junbuml added a subscriber: junbuml.
junbuml added a comment.
It appears that build fails due to this change :
llvm/tools/clang/lib/Sema/CodeCompleteConsumer.cpp:448:3: error: default label
in switch which covers all enumeration values [-Werror,-Wcovered-switch-default]
default: llvm_unreacha
junbuml added a subscriber: junbuml.
junbuml added a comment.
In current trunk, I got build error :
llvm/tools/clang/lib/Basic/Targets.cpp:2090:9: error:
'setSupportedOpenCLOpts' overrides a member function but is not marked
'override' [-Werror,-Winconsistent-missing-override]
void setSupp
junbuml abandoned this revision.
junbuml added a comment.
Abandon this based the last comment in http://reviews.llvm.org/D15289.
http://reviews.llvm.org/D13304
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mai
junbuml added a comment.
Just ping to see if there is any objection about adding the extra check for
CallSites in EHRs in inliner. I will be happy to hear any opinion, suggestion,
or objection.
http://reviews.llvm.org/D13304
___
cfe-commits maili
junbuml added a comment.
If we want to add a check for CallSites in EHRs in inliner, we may be able to
borrow things done in BranchProbabilityInfo::calcColdCallHeuristics, but for
exception handing intrinsics, not for cold, and make getInlineThreshold()
return a lower threshold so that we can
junbuml added a comment.
Thanks Richard for your comment !
If the frond-end is not a good to place for this, I think there are two places
we can consider : inliner or prune-eh.
1. In inliner, we can directly check if a CallSite branches an exception
region, and then make getInlineThreshold() r
junbuml added a comment.
The basic idea of this change is to avoid inlining callsites invoked in
exception handling regions (EHR) so that we can reduce code size blow-up in
very cold regions and indirectly increase inline opportunities for functions
containing exception handling code.
I think
junbuml added a comment.
I see what you mean. Now, I doubt if
BranchProbabilityInfo::calcColdCallHeuristics() set the Cold before inliner.
As far as I check, BranchProbabilityInfo is executed after inliner.
Another issue is that even if we can add the Cold in callsites in exception
handling
junbuml added a comment.
Did you mean to add the Cold in the exception handling region itself instead of
callsite in throw statements ?
http://reviews.llvm.org/D13304
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi
junbuml added a comment.
I just want to ping one more time to see if there is any objection about the
basic idea of this change. If the basic idea itself is acceptable, then I want
to find the best way to get idea in.
Please let me know any new suggestion or any opinion about moving this
imple
junbuml added a comment.
Just want to hear if there is any opinion about moving this implementation back
to PrunceEH.cpp like http://reviews.llvm.org/D12979 and adding the minimum
callee size check to avoid marking noinlines on trivial constrictor calls.
Please let me know any objection or sugg
junbuml added a comment.
Thanks Richard for the comment.
Initially, I intended to implement this in inliner by checking if a callsite is
in exception handling regions. However, I decided not to implement this in
inliner because this kind of check should be performed for all callsites if we
imp
junbuml added a comment.
Is there any comment on this change? Note that with change, I observed about 6%
performance improvement in spec2006/xalancbmk. I will address any comment to
get this in.
http://reviews.llvm.org/D13304
___
cfe-commits maili
junbuml updated this revision to Diff 36551.
junbuml added a comment.
Just minor cleaning. Please let me know any comment.
http://reviews.llvm.org/D13304
Files:
lib/CodeGen/CGCall.cpp
lib/CodeGen/CGException.cpp
lib/CodeGen/CodeGenFunction.cpp
lib/CodeGen/CodeGenFunction.h
test/CodeGe
junbuml abandoned this revision.
junbuml added a comment.
Move this clang change in http://reviews.llvm.org/D13304
http://reviews.llvm.org/D12979
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listin
junbuml created this revision.
junbuml added reviewers: hfinkel, mcrosier, reames, ashutosh.nema, majnemer.
junbuml added subscribers: gberry, cfe-commits.
It might be reasonably to avoid inlining CallSites invoked in
exception handling context so that we can reduce code size blow-up in
EH regions
junbuml added a comment.
In this clang change, I added a state flag (bool IsColdRegion) in
CodeGenFunction and set/reset the flag in EmitCXXThrowExpr(). In EmitCall(),
the NoInline attribute would be added if IsColdRegion is true. As of now, this
change only handles throw statements because I d
junbuml edited subscribers, added: cfe-commits; removed: llvm-commits.
junbuml updated this revision to Diff 36107.
http://reviews.llvm.org/D12979
Files:
lib/CodeGen/CGCall.cpp
lib/CodeGen/CGException.cpp
lib/CodeGen/CodeGenFunction.cpp
lib/CodeGen/CodeGenFunction.h
test/CodeGenCXX/thro
18 matches
Mail list logo
