This enables load merging into x2, x4, which is driven by inline offsets.
6500 shaders are affected:
Code Size in affected shaders: -15.14 %
Differential D42078
AMDGPU: Fold inline offset for loads properly in moveToVALU on GFX9 mareko on Jan 15 2018, 9:41 AM. Authored by
Details
This enables load merging into x2, x4, which is driven by inline offsets. 6500 shaders are affected:
Diff Detail
Event Timeline
|
It's a defect that at this point the e32 version is ever encountered. We shouldn't be emitting VCC uses up front. Where is this coming from? I thought I fixed most of these places already