| Issue |
162832
|
| Summary |
Optimization regression after clang 17
|
| Labels |
clang
|
| Assignees |
|
| Reporter |
alex-s168
|
All clang versions on godbolt after 17.0.1 fail to optimize this lookup table into simple shift instructions
```c
typedef struct {
#define X(a) int x##a : 1;
X(0) X(1) X(2) X(3)
X(4) X(5) X(6) X(7)
X(8) X(9) X(10) X(11)
#undef X
} bits;
int get(bits b, int idx) {
switch(idx){
#define X(a) case a: return b.x##a; break;
X(0) X(1) X(2) X(3)
X(4) X(5) X(6) X(7)
X(8) X(9) X(10) X(11)
#undef X
default: __builtin_unreachable();
}
}
```
with clang 17.0.1 `-target x86_64 -O3`:
```x86asm
get:
mov cl, 15
sub cl, sil
shl edi, cl
movsx eax, di
sar eax, 15
ret
```
but later clang versions don't do this anymore, and instead emit a jump table, with a shift in each case
See: https://godbolt.org/z/c79q7EY4P
(I opened the opt pipeline viewer and noticed that the output of InstCombinerPass is different.
I don't have much knowledge about the internals of llvm tho.
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
