Skip to content

nir/load_store_vectorize: improve vectorization with identical operations

Rhys Perry requested to merge pendingchaos/mesa:nir_mem_vectorize_identical into master

We might have several identical options to vectorize an entry with, but only one might be vectorizable because of writes interfering.

An example of this is a pattern found in some CTS tests:

a = load(0)
b = load(4)
store(0, a)
store(4, b)
a = load(0)
b = load(4)
store(0, a)
store(4, b)
...

It might have attempted to vectorize the first load(0) with the second load(4) without attempting the second load(4) when the first fails. This changes vectorize_entries() to continue even if the first try_vectorize() failed.

fossil-db (Navi):

Totals from 130 (0.10% of 128321) affected shaders:
SGPRs: 8288 -> 8344 (+0.68%); split: -0.10%, +0.77%
VGPRs: 4940 -> 4932 (-0.16%)
CodeSize: 384228 -> 384160 (-0.02%); split: -0.03%, +0.01%
MaxWaves: 1800 -> 1804 (+0.22%)
Instrs: 72299 -> 72287 (-0.02%); split: -0.03%, +0.02%

Merge request reports