1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -march=amdgcn -mcpu=gfx1030 -verify-machineinstrs < %s | FileCheck %s 3; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck %s 4; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1030 -verify-machineinstrs < %s | FileCheck %s 5 6define amdgpu_gs void @test_fptrunc_round_upward(float %a, i32 %data0, <4 x i32> %data1, half addrspace(1)* %out) { 7; CHECK-LABEL: test_fptrunc_round_upward: 8; CHECK: ; %bb.0: 9; CHECK-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 2, 1), 1 10; CHECK-NEXT: v_cvt_f16_f32_e32 v0, v0 11; CHECK-NEXT: global_store_short v[6:7], v0, off 12; CHECK-NEXT: s_endpgm 13 %res = call half @llvm.fptrunc.round(float %a, metadata !"round.upward") 14 store half %res, half addrspace(1)* %out, align 4 15 ret void 16} 17 18define amdgpu_gs void @test_fptrunc_round_downward(float %a, i32 %data0, <4 x i32> %data1, half addrspace(1)* %out) { 19; CHECK-LABEL: test_fptrunc_round_downward: 20; CHECK: ; %bb.0: 21; CHECK-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 3, 1), 1 22; CHECK-NEXT: v_cvt_f16_f32_e32 v0, v0 23; CHECK-NEXT: global_store_short v[6:7], v0, off 24; CHECK-NEXT: s_endpgm 25 %res = call half @llvm.fptrunc.round(float %a, metadata !"round.downward") 26 store half %res, half addrspace(1)* %out, align 4 27 ret void 28} 29 30define amdgpu_gs void @test_fptrunc_round_upward_multiple_calls(float %a, float %b, i32 %data0, <4 x i32> %data1, half addrspace(1)* %out) { 31; CHECK-LABEL: test_fptrunc_round_upward_multiple_calls: 32; CHECK: ; %bb.0: 33; CHECK-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 2, 1), 1 34; CHECK-NEXT: v_cvt_f16_f32_e32 v0, v0 35; CHECK-NEXT: v_cvt_f16_f32_e32 v2, v1 36; CHECK-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 2, 2), 2 37; CHECK-NEXT: v_cvt_f16_f32_e32 v1, v1 38; CHECK-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 3, 1), 0 39; CHECK-NEXT: v_add_f16_e32 v0, v0, v2 40; CHECK-NEXT: v_add_f16_e32 v0, v1, v0 41; CHECK-NEXT: global_store_short v[7:8], v0, off 42; CHECK-NEXT: s_endpgm 43 %res1 = call half @llvm.fptrunc.round(float %a, metadata !"round.upward") 44 %res2 = call half @llvm.fptrunc.round(float %b, metadata !"round.upward") 45 %res3 = call half @llvm.fptrunc.round(float %b, metadata !"round.downward") 46 %res4 = fadd half %res1, %res2 47 %res5 = fadd half %res3, %res4 48 store half %res5, half addrspace(1)* %out, align 4 49 ret void 50} 51 52declare half @llvm.fptrunc.round(float, metadata) 53