diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll | 307 |
1 files changed, 98 insertions, 209 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll index f9a24fe..3856f0c 100644 --- a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll +++ b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll @@ -24,9 +24,9 @@ define amdgpu_kernel void @raw_buffer_atomic_add_noret_f64(<4 x i32> %rsrc, doub ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_buffer_atomic_add_noret_f64: @@ -35,9 +35,9 @@ define amdgpu_kernel void @raw_buffer_atomic_add_noret_f64(<4 x i32> %rsrc, doub ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_buffer_atomic_add_noret_f64: @@ -143,9 +143,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_noret_f64(ptr addrspace(8) ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_ptr_buffer_atomic_add_noret_f64: @@ -154,9 +154,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_noret_f64(ptr addrspace(8) ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_ptr_buffer_atomic_add_noret_f64: @@ -262,9 +262,9 @@ define amdgpu_kernel void @struct_buffer_atomic_add_noret_f64(<4 x i32> %rsrc, d ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_buffer_atomic_add_noret_f64: @@ -273,9 +273,9 @@ define amdgpu_kernel void @struct_buffer_atomic_add_noret_f64(<4 x i32> %rsrc, d ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_buffer_atomic_add_noret_f64: @@ -380,9 +380,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_noret_f64(ptr addrspace( ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_ptr_buffer_atomic_add_noret_f64: @@ -391,9 +391,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_noret_f64(ptr addrspace( ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_ptr_buffer_atomic_add_noret_f64: @@ -498,9 +498,9 @@ define amdgpu_kernel void @raw_buffer_atomic_min_noret_f64(<4 x i32> %rsrc, doub ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_buffer_atomic_min_noret_f64: @@ -509,9 +509,9 @@ define amdgpu_kernel void @raw_buffer_atomic_min_noret_f64(<4 x i32> %rsrc, doub ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_buffer_atomic_min_noret_f64: @@ -617,9 +617,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_noret_f64(ptr addrspace(8) ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_ptr_buffer_atomic_min_noret_f64: @@ -628,9 +628,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_noret_f64(ptr addrspace(8) ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_ptr_buffer_atomic_min_noret_f64: @@ -736,9 +736,9 @@ define amdgpu_kernel void @struct_buffer_atomic_min_noret_f64(<4 x i32> %rsrc, d ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_buffer_atomic_min_noret_f64: @@ -747,9 +747,9 @@ define amdgpu_kernel void @struct_buffer_atomic_min_noret_f64(<4 x i32> %rsrc, d ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_buffer_atomic_min_noret_f64: @@ -854,9 +854,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_noret_f64(ptr addrspace( ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_ptr_buffer_atomic_min_noret_f64: @@ -865,9 +865,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_noret_f64(ptr addrspace( ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_ptr_buffer_atomic_min_noret_f64: @@ -972,9 +972,9 @@ define amdgpu_kernel void @raw_buffer_atomic_max_noret_f64(<4 x i32> %rsrc, doub ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_buffer_atomic_max_noret_f64: @@ -983,9 +983,9 @@ define amdgpu_kernel void @raw_buffer_atomic_max_noret_f64(<4 x i32> %rsrc, doub ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_buffer_atomic_max_noret_f64: @@ -1091,9 +1091,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_noret_f64(ptr addrspace(8) ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 offen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 offen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: raw_ptr_buffer_atomic_max_noret_f64: @@ -1102,9 +1102,9 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_noret_f64(ptr addrspace(8) ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 offen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 offen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: raw_ptr_buffer_atomic_max_noret_f64: @@ -1210,9 +1210,9 @@ define amdgpu_kernel void @struct_buffer_atomic_max_noret_f64(<4 x i32> %rsrc, d ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_buffer_atomic_max_noret_f64: @@ -1221,9 +1221,9 @@ define amdgpu_kernel void @struct_buffer_atomic_max_noret_f64(<4 x i32> %rsrc, d ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_buffer_atomic_max_noret_f64: @@ -1328,9 +1328,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_noret_f64(ptr addrspace( ; GFX90A-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX90A-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1] -; GFX90A-NEXT: v_mov_b32_e32 v2, s8 -; GFX90A-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s8 +; GFX90A-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX90A-NEXT: s_endpgm ; ; GFX942-LABEL: struct_ptr_buffer_atomic_max_noret_f64: @@ -1339,9 +1339,9 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_noret_f64(ptr addrspace( ; GFX942-NEXT: s_load_dword s8, s[4:5], 0x3c ; GFX942-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[6:7] -; GFX942-NEXT: v_mov_b32_e32 v2, s8 -; GFX942-NEXT: buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7] +; GFX942-NEXT: v_mov_b32_e32 v0, s8 +; GFX942-NEXT: buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen ; GFX942-NEXT: s_endpgm ; ; GFX1250-LABEL: struct_ptr_buffer_atomic_max_noret_f64: @@ -2079,9 +2079,9 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret(ptr addrspace(3) %ptr, do ; GFX90A-NEXT: s_load_dword s2, s[4:5], 0x24 ; GFX90A-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x2c ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) -; GFX90A-NEXT: v_mov_b32_e32 v2, s2 -; GFX90A-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1] -; GFX90A-NEXT: ds_add_f64 v2, v[0:1] +; GFX90A-NEXT: v_mov_b32_e32 v0, s2 +; GFX90A-NEXT: v_pk_mov_b32 v[2:3], s[0:1], s[0:1] op_sel:[0,1] +; GFX90A-NEXT: ds_add_f64 v0, v[2:3] ; GFX90A-NEXT: s_waitcnt lgkmcnt(0) ; GFX90A-NEXT: s_endpgm ; @@ -2090,9 +2090,9 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret(ptr addrspace(3) %ptr, do ; GFX942-NEXT: s_load_dword s2, s[4:5], 0x24 ; GFX942-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x2c ; GFX942-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-NEXT: v_mov_b32_e32 v2, s2 -; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[0:1] -; GFX942-NEXT: ds_add_f64 v2, v[0:1] +; GFX942-NEXT: v_mov_b32_e32 v0, s2 +; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[0:1] +; GFX942-NEXT: ds_add_f64 v0, v[2:3] ; GFX942-NEXT: s_waitcnt lgkmcnt(0) ; GFX942-NEXT: s_endpgm ; @@ -2102,23 +2102,10 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret(ptr addrspace(3) %ptr, do ; GFX1250-NEXT: s_load_b32 s2, s[4:5], 0x24 ; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x2c ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v2, s2 -; GFX1250-NEXT: s_mov_b32 s2, 0 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: .LBB51_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 +; GFX1250-NEXT: v_mov_b32_e32 v2, s2 +; GFX1250-NEXT: v_mov_b64_e32 v[0:1], s[0:1] +; GFX1250-NEXT: ds_add_f64 v2, v[0:1] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX1250-NEXT: v_add_f64_e32 v[4:5], s[0:1], v[0:1] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[4:5], v2, v[4:5], v[0:1] -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[4:5], v[0:1] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], v[4:5] -; GFX1250-NEXT: s_or_b32 s2, vcc_lo, s2 -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s2 -; GFX1250-NEXT: s_cbranch_execnz .LBB51_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end ; GFX1250-NEXT: s_endpgm main_body: %ret = call double @llvm.amdgcn.ds.fadd.f64(ptr addrspace(3) %ptr, double %data, i32 0, i32 0, i1 0) @@ -2148,24 +2135,9 @@ define double @local_atomic_fadd_f64_rtn(ptr addrspace(3) %ptr, double %data) { ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v2, v0 -; GFX1250-NEXT: v_mov_b32_e32 v4, v1 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: .LBB52_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_mov_b64_e32 v[6:7], v[0:1] -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1) -; GFX1250-NEXT: v_add_f64_e32 v[0:1], v[6:7], v[4:5] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[0:1], v2, v[0:1], v[6:7] +; GFX1250-NEXT: v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1 +; GFX1250-NEXT: ds_add_rtn_f64 v[0:1], v0, v[2:3] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[0:1], v[6:7] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB52_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end -; GFX1250-NEXT: s_or_b32 exec_lo, exec_lo, s0 ; GFX1250-NEXT: s_set_pc_i64 s[30:31] main_body: %ret = call double @llvm.amdgcn.ds.fadd.f64(ptr addrspace(3) %ptr, double %data, i32 0, i32 0, i1 0) @@ -2197,24 +2169,11 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret_pat(ptr addrspace(3) %ptr ; GFX1250-LABEL: local_atomic_fadd_f64_noret_pat: ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_load_b32 s0, s[4:5], 0x24 +; GFX1250-NEXT: v_mov_b64_e32 v[0:1], 4.0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v2, s0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: .LBB53_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX1250-NEXT: v_add_f64_e32 v[4:5], 4.0, v[0:1] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[4:5], v2, v[4:5], v[0:1] +; GFX1250-NEXT: v_mov_b32_e32 v2, s0 +; GFX1250-NEXT: ds_add_f64 v2, v[0:1] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[4:5], v[0:1] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], v[4:5] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB53_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end ; GFX1250-NEXT: s_endpgm main_body: %ret = atomicrmw fadd ptr addrspace(3) %ptr, double 4.0 seq_cst, !amdgpu.no.fine.grained.memory !0 @@ -2246,24 +2205,11 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret_pat_flush(ptr addrspace(3 ; GFX1250-LABEL: local_atomic_fadd_f64_noret_pat_flush: ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_load_b32 s0, s[4:5], 0x24 +; GFX1250-NEXT: v_mov_b64_e32 v[0:1], 4.0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v2, s0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: .LBB54_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 +; GFX1250-NEXT: v_mov_b32_e32 v2, s0 +; GFX1250-NEXT: ds_add_f64 v2, v[0:1] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX1250-NEXT: v_add_f64_e32 v[4:5], 4.0, v[0:1] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[4:5], v2, v[4:5], v[0:1] -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[4:5], v[0:1] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], v[4:5] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB54_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end ; GFX1250-NEXT: s_endpgm main_body: %ret = atomicrmw fadd ptr addrspace(3) %ptr, double 4.0 seq_cst, !amdgpu.no.fine.grained.memory !0 @@ -2295,24 +2241,11 @@ define amdgpu_kernel void @local_atomic_fadd_f64_noret_pat_flush_safe(ptr addrsp ; GFX1250-LABEL: local_atomic_fadd_f64_noret_pat_flush_safe: ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_load_b32 s0, s[4:5], 0x24 +; GFX1250-NEXT: v_mov_b64_e32 v[0:1], 4.0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v2, s0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: .LBB55_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 +; GFX1250-NEXT: v_mov_b32_e32 v2, s0 +; GFX1250-NEXT: ds_add_f64 v2, v[0:1] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX1250-NEXT: v_add_f64_e32 v[4:5], 4.0, v[0:1] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[4:5], v2, v[4:5], v[0:1] -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[4:5], v[0:1] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], v[4:5] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB55_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end ; GFX1250-NEXT: s_endpgm main_body: %ret = atomicrmw fadd ptr addrspace(3) %ptr, double 4.0 seq_cst @@ -2341,23 +2274,9 @@ define double @local_atomic_fadd_f64_rtn_pat(ptr addrspace(3) %ptr, double %data ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_mov_b32_e32 v2, v0 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: .LBB56_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_mov_b64_e32 v[4:5], v[0:1] -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1) -; GFX1250-NEXT: v_add_f64_e32 v[0:1], 4.0, v[4:5] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[0:1], v2, v[0:1], v[4:5] +; GFX1250-NEXT: v_mov_b64_e32 v[2:3], 4.0 +; GFX1250-NEXT: ds_add_rtn_f64 v[0:1], v0, v[2:3] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[0:1], v[4:5] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB56_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end -; GFX1250-NEXT: s_or_b32 exec_lo, exec_lo, s0 ; GFX1250-NEXT: s_set_pc_i64 s[30:31] main_body: %ret = atomicrmw fadd ptr addrspace(3) %ptr, double 4.0 seq_cst, !amdgpu.no.fine.grained.memory !0 @@ -2387,24 +2306,9 @@ define double @local_atomic_fadd_f64_rtn_ieee_unsafe(ptr addrspace(3) %ptr, doub ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v2, v0 -; GFX1250-NEXT: v_mov_b32_e32 v4, v1 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: .LBB57_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 +; GFX1250-NEXT: v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1 +; GFX1250-NEXT: ds_add_rtn_f64 v[0:1], v0, v[2:3] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_mov_b64_e32 v[6:7], v[0:1] -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1) -; GFX1250-NEXT: v_add_f64_e32 v[0:1], v[6:7], v[4:5] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[0:1], v2, v[0:1], v[6:7] -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[0:1], v[6:7] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB57_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end -; GFX1250-NEXT: s_or_b32 exec_lo, exec_lo, s0 ; GFX1250-NEXT: s_set_pc_i64 s[30:31] main_body: %ret = call double @llvm.amdgcn.ds.fadd.f64(ptr addrspace(3) %ptr, double %data, i32 0, i32 0, i1 0) @@ -2434,24 +2338,9 @@ define double @local_atomic_fadd_f64_rtn_ieee_safe(ptr addrspace(3) %ptr, double ; GFX1250: ; %bb.0: ; %main_body ; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v2, v0 -; GFX1250-NEXT: v_mov_b32_e32 v4, v1 -; GFX1250-NEXT: ds_load_b64 v[0:1], v0 -; GFX1250-NEXT: s_mov_b32 s0, 0 -; GFX1250-NEXT: .LBB58_1: ; %atomicrmw.start -; GFX1250-NEXT: ; =>This Inner Loop Header: Depth=1 -; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_mov_b64_e32 v[6:7], v[0:1] -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1) -; GFX1250-NEXT: v_add_f64_e32 v[0:1], v[6:7], v[4:5] -; GFX1250-NEXT: ds_cmpstore_rtn_b64 v[0:1], v2, v[0:1], v[6:7] +; GFX1250-NEXT: v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1 +; GFX1250-NEXT: ds_add_rtn_f64 v[0:1], v0, v[2:3] ; GFX1250-NEXT: s_wait_dscnt 0x0 -; GFX1250-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[0:1], v[6:7] -; GFX1250-NEXT: s_or_b32 s0, vcc_lo, s0 -; GFX1250-NEXT: s_and_not1_b32 exec_lo, exec_lo, s0 -; GFX1250-NEXT: s_cbranch_execnz .LBB58_1 -; GFX1250-NEXT: ; %bb.2: ; %atomicrmw.end -; GFX1250-NEXT: s_or_b32 exec_lo, exec_lo, s0 ; GFX1250-NEXT: s_set_pc_i64 s[30:31] main_body: %ret = call double @llvm.amdgcn.ds.fadd.f64(ptr addrspace(3) %ptr, double %data, i32 0, i32 0, i1 0) |