On x86-64, I got FAIL: gcc.target/i386/pr88531-2b.c scan-assembler-times vmulps 1 FAIL: gcc.target/i386/pr88531-2c.c scan-assembler-times vmulps 1
I see it too.
https://gcc.gnu.org/pipermail/gcc-patches/2024-September/662523.html
/home/rguenther/src/gcc/gcc/testsuite/gcc.target/i386/pr88531-2a.c:15:28: optimized: loop vectorized using 32 byte vectors /home/rguenther/src/gcc/gcc/testsuite/gcc.target/i386/pr88531-2a.c:15:28: optimized: loop vectorized using 16 byte vectors compared to /home/rguenther/src/gcc-14-branch/gcc/testsuite/gcc.target/i386/pr88531-2a.c:15:28: optimized: loop vectorized using 8 byte vectors this is the known issue around the x86 cost model not being able to see STMT_VINFO_MEMORY_ACCESS_TYPE when using SLP. See https://gcc.gnu.org/pipermail/gcc-patches/2024-June/653814.html I wasn't particularly happy with that transitional change but I guess I'll pick it up now (after re-testing of course).
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>: https://gcc.gnu.org/g:9f74d79afbf54d0e47905de755f41d95e46f178f commit r15-4263-g9f74d79afbf54d0e47905de755f41d95e46f178f Author: Richard Biener <rguenther@suse.de> Date: Thu Jun 6 15:52:02 2024 +0200 tree-optimization/117080 - Add SLP_TREE_MEMORY_ACCESS_TYPE It turns out target costing code looks at STMT_VINFO_MEMORY_ACCESS_TYPE to identify operations from (emulated) gathers for example. This doesn't work for SLP loads since we do not set STMT_VINFO_MEMORY_ACCESS_TYPE there as the vectorization strathegy might differ between different stmt uses. It seems we got away with setting it for stores though. The following adds a memory_access_type field to slp_tree and sets it from load and store vectorization code. All the costing doesn't record the SLP node (that was only done selectively for some corner case). The costing is really in need of a big overhaul, the following just massages the two relevant ops to fix gcc.dg/target/pr88531-2[bc].c FAILs when switching on SLP for non-grouped stores. In particular currently we either have a SLP node or a stmt_info in the cost hook but not both. So the following mitigates this, postponing a rewrite of costing to next stage1. Other targets look possibly affected as well but are left to respective maintainers to update. PR tree-optimization/117080 * tree-vectorizer.h (_slp_tree::memory_access_type): Add. (SLP_TREE_MEMORY_ACCESS_TYPE): New. (record_stmt_cost): Add another overload. * tree-vect-slp.cc (_slp_tree::_slp_tree): Initialize memory_access_type. * tree-vect-stmts.cc (vectorizable_store): Set SLP_TREE_MEMORY_ACCESS_TYPE. (vectorizable_load): Likewise. Also record the SLP node when costing emulated gather offset decompose and vector composition. * config/i386/i386.cc (ix86_vector_costs::add_stmt_cost): Also recognize SLP emulated gather/scatter.
Fixed.