AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel
Currently the default C calling convention functions are treated
the same as compute kernels. Make this explicit so the default
calling convention can be changed to a non-kernel.
Converted with perl -pi -e 's/define void/define amdgpu_kernel void/'
on the relevant test directories (and undoing in one place that actually
wanted a non-kernel).
llvm-svn: 298444
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.ll
index c644288..0ce26d0 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.ll
@@ -6,7 +6,7 @@
; FUNC-LABEL: {{^}}rsq_f32:
; SI: v_rsq_f32_e32 {{v[0-9]+}}, {{s[0-9]+}}
-define void @rsq_f32(float addrspace(1)* %out, float %src) #1 {
+define amdgpu_kernel void @rsq_f32(float addrspace(1)* %out, float %src) #1 {
%rsq = call float @llvm.amdgcn.rsq.f32(float %src) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
@@ -15,7 +15,7 @@
; TODO: Really these should be constant folded
; FUNC-LABEL: {{^}}rsq_f32_constant_4.0
; SI: v_rsq_f32_e32 {{v[0-9]+}}, 4.0
-define void @rsq_f32_constant_4.0(float addrspace(1)* %out) #1 {
+define amdgpu_kernel void @rsq_f32_constant_4.0(float addrspace(1)* %out) #1 {
%rsq = call float @llvm.amdgcn.rsq.f32(float 4.0) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
@@ -23,7 +23,7 @@
; FUNC-LABEL: {{^}}rsq_f32_constant_100.0
; SI: v_rsq_f32_e32 {{v[0-9]+}}, 0x42c80000
-define void @rsq_f32_constant_100.0(float addrspace(1)* %out) #1 {
+define amdgpu_kernel void @rsq_f32_constant_100.0(float addrspace(1)* %out) #1 {
%rsq = call float @llvm.amdgcn.rsq.f32(float 100.0) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
@@ -31,7 +31,7 @@
; FUNC-LABEL: {{^}}rsq_f64:
; SI: v_rsq_f64_e32 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}
-define void @rsq_f64(double addrspace(1)* %out, double %src) #1 {
+define amdgpu_kernel void @rsq_f64(double addrspace(1)* %out, double %src) #1 {
%rsq = call double @llvm.amdgcn.rsq.f64(double %src) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
@@ -40,7 +40,7 @@
; TODO: Really these should be constant folded
; FUNC-LABEL: {{^}}rsq_f64_constant_4.0
; SI: v_rsq_f64_e32 {{v\[[0-9]+:[0-9]+\]}}, 4.0
-define void @rsq_f64_constant_4.0(double addrspace(1)* %out) #1 {
+define amdgpu_kernel void @rsq_f64_constant_4.0(double addrspace(1)* %out) #1 {
%rsq = call double @llvm.amdgcn.rsq.f64(double 4.0) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
@@ -50,7 +50,7 @@
; SI-DAG: s_mov_b32 s{{[0-9]+}}, 0x40590000
; SI-DAG: s_mov_b32 s{{[0-9]+}}, 0{{$}}
; SI: v_rsq_f64_e32 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}
-define void @rsq_f64_constant_100.0(double addrspace(1)* %out) #1 {
+define amdgpu_kernel void @rsq_f64_constant_100.0(double addrspace(1)* %out) #1 {
%rsq = call double @llvm.amdgcn.rsq.f64(double 100.0) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
@@ -58,7 +58,7 @@
; FUNC-LABEL: {{^}}rsq_undef_f32:
; SI-NOT: v_rsq_f32
-define void @rsq_undef_f32(float addrspace(1)* %out) #1 {
+define amdgpu_kernel void @rsq_undef_f32(float addrspace(1)* %out) #1 {
%rsq = call float @llvm.amdgcn.rsq.f32(float undef)
store float %rsq, float addrspace(1)* %out, align 4
ret void