1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GFX11 3; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GFX11 4 5declare float @llvm.amdgcn.fdot2.f32.bf16(<2 x i16> %a, <2 x i16> %b, float %c, i1 %clamp) 6 7define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f32_bf16_clamp( 8; GFX11-LABEL: test_llvm_amdgcn_fdot2_f32_bf16_clamp: 9; GFX11: ; %bb.0: ; %entry 10; GFX11-NEXT: s_load_b256 s[0:7], s[0:1], 0x24 11; GFX11-NEXT: v_mov_b32_e32 v1, 0 12; GFX11-NEXT: s_waitcnt lgkmcnt(0) 13; GFX11-NEXT: s_load_b32 s6, s[6:7], 0x0 14; GFX11-NEXT: s_load_b32 s2, s[2:3], 0x0 15; GFX11-NEXT: s_load_b32 s3, s[4:5], 0x0 16; GFX11-NEXT: s_waitcnt lgkmcnt(0) 17; GFX11-NEXT: v_mov_b32_e32 v0, s6 18; GFX11-NEXT: v_dot2_f32_bf16 v0, s2, s3, v0 clamp 19; GFX11-NEXT: global_store_b32 v1, v0, s[0:1] 20; GFX11-NEXT: s_endpgm 21 float addrspace(1)* %r, 22 <2 x i16> addrspace(1)* %a, 23 <2 x i16> addrspace(1)* %b, 24 float addrspace(1)* %c) { 25entry: 26 %a.val = load <2 x i16>, <2 x i16> addrspace(1)* %a 27 %b.val = load <2 x i16>, <2 x i16> addrspace(1)* %b 28 %c.val = load float, float addrspace(1)* %c 29 %r.val = call float @llvm.amdgcn.fdot2.f32.bf16(<2 x i16> %a.val, <2 x i16> %b.val, float %c.val, i1 1) 30 store float %r.val, float addrspace(1)* %r 31 ret void 32} 33 34 35define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f32_bf16_no_clamp( 36; GFX11-LABEL: test_llvm_amdgcn_fdot2_f32_bf16_no_clamp: 37; GFX11: ; %bb.0: ; %entry 38; GFX11-NEXT: s_load_b256 s[0:7], s[0:1], 0x24 39; GFX11-NEXT: v_mov_b32_e32 v1, 0 40; GFX11-NEXT: s_waitcnt lgkmcnt(0) 41; GFX11-NEXT: s_load_b32 s6, s[6:7], 0x0 42; GFX11-NEXT: s_load_b32 s2, s[2:3], 0x0 43; GFX11-NEXT: s_load_b32 s3, s[4:5], 0x0 44; GFX11-NEXT: s_waitcnt lgkmcnt(0) 45; GFX11-NEXT: v_mov_b32_e32 v0, s6 46; GFX11-NEXT: v_dot2_f32_bf16 v0, s2, s3, v0 47; GFX11-NEXT: global_store_b32 v1, v0, s[0:1] 48; GFX11-NEXT: s_endpgm 49 float addrspace(1)* %r, 50 <2 x i16> addrspace(1)* %a, 51 <2 x i16> addrspace(1)* %b, 52 float addrspace(1)* %c) { 53entry: 54 %a.val = load <2 x i16>, <2 x i16> addrspace(1)* %a 55 %b.val = load <2 x i16>, <2 x i16> addrspace(1)* %b 56 %c.val = load float, float addrspace(1)* %c 57 %r.val = call float @llvm.amdgcn.fdot2.f32.bf16(<2 x i16> %a.val, <2 x i16> %b.val, float %c.val, i1 0) 58 store float %r.val, float addrspace(1)* %r 59 ret void 60} 61