diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll | 42 |
1 files changed, 18 insertions, 24 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll index 198bf839cb1c..3eba106b861c 100644 --- a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll +++ b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll @@ -9090,11 +9090,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset__amdgpu ; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 +; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) +; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v1 ; GFX11-TRUE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v1 +; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 ; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0 ; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-TRUE16-NEXT: .p2align 6 @@ -9146,11 +9145,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset__amdgpu ; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400 ; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0 -; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) -; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 +; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v1 ; GFX11-FAKE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v1 +; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 ; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-FAKE16-NEXT: .p2align 6 ; GFX11-FAKE16-NEXT: .LBB26_1: ; %atomicrmw.start @@ -10698,11 +10696,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset(ptr add ; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 +; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) +; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v1 ; GFX11-TRUE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v1 +; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 ; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0 ; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-TRUE16-NEXT: .p2align 6 @@ -10754,11 +10751,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset(ptr add ; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400 ; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0 -; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) -; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 +; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v1 ; GFX11-FAKE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v1 +; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 ; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-FAKE16-NEXT: .p2align 6 ; GFX11-FAKE16-NEXT: .LBB29_1: ; %atomicrmw.start @@ -11563,11 +11559,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset__amdgpu ; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400 -; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 +; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) +; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v1 ; GFX11-TRUE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v1 +; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1 ; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0 ; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-TRUE16-NEXT: .p2align 6 @@ -11619,11 +11614,10 @@ define <2 x bfloat> @buffer_fat_ptr_agent_atomic_fadd_ret_v2bf16__offset__amdgpu ; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s16 ; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400 ; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0 -; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4 -; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) -; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 +; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v1 ; GFX11-FAKE16-NEXT: buffer_load_b32 v0, v0, s[0:3], 0 offen offset:1024 -; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v1 +; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v1 ; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1 ; GFX11-FAKE16-NEXT: .p2align 6 ; GFX11-FAKE16-NEXT: .LBB31_1: ; %atomicrmw.start |