1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-23 04:52:54 +02:00
llvm-mirror/test/CodeGen/AMDGPU/llvm.amdgcn.rsq.ll
Matt Arsenault 42dd6642c6 AMDGPU: Fold more custom nodes to undef
This will help sneak undefs past GVN into the DAG for
some tests.

Also add missing intrinsic for rsq_legacy, even though the node
was already selected to the instruction. Also start passing
the debug location to intrinsic errors.

llvm-svn: 273181
2016-06-20 18:33:56 +00:00

69 lines
2.4 KiB
LLVM

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare float @llvm.amdgcn.rsq.f32(float) #0
declare double @llvm.amdgcn.rsq.f64(double) #0
; 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 {
%rsq = call float @llvm.amdgcn.rsq.f32(float %src) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
}
; 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 {
%rsq = call float @llvm.amdgcn.rsq.f32(float 4.0) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
}
; 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 {
%rsq = call float @llvm.amdgcn.rsq.f32(float 100.0) #0
store float %rsq, float addrspace(1)* %out, align 4
ret void
}
; 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 {
%rsq = call double @llvm.amdgcn.rsq.f64(double %src) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
}
; 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 {
%rsq = call double @llvm.amdgcn.rsq.f64(double 4.0) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rsq_f64_constant_100.0
; 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 {
%rsq = call double @llvm.amdgcn.rsq.f64(double 100.0) #0
store double %rsq, double addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}rsq_undef_f32:
; SI-NOT: v_rsq_f32
define 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
}
attributes #0 = { nounwind readnone }
attributes #1 = { nounwind }