3 Commits

Author SHA1 Message Date
Adrian Kuegel
da9559d69a
Do not use PerformEXTRACTCombine for v8i8 types (#81242)
Same as with v4i8 types, we should not be using PerformEXTRACTCombine
for v8i8 types.
2024-02-12 07:31:31 +01:00
Artem Belevich
cbafb6f2f5
[NVPTX] Improve lowering of v4i8 (#67866)
Make v4i8 a legal type and plumb through lowering of relevant instructions.
2023-10-09 12:33:22 -07:00
Artem Belevich
d06b3e3b6a
[NVPTX] improve lowering for common byte-extraction operations. (#66945)
Some critical code paths we have depend on efficient byte extraction
from data loaded as integers. By default LLVM tries to extract bytes by 
storing/loading from stack, which is very inefficient on GPU.
2023-09-21 13:48:54 -07:00