[llvm] [X86][LoopVectorize] Add support for arc and hyperbolic trig functions (PR #99383)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 17 13:15:45 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-analysis
@llvm/pr-subscribers-backend-x86
Author: Farzon Lotfi (farzonl)
<details>
<summary>Changes</summary>
This change is part 2 x86 Loop Vectorization of : https://github.com/llvm/llvm-project/pull/96222
This change adds support for AOCL-LibM for x86 Loop vectorization.
In the process three new vectorizations were added for more completeness
- [amd_vrs16_acosf](https://github.com/amd/aocl-libm-ose/blob/9c0b67293ba01e509a6308247d82a8f1adfbbc67/scripts/libalm.def#L221)
- [amd_vrd2_cosh](https://github.com/amd/aocl-libm-ose/blob/9c0b67293ba01e509a6308247d82a8f1adfbbc67/scripts/libalm.def#L124)
- [amd_vrs16_tanhf](https://github.com/amd/aocl-libm-ose/blob/9c0b67293ba01e509a6308247d82a8f1adfbbc67/scripts/libalm.def#L224)
It also has veclib call loop vectorization hence the test cases in `llvm/test/Transforms/LoopVectorize/X86/veclib-calls.ll`
finally the last pr missed tests for `llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll` and `llvm/test/CodeGen/X86/vec-libcalls.ll` so added those aswell.
No evidence was found for arc and hyperbolic trig glibc vector math functions
https://github.com/lattera/glibc/blob/master/sysdeps/x86/fpu/bits/math-vector.h
so no new `_ZGVbN2v_*` and `_ZGVdN4v_*` .
So no new tests in `llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-VF2-VF8.ll`
Also no new svml and no new tests to: `llvm/test/Transforms/LoopVectorize/X86/svml-calls.ll`
There was not enough evidence that there were svml arc and hyperbolic trig vector implementations, Documentation was scares so looked at test cases in [numpy](https://github.com/numpy/SVML/blob/32bf2a98420762a63ab418aaa0a7d6e17eb9627a/linux/avx512/svml_z0_acos_d_la.s#L8). Someone with more experience with svml should investigate.
## History
This change is part of https://github.com/llvm/llvm-project/issues/87367's investigation on supporting IEEE math operations as intrinsics.
Which was discussed in this RFC:
https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294
This change adds loop vectorization for `acos`, `asin`, `atan`, `cosh`, `sinh`, and `tanh`.
https://github.com/llvm/llvm-project/issues/70079
https://github.com/llvm/llvm-project/issues/70080
https://github.com/llvm/llvm-project/issues/70081
https://github.com/llvm/llvm-project/issues/70083
https://github.com/llvm/llvm-project/issues/70084
https://github.com/llvm/llvm-project/issues/95966
---
Patch is 87.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/99383.diff
5 Files Affected:
- (modified) llvm/include/llvm/Analysis/VecFuncs.def (+35)
- (modified) llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll (+90)
- (modified) llvm/test/CodeGen/X86/vec-libcalls.ll (+1212)
- (modified) llvm/test/Transforms/LoopVectorize/X86/amdlibm-calls.ll (+446)
- (modified) llvm/test/Transforms/LoopVectorize/X86/veclib-calls.ll (+151)
``````````diff
diff --git a/llvm/include/llvm/Analysis/VecFuncs.def b/llvm/include/llvm/Analysis/VecFuncs.def
index ffdf8b8c3bc79..19e0058b4c3dd 100644
--- a/llvm/include/llvm/Analysis/VecFuncs.def
+++ b/llvm/include/llvm/Analysis/VecFuncs.def
@@ -51,13 +51,19 @@ TLI_DEFINE_VECFUNC("llvm.cos.f32", "vcosf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("tanf", "vtanf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("llvm.tan.f32", "vtanf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("asinf", "vasinf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.asin.f32", "vasinf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("acosf", "vacosf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.acos.f32", "vacosf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("atanf", "vatanf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.atan.f32", "vatanf", FIXED(4), "_ZGV_LLVM_N4v")
// Hyperbolic Functions
TLI_DEFINE_VECFUNC("sinhf", "vsinhf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.sinh.f32", "vsinhf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("coshf", "vcoshf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.cosh.f32", "vcoshf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("tanhf", "vtanhf", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.tanh.f32", "vtanhf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("asinhf", "vasinhf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("acoshf", "vacoshf", FIXED(4), "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("atanhf", "vatanhf", FIXED(4), "_ZGV_LLVM_N4v")
@@ -1292,8 +1298,20 @@ TLI_DEFINE_VECFUNC("asinf", "amd_vrs4_asinf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("asinf", "amd_vrs8_asinf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
TLI_DEFINE_VECFUNC("asinf", "amd_vrs16_asinf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+TLI_DEFINE_VECFUNC("llvm.asin.f64", "amd_vrd8_asin", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.asin.f32", "amd_vrs4_asinf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.asin.f32", "amd_vrs8_asinf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.asin.f32", "amd_vrs16_asinf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+
+
TLI_DEFINE_VECFUNC("acosf", "amd_vrs4_acosf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("acosf", "amd_vrs8_acosf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("acosf", "amd_vrs16_acosf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+
+TLI_DEFINE_VECFUNC("llvm.acos.f32", "amd_vrs16_acosf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+TLI_DEFINE_VECFUNC("llvm.acos.f32", "amd_vrs8_acosf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.acos.f32", "amd_vrs4_acosf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+
TLI_DEFINE_VECFUNC("atan", "amd_vrd2_atan", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
TLI_DEFINE_VECFUNC("atan", "amd_vrd4_atan", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
@@ -1302,11 +1320,28 @@ TLI_DEFINE_VECFUNC("atanf", "amd_vrs4_atanf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("atanf", "amd_vrs8_atanf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
TLI_DEFINE_VECFUNC("atanf", "amd_vrs16_atanf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+TLI_DEFINE_VECFUNC("llvm.atan.f64", "amd_vrd2_atan", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC("llvm.atan.f64", "amd_vrd4_atan", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.atan.f64", "amd_vrd8_atan", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.atan.f32", "amd_vrs4_atanf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.atan.f32", "amd_vrs8_atanf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.atan.f32", "amd_vrs16_atanf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+
+TLI_DEFINE_VECFUNC("cosh", "amd_vrd2_cosh" , FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
TLI_DEFINE_VECFUNC("coshf", "amd_vrs4_coshf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("coshf", "amd_vrs8_coshf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.cosh.f64", "amd_vrd2_cosh" , FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC("llvm.cosh.f32", "amd_vrs4_coshf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.cosh.f32", "amd_vrs8_coshf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+
TLI_DEFINE_VECFUNC("tanhf", "amd_vrs4_tanhf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
TLI_DEFINE_VECFUNC("tanhf", "amd_vrs8_tanhf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("tanhf", "amd_vrs16_tanhf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
+
+TLI_DEFINE_VECFUNC("llvm.tanh.f32", "amd_vrs4_tanhf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.tanh.f32", "amd_vrs8_tanhf", FIXED(8), NOMASK, "_ZGV_LLVM_N8v")
+TLI_DEFINE_VECFUNC("llvm.tanh.f32", "amd_vrs16_tanhf", FIXED(16), NOMASK, "_ZGV_LLVM_N16v")
TLI_DEFINE_VECFUNC("cbrt", "amd_vrd2_cbrt", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
TLI_DEFINE_VECFUNC("cbrtf", "amd_vrs4_cbrtf", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
diff --git a/llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll b/llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll
index cfec52c0e6886..835cd9f509b0d 100644
--- a/llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll
+++ b/llvm/test/CodeGen/X86/fp-strict-libcalls-msvc32.ll
@@ -177,6 +177,90 @@ define float @tan(float %x) #0 {
ret float %result
}
+define float @acos(float %x) #0 {
+; CHECK-LABEL: acos:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _acosf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.acos.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
+define float @asin(float %x) #0 {
+; CHECK-LABEL: asin:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _asinf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.asin.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
+define float @atan(float %x) #0 {
+; CHECK-LABEL: atan:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _atanf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.atan.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
+define float @cosh(float %x) #0 {
+; CHECK-LABEL: cosh:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _coshf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.cosh.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
+define float @sinh(float %x) #0 {
+; CHECK-LABEL: sinh:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _sinhf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.sinh.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
+define float @tanh(float %x) #0 {
+; CHECK-LABEL: tanh:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: flds {{[0-9]+}}(%esp)
+; CHECK-NEXT: fstps (%esp)
+; CHECK-NEXT: wait
+; CHECK-NEXT: calll _tanhf
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
+ %result = call float @llvm.experimental.constrained.tanh.f32(float %x, metadata !"round.dynamic", metadata !"fpexcept.strict") #0
+ ret float %result
+}
+
attributes #0 = { strictfp }
declare float @llvm.experimental.constrained.ceil.f32(float, metadata)
@@ -189,3 +273,9 @@ declare float @llvm.experimental.constrained.log10.f32(float, metadata, metadata
declare float @llvm.experimental.constrained.pow.f32(float, float, metadata, metadata)
declare float @llvm.experimental.constrained.sin.f32(float, metadata, metadata)
declare float @llvm.experimental.constrained.tan.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.acos.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.asin.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.atan.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.cosh.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.sinh.f32(float, metadata, metadata)
+declare float @llvm.experimental.constrained.tanh.f32(float, metadata, metadata)
diff --git a/llvm/test/CodeGen/X86/vec-libcalls.ll b/llvm/test/CodeGen/X86/vec-libcalls.ll
index 6857101d3d75b..b107b1c2749cc 100644
--- a/llvm/test/CodeGen/X86/vec-libcalls.ll
+++ b/llvm/test/CodeGen/X86/vec-libcalls.ll
@@ -25,6 +25,54 @@ declare <5 x float> @llvm.tan.v5f32(<5 x float>)
declare <6 x float> @llvm.tan.v6f32(<6 x float>)
declare <3 x double> @llvm.tan.v3f64(<3 x double>)
+declare <1 x float> @llvm.acos.v1f32(<1 x float>)
+declare <2 x float> @llvm.acos.v2f32(<2 x float>)
+declare <3 x float> @llvm.acos.v3f32(<3 x float>)
+declare <4 x float> @llvm.acos.v4f32(<4 x float>)
+declare <5 x float> @llvm.acos.v5f32(<5 x float>)
+declare <6 x float> @llvm.acos.v6f32(<6 x float>)
+declare <3 x double> @llvm.acos.v3f64(<3 x double
+>)
+declare <1 x float> @llvm.asin.v1f32(<1 x float>)
+declare <2 x float> @llvm.asin.v2f32(<2 x float>)
+declare <3 x float> @llvm.asin.v3f32(<3 x float>)
+declare <4 x float> @llvm.asin.v4f32(<4 x float>)
+declare <5 x float> @llvm.asin.v5f32(<5 x float>)
+declare <6 x float> @llvm.asin.v6f32(<6 x float>)
+declare <3 x double> @llvm.asin.v3f64(<3 x double>)
+
+declare <1 x float> @llvm.atan.v1f32(<1 x float>)
+declare <2 x float> @llvm.atan.v2f32(<2 x float>)
+declare <3 x float> @llvm.atan.v3f32(<3 x float>)
+declare <4 x float> @llvm.atan.v4f32(<4 x float>)
+declare <5 x float> @llvm.atan.v5f32(<5 x float>)
+declare <6 x float> @llvm.atan.v6f32(<6 x float>)
+declare <3 x double> @llvm.atan.v3f64(<3 x double>)
+
+declare <1 x float> @llvm.cosh.v1f32(<1 x float>)
+declare <2 x float> @llvm.cosh.v2f32(<2 x float>)
+declare <3 x float> @llvm.cosh.v3f32(<3 x float>)
+declare <4 x float> @llvm.cosh.v4f32(<4 x float>)
+declare <5 x float> @llvm.cosh.v5f32(<5 x float>)
+declare <6 x float> @llvm.cosh.v6f32(<6 x float>)
+declare <3 x double> @llvm.cosh.v3f64(<3 x double>)
+
+declare <1 x float> @llvm.sinh.v1f32(<1 x float>)
+declare <2 x float> @llvm.sinh.v2f32(<2 x float>)
+declare <3 x float> @llvm.sinh.v3f32(<3 x float>)
+declare <4 x float> @llvm.sinh.v4f32(<4 x float>)
+declare <5 x float> @llvm.sinh.v5f32(<5 x float>)
+declare <6 x float> @llvm.sinh.v6f32(<6 x float>)
+declare <3 x double> @llvm.sinh.v3f64(<3 x double>)
+
+declare <1 x float> @llvm.tanh.v1f32(<1 x float>)
+declare <2 x float> @llvm.tanh.v2f32(<2 x float>)
+declare <3 x float> @llvm.tanh.v3f32(<3 x float>)
+declare <4 x float> @llvm.tanh.v4f32(<4 x float>)
+declare <5 x float> @llvm.tanh.v5f32(<5 x float>)
+declare <6 x float> @llvm.tanh.v6f32(<6 x float>)
+declare <3 x double> @llvm.tanh.v3f64(<3 x double>)
+
; Verify that all of the potential libcall candidates are handled.
; Some of these have custom lowering, so those cases won't have
; libcalls.
@@ -432,6 +480,1170 @@ define <3 x double> @tan_v3f64(<3 x double> %x) nounwind {
ret <3 x double> %r
}
+define <1 x float> @acos_v1f32(<1 x float> %x) nounwind {
+; CHECK-LABEL: acos_v1f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushq %rax
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: popq %rax
+; CHECK-NEXT: retq
+ %r = call <1 x float> @llvm.acos.v1f32(<1 x float> %x)
+ ret <1 x float> %r
+}
+
+define <2 x float> @acos_v2f32(<2 x float> %x) nounwind {
+; CHECK-LABEL: acos_v2f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $40, %rsp
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-NEXT: vmovshdup (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: addq $40, %rsp
+; CHECK-NEXT: retq
+ %r = call <2 x float> @llvm.acos.v2f32(<2 x float> %x)
+ ret <2 x float> %r
+}
+
+define <3 x float> @acos_v3f32(<3 x float> %x) nounwind {
+; CHECK-LABEL: acos_v3f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $40, %rsp
+; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,0]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
+; CHECK-NEXT: addq $40, %rsp
+; CHECK-NEXT: retq
+ %r = call <3 x float> @llvm.acos.v3f32(<3 x float> %x)
+ ret <3 x float> %r
+}
+
+define <4 x float> @acos_v4f32(<4 x float> %x) nounwind {
+; CHECK-LABEL: acos_v4f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $40, %rsp
+; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,0]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[3,3,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
+; CHECK-NEXT: addq $40, %rsp
+; CHECK-NEXT: retq
+ %r = call <4 x float> @llvm.acos.v4f32(<4 x float> %x)
+ ret <4 x float> %r
+}
+
+define <5 x float> @acos_v5f32(<5 x float> %x) nounwind {
+; CHECK-LABEL: acos_v5f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $72, %rsp
+; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,0]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[3,3,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
+; CHECK-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; CHECK-NEXT: addq $72, %rsp
+; CHECK-NEXT: retq
+ %r = call <5 x float> @llvm.acos.v5f32(<5 x float> %x)
+ ret <5 x float> %r
+}
+
+define <6 x float> @acos_v6f32(<6 x float> %x) nounwind {
+; CHECK-LABEL: acos_v6f32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $72, %rsp
+; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-NEXT: vmovshdup (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,1,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,0]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[3,3,3,3]
+; CHECK-NEXT: callq acosf at PLT
+; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
+; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; CHECK-NEXT: addq $72, %rsp
+; CHECK-NEXT: retq
+ %r = call <6 x float> @llvm.acos.v6f32(<6 x float> %x)
+ ret <6 x float> %r
+}
+
+define <3 x double> @acos_v3f64(<3 x double> %x) nounwind {
+; CHECK-LABEL: acos_v3f64:
+; CHECK: # %bb.0:
+; CHECK-NEXT: subq $72, %rsp
+; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acos at PLT
+; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-NEXT: # xmm0 = mem[1,0]
+; CHECK-NEXT: callq acos at PLT
+; CHECK-NEXT: vmovapd (%rsp), %xmm1 # 16-byte Reload
+; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; CHECK-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
+; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: callq acos at PLT
+; CHECK-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; CHECK-NEXT: addq $72, %rsp
+; CHECK-NEXT: retq
+ %r = call <...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/99383
More information about the llvm-commits
mailing list