118be88e2SFraser Cormack; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 2*b279f6b0SFangrui Song; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx71 --enable-unsafe-fp-math | FileCheck --check-prefixes=CHECK %s 3*b279f6b0SFangrui Song; RUN: %if ptxas-11.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx71 --enable-unsafe-fp-math | %ptxas-verify -arch=sm_80 %} 4ca10e3b2SHan Shen 5ca10e3b2SHan Shentarget datalayout = "e-m:o-i64:64-i128:128-n32:64-S128" 6ca10e3b2SHan Shen 7ca10e3b2SHan Shendeclare <2 x bfloat> @llvm.sin.f16(<2 x bfloat> %a) #0 8ca10e3b2SHan Shendeclare <2 x bfloat> @llvm.cos.f16(<2 x bfloat> %a) #0 9ca10e3b2SHan Shen 10ca10e3b2SHan Shendefine <2 x bfloat> @test_sin(<2 x bfloat> %a) #0 #1 { 1118be88e2SFraser Cormack; CHECK-LABEL: test_sin( 1218be88e2SFraser Cormack; CHECK: { 134b24ab4bSAlex MacLean; CHECK-NEXT: .reg .b16 %rs<3>; 1418be88e2SFraser Cormack; CHECK-NEXT: .reg .b32 %r<3>; 1518be88e2SFraser Cormack; CHECK-NEXT: .reg .f32 %f<5>; 1618be88e2SFraser Cormack; CHECK-EMPTY: 1718be88e2SFraser Cormack; CHECK-NEXT: // %bb.0: 1818be88e2SFraser Cormack; CHECK-NEXT: ld.param.b32 %r1, [test_sin_param_0]; 1918be88e2SFraser Cormack; CHECK-NEXT: mov.b32 {%rs1, %rs2}, %r1; 204b24ab4bSAlex MacLean; CHECK-NEXT: cvt.f32.bf16 %f1, %rs1; 2118be88e2SFraser Cormack; CHECK-NEXT: sin.approx.f32 %f2, %f1; 224b24ab4bSAlex MacLean; CHECK-NEXT: cvt.f32.bf16 %f3, %rs2; 2318be88e2SFraser Cormack; CHECK-NEXT: sin.approx.f32 %f4, %f3; 244b24ab4bSAlex MacLean; CHECK-NEXT: cvt.rn.bf16x2.f32 %r2, %f4, %f2; 2518be88e2SFraser Cormack; CHECK-NEXT: st.param.b32 [func_retval0], %r2; 2618be88e2SFraser Cormack; CHECK-NEXT: ret; 27ca10e3b2SHan Shen %r = call <2 x bfloat> @llvm.sin.f16(<2 x bfloat> %a) 28ca10e3b2SHan Shen ret <2 x bfloat> %r 29ca10e3b2SHan Shen} 30ca10e3b2SHan Shen 31ca10e3b2SHan Shendefine <2 x bfloat> @test_cos(<2 x bfloat> %a) #0 #1 { 3218be88e2SFraser Cormack; CHECK-LABEL: test_cos( 3318be88e2SFraser Cormack; CHECK: { 344b24ab4bSAlex MacLean; CHECK-NEXT: .reg .b16 %rs<3>; 3518be88e2SFraser Cormack; CHECK-NEXT: .reg .b32 %r<3>; 3618be88e2SFraser Cormack; CHECK-NEXT: .reg .f32 %f<5>; 3718be88e2SFraser Cormack; CHECK-EMPTY: 3818be88e2SFraser Cormack; CHECK-NEXT: // %bb.0: 3918be88e2SFraser Cormack; CHECK-NEXT: ld.param.b32 %r1, [test_cos_param_0]; 4018be88e2SFraser Cormack; CHECK-NEXT: mov.b32 {%rs1, %rs2}, %r1; 414b24ab4bSAlex MacLean; CHECK-NEXT: cvt.f32.bf16 %f1, %rs1; 4218be88e2SFraser Cormack; CHECK-NEXT: cos.approx.f32 %f2, %f1; 434b24ab4bSAlex MacLean; CHECK-NEXT: cvt.f32.bf16 %f3, %rs2; 4418be88e2SFraser Cormack; CHECK-NEXT: cos.approx.f32 %f4, %f3; 454b24ab4bSAlex MacLean; CHECK-NEXT: cvt.rn.bf16x2.f32 %r2, %f4, %f2; 4618be88e2SFraser Cormack; CHECK-NEXT: st.param.b32 [func_retval0], %r2; 4718be88e2SFraser Cormack; CHECK-NEXT: ret; 48ca10e3b2SHan Shen %r = call <2 x bfloat> @llvm.cos.f16(<2 x bfloat> %a) 49ca10e3b2SHan Shen ret <2 x bfloat> %r 50ca10e3b2SHan Shen} 51ca10e3b2SHan Shen 52