1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=SI -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=CI -check-prefix=FUNC %s
3 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=CI -check-prefix=FUNC %s
5 declare double @llvm.ceil.f64(double) nounwind readnone
6 declare <2 x double> @llvm.ceil.v2f64(<2 x double>) nounwind readnone
7 declare <3 x double> @llvm.ceil.v3f64(<3 x double>) nounwind readnone
8 declare <4 x double> @llvm.ceil.v4f64(<4 x double>) nounwind readnone
9 declare <8 x double> @llvm.ceil.v8f64(<8 x double>) nounwind readnone
10 declare <16 x double> @llvm.ceil.v16f64(<16 x double>) nounwind readnone
12 ; FUNC-LABEL: {{^}}fceil_f64:
14 ; SI: s_bfe_u32 [[SEXP:s[0-9]+]], {{s[0-9]+}}, 0xb0014
15 ; SI-DAG: s_and_b32 s{{[0-9]+}}, s{{[0-9]+}}, 0x80000000
16 ; FIXME: We should be using s_addk_i32 here, but the reg allocation hints
17 ; are not always followed.
18 ; SI-DAG: s_add_i32 [[SEXP0:s[0-9]+]], [[SEXP]], 0xfffffc01
19 ; SI-DAG: s_lshr_b64 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], [[SEXP0]]
27 ; SI-DAG: v_cmp_gt_f64
28 ; SI-DAG: v_cmp_lg_f64
29 ; SI-DAG: v_cndmask_b32
33 define amdgpu_kernel void @fceil_f64(double addrspace(1)* %out, double %x) {
34 %y = call double @llvm.ceil.f64(double %x) nounwind readnone
35 store double %y, double addrspace(1)* %out
39 ; FUNC-LABEL: {{^}}fceil_v2f64:
42 define amdgpu_kernel void @fceil_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
43 %y = call <2 x double> @llvm.ceil.v2f64(<2 x double> %x) nounwind readnone
44 store <2 x double> %y, <2 x double> addrspace(1)* %out
48 ; FIXME-FUNC-LABEL: {{^}}fceil_v3f64:
49 ; FIXME-CI: v_ceil_f64_e32
50 ; FIXME-CI: v_ceil_f64_e32
51 ; FIXME-CI: v_ceil_f64_e32
52 ; define amdgpu_kernel void @fceil_v3f64(<3 x double> addrspace(1)* %out, <3 x double> %x) {
53 ; %y = call <3 x double> @llvm.ceil.v3f64(<3 x double> %x) nounwind readnone
54 ; store <3 x double> %y, <3 x double> addrspace(1)* %out
58 ; FUNC-LABEL: {{^}}fceil_v4f64:
63 define amdgpu_kernel void @fceil_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %x) {
64 %y = call <4 x double> @llvm.ceil.v4f64(<4 x double> %x) nounwind readnone
65 store <4 x double> %y, <4 x double> addrspace(1)* %out
69 ; FUNC-LABEL: {{^}}fceil_v8f64:
78 define amdgpu_kernel void @fceil_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %x) {
79 %y = call <8 x double> @llvm.ceil.v8f64(<8 x double> %x) nounwind readnone
80 store <8 x double> %y, <8 x double> addrspace(1)* %out
84 ; FUNC-LABEL: {{^}}fceil_v16f64:
101 define amdgpu_kernel void @fceil_v16f64(<16 x double> addrspace(1)* %out, <16 x double> %x) {
102 %y = call <16 x double> @llvm.ceil.v16f64(<16 x double> %x) nounwind readnone
103 store <16 x double> %y, <16 x double> addrspace(1)* %out