Skip to content
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

[TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF. #72990

Merged
merged 1 commit into from
Nov 24, 2023
Merged

[TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF. #72990

merged 1 commit into from
Nov 24, 2023

Conversation

paschalis-mpeis
Copy link
Member

@paschalis-mpeis paschalis-mpeis commented Nov 21, 2023 •
edited
Loading

The first commit adds tests to shows how the mappings are missing.
The second commit adds the mappings and updates those tests.

  • Add TLI Mappings for @llvm.exp10 on ArmPL, SLEEF.
  • [TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF.

@llvmbot
Copy link
Member

llvmbot commented Nov 21, 2023 •
edited
Loading

@llvm/pr-subscribers-backend-aarch64

@llvm/pr-subscribers-llvm-transforms

Author: Paschalis Mpeis (paschalis-mpeis)

Changes
  • Add TLI Mappings for @llvm.exp10 on ArmPL, SLEEF.
  • [TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF.

Patch is 27.38 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/72990.diff

6 Files Affected:

  • (modified) llvm/include/llvm/Analysis/VecFuncs.def (+9)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll (+45-1)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll (+20)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll (+21-1)
  • (modified) llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll (+51)
  • (modified) llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll (+85-20)
diff --git a/llvm/include/llvm/Analysis/VecFuncs.def b/llvm/include/llvm/Analysis/VecFuncs.def
index b2e989e4013ea00..c628e72b24d12b0 100644
--- a/llvm/include/llvm/Analysis/VecFuncs.def
+++ b/llvm/include/llvm/Analysis/VecFuncs.def
@@ -492,6 +492,7 @@ TLI_DEFINE_VECFUNC( "exp2", "_ZGVnN2v_exp2", FIXED(2), "_ZGV_LLVM_N2v")
 TLI_DEFINE_VECFUNC( "llvm.exp2.f64", "_ZGVnN2v_exp2", FIXED(2), "_ZGV_LLVM_N2v")
 
 TLI_DEFINE_VECFUNC( "exp10", "_ZGVnN2v_exp10", FIXED(2), "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC( "llvm.exp10.f64", "_ZGVnN2v_exp10", FIXED(2), "_ZGV_LLVM_N2v")
 
 TLI_DEFINE_VECFUNC( "lgamma", "_ZGVnN2v_lgamma", FIXED(2), "_ZGV_LLVM_N2v")
 
@@ -544,6 +545,7 @@ TLI_DEFINE_VECFUNC( "exp2f", "_ZGVnN4v_exp2f", FIXED(4), "_ZGV_LLVM_N4v")
 TLI_DEFINE_VECFUNC( "llvm.exp2.f32", "_ZGVnN4v_exp2f", FIXED(4), "_ZGV_LLVM_N4v")
 
 TLI_DEFINE_VECFUNC( "exp10f", "_ZGVnN4v_exp10f", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC( "llvm.exp10.f32", "_ZGVnN4v_exp10f", FIXED(4), "_ZGV_LLVM_N4v")
 
 TLI_DEFINE_VECFUNC( "lgammaf", "_ZGVnN4v_lgammaf", FIXED(4), "_ZGV_LLVM_N4v")
 
@@ -609,6 +611,8 @@ TLI_DEFINE_VECFUNC("llvm.exp2.f32", "_ZGVsMxv_exp2f", SCALABLE(4), MASKED, "_ZGV
 
 TLI_DEFINE_VECFUNC("exp10", "_ZGVsMxv_exp10",  SCALABLE(2), MASKED, "_ZGVsMxv")
 TLI_DEFINE_VECFUNC("exp10f", "_ZGVsMxv_exp10f", SCALABLE(4), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "_ZGVsMxv_exp10",  SCALABLE(2), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "_ZGVsMxv_exp10f", SCALABLE(4), MASKED, "_ZGVsMxv")
 
 TLI_DEFINE_VECFUNC("fmod", "_ZGVsMxvv_fmod", SCALABLE(2), MASKED, "_ZGVsMxvv")
 TLI_DEFINE_VECFUNC("fmodf", "_ZGVsMxvv_fmodf", SCALABLE(4), MASKED, "_ZGVsMxvv")
@@ -753,6 +757,11 @@ TLI_DEFINE_VECFUNC("exp10f", "armpl_vexp10q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N
 TLI_DEFINE_VECFUNC("exp10", "armpl_svexp10_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
 TLI_DEFINE_VECFUNC("exp10f", "armpl_svexp10_f32_x", SCALABLE(4), MASKED, "_ZGVsMxv")
 
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "armpl_vexp10q_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "armpl_vexp10q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "armpl_svexp10_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "armpl_svexp10_f32_x", SCALABLE(4), MASKED, "_ZGVsMxv")
+
 TLI_DEFINE_VECFUNC("expm1", "armpl_vexpm1q_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
 TLI_DEFINE_VECFUNC("expm1f", "armpl_vexpm1q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
 TLI_DEFINE_VECFUNC("expm1", "armpl_svexpm1_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
index 127514961f48dfd..a38d4a53407c5d2 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
@@ -15,7 +15,7 @@ declare <vscale x 2 x double> @llvm.cos.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.cos.nxv4f32(<vscale x 4 x float>)
 
 ;.
-; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [14 x ptr] [ptr @armpl_vcosq_f64, ptr @armpl_vcosq_f32, ptr @armpl_vsinq_f64, ptr @armpl_vsinq_f32, ptr @armpl_vexpq_f64, ptr @armpl_vexpq_f32, ptr @armpl_vexp2q_f64, ptr @armpl_vexp2q_f32, ptr @armpl_vlogq_f64, ptr @armpl_vlogq_f32, ptr @armpl_vlog2q_f64, ptr @armpl_vlog2q_f32, ptr @armpl_vlog10q_f64, ptr @armpl_vlog10q_f32], section "llvm.metadata"
+; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [16 x ptr] [ptr @armpl_vcosq_f64, ptr @armpl_vcosq_f32, ptr @armpl_vsinq_f64, ptr @armpl_vsinq_f32, ptr @armpl_vexpq_f64, ptr @armpl_vexpq_f32, ptr @armpl_vexp2q_f64, ptr @armpl_vexp2q_f32, ptr @armpl_vexp10q_f64, ptr @armpl_vexp10q_f32, ptr @armpl_vlogq_f64, ptr @armpl_vlogq_f32, ptr @armpl_vlog2q_f64, ptr @armpl_vlog2q_f32, ptr @armpl_vlog10q_f64, ptr @armpl_vlog10q_f32], section "llvm.metadata"
 ;.
 define <2 x double> @llvm_cos_f64(<2 x double> %in) {
 ; CHECK-LABEL: define <2 x double> @llvm_cos_f64
@@ -192,6 +192,50 @@ define <vscale x 4 x float> @llvm_exp2_vscale_f32(<vscale x 4 x float> %in) #0 {
   ret <vscale x 4 x float> %1
 }
 
+declare <2 x double> @llvm.exp10.v2f64(<2 x double>)
+declare <4 x float> @llvm.exp10.v4f32(<4 x float>)
+declare <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double>)
+declare <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float>)
+
+define <2 x double> @llvm_exp10_f64(<2 x double> %in) {
+; CHECK-LABEL: define <2 x double> @llvm_exp10_f64
+; CHECK-SAME: (<2 x double> [[IN:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @armpl_vexp10q_f64(<2 x double> [[IN]])
+; CHECK-NEXT:    ret <2 x double> [[TMP1]]
+;
+  %1 = call fast <2 x double> @llvm.exp10.v2f64(<2 x double> %in)
+  ret <2 x double> %1
+}
+
+define <4 x float> @llvm_exp10_f32(<4 x float> %in) {
+; CHECK-LABEL: define <4 x float> @llvm_exp10_f32
+; CHECK-SAME: (<4 x float> [[IN:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <4 x float> @armpl_vexp10q_f32(<4 x float> [[IN]])
+; CHECK-NEXT:    ret <4 x float> [[TMP1]]
+;
+  %1 = call fast <4 x float> @llvm.exp10.v4f32(<4 x float> %in)
+  ret <4 x float> %1
+}
+
+define <vscale x 2 x double> @llvm_exp10_vscale_f64(<vscale x 2 x double> %in) #0 {
+; CHECK-LABEL: define <vscale x 2 x double> @llvm_exp10_vscale_f64
+; CHECK-SAME: (<vscale x 2 x double> [[IN:%.*]]) #[[ATTR1]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> [[IN]])
+; CHECK-NEXT:    ret <vscale x 2 x double> [[TMP1]]
+;
+  %1 = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> %in)
+  ret <vscale x 2 x double> %1
+}
+
+define <vscale x 4 x float> @llvm_exp10_vscale_f32(<vscale x 4 x float> %in) #0 {
+; CHECK-LABEL: define <vscale x 4 x float> @llvm_exp10_vscale_f32
+; CHECK-SAME: (<vscale x 4 x float> [[IN:%.*]]) #[[ATTR1]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> [[IN]])
+; CHECK-NEXT:    ret <vscale x 4 x float> [[TMP1]]
+;
+  %1 = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> %in)
+  ret <vscale x 4 x float> %1
+}
 
 declare <2 x double> @llvm.log.v2f64(<2 x double>)
 declare <4 x float> @llvm.log.v4f32(<4 x float>)
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
index f05af5268e957e1..8b06c41bcb1a6d1 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
@@ -95,6 +95,24 @@ define <vscale x 4 x float> @llvm_exp2_vscale_f32(<vscale x 4 x float> %in) {
   ret <vscale x 4 x float> %1
 }
 
+define <vscale x 2 x double> @llvm_exp10_vscale_f64(<vscale x 2 x double> %in) {
+; CHECK-LABEL: @llvm_exp10_vscale_f64(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> [[IN:%.*]])
+; CHECK-NEXT:    ret <vscale x 2 x double> [[TMP1]]
+;
+  %1 = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> %in)
+  ret <vscale x 2 x double> %1
+}
+
+define <vscale x 4 x float> @llvm_exp10_vscale_f32(<vscale x 4 x float> %in) {
+; CHECK-LABEL: @llvm_exp10_vscale_f32(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> [[IN:%.*]])
+; CHECK-NEXT:    ret <vscale x 4 x float> [[TMP1]]
+;
+  %1 = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> %in)
+  ret <vscale x 4 x float> %1
+}
+
 define <vscale x 2 x double> @llvm_fabs_vscale_f64(<vscale x 2 x double> %in) {
 ; CHECK-LABEL: @llvm_fabs_vscale_f64(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.fabs.nxv2f64(<vscale x 2 x double> [[IN:%.*]])
@@ -375,6 +393,8 @@ declare <vscale x 2 x double> @llvm.exp.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.exp.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.exp2.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.exp2.nxv4f32(<vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double>)
+declare <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.fabs.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.fabs.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.floor.nxv2f64(<vscale x 2 x double>)
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
index 5dd87a4bb29550c..cedb7dd85149d00 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
@@ -4,7 +4,7 @@
 target triple = "aarch64-unknown-linux-gnu"
 
 ;.
-; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [14 x ptr] [ptr @_ZGVnN2v_cos, ptr @_ZGVnN4v_cosf, ptr @_ZGVnN2v_exp, ptr @_ZGVnN4v_expf, ptr @_ZGVnN2v_exp2, ptr @_ZGVnN4v_exp2f, ptr @_ZGVnN2v_log, ptr @_ZGVnN4v_logf, ptr @_ZGVnN2v_log10, ptr @_ZGVnN4v_log10f, ptr @_ZGVnN2v_log2, ptr @_ZGVnN4v_log2f, ptr @_ZGVnN2v_sin, ptr @_ZGVnN4v_sinf], section "llvm.metadata"
+; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [16 x ptr] [ptr @_ZGVnN2v_cos, ptr @_ZGVnN4v_cosf, ptr @_ZGVnN2v_exp, ptr @_ZGVnN4v_expf, ptr @_ZGVnN2v_exp2, ptr @_ZGVnN4v_exp2f, ptr @_ZGVnN2v_exp10, ptr @_ZGVnN4v_exp10f, ptr @_ZGVnN2v_log, ptr @_ZGVnN4v_logf, ptr @_ZGVnN2v_log10, ptr @_ZGVnN4v_log10f, ptr @_ZGVnN2v_log2, ptr @_ZGVnN4v_log2f, ptr @_ZGVnN2v_sin, ptr @_ZGVnN4v_sinf], section "llvm.metadata"
 ;.
 define <2 x double> @llvm_ceil_f64(<2 x double> %in) {
 ; CHECK-LABEL: @llvm_ceil_f64(
@@ -96,6 +96,24 @@ define <4 x float> @llvm_exp2_f32(<4 x float> %in) {
   ret <4 x float> %1
 }
 
+define <2 x double> @llvm_exp10_f64(<2 x double> %in) {
+; CHECK-LABEL: @llvm_exp10_f64(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @_ZGVnN2v_exp10(<2 x double> [[IN:%.*]])
+; CHECK-NEXT:    ret <2 x double> [[TMP1]]
+;
+  %1 = call fast <2 x double> @llvm.exp10.v2f64(<2 x double> %in)
+  ret <2 x double> %1
+}
+
+define <4 x float> @llvm_exp10_f32(<4 x float> %in) {
+; CHECK-LABEL: @llvm_exp10_f32(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <4 x float> @_ZGVnN4v_exp10f(<4 x float> [[IN:%.*]])
+; CHECK-NEXT:    ret <4 x float> [[TMP1]]
+;
+  %1 = call fast <4 x float> @llvm.exp10.v4f32(<4 x float> %in)
+  ret <4 x float> %1
+}
+
 define <2 x double> @llvm_fabs_f64(<2 x double> %in) {
 ; CHECK-LABEL: @llvm_fabs_f64(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @llvm.fabs.v2f64(<2 x double> [[IN:%.*]])
@@ -376,6 +394,8 @@ declare <2 x double> @llvm.exp.v2f64(<2 x double>)
 declare <4 x float> @llvm.exp.v4f32(<4 x float>)
 declare <2 x double> @llvm.exp2.v2f64(<2 x double>)
 declare <4 x float> @llvm.exp2.v4f32(<4 x float>)
+declare <2 x double> @llvm.exp10.v2f64(<2 x double>)
+declare <4 x float> @llvm.exp10.v4f32(<4 x float>)
 declare <2 x double> @llvm.fabs.v2f64(<2 x double>)
 declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
 declare <2 x double> @llvm.floor.v2f64(<2 x double>)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
index 0a27c74782ccba1..03d959c928577d5 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
@@ -161,6 +161,57 @@ define void @exp2_f32(ptr nocapture %in.ptr, ptr %out.ptr) {
   ret void
 }
 
+declare double @llvm.exp10.f64(double)
+declare float @llvm.exp10.f32(float)
+
+define void @exp10_f64(ptr nocapture %in.ptr, ptr %out.ptr) {
+; CHECK-LABEL: @exp10_f64(
+; NEON:     [[TMP5:%.*]] = call <2 x double> @armpl_vexp10q_f64(<2 x double> [[TMP4:%.*]])
+; SVE:      [[TMP5:%.*]] = call <vscale x 2 x double> @armpl_svexp10_f64_x(<vscale x 2 x double> [[TMP4:%.*]], <vscale x 2 x i1> {{.*}})
+; CHECK:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %in.gep = getelementptr inbounds double, ptr %in.ptr, i64 %iv
+  %in = load double, ptr %in.gep, align 8
+  %call = tail call double @llvm.exp10.f64(double %in)
+  %out.gep = getelementptr inbounds double, ptr %out.ptr, i64 %iv
+  store double %call, ptr %out.gep, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
+define void @exp10_f32(ptr nocapture %in.ptr, ptr %out.ptr) {
+; CHECK-LABEL: @exp10_f32(
+; NEON:     [[TMP5:%.*]] = call <4 x float> @armpl_vexp10q_f32(<4 x float> [[TMP4:%.*]])
+; SVE:      [[TMP5:%.*]] = call <vscale x 4 x float> @armpl_svexp10_f32_x(<vscale x 4 x float> [[TMP4:%.*]], <vscale x 4 x i1> {{.*}})
+; CHECK:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %in.gep = getelementptr inbounds float, ptr %in.ptr, i64 %iv
+  %in = load float, ptr %in.gep, align 8
+  %call = tail call float @llvm.exp10.f32(float %in)
+  %out.gep = getelementptr inbounds float, ptr %out.ptr, i64 %iv
+  store float %call, ptr %out.gep, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
 declare double @llvm.log.f64(double)
 declare float @llvm.log.f32(float)
 
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
index 715c2c352b7762b..14b7fd28b38e197 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
@@ -334,6 +334,71 @@ define void @llvm_exp2_f32(float* nocapture %varray) {
   ret void
 }
 
+declare double @llvm.exp10.f64(double)
+declare float @llvm.exp10.f32(float)
+
+define void @llvm_exp10_f64(double* nocapture %varray) {
+; NEON-LABEL: define void @llvm_exp10_f64
+; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
+; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_exp10(<2 x double> [[TMP1:%.*]])
+; NEON:    [[CALL:%.*]] = tail call double @llvm.exp10.f64(double [[CONV:%.*]]) #[[ATTR7:[0-9]+]]
+; NEON:    ret void
+;
+; SVE-LABEL: define void @llvm_exp10_f64
+; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
+; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_exp10(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
+; SVE:    [[CALL:%.*]] = tail call double @llvm.exp10.f64(double [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
+; SVE:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to double
+  %call = tail call double @llvm.exp10.f64(double %conv)
+  %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
+  store double %call, double* %arrayidx, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
+define void @llvm_exp10_f32(float* nocapture %varray) {
+; NEON-LABEL: define void @llvm_exp10_f32
+; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
+; NEON:    [[TMP2:%.*]] = call <4 x float> @_ZGVnN4v_exp10f(<4 x float> [[TMP1:%.*]])
+; NEON:    [[CALL:%.*]] = tail call float @llvm.exp10.f32(float [[CONV:%.*]]) #[[ATTR8:[0-9]+]]
+; NEON:    ret void
+;
+; SVE-LABEL: define void @llvm_exp10_f32
+; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
+; SVE:    [[TMP12:%.*]] = call <vscale x 4 x float> @_ZGVsMxv_exp10f(<vscale x 4 x float> [[TMP11:%.*]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer))
+; SVE:    [[CALL:%.*]] = tail call float @llvm.exp10.f32(float [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
+; SVE:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to float
+  %call = tail call float @llvm.exp10.f32(float %conv)
+  %arrayidx = getelementptr inbounds float, float* %varray, i64 %iv
+  store float %call, float* %arrayidx, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
 declare double @llvm.fabs.f64(double)
 declare float @llvm.fabs.f32(float)
 
@@ -537,13 +602,13 @@ define void @llvm_log_f64(double* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log_f64
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_log(<2 x double> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR7:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR9:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log_f64
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_log(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR12:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -568,13 +633,13 @@ define void @llvm_log_f32(float* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log_f32
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <4 x float> @_ZGVnN4v_logf(<4 x float> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR8:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log_f32
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 4 x float> @_ZGVsMxv_logf(<vscale x 4 x float> [[TMP11:%.*]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR13:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -602,13 +667,13 @@ define void @llvm_log10_f64(double* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log10_f64
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_log10(<2 x double> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR9:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log10_f64
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_log10(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR12:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR14:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -633,13 +698,13 @@ define void @llvm_log10_f32(float* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log10_f32
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 21, 2023

@llvm/pr-subscribers-llvm-analysis

Author: Paschalis Mpeis (paschalis-mpeis)

Changes
  • Add TLI Mappings for @llvm.exp10 on ArmPL, SLEEF.
  • [TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF.

Patch is 27.38 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/72990.diff

6 Files Affected:

  • (modified) llvm/include/llvm/Analysis/VecFuncs.def (+9)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll (+45-1)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll (+20)
  • (modified) llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll (+21-1)
  • (modified) llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll (+51)
  • (modified) llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll (+85-20)
diff --git a/llvm/include/llvm/Analysis/VecFuncs.def b/llvm/include/llvm/Analysis/VecFuncs.def
index b2e989e4013ea00..c628e72b24d12b0 100644
--- a/llvm/include/llvm/Analysis/VecFuncs.def
+++ b/llvm/include/llvm/Analysis/VecFuncs.def
@@ -492,6 +492,7 @@ TLI_DEFINE_VECFUNC( "exp2", "_ZGVnN2v_exp2", FIXED(2), "_ZGV_LLVM_N2v")
 TLI_DEFINE_VECFUNC( "llvm.exp2.f64", "_ZGVnN2v_exp2", FIXED(2), "_ZGV_LLVM_N2v")
 
 TLI_DEFINE_VECFUNC( "exp10", "_ZGVnN2v_exp10", FIXED(2), "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC( "llvm.exp10.f64", "_ZGVnN2v_exp10", FIXED(2), "_ZGV_LLVM_N2v")
 
 TLI_DEFINE_VECFUNC( "lgamma", "_ZGVnN2v_lgamma", FIXED(2), "_ZGV_LLVM_N2v")
 
@@ -544,6 +545,7 @@ TLI_DEFINE_VECFUNC( "exp2f", "_ZGVnN4v_exp2f", FIXED(4), "_ZGV_LLVM_N4v")
 TLI_DEFINE_VECFUNC( "llvm.exp2.f32", "_ZGVnN4v_exp2f", FIXED(4), "_ZGV_LLVM_N4v")
 
 TLI_DEFINE_VECFUNC( "exp10f", "_ZGVnN4v_exp10f", FIXED(4), "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC( "llvm.exp10.f32", "_ZGVnN4v_exp10f", FIXED(4), "_ZGV_LLVM_N4v")
 
 TLI_DEFINE_VECFUNC( "lgammaf", "_ZGVnN4v_lgammaf", FIXED(4), "_ZGV_LLVM_N4v")
 
@@ -609,6 +611,8 @@ TLI_DEFINE_VECFUNC("llvm.exp2.f32", "_ZGVsMxv_exp2f", SCALABLE(4), MASKED, "_ZGV
 
 TLI_DEFINE_VECFUNC("exp10", "_ZGVsMxv_exp10",  SCALABLE(2), MASKED, "_ZGVsMxv")
 TLI_DEFINE_VECFUNC("exp10f", "_ZGVsMxv_exp10f", SCALABLE(4), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "_ZGVsMxv_exp10",  SCALABLE(2), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "_ZGVsMxv_exp10f", SCALABLE(4), MASKED, "_ZGVsMxv")
 
 TLI_DEFINE_VECFUNC("fmod", "_ZGVsMxvv_fmod", SCALABLE(2), MASKED, "_ZGVsMxvv")
 TLI_DEFINE_VECFUNC("fmodf", "_ZGVsMxvv_fmodf", SCALABLE(4), MASKED, "_ZGVsMxvv")
@@ -753,6 +757,11 @@ TLI_DEFINE_VECFUNC("exp10f", "armpl_vexp10q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N
 TLI_DEFINE_VECFUNC("exp10", "armpl_svexp10_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
 TLI_DEFINE_VECFUNC("exp10f", "armpl_svexp10_f32_x", SCALABLE(4), MASKED, "_ZGVsMxv")
 
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "armpl_vexp10q_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "armpl_vexp10q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("llvm.exp10.f64", "armpl_svexp10_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("llvm.exp10.f32", "armpl_svexp10_f32_x", SCALABLE(4), MASKED, "_ZGVsMxv")
+
 TLI_DEFINE_VECFUNC("expm1", "armpl_vexpm1q_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
 TLI_DEFINE_VECFUNC("expm1f", "armpl_vexpm1q_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
 TLI_DEFINE_VECFUNC("expm1", "armpl_svexpm1_f64_x",  SCALABLE(2), MASKED, "_ZGVsMxv")
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
index 127514961f48dfd..a38d4a53407c5d2 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-armpl.ll
@@ -15,7 +15,7 @@ declare <vscale x 2 x double> @llvm.cos.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.cos.nxv4f32(<vscale x 4 x float>)
 
 ;.
-; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [14 x ptr] [ptr @armpl_vcosq_f64, ptr @armpl_vcosq_f32, ptr @armpl_vsinq_f64, ptr @armpl_vsinq_f32, ptr @armpl_vexpq_f64, ptr @armpl_vexpq_f32, ptr @armpl_vexp2q_f64, ptr @armpl_vexp2q_f32, ptr @armpl_vlogq_f64, ptr @armpl_vlogq_f32, ptr @armpl_vlog2q_f64, ptr @armpl_vlog2q_f32, ptr @armpl_vlog10q_f64, ptr @armpl_vlog10q_f32], section "llvm.metadata"
+; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [16 x ptr] [ptr @armpl_vcosq_f64, ptr @armpl_vcosq_f32, ptr @armpl_vsinq_f64, ptr @armpl_vsinq_f32, ptr @armpl_vexpq_f64, ptr @armpl_vexpq_f32, ptr @armpl_vexp2q_f64, ptr @armpl_vexp2q_f32, ptr @armpl_vexp10q_f64, ptr @armpl_vexp10q_f32, ptr @armpl_vlogq_f64, ptr @armpl_vlogq_f32, ptr @armpl_vlog2q_f64, ptr @armpl_vlog2q_f32, ptr @armpl_vlog10q_f64, ptr @armpl_vlog10q_f32], section "llvm.metadata"
 ;.
 define <2 x double> @llvm_cos_f64(<2 x double> %in) {
 ; CHECK-LABEL: define <2 x double> @llvm_cos_f64
@@ -192,6 +192,50 @@ define <vscale x 4 x float> @llvm_exp2_vscale_f32(<vscale x 4 x float> %in) #0 {
   ret <vscale x 4 x float> %1
 }
 
+declare <2 x double> @llvm.exp10.v2f64(<2 x double>)
+declare <4 x float> @llvm.exp10.v4f32(<4 x float>)
+declare <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double>)
+declare <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float>)
+
+define <2 x double> @llvm_exp10_f64(<2 x double> %in) {
+; CHECK-LABEL: define <2 x double> @llvm_exp10_f64
+; CHECK-SAME: (<2 x double> [[IN:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @armpl_vexp10q_f64(<2 x double> [[IN]])
+; CHECK-NEXT:    ret <2 x double> [[TMP1]]
+;
+  %1 = call fast <2 x double> @llvm.exp10.v2f64(<2 x double> %in)
+  ret <2 x double> %1
+}
+
+define <4 x float> @llvm_exp10_f32(<4 x float> %in) {
+; CHECK-LABEL: define <4 x float> @llvm_exp10_f32
+; CHECK-SAME: (<4 x float> [[IN:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <4 x float> @armpl_vexp10q_f32(<4 x float> [[IN]])
+; CHECK-NEXT:    ret <4 x float> [[TMP1]]
+;
+  %1 = call fast <4 x float> @llvm.exp10.v4f32(<4 x float> %in)
+  ret <4 x float> %1
+}
+
+define <vscale x 2 x double> @llvm_exp10_vscale_f64(<vscale x 2 x double> %in) #0 {
+; CHECK-LABEL: define <vscale x 2 x double> @llvm_exp10_vscale_f64
+; CHECK-SAME: (<vscale x 2 x double> [[IN:%.*]]) #[[ATTR1]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> [[IN]])
+; CHECK-NEXT:    ret <vscale x 2 x double> [[TMP1]]
+;
+  %1 = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> %in)
+  ret <vscale x 2 x double> %1
+}
+
+define <vscale x 4 x float> @llvm_exp10_vscale_f32(<vscale x 4 x float> %in) #0 {
+; CHECK-LABEL: define <vscale x 4 x float> @llvm_exp10_vscale_f32
+; CHECK-SAME: (<vscale x 4 x float> [[IN:%.*]]) #[[ATTR1]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> [[IN]])
+; CHECK-NEXT:    ret <vscale x 4 x float> [[TMP1]]
+;
+  %1 = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> %in)
+  ret <vscale x 4 x float> %1
+}
 
 declare <2 x double> @llvm.log.v2f64(<2 x double>)
 declare <4 x float> @llvm.log.v4f32(<4 x float>)
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
index f05af5268e957e1..8b06c41bcb1a6d1 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef-scalable.ll
@@ -95,6 +95,24 @@ define <vscale x 4 x float> @llvm_exp2_vscale_f32(<vscale x 4 x float> %in) {
   ret <vscale x 4 x float> %1
 }
 
+define <vscale x 2 x double> @llvm_exp10_vscale_f64(<vscale x 2 x double> %in) {
+; CHECK-LABEL: @llvm_exp10_vscale_f64(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> [[IN:%.*]])
+; CHECK-NEXT:    ret <vscale x 2 x double> [[TMP1]]
+;
+  %1 = call fast <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double> %in)
+  ret <vscale x 2 x double> %1
+}
+
+define <vscale x 4 x float> @llvm_exp10_vscale_f32(<vscale x 4 x float> %in) {
+; CHECK-LABEL: @llvm_exp10_vscale_f32(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> [[IN:%.*]])
+; CHECK-NEXT:    ret <vscale x 4 x float> [[TMP1]]
+;
+  %1 = call fast <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float> %in)
+  ret <vscale x 4 x float> %1
+}
+
 define <vscale x 2 x double> @llvm_fabs_vscale_f64(<vscale x 2 x double> %in) {
 ; CHECK-LABEL: @llvm_fabs_vscale_f64(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call fast <vscale x 2 x double> @llvm.fabs.nxv2f64(<vscale x 2 x double> [[IN:%.*]])
@@ -375,6 +393,8 @@ declare <vscale x 2 x double> @llvm.exp.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.exp.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.exp2.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.exp2.nxv4f32(<vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.exp10.nxv2f64(<vscale x 2 x double>)
+declare <vscale x 4 x float> @llvm.exp10.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.fabs.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x float> @llvm.fabs.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.floor.nxv2f64(<vscale x 2 x double>)
diff --git a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
index 5dd87a4bb29550c..cedb7dd85149d00 100644
--- a/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
+++ b/llvm/test/CodeGen/AArch64/replace-intrinsics-with-veclib-sleef.ll
@@ -4,7 +4,7 @@
 target triple = "aarch64-unknown-linux-gnu"
 
 ;.
-; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [14 x ptr] [ptr @_ZGVnN2v_cos, ptr @_ZGVnN4v_cosf, ptr @_ZGVnN2v_exp, ptr @_ZGVnN4v_expf, ptr @_ZGVnN2v_exp2, ptr @_ZGVnN4v_exp2f, ptr @_ZGVnN2v_log, ptr @_ZGVnN4v_logf, ptr @_ZGVnN2v_log10, ptr @_ZGVnN4v_log10f, ptr @_ZGVnN2v_log2, ptr @_ZGVnN4v_log2f, ptr @_ZGVnN2v_sin, ptr @_ZGVnN4v_sinf], section "llvm.metadata"
+; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [16 x ptr] [ptr @_ZGVnN2v_cos, ptr @_ZGVnN4v_cosf, ptr @_ZGVnN2v_exp, ptr @_ZGVnN4v_expf, ptr @_ZGVnN2v_exp2, ptr @_ZGVnN4v_exp2f, ptr @_ZGVnN2v_exp10, ptr @_ZGVnN4v_exp10f, ptr @_ZGVnN2v_log, ptr @_ZGVnN4v_logf, ptr @_ZGVnN2v_log10, ptr @_ZGVnN4v_log10f, ptr @_ZGVnN2v_log2, ptr @_ZGVnN4v_log2f, ptr @_ZGVnN2v_sin, ptr @_ZGVnN4v_sinf], section "llvm.metadata"
 ;.
 define <2 x double> @llvm_ceil_f64(<2 x double> %in) {
 ; CHECK-LABEL: @llvm_ceil_f64(
@@ -96,6 +96,24 @@ define <4 x float> @llvm_exp2_f32(<4 x float> %in) {
   ret <4 x float> %1
 }
 
+define <2 x double> @llvm_exp10_f64(<2 x double> %in) {
+; CHECK-LABEL: @llvm_exp10_f64(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @_ZGVnN2v_exp10(<2 x double> [[IN:%.*]])
+; CHECK-NEXT:    ret <2 x double> [[TMP1]]
+;
+  %1 = call fast <2 x double> @llvm.exp10.v2f64(<2 x double> %in)
+  ret <2 x double> %1
+}
+
+define <4 x float> @llvm_exp10_f32(<4 x float> %in) {
+; CHECK-LABEL: @llvm_exp10_f32(
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast <4 x float> @_ZGVnN4v_exp10f(<4 x float> [[IN:%.*]])
+; CHECK-NEXT:    ret <4 x float> [[TMP1]]
+;
+  %1 = call fast <4 x float> @llvm.exp10.v4f32(<4 x float> %in)
+  ret <4 x float> %1
+}
+
 define <2 x double> @llvm_fabs_f64(<2 x double> %in) {
 ; CHECK-LABEL: @llvm_fabs_f64(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @llvm.fabs.v2f64(<2 x double> [[IN:%.*]])
@@ -376,6 +394,8 @@ declare <2 x double> @llvm.exp.v2f64(<2 x double>)
 declare <4 x float> @llvm.exp.v4f32(<4 x float>)
 declare <2 x double> @llvm.exp2.v2f64(<2 x double>)
 declare <4 x float> @llvm.exp2.v4f32(<4 x float>)
+declare <2 x double> @llvm.exp10.v2f64(<2 x double>)
+declare <4 x float> @llvm.exp10.v4f32(<4 x float>)
 declare <2 x double> @llvm.fabs.v2f64(<2 x double>)
 declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
 declare <2 x double> @llvm.floor.v2f64(<2 x double>)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
index 0a27c74782ccba1..03d959c928577d5 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/armpl-intrinsics.ll
@@ -161,6 +161,57 @@ define void @exp2_f32(ptr nocapture %in.ptr, ptr %out.ptr) {
   ret void
 }
 
+declare double @llvm.exp10.f64(double)
+declare float @llvm.exp10.f32(float)
+
+define void @exp10_f64(ptr nocapture %in.ptr, ptr %out.ptr) {
+; CHECK-LABEL: @exp10_f64(
+; NEON:     [[TMP5:%.*]] = call <2 x double> @armpl_vexp10q_f64(<2 x double> [[TMP4:%.*]])
+; SVE:      [[TMP5:%.*]] = call <vscale x 2 x double> @armpl_svexp10_f64_x(<vscale x 2 x double> [[TMP4:%.*]], <vscale x 2 x i1> {{.*}})
+; CHECK:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %in.gep = getelementptr inbounds double, ptr %in.ptr, i64 %iv
+  %in = load double, ptr %in.gep, align 8
+  %call = tail call double @llvm.exp10.f64(double %in)
+  %out.gep = getelementptr inbounds double, ptr %out.ptr, i64 %iv
+  store double %call, ptr %out.gep, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
+define void @exp10_f32(ptr nocapture %in.ptr, ptr %out.ptr) {
+; CHECK-LABEL: @exp10_f32(
+; NEON:     [[TMP5:%.*]] = call <4 x float> @armpl_vexp10q_f32(<4 x float> [[TMP4:%.*]])
+; SVE:      [[TMP5:%.*]] = call <vscale x 4 x float> @armpl_svexp10_f32_x(<vscale x 4 x float> [[TMP4:%.*]], <vscale x 4 x i1> {{.*}})
+; CHECK:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %in.gep = getelementptr inbounds float, ptr %in.ptr, i64 %iv
+  %in = load float, ptr %in.gep, align 8
+  %call = tail call float @llvm.exp10.f32(float %in)
+  %out.gep = getelementptr inbounds float, ptr %out.ptr, i64 %iv
+  store float %call, ptr %out.gep, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
 declare double @llvm.log.f64(double)
 declare float @llvm.log.f32(float)
 
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
index 715c2c352b7762b..14b7fd28b38e197 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sleef-intrinsic-calls-aarch64.ll
@@ -334,6 +334,71 @@ define void @llvm_exp2_f32(float* nocapture %varray) {
   ret void
 }
 
+declare double @llvm.exp10.f64(double)
+declare float @llvm.exp10.f32(float)
+
+define void @llvm_exp10_f64(double* nocapture %varray) {
+; NEON-LABEL: define void @llvm_exp10_f64
+; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
+; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_exp10(<2 x double> [[TMP1:%.*]])
+; NEON:    [[CALL:%.*]] = tail call double @llvm.exp10.f64(double [[CONV:%.*]]) #[[ATTR7:[0-9]+]]
+; NEON:    ret void
+;
+; SVE-LABEL: define void @llvm_exp10_f64
+; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
+; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_exp10(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
+; SVE:    [[CALL:%.*]] = tail call double @llvm.exp10.f64(double [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
+; SVE:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to double
+  %call = tail call double @llvm.exp10.f64(double %conv)
+  %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
+  store double %call, double* %arrayidx, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
+define void @llvm_exp10_f32(float* nocapture %varray) {
+; NEON-LABEL: define void @llvm_exp10_f32
+; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
+; NEON:    [[TMP2:%.*]] = call <4 x float> @_ZGVnN4v_exp10f(<4 x float> [[TMP1:%.*]])
+; NEON:    [[CALL:%.*]] = tail call float @llvm.exp10.f32(float [[CONV:%.*]]) #[[ATTR8:[0-9]+]]
+; NEON:    ret void
+;
+; SVE-LABEL: define void @llvm_exp10_f32
+; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
+; SVE:    [[TMP12:%.*]] = call <vscale x 4 x float> @_ZGVsMxv_exp10f(<vscale x 4 x float> [[TMP11:%.*]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer))
+; SVE:    [[CALL:%.*]] = tail call float @llvm.exp10.f32(float [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
+; SVE:    ret void
+;
+  entry:
+  br label %for.body
+
+  for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to float
+  %call = tail call float @llvm.exp10.f32(float %conv)
+  %arrayidx = getelementptr inbounds float, float* %varray, i64 %iv
+  store float %call, float* %arrayidx, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond, label %for.end, label %for.body
+
+  for.end:
+  ret void
+}
+
 declare double @llvm.fabs.f64(double)
 declare float @llvm.fabs.f32(float)
 
@@ -537,13 +602,13 @@ define void @llvm_log_f64(double* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log_f64
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_log(<2 x double> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR7:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR9:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log_f64
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_log(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call double @llvm.log.f64(double [[CONV:%.*]]) #[[ATTR12:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -568,13 +633,13 @@ define void @llvm_log_f32(float* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log_f32
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <4 x float> @_ZGVnN4v_logf(<4 x float> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR8:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR10:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log_f32
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 4 x float> @_ZGVsMxv_logf(<vscale x 4 x float> [[TMP11:%.*]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call float @llvm.log.f32(float [[CONV:%.*]]) #[[ATTR13:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -602,13 +667,13 @@ define void @llvm_log10_f64(double* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log10_f64
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2:%.*]] = call <2 x double> @_ZGVnN2v_log10(<2 x double> [[TMP1:%.*]])
-; NEON:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR9:[0-9]+]]
+; NEON:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR11:[0-9]+]]
 ; NEON:    ret void
 ;
 ; SVE-LABEL: define void @llvm_log10_f64
 ; SVE-SAME: (ptr nocapture [[VARRAY:%.*]]) #[[ATTR1]] {
 ; SVE:    [[TMP12:%.*]] = call <vscale x 2 x double> @_ZGVsMxv_log10(<vscale x 2 x double> [[TMP11:%.*]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer))
-; SVE:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR12:[0-9]+]]
+; SVE:    [[CALL:%.*]] = tail call double @llvm.log10.f64(double [[CONV:%.*]]) #[[ATTR14:[0-9]+]]
 ; SVE:    ret void
 ;
   entry:
@@ -633,13 +698,13 @@ define void @llvm_log10_f32(float* nocapture %varray) {
 ; NEON-LABEL: define void @llvm_log10_f32
 ; NEON-SAME: (ptr nocapture [[VARRAY:%.*]]) {
 ; NEON:    [[TMP2...
[truncated]

mgabka
mgabka approved these changes Nov 23, 2023
View reviewed changes
Copy link
Contributor

@mgabka mgabka left a comment

Choose a reason for hiding this comment

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

LGTM

@paschalis-mpeis
[TLI][AArch64] Add TLI Mappings of @llvm.exp10 for ArmPL and SLEEF.
de6c9c8 
Update regex to _explicitly_ show which exp versions are added.
The previous regex used `exp[^e]` to avoid matching calls like:
`@llvm.experimental.stepvector`.

Note: ArmPL Mappings for scalable types are not yet utilized
(eg, `llvm.exp10.nxv2f64`, `llvm.exp10.nxv4f32`), as `replace-with-veclib`
pass needs improvements.
@paschalis-mpeis paschalis-mpeis merged commit de6c9c8 into llvm:main Nov 24, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mgabka mgabka mgabka approved these changes

@paulwalker-arm paulwalker-arm Awaiting requested review from paulwalker-arm

@huntergr-arm huntergr-arm Awaiting requested review from huntergr-arm

Assignees
No one assigned
Labels
backend:AArch64 llvm:analysis llvm:transforms
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@paschalis-mpeis @llvmbot @paulwalker-arm @mgabka