diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll')
-rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll | 16 |
1 files changed, 6 insertions, 10 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll index 9e82b41..2482d10 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.minimum.f16.ll @@ -866,19 +866,15 @@ define void @s_minimum_v2f16(<2 x half> inreg %src0, <2 x half> inreg %src1) { ; GFX11-TRUE16-LABEL: s_minimum_v2f16: ; GFX11-TRUE16: ; %bb.0: ; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX11-TRUE16-NEXT: v_cmp_o_f16_e64 s2, s0, s1 ; GFX11-TRUE16-NEXT: v_pk_min_f16 v0, s0, s1 -; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16 -; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) +; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s1, 16 +; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s0, 16 ; GFX11-TRUE16-NEXT: v_cmp_o_f16_e64 s0, s0, s1 +; GFX11-TRUE16-NEXT: v_cmp_o_f16_e64 s1, s3, s2 ; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0 -; GFX11-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v0.l, s2 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-TRUE16-NEXT: v_cndmask_b16 v1.l, 0x7e00, v1.l, s0 -; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v1, 16, v0 +; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) +; GFX11-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v0.l, s0 +; GFX11-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.l, s1 ; GFX11-TRUE16-NEXT: ;;#ASMSTART ; GFX11-TRUE16-NEXT: ; use v0 ; GFX11-TRUE16-NEXT: ;;#ASMEND |