[llvm] AMDGPU: Directly emit sqrt intrinsic when folding rootn(x, 2) (PR #92598)
via llvm-commits
llvm-commits at lists.llvm.org
Fri May 17 13:37:40 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-amdgpu
Author: Matt Arsenault (arsenm)
<details>
<summary>Changes</summary>
This avoids depending on pre/post link runs.
Depends #<!-- -->92595
---
Patch is 28.52 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/92598.diff
3 Files Affected:
- (modified) llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp (+34-19)
- (modified) llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-rootn.ll (+66-69)
- (modified) llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll (+1-2)
``````````diff
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
index faf04c3c7e709..aab79ceb57f22 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
@@ -22,6 +22,7 @@
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/IntrinsicsAMDGPU.h"
+#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/InitializePasses.h"
#include <cmath>
@@ -1156,35 +1157,49 @@ bool AMDGPULibCalls::fold_pow(FPMathOperator *FPOp, IRBuilder<> &B,
bool AMDGPULibCalls::fold_rootn(FPMathOperator *FPOp, IRBuilder<> &B,
const FuncInfo &FInfo) {
- // skip vector function
- if (getVecSize(FInfo) != 1)
- return false;
-
Value *opr0 = FPOp->getOperand(0);
Value *opr1 = FPOp->getOperand(1);
- ConstantInt *CINT = dyn_cast<ConstantInt>(opr1);
- if (!CINT) {
+ const APInt *CINT = nullptr;
+ if (!match(opr1, m_APIntAllowPoison(CINT)))
return false;
- }
+
+ Function *Parent = B.GetInsertBlock()->getParent();
+
int ci_opr1 = (int)CINT->getSExtValue();
- if (ci_opr1 == 1) { // rootn(x, 1) = x
- LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> " << *opr0 << "\n");
+ if (ci_opr1 == 1 && !Parent->hasFnAttribute(Attribute::StrictFP)) {
+ // rootn(x, 1) = x
+ //
+ // TODO: Insert constrained canonicalize for strictfp case.
+ LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> " << *opr0 << '\n');
replaceCall(FPOp, opr0);
return true;
}
Module *M = B.GetInsertBlock()->getModule();
- if (ci_opr1 == 2) { // rootn(x, 2) = sqrt(x)
- if (FunctionCallee FPExpr =
- getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_SQRT, FInfo))) {
- LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> sqrt(" << *opr0
- << ")\n");
- Value *nval = CreateCallEx(B,FPExpr, opr0, "__rootn2sqrt");
- replaceCall(FPOp, nval);
- return true;
- }
- } else if (ci_opr1 == 3) { // rootn(x, 3) = cbrt(x)
+
+ CallInst *CI = cast<CallInst>(FPOp);
+ if (ci_opr1 == 2 &&
+ shouldReplaceLibcallWithIntrinsic(CI,
+ /*AllowMinSizeF32=*/true,
+ /*AllowF64=*/true)) {
+ // rootn(x, 2) = sqrt(x)
+ LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> sqrt(" << *opr0 << ")\n");
+
+ CallInst *NewCall = B.CreateUnaryIntrinsic(Intrinsic::sqrt, opr0, CI);
+ NewCall->takeName(CI);
+
+ // OpenCL rootn has a looser ulp of 2 requirement than sqrt, so add some
+ // metadata.
+ MDBuilder MDHelper(M->getContext());
+ MDNode *FPMD = MDHelper.createFPMath(std::max(FPOp->getFPAccuracy(), 2.0f));
+ NewCall->setMetadata(LLVMContext::MD_fpmath, FPMD);
+
+ replaceCall(CI, NewCall);
+ return true;
+ }
+
+ if (ci_opr1 == 3) { // rootn(x, 3) = cbrt(x)
if (FunctionCallee FPExpr =
getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_CBRT, FInfo))) {
LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> cbrt(" << *opr0
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-rootn.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-rootn.ll
index 2e64a3456c242..c105ad7590e69 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-rootn.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-rootn.ll
@@ -272,8 +272,8 @@ define half @test_rootn_f16_1(half %x) {
define half @test_rootn_f16_2(half %x) {
; CHECK-LABEL: define half @test_rootn_f16_2(
; CHECK-SAME: half [[X:%.*]]) {
-; CHECK-NEXT: [[__ROOTN2SQRT:%.*]] = call half @_Z4sqrtDh(half [[X]])
-; CHECK-NEXT: ret half [[__ROOTN2SQRT]]
+; CHECK-NEXT: [[CALL:%.*]] = call half @llvm.sqrt.f16(half [[X]]), !fpmath [[META0:![0-9]+]]
+; CHECK-NEXT: ret half [[CALL]]
;
%call = tail call half @_Z5rootnDhi(half %x, i32 2)
ret half %call
@@ -342,8 +342,7 @@ define <2 x half> @test_rootn_v2f16_0(<2 x half> %x) {
define <2 x half> @test_rootn_v2f16_1(<2 x half> %x) {
; CHECK-LABEL: define <2 x half> @test_rootn_v2f16_1(
; CHECK-SAME: <2 x half> [[X:%.*]]) {
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> [[X]], <2 x i32> <i32 1, i32 1>)
-; CHECK-NEXT: ret <2 x half> [[CALL]]
+; CHECK-NEXT: ret <2 x half> [[X]]
;
%call = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> %x, <2 x i32> <i32 1, i32 1>)
ret <2 x half> %call
@@ -352,7 +351,7 @@ define <2 x half> @test_rootn_v2f16_1(<2 x half> %x) {
define <2 x half> @test_rootn_v2f16_2(<2 x half> %x) {
; CHECK-LABEL: define <2 x half> @test_rootn_v2f16_2(
; CHECK-SAME: <2 x half> [[X:%.*]]) {
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> [[X]], <2 x i32> <i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <2 x half> @llvm.sqrt.v2f16(<2 x half> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <2 x half> [[CALL]]
;
%call = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> %x, <2 x i32> <i32 2, i32 2>)
@@ -362,8 +361,8 @@ define <2 x half> @test_rootn_v2f16_2(<2 x half> %x) {
define <2 x half> @test_rootn_v2f16_neg1(<2 x half> %x) {
; CHECK-LABEL: define <2 x half> @test_rootn_v2f16_neg1(
; CHECK-SAME: <2 x half> [[X:%.*]]) {
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> [[X]], <2 x i32> <i32 -1, i32 -1>)
-; CHECK-NEXT: ret <2 x half> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <2 x half> <half 0xH3C00, half 0xH3C00>, [[X]]
+; CHECK-NEXT: ret <2 x half> [[__ROOTN2DIV]]
;
%call = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> %x, <2 x i32> <i32 -1, i32 -1>)
ret <2 x half> %call
@@ -372,8 +371,8 @@ define <2 x half> @test_rootn_v2f16_neg1(<2 x half> %x) {
define <2 x half> @test_rootn_v2f16_neg2(<2 x half> %x) {
; CHECK-LABEL: define <2 x half> @test_rootn_v2f16_neg2(
; CHECK-SAME: <2 x half> [[X:%.*]]) {
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> [[X]], <2 x i32> <i32 -2, i32 -2>)
-; CHECK-NEXT: ret <2 x half> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2RSQRT:%.*]] = call <2 x half> @_Z5rsqrtDv2_Dh(<2 x half> [[X]])
+; CHECK-NEXT: ret <2 x half> [[__ROOTN2RSQRT]]
;
%call = tail call <2 x half> @_Z5rootnDv2_DhDv2_i(<2 x half> %x, <2 x i32> <i32 -2, i32 -2>)
ret <2 x half> %call
@@ -512,7 +511,8 @@ define float @test_rootn_f32__y_1__strictfp(float %x) #1 {
; CHECK-LABEL: define float @test_rootn_f32__y_1__strictfp(
; CHECK-SAME: float [[X:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: entry:
-; CHECK-NEXT: ret float [[X]]
+; CHECK-NEXT: [[CALL:%.*]] = tail call float @_Z5rootnfi(float [[X]], i32 1) #[[ATTR0]]
+; CHECK-NEXT: ret float [[CALL]]
;
entry:
%call = tail call float @_Z5rootnfi(float %x, i32 1) #1
@@ -523,8 +523,7 @@ define <2 x float> @test_rootn_v2f32__y_1(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_1(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 1, i32 1>)
-; CHECK-NEXT: ret <2 x float> [[CALL]]
+; CHECK-NEXT: ret <2 x float> [[X]]
;
entry:
%call = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> %x, <2 x i32> <i32 1, i32 1>)
@@ -547,8 +546,7 @@ define <2 x float> @test_rootn_v2f32__y_1_undef(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_1_undef(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 1, i32 poison>)
-; CHECK-NEXT: ret <2 x float> [[CALL]]
+; CHECK-NEXT: ret <2 x float> [[X]]
;
entry:
%call = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> %x, <2 x i32> <i32 1, i32 poison>)
@@ -559,8 +557,7 @@ define <3 x float> @test_rootn_v3f32__y_1(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_1(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 1, i32 1, i32 1>)
-; CHECK-NEXT: ret <3 x float> [[CALL]]
+; CHECK-NEXT: ret <3 x float> [[X]]
;
entry:
%call = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> %x, <3 x i32> <i32 1, i32 1, i32 1>)
@@ -571,8 +568,7 @@ define <3 x float> @test_rootn_v3f32__y_1_undef(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_1_undef(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 1, i32 1, i32 poison>)
-; CHECK-NEXT: ret <3 x float> [[CALL]]
+; CHECK-NEXT: ret <3 x float> [[X]]
;
entry:
%call = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> %x, <3 x i32> <i32 1, i32 1, i32 poison>)
@@ -583,8 +579,7 @@ define <4 x float> @test_rootn_v4f32__y_1(<4 x float> %x) {
; CHECK-LABEL: define <4 x float> @test_rootn_v4f32__y_1(
; CHECK-SAME: <4 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <4 x float> @_Z5rootnDv4_fDv4_i(<4 x float> [[X]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>)
-; CHECK-NEXT: ret <4 x float> [[CALL]]
+; CHECK-NEXT: ret <4 x float> [[X]]
;
entry:
%call = tail call <4 x float> @_Z5rootnDv4_fDv4_i(<4 x float> %x, <4 x i32> <i32 1, i32 1, i32 1, i32 1>)
@@ -595,8 +590,7 @@ define <8 x float> @test_rootn_v8f32__y_1(<8 x float> %x) {
; CHECK-LABEL: define <8 x float> @test_rootn_v8f32__y_1(
; CHECK-SAME: <8 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <8 x float> @_Z5rootnDv8_fDv8_i(<8 x float> [[X]], <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>)
-; CHECK-NEXT: ret <8 x float> [[CALL]]
+; CHECK-NEXT: ret <8 x float> [[X]]
;
entry:
%call = tail call <8 x float> @_Z5rootnDv8_fDv8_i(<8 x float> %x, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>)
@@ -607,8 +601,7 @@ define <16 x float> @test_rootn_v16f32__y_1(<16 x float> %x) {
; CHECK-LABEL: define <16 x float> @test_rootn_v16f32__y_1(
; CHECK-SAME: <16 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <16 x float> @_Z5rootnDv16_fDv16_i(<16 x float> [[X]], <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>)
-; CHECK-NEXT: ret <16 x float> [[CALL]]
+; CHECK-NEXT: ret <16 x float> [[X]]
;
entry:
%call = tail call <16 x float> @_Z5rootnDv16_fDv16_i(<16 x float> %x, <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>)
@@ -619,8 +612,8 @@ define float @test_rootn_f32__y_2(float %x) {
; CHECK-LABEL: define float @test_rootn_f32__y_2(
; CHECK-SAME: float [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[__ROOTN2SQRT:%.*]] = call float @_Z4sqrtf(float [[X]])
-; CHECK-NEXT: ret float [[__ROOTN2SQRT]]
+; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.sqrt.f32(float [[X]]), !fpmath [[META0]]
+; CHECK-NEXT: ret float [[CALL]]
;
entry:
%call = tail call float @_Z5rootnfi(float %x, i32 2)
@@ -631,8 +624,8 @@ define float @test_rootn_f32__y_2_flags(float %x) {
; CHECK-LABEL: define float @test_rootn_f32__y_2_flags(
; CHECK-SAME: float [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[__ROOTN2SQRT:%.*]] = call nnan nsz float @_Z4sqrtf(float [[X]])
-; CHECK-NEXT: ret float [[__ROOTN2SQRT]]
+; CHECK-NEXT: [[CALL:%.*]] = call nnan nsz float @llvm.sqrt.f32(float [[X]]), !fpmath [[META0]]
+; CHECK-NEXT: ret float [[CALL]]
;
entry:
%call = tail call nnan nsz float @_Z5rootnfi(float %x, i32 2)
@@ -644,8 +637,8 @@ define float @test_rootn_f32__y_2_fpmath_3(float %x) {
; CHECK-LABEL: define float @test_rootn_f32__y_2_fpmath_3(
; CHECK-SAME: float [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[__ROOTN2SQRT:%.*]] = call nnan nsz float @_Z4sqrtf(float [[X]])
-; CHECK-NEXT: ret float [[__ROOTN2SQRT]]
+; CHECK-NEXT: [[CALL:%.*]] = call nnan nsz float @llvm.sqrt.f32(float [[X]]), !fpmath [[META1:![0-9]+]]
+; CHECK-NEXT: ret float [[CALL]]
;
entry:
%call = tail call nnan nsz float @_Z5rootnfi(float %x, i32 2), !fpmath !0
@@ -656,7 +649,7 @@ define <2 x float> @test_rootn_v2f32__y_2_flags(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_2_flags(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call nnan nsz <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call nnan nsz <2 x float> @llvm.sqrt.v2f32(<2 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <2 x float> [[CALL]]
;
entry:
@@ -668,7 +661,7 @@ define <3 x float> @test_rootn_v3f32__y_2(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_2(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 2, i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <3 x float> @llvm.sqrt.v3f32(<3 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <3 x float> [[CALL]]
;
entry:
@@ -680,7 +673,7 @@ define <3 x float> @test_rootn_v3f32__y_2_undef(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_2_undef(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 2, i32 poison, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <3 x float> @llvm.sqrt.v3f32(<3 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <3 x float> [[CALL]]
;
entry:
@@ -692,7 +685,7 @@ define <4 x float> @test_rootn_v4f32__y_2(<4 x float> %x) {
; CHECK-LABEL: define <4 x float> @test_rootn_v4f32__y_2(
; CHECK-SAME: <4 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <4 x float> @_Z5rootnDv4_fDv4_i(<4 x float> [[X]], <4 x i32> <i32 2, i32 2, i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <4 x float> [[CALL]]
;
entry:
@@ -704,7 +697,7 @@ define <8 x float> @test_rootn_v8f32__y_2(<8 x float> %x) {
; CHECK-LABEL: define <8 x float> @test_rootn_v8f32__y_2(
; CHECK-SAME: <8 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <8 x float> @_Z5rootnDv8_fDv8_i(<8 x float> [[X]], <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <8 x float> @llvm.sqrt.v8f32(<8 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <8 x float> [[CALL]]
;
entry:
@@ -716,7 +709,7 @@ define <16 x float> @test_rootn_v16f32__y_2(<16 x float> %x) {
; CHECK-LABEL: define <16 x float> @test_rootn_v16f32__y_2(
; CHECK-SAME: <16 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <16 x float> @_Z5rootnDv16_fDv16_i(<16 x float> [[X]], <16 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>)
+; CHECK-NEXT: [[CALL:%.*]] = call <16 x float> @llvm.sqrt.v16f32(<16 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <16 x float> [[CALL]]
;
entry:
@@ -740,8 +733,8 @@ define <2 x float> @test_rootn_v2f32__y_3(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_3(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 3, i32 3>)
-; CHECK-NEXT: ret <2 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2CBRT:%.*]] = call <2 x float> @_Z4cbrtDv2_f(<2 x float> [[X]])
+; CHECK-NEXT: ret <2 x float> [[__ROOTN2CBRT]]
;
entry:
%call = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> %x, <2 x i32> <i32 3, i32 3>)
@@ -764,7 +757,7 @@ define <2 x float> @test_rootn_v2f32__y_nonsplat_2_poison(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_nonsplat_2_poison(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 2, i32 poison>)
+; CHECK-NEXT: [[CALL:%.*]] = call <2 x float> @llvm.sqrt.v2f32(<2 x float> [[X]]), !fpmath [[META0]]
; CHECK-NEXT: ret <2 x float> [[CALL]]
;
entry:
@@ -800,8 +793,8 @@ define <2 x float> @test_rootn_v2f32__y_neg1(<2 x float> %x) {
; CHECK-LABEL: define <2 x float> @test_rootn_v2f32__y_neg1(
; CHECK-SAME: <2 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> [[X]], <2 x i32> <i32 -1, i32 -1>)
-; CHECK-NEXT: ret <2 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, [[X]]
+; CHECK-NEXT: ret <2 x float> [[__ROOTN2DIV]]
;
entry:
%call = tail call <2 x float> @_Z5rootnDv2_fDv2_i(<2 x float> %x, <2 x i32> <i32 -1, i32 -1>)
@@ -812,8 +805,8 @@ define <3 x float> @test_rootn_v3f32__y_neg1(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_neg1(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 -1, i32 -1, i32 -1>)
-; CHECK-NEXT: ret <3 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[X]]
+; CHECK-NEXT: ret <3 x float> [[__ROOTN2DIV]]
;
entry:
%call = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> %x, <3 x i32> <i32 -1, i32 -1, i32 -1>)
@@ -824,8 +817,8 @@ define <3 x float> @test_rootn_v3f32__y_neg1_undef(<3 x float> %x) {
; CHECK-LABEL: define <3 x float> @test_rootn_v3f32__y_neg1_undef(
; CHECK-SAME: <3 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> [[X]], <3 x i32> <i32 -1, i32 -1, i32 poison>)
-; CHECK-NEXT: ret <3 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[X]]
+; CHECK-NEXT: ret <3 x float> [[__ROOTN2DIV]]
;
entry:
%call = tail call <3 x float> @_Z5rootnDv3_fDv3_i(<3 x float> %x, <3 x i32> <i32 -1, i32 -1, i32 poison>)
@@ -836,8 +829,8 @@ define <4 x float> @test_rootn_v4f32__y_neg1(<4 x float> %x) {
; CHECK-LABEL: define <4 x float> @test_rootn_v4f32__y_neg1(
; CHECK-SAME: <4 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <4 x float> @_Z5rootnDv4_fDv4_i(<4 x float> [[X]], <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>)
-; CHECK-NEXT: ret <4 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[X]]
+; CHECK-NEXT: ret <4 x float> [[__ROOTN2DIV]]
;
entry:
%call = tail call <4 x float> @_Z5rootnDv4_fDv4_i(<4 x float> %x, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>)
@@ -848,8 +841,8 @@ define <8 x float> @test_rootn_v8f32__y_neg1(<8 x float> %x) {
; CHECK-LABEL: define <8 x float> @test_rootn_v8f32__y_neg1(
; CHECK-SAME: <8 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <8 x float> @_Z5rootnDv8_fDv8_i(<8 x float> [[X]], <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>)
-; CHECK-NEXT: ret <8 x float> [[CALL]]
+; CHECK-NEXT: [[__ROOTN2DIV:%.*]] = fdiv <8 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[X]]
+; CHECK-NEXT: ret <8 x float> [[__ROOTN2DIV]]
;
entry:
%call = tail call <8 x float> @_Z5rootnDv8_fDv8_i(<8 x float> %x, <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>)
@@ -860,8 +853,8 @@ define <16 x float> @test_rootn_v16f32__y_neg1(<16 x float> %x) {
; CHECK-LABEL: define <16 x float> @test_rootn_v16f32__y_neg1(
; CHECK-SAME: <16 x float> [[X:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[CALL:%.*]] = tail call <16 x float> @_Z5rootnDv16_fDv16_i(<16 x float> [[X]], <16 x i32> <i32 -1, i32 -1...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/92598
More information about the llvm-commits
mailing list