llvm-6502/test/CodeGen/R600/uint_to_fp.f64.ll
Tom Stellard 1f996fa36b R600/SI: Add a stub GCNTargetMachine
This is equivalent to the AMDGPUTargetMachine now, but it is the
starting point for separating R600 and GCN functionality into separate
targets.

It is recommened that users start using the gcn triple for GCN-based
GPUs, because using the r600 triple for these GPUs will be deprecated in
the future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225277 91177308-0d34-0410-b5e6-96231b3b80d8
2015-01-06 18:00:21 +00:00

98 lines
3.6 KiB
LLVM

; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; SI-LABEL: {{^}}v_uint_to_fp_i64_to_f64
; SI: buffer_load_dwordx2 v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}
; SI-DAG: v_cvt_f64_u32_e32 [[LO_CONV:v\[[0-9]+:[0-9]+\]]], v[[LO]]
; SI-DAG: v_cvt_f64_u32_e32 [[HI_CONV:v\[[0-9]+:[0-9]+\]]], v[[HI]]
; SI: v_ldexp_f64 [[LDEXP:v\[[0-9]+:[0-9]+\]]], [[HI_CONV]], 32
; SI: v_add_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[LDEXP]], [[LO_CONV]]
; SI: buffer_store_dwordx2 [[RESULT]]
define void @v_uint_to_fp_i64_to_f64(double addrspace(1)* %out, i64 addrspace(1)* %in) {
%tid = call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep = getelementptr i64 addrspace(1)* %in, i32 %tid
%val = load i64 addrspace(1)* %gep, align 8
%result = uitofp i64 %val to double
store double %result, double addrspace(1)* %out
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_i64_to_f64
define void @s_uint_to_fp_i64_to_f64(double addrspace(1)* %out, i64 %in) {
%cast = uitofp i64 %in to double
store double %cast, double addrspace(1)* %out, align 8
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_v2i64_to_v2f64
define void @s_uint_to_fp_v2i64_to_v2f64(<2 x double> addrspace(1)* %out, <2 x i64> %in) {
%cast = uitofp <2 x i64> %in to <2 x double>
store <2 x double> %cast, <2 x double> addrspace(1)* %out, align 16
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_v4i64_to_v4f64
define void @s_uint_to_fp_v4i64_to_v4f64(<4 x double> addrspace(1)* %out, <4 x i64> %in) {
%cast = uitofp <4 x i64> %in to <4 x double>
store <4 x double> %cast, <4 x double> addrspace(1)* %out, align 16
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_i32_to_f64
; SI: v_cvt_f64_u32_e32
; SI: s_endpgm
define void @s_uint_to_fp_i32_to_f64(double addrspace(1)* %out, i32 %in) {
%cast = uitofp i32 %in to double
store double %cast, double addrspace(1)* %out, align 8
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_v2i32_to_v2f64
; SI: v_cvt_f64_u32_e32
; SI: v_cvt_f64_u32_e32
; SI: s_endpgm
define void @s_uint_to_fp_v2i32_to_v2f64(<2 x double> addrspace(1)* %out, <2 x i32> %in) {
%cast = uitofp <2 x i32> %in to <2 x double>
store <2 x double> %cast, <2 x double> addrspace(1)* %out, align 16
ret void
}
; SI-LABEL: {{^}}s_uint_to_fp_v4i32_to_v4f64
; SI: v_cvt_f64_u32_e32
; SI: v_cvt_f64_u32_e32
; SI: v_cvt_f64_u32_e32
; SI: v_cvt_f64_u32_e32
; SI: s_endpgm
define void @s_uint_to_fp_v4i32_to_v4f64(<4 x double> addrspace(1)* %out, <4 x i32> %in) {
%cast = uitofp <4 x i32> %in to <4 x double>
store <4 x double> %cast, <4 x double> addrspace(1)* %out, align 16
ret void
}
; SI-LABEL: {{^}}uint_to_fp_i1_to_f64:
; SI: v_cmp_eq_i32_e64 [[CMP:s\[[0-9]+:[0-9]\]]],
; FIXME: We should the VGPR sources for V_CNDMASK are copied from SGPRs,
; we should be able to fold the SGPRs into the V_CNDMASK instructions.
; SI: v_cndmask_b32_e64 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, [[CMP]]
; SI: v_cndmask_b32_e64 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, [[CMP]]
; SI: buffer_store_dwordx2
; SI: s_endpgm
define void @uint_to_fp_i1_to_f64(double addrspace(1)* %out, i32 %in) {
%cmp = icmp eq i32 %in, 0
%fp = uitofp i1 %cmp to double
store double %fp, double addrspace(1)* %out, align 4
ret void
}
; SI-LABEL: {{^}}uint_to_fp_i1_to_f64_load:
; SI: v_cndmask_b32_e64 [[IRESULT:v[0-9]]], 0, 1
; SI-NEXT: v_cvt_f64_u32_e32 [[RESULT:v\[[0-9]+:[0-9]\]]], [[IRESULT]]
; SI: buffer_store_dwordx2 [[RESULT]]
; SI: s_endpgm
define void @uint_to_fp_i1_to_f64_load(double addrspace(1)* %out, i1 %in) {
%fp = uitofp i1 %in to double
store double %fp, double addrspace(1)* %out, align 8
ret void
}