On 7/22/25 4:19 AM, Konstantinos Eleftheriou wrote:
While scanning the instructions and upon reaching an instruction that
doesn't satisfy the constraints that we have set, we were removing the
already detected stores, but we were continuing adding stores from that
point onward. This was causing issues when the address ranges from later
stores overlapped with the load's address, leading to partial and wrong
update of the register containing the loaded value.
With this patch, we are skipping the tranformation for stores that operate
on the load's address range, when stores that operate on the same range
have been deleted due to constraint violations.
PR rtl-optimization/119795
gcc/ChangeLog:
* avoid-store-forwarding.cc
(store_forwarding_analyzer::avoid_store_forwarding): Skip
transformations for stores that operate on the same address
range as deleted ones.
gcc/testsuite/ChangeLog:
+ /* Check if another store in the load's address range has
+ been deleted due to a constraint violation. In this case
+ we can't forward any other stores that operate in this
+ range, as it would lead to partial update of the register
+ that holds the loaded value. */
+ FOR_EACH_VEC_ELT (store_exprs_del, j, del_it)
+ {
+ rtx del_store_mem = *del_it;
+ same_range_as_removed
+ = is_store_forwarding (del_store_mem, load_mem, NULL);
+ break;
+ }
Doesn't this break the loop after the first element every time? What am
I missing here?
Jeff
