mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-22 20:43:44 +02:00
dd9ab77318
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
29 lines
1.2 KiB
LLVM
29 lines
1.2 KiB
LLVM
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG %s
|
|
|
|
declare float @llvm.r600.recipsqrt.ieee.f32(float) nounwind readnone
|
|
|
|
; EG-LABEL: {{^}}recipsqrt.ieee_f32:
|
|
; EG: RECIPSQRT_IEEE
|
|
define amdgpu_kernel void @recipsqrt.ieee_f32(float addrspace(1)* %out, float %src) nounwind {
|
|
%recipsqrt.ieee = call float @llvm.r600.recipsqrt.ieee.f32(float %src) nounwind readnone
|
|
store float %recipsqrt.ieee, float addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|
|
|
|
; TODO: Really these should be constant folded
|
|
; EG-LABEL: {{^}}recipsqrt.ieee_f32_constant_4.0
|
|
; EG: RECIPSQRT_IEEE
|
|
define amdgpu_kernel void @recipsqrt.ieee_f32_constant_4.0(float addrspace(1)* %out) nounwind {
|
|
%recipsqrt.ieee = call float @llvm.r600.recipsqrt.ieee.f32(float 4.0) nounwind readnone
|
|
store float %recipsqrt.ieee, float addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|
|
|
|
; EG-LABEL: {{^}}recipsqrt.ieee_f32_constant_100.0
|
|
; EG: RECIPSQRT_IEEE
|
|
define amdgpu_kernel void @recipsqrt.ieee_f32_constant_100.0(float addrspace(1)* %out) nounwind {
|
|
%recipsqrt.ieee = call float @llvm.r600.recipsqrt.ieee.f32(float 100.0) nounwind readnone
|
|
store float %recipsqrt.ieee, float addrspace(1)* %out, align 4
|
|
ret void
|
|
}
|