radv: skip shaders cache for fast-linked pipelines with GPL
In addition to !21008 (merged), this makes GPL fast-linking as fast as it should be.
Next big optimization is to stop compiling a noop FS every time, MR coming shortly...
In addition to !21008 (merged), this makes GPL fast-linking as fast as it should be.
Next big optimization is to stop compiling a noop FS every time, MR coming shortly...