@@ -30,12 +30,10 @@ define protected amdgpu_kernel void @func_int8(ptr addrspace(1) %p_a_grid.coerce
30
30
; CHECK-NEXT: [[TMP1:%.*]] = add nsw i32 [[INDEX]], [[MUL17]]
31
31
; CHECK-NEXT: [[TMP2:%.*]] = sext i32 [[TMP0]] to i64
32
32
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[P_A_GRID_COERCE]], i64 [[TMP2]]
33
- ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP3]], i32 0
34
- ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr addrspace(1) [[TMP4]], align 1
33
+ ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr addrspace(1) [[TMP3]], align 1
35
34
; CHECK-NEXT: [[TMP5:%.*]] = sext i32 [[TMP1]] to i64
36
35
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[P_B_GRID_COERCE]], i64 [[TMP5]]
37
- ; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP6]], i32 0
38
- ; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x i8>, ptr addrspace(1) [[TMP7]], align 1
36
+ ; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x i8>, ptr addrspace(1) [[TMP6]], align 1
39
37
; CHECK-NEXT: [[TMP8:%.*]] = zext <2 x i8> [[WIDE_LOAD]] to <2 x i32>
40
38
; CHECK-NEXT: [[TMP9:%.*]] = zext <2 x i8> [[WIDE_LOAD1]] to <2 x i32>
41
39
; CHECK-NEXT: [[TMP10:%.*]] = mul nuw nsw <2 x i32> [[TMP9]], [[TMP8]]
0 commit comments