blob: 579a1454dd9ae326cb0e30f219b28ed4ebedc706 [file] [log] [blame]
Alexander Timofeev982aee62017-07-04 17:32:00 +00001; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
2; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
3; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
Matt Arsenaultb0df9252014-07-10 03:22:20 +00004
Tim Northoverfd7e4242014-07-17 10:51:23 +00005declare i16 @llvm.convert.to.fp16.f32(float) nounwind readnone
Matt Arsenaultb0df9252014-07-10 03:22:20 +00006
Jan Vesely0d6cb1c2017-01-11 00:12:39 +00007; FUNC-LABEL: {{^}}test_convert_fp32_to_fp16:
8; GCN: buffer_load_dword [[VAL:v[0-9]+]]
9; GCN: v_cvt_f16_f32_e32 [[RESULT:v[0-9]+]], [[VAL]]
10; GCN: buffer_store_short [[RESULT]]
11
12; EG: MEM_RAT MSKOR
13; EG: VTX_READ_32
14; EG: FLT32_TO_FLT16
Matt Arsenault3dbeefa2017-03-21 21:39:51 +000015define amdgpu_kernel void @test_convert_fp32_to_fp16(i16 addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
David Blaikiea79ac142015-02-27 21:17:42 +000016 %val = load float, float addrspace(1)* %in, align 4
Tim Northoverfd7e4242014-07-17 10:51:23 +000017 %cvt = call i16 @llvm.convert.to.fp16.f32(float %val) nounwind readnone
Matt Arsenaultb0df9252014-07-10 03:22:20 +000018 store i16 %cvt, i16 addrspace(1)* %out, align 2
19 ret void
20}