On 4/4/25 3:21 AM, Richard Sandiford wrote:
One of the problems in PR101523 was that, after each successful 2->2 combination attempt, try_combine would restart combination attempts at i2 even if i2 hadn't changed. This led to quadratic behaviour as the same failed combinations between i2 and i3 were tried repeatedly. The original patch for the PR dealt with that by disallowing 2->2 combinations. However, that led to various optimisation regressions, so there was interest in allowing the combinations again, at least until an alternative way of getting the same results is in place. This patch is a variant of Richi's in: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101523#c53 but limited to when we're combining 2 instructions. This speeds up combine by 10x on the original PR101523 testcase and reduces combine's memory footprint by 100x. gcc/ PR testsuite/116398 * combine.cc (try_combine): Reallow 2->2 combinations. Detect when only i3 has changed and restart from i3 in that case. gcc/testsuite/ * gcc.target/aarch64/popcnt-le-1.c: Account for commutativity of TST. * gcc.target/aarch64/popcnt-le-3.c: Likewise AND. * gcc.target/aarch64/sve/pred-not-gen-1.c: Revert previous patch. * gcc.target/aarch64/sve/pred-not-gen-4.c: Likewise. * gcc.target/aarch64/sve/var_stride_2.c: Likewise. * gcc.target/aarch64/sve/var_stride_4.c: Likewise.
Looks quite safe, even at this stage in the game. I'll ACK as well. jeff
