blob: fe0e4409f16f235c6f6fc0379a1bcd7c23c91e63 [file] [log] [blame]
Matt Arsenault8728c5f2017-08-07 14:58:04 +00001; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
Matt Arsenault7aad8fd2017-01-24 22:02:15 +00002; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
Tom Stellard7c840bc2015-03-16 15:53:55 +00003
4declare double @llvm.maxnum.f64(double, double) nounwind readnone
5
6; SI-LABEL: {{^}}test_fmax3_f64:
Matt Arsenault687ec752018-10-22 16:27:27 +00007; SI: buffer_load_dwordx2 [[REGA:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+:[0-9]+}}], 0{{$}}
8; SI: buffer_load_dwordx2 [[REGB:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:8
Tom Stellard0d23ebe2016-08-29 19:42:52 +00009; SI: buffer_load_dwordx2 [[REGC:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:16
Matt Arsenault687ec752018-10-22 16:27:27 +000010; SI: v_max_f64 [[QUIET_A:v\[[0-9]+:[0-9]+\]]], [[REGA]], [[REGA]]
11; SI: v_max_f64 [[QUIET_B:v\[[0-9]+:[0-9]+\]]], [[REGB]], [[REGB]]
12; SI: v_max_f64 [[MAX0:v\[[0-9]+:[0-9]+\]]], [[QUIET_A]], [[QUIET_B]]
13; SI: v_max_f64 [[QUIET_C:v\[[0-9]+:[0-9]+\]]], [[REGC]], [[REGC]]
14; SI: v_max_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[MAX0]], [[QUIET_C]]
Tom Stellard7c840bc2015-03-16 15:53:55 +000015; SI: buffer_store_dwordx2 [[RESULT]],
16; SI: s_endpgm
Matt Arsenault3dbeefa2017-03-21 21:39:51 +000017define amdgpu_kernel void @test_fmax3_f64(double addrspace(1)* %out, double addrspace(1)* %aptr) nounwind {
Tom Stellard7c840bc2015-03-16 15:53:55 +000018 %bptr = getelementptr double, double addrspace(1)* %aptr, i32 1
19 %cptr = getelementptr double, double addrspace(1)* %aptr, i32 2
Matt Arsenault44e54832016-04-12 13:38:18 +000020 %a = load volatile double, double addrspace(1)* %aptr, align 8
21 %b = load volatile double, double addrspace(1)* %bptr, align 8
22 %c = load volatile double, double addrspace(1)* %cptr, align 8
Tom Stellard7c840bc2015-03-16 15:53:55 +000023 %f0 = call double @llvm.maxnum.f64(double %a, double %b) nounwind readnone
24 %f1 = call double @llvm.maxnum.f64(double %f0, double %c) nounwind readnone
25 store double %f1, double addrspace(1)* %out, align 8
26 ret void
27}