diff options
Diffstat (limited to 'llvm/test/CodeGen/R600/local-atomics.ll')
| -rw-r--r-- | llvm/test/CodeGen/R600/local-atomics.ll | 276 |
1 files changed, 138 insertions, 138 deletions
diff --git a/llvm/test/CodeGen/R600/local-atomics.ll b/llvm/test/CodeGen/R600/local-atomics.ll index 354566a983f..2ac811f26d8 100644 --- a/llvm/test/CodeGen/R600/local-atomics.ll +++ b/llvm/test/CodeGen/R600/local-atomics.ll @@ -4,12 +4,12 @@ ; FUNC-LABEL: {{^}}lds_atomic_xchg_ret_i32: ; EG: LDS_WRXCHG_RET * -; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]], -; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4 -; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]] -; SI: DS_WRXCHG_RTN_B32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] -; SI: BUFFER_STORE_DWORD [[RESULT]], -; SI: S_ENDPGM +; SI: s_load_dword [[SPTR:s[0-9]+]], +; SI: v_mov_b32_e32 [[DATA:v[0-9]+]], 4 +; SI: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[SPTR]] +; SI: ds_wrxchg_rtn_b32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] +; SI: buffer_store_dword [[RESULT]], +; SI: s_endpgm define void @lds_atomic_xchg_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw xchg i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -18,8 +18,8 @@ define void @lds_atomic_xchg_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* % ; FUNC-LABEL: {{^}}lds_atomic_xchg_ret_i32_offset: ; EG: LDS_WRXCHG_RET * -; SI: DS_WRXCHG_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_wrxchg_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_xchg_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw xchg i32 addrspace(3)* %gep, i32 4 seq_cst @@ -30,12 +30,12 @@ define void @lds_atomic_xchg_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspac ; XXX - Is it really necessary to load 4 into VGPR? ; FUNC-LABEL: {{^}}lds_atomic_add_ret_i32: ; EG: LDS_ADD_RET * -; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]], -; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4 -; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]] -; SI: DS_ADD_RTN_U32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] -; SI: BUFFER_STORE_DWORD [[RESULT]], -; SI: S_ENDPGM +; SI: s_load_dword [[SPTR:s[0-9]+]], +; SI: v_mov_b32_e32 [[DATA:v[0-9]+]], 4 +; SI: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[SPTR]] +; SI: ds_add_rtn_u32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] +; SI: buffer_store_dword [[RESULT]], +; SI: s_endpgm define void @lds_atomic_add_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw add i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -44,8 +44,8 @@ define void @lds_atomic_add_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_add_ret_i32_offset: ; EG: LDS_ADD_RET * -; SI: DS_ADD_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_add_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw add i32 addrspace(3)* %gep, i32 4 seq_cst @@ -55,9 +55,9 @@ define void @lds_atomic_add_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_add_ret_i32_bad_si_offset: ; EG: LDS_ADD_RET * -; SI: DS_ADD_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} [M0] -; CI: DS_ADD_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} [M0] +; CI: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_add_ret_i32_bad_si_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr, i32 %a, i32 %b) nounwind { %sub = sub i32 %a, %b %add = add i32 %sub, 4 @@ -69,10 +69,10 @@ define void @lds_atomic_add_ret_i32_bad_si_offset(i32 addrspace(1)* %out, i32 ad ; FUNC-LABEL: {{^}}lds_atomic_inc_ret_i32: ; EG: LDS_ADD_RET * -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_INC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] [M0] -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_inc_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] [M0] +; SI: s_endpgm define void @lds_atomic_inc_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw add i32 addrspace(3)* %ptr, i32 1 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -81,10 +81,10 @@ define void @lds_atomic_inc_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_inc_ret_i32_offset: ; EG: LDS_ADD_RET * -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_INC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] offset:16 -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_inc_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] offset:16 +; SI: s_endpgm define void @lds_atomic_inc_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw add i32 addrspace(3)* %gep, i32 1 seq_cst @@ -94,9 +94,9 @@ define void @lds_atomic_inc_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_inc_ret_i32_bad_si_offset: ; EG: LDS_ADD_RET * -; SI: DS_INC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} [M0] -; CI: DS_INC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_inc_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} [M0] +; CI: ds_inc_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_inc_ret_i32_bad_si_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr, i32 %a, i32 %b) nounwind { %sub = sub i32 %a, %b %add = add i32 %sub, 4 @@ -108,8 +108,8 @@ define void @lds_atomic_inc_ret_i32_bad_si_offset(i32 addrspace(1)* %out, i32 ad ; FUNC-LABEL: {{^}}lds_atomic_sub_ret_i32: ; EG: LDS_SUB_RET * -; SI: DS_SUB_RTN_U32 -; SI: S_ENDPGM +; SI: ds_sub_rtn_u32 +; SI: s_endpgm define void @lds_atomic_sub_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw sub i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -118,8 +118,8 @@ define void @lds_atomic_sub_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_sub_ret_i32_offset: ; EG: LDS_SUB_RET * -; SI: DS_SUB_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_sub_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw sub i32 addrspace(3)* %gep, i32 4 seq_cst @@ -129,10 +129,10 @@ define void @lds_atomic_sub_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_dec_ret_i32: ; EG: LDS_SUB_RET * -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_DEC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] [M0] -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_dec_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] [M0] +; SI: s_endpgm define void @lds_atomic_dec_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw sub i32 addrspace(3)* %ptr, i32 1 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -141,10 +141,10 @@ define void @lds_atomic_dec_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_dec_ret_i32_offset: ; EG: LDS_SUB_RET * -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_DEC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] offset:16 -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_dec_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, [[NEGONE]] offset:16 +; SI: s_endpgm define void @lds_atomic_dec_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw sub i32 addrspace(3)* %gep, i32 1 seq_cst @@ -154,8 +154,8 @@ define void @lds_atomic_dec_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_and_ret_i32: ; EG: LDS_AND_RET * -; SI: DS_AND_RTN_B32 -; SI: S_ENDPGM +; SI: ds_and_rtn_b32 +; SI: s_endpgm define void @lds_atomic_and_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw and i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -164,8 +164,8 @@ define void @lds_atomic_and_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_and_ret_i32_offset: ; EG: LDS_AND_RET * -; SI: DS_AND_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_and_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_and_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw and i32 addrspace(3)* %gep, i32 4 seq_cst @@ -175,8 +175,8 @@ define void @lds_atomic_and_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_or_ret_i32: ; EG: LDS_OR_RET * -; SI: DS_OR_RTN_B32 -; SI: S_ENDPGM +; SI: ds_or_rtn_b32 +; SI: s_endpgm define void @lds_atomic_or_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw or i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -185,8 +185,8 @@ define void @lds_atomic_or_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %pt ; FUNC-LABEL: {{^}}lds_atomic_or_ret_i32_offset: ; EG: LDS_OR_RET * -; SI: DS_OR_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_or_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_or_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw or i32 addrspace(3)* %gep, i32 4 seq_cst @@ -196,8 +196,8 @@ define void @lds_atomic_or_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace( ; FUNC-LABEL: {{^}}lds_atomic_xor_ret_i32: ; EG: LDS_XOR_RET * -; SI: DS_XOR_RTN_B32 -; SI: S_ENDPGM +; SI: ds_xor_rtn_b32 +; SI: s_endpgm define void @lds_atomic_xor_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw xor i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -206,8 +206,8 @@ define void @lds_atomic_xor_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_xor_ret_i32_offset: ; EG: LDS_XOR_RET * -; SI: DS_XOR_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_xor_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_xor_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw xor i32 addrspace(3)* %gep, i32 4 seq_cst @@ -225,8 +225,8 @@ define void @lds_atomic_xor_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_min_ret_i32: ; EG: LDS_MIN_INT_RET * -; SI: DS_MIN_RTN_I32 -; SI: S_ENDPGM +; SI: ds_min_rtn_i32 +; SI: s_endpgm define void @lds_atomic_min_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw min i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -235,8 +235,8 @@ define void @lds_atomic_min_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_min_ret_i32_offset: ; EG: LDS_MIN_INT_RET * -; SI: DS_MIN_RTN_I32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_min_rtn_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_min_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw min i32 addrspace(3)* %gep, i32 4 seq_cst @@ -246,8 +246,8 @@ define void @lds_atomic_min_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_max_ret_i32: ; EG: LDS_MAX_INT_RET * -; SI: DS_MAX_RTN_I32 -; SI: S_ENDPGM +; SI: ds_max_rtn_i32 +; SI: s_endpgm define void @lds_atomic_max_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw max i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -256,8 +256,8 @@ define void @lds_atomic_max_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %p ; FUNC-LABEL: {{^}}lds_atomic_max_ret_i32_offset: ; EG: LDS_MAX_INT_RET * -; SI: DS_MAX_RTN_I32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_max_rtn_i32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_max_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw max i32 addrspace(3)* %gep, i32 4 seq_cst @@ -267,8 +267,8 @@ define void @lds_atomic_max_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace ; FUNC-LABEL: {{^}}lds_atomic_umin_ret_i32: ; EG: LDS_MIN_UINT_RET * -; SI: DS_MIN_RTN_U32 -; SI: S_ENDPGM +; SI: ds_min_rtn_u32 +; SI: s_endpgm define void @lds_atomic_umin_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw umin i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -277,8 +277,8 @@ define void @lds_atomic_umin_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* % ; FUNC-LABEL: {{^}}lds_atomic_umin_ret_i32_offset: ; EG: LDS_MIN_UINT_RET * -; SI: DS_MIN_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_min_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_umin_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw umin i32 addrspace(3)* %gep, i32 4 seq_cst @@ -288,8 +288,8 @@ define void @lds_atomic_umin_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspac ; FUNC-LABEL: {{^}}lds_atomic_umax_ret_i32: ; EG: LDS_MAX_UINT_RET * -; SI: DS_MAX_RTN_U32 -; SI: S_ENDPGM +; SI: ds_max_rtn_u32 +; SI: s_endpgm define void @lds_atomic_umax_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw umax i32 addrspace(3)* %ptr, i32 4 seq_cst store i32 %result, i32 addrspace(1)* %out, align 4 @@ -298,8 +298,8 @@ define void @lds_atomic_umax_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* % ; FUNC-LABEL: {{^}}lds_atomic_umax_ret_i32_offset: ; EG: LDS_MAX_UINT_RET * -; SI: DS_MAX_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_max_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_umax_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw umax i32 addrspace(3)* %gep, i32 4 seq_cst @@ -308,19 +308,19 @@ define void @lds_atomic_umax_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspac } ; FUNC-LABEL: {{^}}lds_atomic_xchg_noret_i32: -; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]], -; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4 -; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]] -; SI: DS_WRXCHG_RTN_B32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] -; SI: S_ENDPGM +; SI: s_load_dword [[SPTR:s[0-9]+]], +; SI: v_mov_b32_e32 [[DATA:v[0-9]+]], 4 +; SI: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[SPTR]] +; SI: ds_wrxchg_rtn_b32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]] [M0] +; SI: s_endpgm define void @lds_atomic_xchg_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw xchg i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_xchg_noret_i32_offset: -; SI: DS_WRXCHG_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_wrxchg_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_xchg_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw xchg i32 addrspace(3)* %gep, i32 4 seq_cst @@ -329,19 +329,19 @@ define void @lds_atomic_xchg_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { ; XXX - Is it really necessary to load 4 into VGPR? ; FUNC-LABEL: {{^}}lds_atomic_add_noret_i32: -; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]], -; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4 -; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]] -; SI: DS_ADD_U32 [[VPTR]], [[DATA]] [M0] -; SI: S_ENDPGM +; SI: s_load_dword [[SPTR:s[0-9]+]], +; SI: v_mov_b32_e32 [[DATA:v[0-9]+]], 4 +; SI: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[SPTR]] +; SI: ds_add_u32 [[VPTR]], [[DATA]] [M0] +; SI: s_endpgm define void @lds_atomic_add_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw add i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_add_noret_i32_offset: -; SI: DS_ADD_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_add_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_add_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw add i32 addrspace(3)* %gep, i32 4 seq_cst @@ -349,9 +349,9 @@ define void @lds_atomic_add_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_add_noret_i32_bad_si_offset -; SI: DS_ADD_U32 v{{[0-9]+}}, v{{[0-9]+}} [M0] -; CI: DS_ADD_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 [M0] -; SI: S_ENDPGM +; SI: ds_add_u32 v{{[0-9]+}}, v{{[0-9]+}} [M0] +; CI: ds_add_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 [M0] +; SI: s_endpgm define void @lds_atomic_add_noret_i32_bad_si_offset(i32 addrspace(3)* %ptr, i32 %a, i32 %b) nounwind { %sub = sub i32 %a, %b %add = add i32 %sub, 4 @@ -361,20 +361,20 @@ define void @lds_atomic_add_noret_i32_bad_si_offset(i32 addrspace(3)* %ptr, i32 } ; FUNC-LABEL: {{^}}lds_atomic_inc_noret_i32: -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_INC_U32 v{{[0-9]+}}, [[NEGONE]] [M0] -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_inc_u32 v{{[0-9]+}}, [[NEGONE]] [M0] +; SI: s_endpgm define void @lds_atomic_inc_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw add i32 addrspace(3)* %ptr, i32 1 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_inc_noret_i32_offset: -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_INC_U32 v{{[0-9]+}}, [[NEGONE]] offset:16 -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_inc_u32 v{{[0-9]+}}, [[NEGONE]] offset:16 +; SI: s_endpgm define void @lds_atomic_inc_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw add i32 addrspace(3)* %gep, i32 1 seq_cst @@ -382,9 +382,9 @@ define void @lds_atomic_inc_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_inc_noret_i32_bad_si_offset: -; SI: DS_INC_U32 v{{[0-9]+}}, v{{[0-9]+}} -; CI: DS_INC_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_inc_u32 v{{[0-9]+}}, v{{[0-9]+}} +; CI: ds_inc_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_inc_noret_i32_bad_si_offset(i32 addrspace(3)* %ptr, i32 %a, i32 %b) nounwind { %sub = sub i32 %a, %b %add = add i32 %sub, 4 @@ -394,16 +394,16 @@ define void @lds_atomic_inc_noret_i32_bad_si_offset(i32 addrspace(3)* %ptr, i32 } ; FUNC-LABEL: {{^}}lds_atomic_sub_noret_i32: -; SI: DS_SUB_U32 -; SI: S_ENDPGM +; SI: ds_sub_u32 +; SI: s_endpgm define void @lds_atomic_sub_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw sub i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_sub_noret_i32_offset: -; SI: DS_SUB_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_sub_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_sub_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw sub i32 addrspace(3)* %gep, i32 4 seq_cst @@ -411,20 +411,20 @@ define void @lds_atomic_sub_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_dec_noret_i32: -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_DEC_U32 v{{[0-9]+}}, [[NEGONE]] -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_dec_u32 v{{[0-9]+}}, [[NEGONE]] +; SI: s_endpgm define void @lds_atomic_dec_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw sub i32 addrspace(3)* %ptr, i32 1 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_dec_noret_i32_offset: -; SI: S_MOV_B32 [[SNEGONE:s[0-9]+]], -1 -; SI: V_MOV_B32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] -; SI: DS_DEC_U32 v{{[0-9]+}}, [[NEGONE]] offset:16 -; SI: S_ENDPGM +; SI: s_mov_b32 [[SNEGONE:s[0-9]+]], -1 +; SI: v_mov_b32_e32 [[NEGONE:v[0-9]+]], [[SNEGONE]] +; SI: ds_dec_u32 v{{[0-9]+}}, [[NEGONE]] offset:16 +; SI: s_endpgm define void @lds_atomic_dec_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw sub i32 addrspace(3)* %gep, i32 1 seq_cst @@ -432,16 +432,16 @@ define void @lds_atomic_dec_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_and_noret_i32: -; SI: DS_AND_B32 -; SI: S_ENDPGM +; SI: ds_and_b32 +; SI: s_endpgm define void @lds_atomic_and_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw and i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_and_noret_i32_offset: -; SI: DS_AND_B32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_and_b32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_and_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw and i32 addrspace(3)* %gep, i32 4 seq_cst @@ -449,16 +449,16 @@ define void @lds_atomic_and_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_or_noret_i32: -; SI: DS_OR_B32 -; SI: S_ENDPGM +; SI: ds_or_b32 +; SI: s_endpgm define void @lds_atomic_or_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw or i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_or_noret_i32_offset: -; SI: DS_OR_B32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_or_b32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_or_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw or i32 addrspace(3)* %gep, i32 4 seq_cst @@ -466,16 +466,16 @@ define void @lds_atomic_or_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_xor_noret_i32: -; SI: DS_XOR_B32 -; SI: S_ENDPGM +; SI: ds_xor_b32 +; SI: s_endpgm define void @lds_atomic_xor_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw xor i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_xor_noret_i32_offset: -; SI: DS_XOR_B32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_xor_b32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_xor_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw xor i32 addrspace(3)* %gep, i32 4 seq_cst @@ -490,16 +490,16 @@ define void @lds_atomic_xor_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { ; } ; FUNC-LABEL: {{^}}lds_atomic_min_noret_i32: -; SI: DS_MIN_I32 -; SI: S_ENDPGM +; SI: ds_min_i32 +; SI: s_endpgm define void @lds_atomic_min_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw min i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_min_noret_i32_offset: -; SI: DS_MIN_I32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_min_i32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_min_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw min i32 addrspace(3)* %gep, i32 4 seq_cst @@ -507,16 +507,16 @@ define void @lds_atomic_min_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_max_noret_i32: -; SI: DS_MAX_I32 -; SI: S_ENDPGM +; SI: ds_max_i32 +; SI: s_endpgm define void @lds_atomic_max_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw max i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_max_noret_i32_offset: -; SI: DS_MAX_I32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_max_i32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_max_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw max i32 addrspace(3)* %gep, i32 4 seq_cst @@ -524,16 +524,16 @@ define void @lds_atomic_max_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_umin_noret_i32: -; SI: DS_MIN_U32 -; SI: S_ENDPGM +; SI: ds_min_u32 +; SI: s_endpgm define void @lds_atomic_umin_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw umin i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_umin_noret_i32_offset: -; SI: DS_MIN_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_min_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_umin_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw umin i32 addrspace(3)* %gep, i32 4 seq_cst @@ -541,16 +541,16 @@ define void @lds_atomic_umin_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { } ; FUNC-LABEL: {{^}}lds_atomic_umax_noret_i32: -; SI: DS_MAX_U32 -; SI: S_ENDPGM +; SI: ds_max_u32 +; SI: s_endpgm define void @lds_atomic_umax_noret_i32(i32 addrspace(3)* %ptr) nounwind { %result = atomicrmw umax i32 addrspace(3)* %ptr, i32 4 seq_cst ret void } ; FUNC-LABEL: {{^}}lds_atomic_umax_noret_i32_offset: -; SI: DS_MAX_U32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 -; SI: S_ENDPGM +; SI: ds_max_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16 +; SI: s_endpgm define void @lds_atomic_umax_noret_i32_offset(i32 addrspace(3)* %ptr) nounwind { %gep = getelementptr i32 addrspace(3)* %ptr, i32 4 %result = atomicrmw umax i32 addrspace(3)* %gep, i32 4 seq_cst |
