GCN: Workaround Julia's broken alloca handling #47
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jpsamaroo
added
bug
Codecov Report
@@ Coverage Diff @@
## master #47 +/- ##
==========================================
+ Coverage 58.91% 65.73% +6.81%
==========================================
Files 30 25 -5
Lines 1570 1430 -140
==========================================
+ Hits 925 940 +15
+ Misses 645 490 -155
Continue to review full report at Codecov.
|
Member
Yeah I'd rather not have this globally. |
jpsamaroo
force-pushed
the
jps/gcn-workaround-allocas
branch
from
8bb863a to
c44dac6
Compare
jpsamaroo
force-pushed
the
jps/gcn-workaround-allocas
branch
from
c44dac6 to
440e982
Compare
Member
Author
|
Done right in #71 |
This was referenced Aug 21, 2020
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR works around the fact that allocas in an AS other than 0 (AS 5 in the case of GCN) appears to break Julia's
llvm-propagate-addrspacespass. That pass attempts to perform a bitcast between two types in different addrspaces, which is apparently incorrect. I have a closed PR which was attempting to fix this, which I plan to reopen once I get it working properly, and have found solid justification for why it's needed.@maleadt I know this looks ugly, but I'd like to have some way to work around this pass until I've gotten the pass fixed in Julia. I plan to make this conditional on a
CompilerJob{GCNCompilerTarget}, unless you somehow want it for all backends.