| Issue |
142961
|
| Summary |
[aarch64] gcc get better result than llvm for abs
|
| Labels |
NEON
|
| Assignees |
|
| Reporter |
vfdff
|
* test: https://godbolt.org/z/cTcjvPE1z
```
#include <stdlib.h>
int square (unsigned char *pix1, unsigned char *pix2) {
int sum=0;
for (int i=0; i< 16; i++)
sum += abs (pix1[i]-pix2[i]);
return sum;
}
```
* gcc: uabdl2 + uabal + uaddlv
```
square:
ldr q31, [x0]
ldr q30, [x1]
uabdl2 v29.8h, v31.16b, v30.16b
uabal v29.8h, v31.8b, v30.8b
uaddlv s31, v29.8h
fmov w0, s31
ret
```
* llvm:
```
square:
ldr q0, [x0]
ldr q1, [x1]
uabd v0.16b, v0.16b, v1.16b
ushll2 v1.8h, v0.16b, #0
ushll v0.8h, v0.8b, #0
uaddl2 v2.4s, v0.8h, v1.8h
uaddl v0.4s, v0.4h, v1.4h
add v0.4s, v0.4s, v2.4s
addv s0, v0.4s
fmov w0, s0
ret
```
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
