]> git.ipfire.org Git - thirdparty/gcc.git/commit
projects / thirdparty / gcc.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e55de74) | patch
aarch64: Adjust SVE vector costs for NVIDIA Olympus
authorKyrylo Tkachov <ktkachov@nvidia.com>
Fri, 6 Feb 2026 10:34:29 +0000 (02:34 -0800)
committerKyrylo Tkachov <ktkachov@nvidia.com>
Tue, 10 Feb 2026 10:02:10 +0000 (11:02 +0100)
commit4d2af0735063ab7d38c4da0ec6f992276af855af
tree6afcd055596f770f6e220d9e1eb5e1a2d56ac879tree
parente55de74d50d236db0472431fe242f3244e41b884commit | diff
aarch64: Adjust SVE vector costs for NVIDIA Olympus

After some more analysis, I'd like to adjust the SVE load
cost for NVIDIA Olympus to more accurately reflect the core SWOG.
This leads to more sensible Advanced SIMD vs SVE autovec decisions.

Bootstrapped and tested on aarch64-none-linux-gnu.

Signed-off-by: Kyrylo Tkachov <ktkachov@nvidia.com>
gcc/ChangeLog

* config/aarch64/tuning_models/olympus.h (olympus_sve_vector_cost):
Set align_load_cost to 6.
gcc/config/aarch64/tuning_models/olympus.h diff | blob | blame | history
Mirror of https://gcc.gnu.org/git/gcc.git