/arsenm/runtime-libcalls/riscv-sincos

arsenm commented Aug 9, 2026

Uh oh!

arsenm commented Aug 9, 2026

Uh oh!

llvmorg-github-actions Bot commented Aug 9, 2026 •

edited

Loading

Uh oh!

nemanjai Aug 10, 2026

Uh oh!

arsenm Aug 10, 2026

Uh oh!

nemanjai Aug 10, 2026

Uh oh!

arsenm Aug 10, 2026

Uh oh!

Uh oh!

lenary left a comment

Uh oh!

topperc left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Uh oh!

Conversation

arsenm commented Aug 9, 2026

Uh oh!

arsenm commented Aug 9, 2026

Uh oh!

llvmorg-github-actions Bot commented Aug 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nemanjai Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

arsenm Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

nemanjai Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

arsenm Aug 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

lenary left a comment

Choose a reason for hiding this comment

Uh oh!

topperc left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

llvmorg-github-actions Bot commented Aug 9, 2026 •

edited

Loading