mirror of
https://github.com/RPCS3/llvm.git
synced 2025-02-23 20:22:09 +00:00
2e48864110
Try to avoid mutually exclusive features. Don't use a real default GPU, and use a fake "generic". The goal is to make it easier to see which set of features are incompatible between feature strings. Most of the test changes are due to random scheduling changes from not having a default fullspeed model. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@310258 91177308-0d34-0410-b5e6-96231b3b80d8
64 lines
2.8 KiB
LLVM
64 lines
2.8 KiB
LLVM
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
|
|
|
|
declare half @llvm.cos.f16(half %a)
|
|
declare <2 x half> @llvm.cos.v2f16(<2 x half> %a)
|
|
|
|
; GCN-LABEL: {{^}}cos_f16
|
|
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
|
|
; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
|
|
; GCN: v_mul_f32_e32 v[[M_F32:[0-9]+]], {{0.15915494|0x3e22f983}}, v[[A_F32]]
|
|
; GCN: v_fract_f32_e32 v[[F_F32:[0-9]+]], v[[M_F32]]
|
|
; GCN: v_cos_f32_e32 v[[R_F32:[0-9]+]], v[[F_F32]]
|
|
; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
|
|
; GCN: buffer_store_short v[[R_F16]]
|
|
; GCN: s_endpgm
|
|
define amdgpu_kernel void @cos_f16(
|
|
half addrspace(1)* %r,
|
|
half addrspace(1)* %a) {
|
|
entry:
|
|
%a.val = load half, half addrspace(1)* %a
|
|
%r.val = call half @llvm.cos.f16(half %a.val)
|
|
store half %r.val, half addrspace(1)* %r
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}cos_v2f16
|
|
; GCN: buffer_load_dword v[[A_V2_F16:[0-9]+]]
|
|
; SI-DAG: v_mov_b32_e32 v[[HALF_PIE:[0-9]+]], 0x3e22f983{{$}}
|
|
; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
|
|
; SI: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
|
|
; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
|
|
; SI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], v[[A_F32_0]], v[[HALF_PIE]]
|
|
; SI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], v[[A_F32_1]], v[[HALF_PIE]]
|
|
|
|
; VI-DAG: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
|
|
; VI-DAG: v_cvt_f32_f16_sdwa v[[A_F32_1:[0-9]+]], v[[A_V2_F16]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
|
|
; VI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], 0.15915494, v[[A_F32_0]]
|
|
; VI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], 0.15915494, v[[A_F32_1]]
|
|
|
|
; GCN-DAG: v_fract_f32_e32 v[[F_F32_0:[0-9]+]], v[[M_F32_0]]
|
|
; GCN-DAG: v_fract_f32_e32 v[[F_F32_1:[0-9]+]], v[[M_F32_1]]
|
|
; GCN-DAG: v_cos_f32_e32 v[[R_F32_0:[0-9]+]], v[[F_F32_0]]
|
|
; GCN-DAG: v_cos_f32_e32 v[[R_F32_1:[0-9]+]], v[[F_F32_1]]
|
|
|
|
; GCN-DAG: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
|
|
; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
|
|
; VI-DAG: v_cvt_f16_f32_sdwa v[[R_F16_1:[0-9]+]], v[[R_F32_1]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
|
|
; GCN-NOT: and
|
|
|
|
; SI: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
|
|
; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_0]], v[[R_F16_HI]]
|
|
; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_0]], v[[R_F16_1]]
|
|
; GCN: buffer_store_dword v[[R_V2_F16]]
|
|
; GCN: s_endpgm
|
|
define amdgpu_kernel void @cos_v2f16(
|
|
<2 x half> addrspace(1)* %r,
|
|
<2 x half> addrspace(1)* %a) {
|
|
entry:
|
|
%a.val = load <2 x half>, <2 x half> addrspace(1)* %a
|
|
%r.val = call <2 x half> @llvm.cos.v2f16(<2 x half> %a.val)
|
|
store <2 x half> %r.val, <2 x half> addrspace(1)* %r
|
|
ret void
|
|
}
|