rj-jesus wrote:
Hi @paulwalker-arm, I think the alignment requirements of LD1 and LDR are
indeed different, but this only matters if `AlignmentEnforced()` is enabled,
right? I thought `AlignmentEnforced` wasn't generally a concern, otherwise even
the current lowering we have for `vld1q_u8(uint8_t const *ptr)`, for example,
seems too permissive (https://godbolt.org/z/coYefno3j):
```cpp
#include <arm_neon.h>
uint8x16_t foo(uint8_t *ptr) {
return vld1q_u8(ptr);
}
```
Currently gets lowered to:
```llvm
define <16 x i8> @foo(ptr %0) {
%2 = load <16 x i8>, ptr %0, align 1
ret <16 x i8> %2
}
```
Which finally lowers to:
```gas
foo:
ldr q0, [x0]
ret
```
`ptr` isn't necessarily aligned to 16 (in the IR, it's only guaranteed to be
aligned to 1), but, unless I'm missing something in the docs, LDR.Q also seems
to expect an alignment of 16 if `AlignmentEnforced` is enabled, and will fault
if not.
Am I missing anything?
Also, even if we can't indeed lower LD1/ST1 to LDR/STR generally, do you think
it would be worth trying to do it in some other more restricted way (for
example only for SP, which I believe should be aligned to 16), or should we
drop the idea entirely?
https://github.com/llvm/llvm-project/pull/127837
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
