Enable using upstream jitify2 #11287
Conversation
github-actions
bot
added
CMake
This reverts commit 812eea5.
shwina
added
Python
Codecov Report
@@ Coverage Diff @@
## branch-22.10 #11287 +/- ##
===============================================
Coverage ? 86.40%
===============================================
Files ? 145
Lines ? 22959
Branches ? 0
===============================================
Hits ? 19838
Misses ? 3121
Partials ? 0 Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. |
| GIT_REPOSITORY https://github.com/rapidsai/jitify.git | ||
| GIT_TAG cudf_0.19 | ||
| GIT_REPOSITORY https://github.com/NVIDIA/jitify.git | ||
| GIT_TAG jitify2 |
There was a problem hiding this comment.
Should we ask the Jitify team to make a tag so that we're not pinned to a "floating" branch?
There was a problem hiding this comment.
Yes, I think that'd be great
| @@ -22,6 +22,7 @@ | |||
|
|
|||
| // Note: The <cuda/std/*> versions are used in order for Jitify to work with our fixed_point type. | |||
There was a problem hiding this comment.
Should this comment be copied to the other files where these includes occur?
There was a problem hiding this comment.
I don't know -- this comment precedes this PR.
Co-authored-by: Robert Maynard <[email protected]>
|
@gpucibot merge |
This PR enables using upstream jitify2 rather than RAPIDS' fork of jitify2.
This enables us to take advantage of the latest additions/improvements to jitify. Most notably: upstream jitify2 dlsym/dlopens
libcuda.sowhich enables us to drop our shared library dependency onlibcuda.so.Two major issues came up when making the switch:
<limits>and<cuda/std/limits>in the same source file NVIDIA/jitify#107 - I used the workaround mentioned in that issue. Hopefully it is fixed soon and we can eliminate the workaround.-D_FILE_OFFSET_BITS=64to jitify. Due to limitations in the way conda-forge builds glibc, we must explicitly state we require 64bit file offset support.