tree 2a052a9c109c1b4025d549a87671a92fc13fd5bc
parent bf84e445b846f58295b0e903383ae2c82e642707
author Lei Zhang <antiagainst@google.com> 1605904035 -0500
committer GitHub <noreply@github.com> 1605904035 -0500
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJfuCaiCRBK7hj4Ov3rIwAAdHIIAGYZr0yjfwjCi09iA4GzCYKv
 WcJ204oTFySXAlauH1G4OLsSkG2n6vqsCF4VwNWPnrG6kIrU/DiJDromcLhJ0BLG
 mugwA08Xr1OEWc0AG427X8cG6hqcK0Ykc7pyBqChUoAtDX3gqQFXqXc/tl2IITjq
 NnWmIewxTLqgH98ALQ5kjR0QhZoN2/T6rw/jdKNvlPxlApnCMk5uLElwy7Bn3Pm6
 ytZxoPljMSYpuhumvg8wTSRT+4jGnOrvQe7aulJ2n199kqc+srB105wsyoi1eZ1j
 CpmQT7t0jFqOMZ7d4DTOSJ2cEma0WthmAmCypJUHUpk0edewzZsslFeDHq2geZU=
 =5YdI
 -----END PGP SIGNATURE-----
 

[spirv] Add patterns for folding processor ID uses (#3877)

For processor IDs we have their upper bound in the workgroup
count function or SPIR-V entry point ABI, which can be used
to fold processor ID uses where possible.

This commit add a pattern to fold affine.min ops using processor
IDs. It allows us to get static constant value, which can then
be used for memref subviews.