Page MenuHomePhabricator
Feed Advanced Search

Fri, Nov 15

tejohnson accepted D70195: Fix error message missed in commit dde589389fcb8b5098f7a47f1b781b27d29a0cac..

lgtm, thanks

Fri, Nov 15, 8:28 PM · Restricted Project
tejohnson created D70337: [ThinLTO] Avoid extra index lookup during promotion.
Fri, Nov 15, 1:12 PM · Restricted Project
tejohnson accepted D70323: ThinLTO: Fix assembler to emit alwaysInline in the summary.

lgtm

Fri, Nov 15, 12:01 PM · Restricted Project
tejohnson created D70327: [ThinLTO] Promotion handling cleanup (NFC).
Fri, Nov 15, 10:59 AM · Restricted Project

Wed, Nov 13

tejohnson added a comment to D70153: Add support for multi-module bitcode files to llvm-dis.
In D70153#1744929, @pcc wrote:

The way I imagined this might work is that llvm-dis would write out the modules concatenated with some kind of delimiter between them. Then llvm-as would be taught to expect the same input and produce a multi-module file from it, which would allow an easy way to produce multi-module files. Does that sound like it might work better?

Wed, Nov 13, 8:27 PM · Restricted Project
tejohnson accepted D70153: Add support for multi-module bitcode files to llvm-dis.

LGTM - nice improvement!

Wed, Nov 13, 4:22 PM · Restricted Project
tejohnson added a comment to D70195: Fix error message missed in commit dde589389fcb8b5098f7a47f1b781b27d29a0cac..

Thanks for fixing this and for the cleanup. Since we will have the same const INDEX_VERSION variable defined in both the bitcode reader and writer, I'm thinking it will make sense to move this somewhere common, which I guess would be ModuleSummaryIndex.h, maybe as a static variable in the ModuleSummaryIndex class itself.

Wed, Nov 13, 11:45 AM · Restricted Project
tejohnson accepted D70128: [ThinLTO] Add correctness check for RO/WO variable import.

Thanks for doing this! A minor suggestion and a spelling correction on the variable below, but otherwise lgtm. Please commit the ExportList type change as a separate prior NFC commit and the new assertion checking after as a distinct commit.

Wed, Nov 13, 8:46 AM · Restricted Project

Tue, Nov 12

tejohnson added a comment to D70128: [ThinLTO] Add correctness check for RO/WO variable import.

Thanks for doing this. I have some ideas on how to reduce the overhead of this checking below.

Tue, Nov 12, 9:50 AM · Restricted Project

Mon, Nov 11

tejohnson added a comment to D69947: [ThinLTO] Simplify attribute propagation.

Given the write only bug that popped up with the earlier patch, I'm wondering if this is safe. I.e. we must import a global variable if we have marked it as write only, so if we don't import it later (because it fails one of the legality checks in canImportGlobalVar), then we can end up in trouble.

Mon, Nov 11, 11:08 AM · Restricted Project

Fri, Nov 8

tejohnson committed rGb11391bb47d6: ThinLTO : Import always_inline functions irrespective of the threshold (authored by tejohnson).
ThinLTO : Import always_inline functions irrespective of the threshold
Fri, Nov 8, 5:09 PM
tejohnson closed D70014: ThinLTO : Import always_inline functions irrespective of the threshold.
Fri, Nov 8, 5:09 PM · Restricted Project
tejohnson accepted D70014: ThinLTO : Import always_inline functions irrespective of the threshold.

LGTM thanks!

Fri, Nov 8, 11:57 AM · Restricted Project
tejohnson added a comment to D70014: ThinLTO : Import always_inline functions irrespective of the threshold.

Thanks for working on this! One suggestion for test cleanup below.

Fri, Nov 8, 11:21 AM · Restricted Project
tejohnson added inline comments to D70014: ThinLTO : Import always_inline functions irrespective of the threshold.
Fri, Nov 8, 10:44 AM · Restricted Project
tejohnson added a comment to D70014: ThinLTO : Import always_inline functions irrespective of the threshold.

I just started to look at this (added a minor comment below) - but could you please re-upload this with full context so that it is easier to review?

Fri, Nov 8, 9:49 AM · Restricted Project
tejohnson accepted D70006: [ThinLTO] Fix bug when importing writeonly variables.

Let's get this one in (I have a couple more comment nits below), to fix the immediate problem. But please send a follow on patch soon to do the assertion checking I suggested in my earlier comment, to make sure we catch any regressions to the force import of these write only variables.

Fri, Nov 8, 9:49 AM · Restricted Project
tejohnson added a comment to D70006: [ThinLTO] Fix bug when importing writeonly variables.

Ah, ok, I wasn't thinking about the part where the flag propagation was changed. I'm a little concerned that due to that change we are now required to import the variable as a definition. I guess there is no current way to disable importing of references while leaving on importing of functions (e.g. through some kind of internal option for debugging), but since such a mechanism might be useful at some point for debugging, my concern is that it will cause failures. Please add some very clear comments to the code where you are checking the write only flag as to why and what is going on. I.e. why needed for correctness and how we can ignore its references (because it will eventually be made zeroinit). Also, it would be good to add an assert somewhere to ensure that we don't export a reference but not a definition for a write only variable, to guard against any regression if someone does add a debugging mechanism to disable variable importing (which would need to affect the flag propagation). Maybe down at the bottom of ComputeCrossModuleImport where we prune the ExportLists.

Fri, Nov 8, 8:26 AM · Restricted Project
tejohnson added a comment to D70006: [ThinLTO] Fix bug when importing writeonly variables.

A few suggestions for beefing up the comments

Fri, Nov 8, 8:26 AM · Restricted Project
tejohnson added a comment to D70006: [ThinLTO] Fix bug when importing writeonly variables.

We hit a linker undef also because of D69561. I just confirmed this patch fixes the issue. It would be good to get this patch in asap or revert the original change. I have some questions on this patch though.

Fri, Nov 8, 7:49 AM · Restricted Project

Tue, Nov 5

tejohnson committed rGb36e3a8bac3e: [IRMover] Set Address Space for moved global values (authored by tejohnson).
[IRMover] Set Address Space for moved global values
Tue, Nov 5, 4:36 PM
tejohnson closed D69361: [IRMover] Set Address Space for moved global values.
Tue, Nov 5, 4:36 PM · Restricted Project
tejohnson committed rGdce53d71987d: [IRMover] Use GlobalValue::getAddressSpace instead of directly from its type… (authored by tejohnson).
[IRMover] Use GlobalValue::getAddressSpace instead of directly from its type…
Tue, Nov 5, 2:01 PM
tejohnson closed D69550: [IRMover] Use GlobalValue::getAddressSpace instead of directly from its type [NFC].
Tue, Nov 5, 2:01 PM · Restricted Project
tejohnson accepted D69561: [ThinLTO] Import readonly vars with refs.

LGTM with some minor comment changes suggested. Thanks!

Tue, Nov 5, 8:47 AM · Restricted Project
tejohnson added a comment to D69736: Keep import function list for inlinee profile update.

Thanks for splitting the patch. This one lgtm, but please wait to see if @wmi has any more comments. Also the summary needs an update (to remove the bit about the CS profile fix).

Tue, Nov 5, 6:37 AM · Restricted Project
tejohnson added a comment to D69732: [WIP][LTO] Apply SamplePGO pipeline tunes for ThinLTO pre-link to full LTO.

This probably needs to be taken over by someone who cares about full LTO performance

We at PlayStation are definitely interested in full LTO performance, so we're looking into this. We certainly agree with the rationale that if suppressing some optimizations is useful to allow better SamplePGO matching, then we'd expect that would apply equally to both ThinLTO and full LTO.

I guess much of this comes down to a balancing act between:

  1. The amount of the runtime benefit with Sample PGO if these loop optimizations are deferred to the full LTO back-end (like they are for ThinLTO).
  2. The cost in compile-time resources in the full LTO back-end to do these loop optimizations at that later stage.

    From the discussion here, the Sample PGO runtime win (point 1) seems more or less to be a given. If we find the compile-time cost in the full LTO back-end (point 2) is not significant, then the decision should be easy. So after seeing this patch, we're doing some experiments to at least try to get a handle on this. (I'm a bit concerned we won't be able to draw any hard conclusions from the results of our experiments, but at least we'll be able to make a better informed assessment.)

    FTR, for PlayStation, we're using the old PM. But we'll do some experiments for both the old and new PM, to get a sense of the answers to the (old PM) LoopUnrollAndJam point, and the (new PM) FIXME comment.
Tue, Nov 5, 6:28 AM · Restricted Project, Restricted Project

Mon, Nov 4

tejohnson added a comment to D69736: Keep import function list for inlinee profile update.

Nice catch on the dropped import lists!

Mon, Nov 4, 2:17 PM · Restricted Project
tejohnson added a comment to D61634: [clang/llvm] Allow efficient implementation of libc's memory functions in C/C++.

I had some time to work on this finally late last week. I decided the most straightforward thing was to implement the necessary interface changes to the TLI analysis to make it require a Function (without any changes yet to how that analysis operates). See D66428 that I just mailed for review. That takes care of the most widespread changes needed for this migration, and afterwards we can change the analysis to look at the function attributes and make a truly per-function TLI.

D66428 went in a few weeks ago at r371284, and I just mailed the follow on patch D67923 which will adds the support into the TLI analysis to use the Function to override the available builtins (with some of the code stubbed out since we don't yet have those per-Function attributes finalized).

@gchatelet where are you at on finalizing this patch? Also, I mentioned this offline but to follow up here: I think we will want an attribute to represent -fno-builtins (so that it doesn't need to be expanded out into the full list of individual no-builtin-{func} attributes, which would be both more verbose and less efficient, as well as being less backward compatible when new builtin funcs are added).

I'll break this patch in several pieces. The first one is to add the no_builtin attribute, see https://reviews.llvm.org/D61634.

Mon, Nov 4, 11:54 AM · Restricted Project, Restricted Project

Fri, Nov 1

tejohnson added a comment to D69732: [WIP][LTO] Apply SamplePGO pipeline tunes for ThinLTO pre-link to full LTO.

The comments indicate that this is in part due to issues with
the new PM loop pass manager

Wondering how different it is for these loop passes to be enabled for MonoLTO vs ThinLTO? If it's due to problems with the newPM, I guess ThinLTO would have the same problems?

Fri, Nov 1, 3:58 PM · Restricted Project, Restricted Project
tejohnson committed rG16ec00eee7e3: Recommit "[ThinLTO] Handle GUID collision in import global processing"" (authored by tejohnson).
Recommit "[ThinLTO] Handle GUID collision in import global processing""
Fri, Nov 1, 2:06 PM
tejohnson closed D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

Recommitted in 16ec00eee7e348767f4393f189044f87f6374031 with fix for the failure I saw.

Fri, Nov 1, 2:04 PM · Restricted Project
tejohnson added a comment to D69732: [WIP][LTO] Apply SamplePGO pipeline tunes for ThinLTO pre-link to full LTO.

This probably needs to be taken over by someone who cares about full LTO performance (@wristow or @ormris ?). This patch was some cleanup of the full LTO sample PGO pipeline, but has a number of issues I enumerate in the summary.

Fri, Nov 1, 1:09 PM · Restricted Project, Restricted Project
tejohnson created D69732: [WIP][LTO] Apply SamplePGO pipeline tunes for ThinLTO pre-link to full LTO.
Fri, Nov 1, 1:04 PM · Restricted Project, Restricted Project
tejohnson committed rGd39d1a2f87ac: Revert "[LLD][ThinLTO] Handle GUID collision in import global processing" (authored by tejohnson).
Revert "[LLD][ThinLTO] Handle GUID collision in import global processing"
Fri, Nov 1, 10:10 AM
tejohnson added a reverting change for rGcc0b9647b761: [LLD][ThinLTO] Handle GUID collision in import global processing: rGd39d1a2f87ac: Revert "[LLD][ThinLTO] Handle GUID collision in import global processing".
Fri, Nov 1, 10:09 AM
tejohnson reopened D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

I had to revert this as it was causing a failure in my testing. Looking into it...

Fri, Nov 1, 9:58 AM · Restricted Project
tejohnson added inline comments to D69561: [ThinLTO] Import readonly vars with refs.
Fri, Nov 1, 7:21 AM · Restricted Project

Thu, Oct 31

tejohnson accepted D69315: [LegacyPM] Fix structure dumper.

lgtm but please add patch summary

Thu, Oct 31, 6:26 AM · Restricted Project
tejohnson accepted D68593: [NewPM] Add an SROA pass after loop unroll.

lgtm

Thu, Oct 31, 6:16 AM · Restricted Project

Wed, Oct 30

tejohnson committed rGc844f8846aab: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables (authored by tejohnson).
[ThinLTO/WPD] Fix index-based WPD for available_externally vtables
Wed, Oct 30, 6:04 PM
tejohnson closed D69452: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables.
Wed, Oct 30, 6:03 PM · Restricted Project
tejohnson requested changes to D69561: [ThinLTO] Import readonly vars with refs.

Thanks for working on this! I had been thinking about adding this recently but didn't have a chance yet. I have a few suggestions below. One thing that should be considered for a follow-up enhancement is to consider the refs themselves for importing (similar to the way we follow chains of calls when doing function importing).

Wed, Oct 30, 11:31 AM · Restricted Project

Tue, Oct 29

tejohnson updated the diff for D69452: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables.

Update patch per comments.

Tue, Oct 29, 6:32 PM · Restricted Project
tejohnson added inline comments to D69452: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables.
Tue, Oct 29, 6:29 PM · Restricted Project
tejohnson accepted D69550: [IRMover] Use GlobalValue::getAddressSpace instead of directly from its type [NFC].

LGTM

Tue, Oct 29, 7:02 AM · Restricted Project
tejohnson accepted D69361: [IRMover] Set Address Space for moved global values.

LGTM

Tue, Oct 29, 6:55 AM · Restricted Project

Mon, Oct 28

tejohnson added inline comments to D69452: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables.
Mon, Oct 28, 7:00 AM · Restricted Project

Fri, Oct 25

tejohnson created D69452: [ThinLTO/WPD] Fix index-based WPD for available_externally vtables.
Fri, Oct 25, 3:06 PM · Restricted Project
tejohnson committed rGcc0b9647b761: [LLD][ThinLTO] Handle GUID collision in import global processing (authored by tejohnson).
[LLD][ThinLTO] Handle GUID collision in import global processing
Fri, Oct 25, 12:38 PM
tejohnson closed D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.
Fri, Oct 25, 12:38 PM · Restricted Project
tejohnson added a comment to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

I don't have commit access so I'd need someone to push this, thanks.

I'll take care of that for you today.

Fri, Oct 25, 8:24 AM · Restricted Project
tejohnson added a comment to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

I don't have commit access so I'd need someone to push this, thanks.

Fri, Oct 25, 7:38 AM · Restricted Project
tejohnson added a comment to D69428: [GlobalOpt] Remove valgrind specific hacks (revert r160529).

Added the author of r160529 as a reviewer for thoughts.

Fri, Oct 25, 7:10 AM
tejohnson added a reviewer for D69428: [GlobalOpt] Remove valgrind specific hacks (revert r160529): nlewycky.
Fri, Oct 25, 7:10 AM
tejohnson accepted D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

LGTM. Small suggestion to expand comment a bit.

Fri, Oct 25, 6:50 AM · Restricted Project

Thu, Oct 24

tejohnson added a comment to D69406: [clang][ThinLTO] Promote cc1 -fthin_link_bitcode to driver -fthinlto_link_bitcode.

Thanks. Prefer the old name since it is already in use and since this is for the "thin link" phase.

Thu, Oct 24, 3:15 PM · Restricted Project
tejohnson added a comment to D69384: Add unlikely attribute to outlined function.

btw, I had this patch in mind:

diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index 5edc696c29a..60726320cf1 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -434,11 +434,11 @@ bool CodeGenPrepare::runOnFunction(Function &F) {
   ProfileSummaryInfo *PSI =
       &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
 

+  if (!PSI->hasProfileSummary()) {
+    auto EC = F.getEntryCount();
+     if (EC.hasValue() && (F.getEntryCount().getCount() == 0))
+      F.setSectionPrefix(".unlikely");
+  }
+  }
Thu, Oct 24, 3:14 PM · Restricted Project
tejohnson added a comment to D69384: Add unlikely attribute to outlined function.

Because my goal is to improve outlining without PGO (with static analysis), I'm wondering what would be the best place to put this other than HotColdSplitting.

Thu, Oct 24, 2:26 PM · Restricted Project
tejohnson added a comment to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

Another place that already handles multiple locals with the same GUID is in the same file: bool FunctionImportGlobalProcessing::shouldPromoteLocalToGlobal(), from a comment there:

// When exporting, consult the index. We can have more than one local
// with the same GUID, in the case of same-named locals in different but
// same-named source files that were compiled in their respective directories
// (so the source file name and resulting GUID is the same). Find the one
// in this module.

so I think raising an error would cause surprises for different setups.
I'll gladly add more testcases if you can give me some ideas/tips on possible problems.

Thu, Oct 24, 1:50 PM · Restricted Project
tejohnson added a comment to D69384: Add unlikely attribute to outlined function.
In D69384#1720142, @vsk wrote:

Yes, this does need a test, but I'm also not sure why this belongs in the hot/cold splitting pass.

It's possible to mark functions with the cold attribute outside of the splitting pass. Shouldn't all of these cold functions should have a .unlikely section prefix? If so, setting the section prefix should be done later (say, in PreISelLowering).

Thu, Oct 24, 10:56 AM · Restricted Project
tejohnson added a comment to D69361: [IRMover] Set Address Space for moved global values.

This seems fine to me, but can you pull out and commit the NFC parts first, and update the patch to just contain the functional change? It's harder to see with all the NFC changes in here. The NFC part can go in without a phabricator patch.

Thu, Oct 24, 6:50 AM · Restricted Project
tejohnson added a comment to D69327: [Clang][ThinLTO] Add a cache for compile phase output..

We rely on the minimized bitcode from that option in our builds, so it won't be going away. We could add it as a driver option, but it doesn't sound like that will solve the ccache issue. I'm not very familiar with that cache support, or if there is a way to express other output files (in our build system we identify all the outputs of a build action). At least with something like make and ninja the compile action should not be done again if none of the inputs change.

Thu, Oct 24, 6:27 AM · Restricted Project, Restricted Project

Wed, Oct 23

tejohnson added a comment to D69327: [Clang][ThinLTO] Add a cache for compile phase output..

I was just typing up a response similar to @steven_wu 's response... I don't think clang should be in the business of caching the outputs of a prior clang invocation, the build system should and usually does avoid re-executing if the inputs have not changed. Note that this is very different from the caching of objects LTO is doing - because those are not otherwise emitted at all, the cache has to be built in.

Wed, Oct 23, 3:12 PM · Restricted Project, Restricted Project
tejohnson added a comment to D69327: [Clang][ThinLTO] Add a cache for compile phase output..

I was just typing up a response similar to @steven_wu 's response... I don't think clang should be in the business of caching the outputs of a prior clang invocation, the build system should and usually does avoid re-executing if the inputs have not changed. Note that this is very different from the caching of objects LTO is doing - because those are not otherwise emitted at all, the cache has to be built in.

Wed, Oct 23, 3:12 PM · Restricted Project, Restricted Project
tejohnson added a comment to D69327: [Clang][ThinLTO] Add a cache for compile phase output..

I haven't read through the patch in detail yet, but from the description it sounds like a cache is being added for the compile step outputs, e.g. the bitcode output object of the "clang -flto=thin -c" step? Typically the build system takes care of incremental build handling for these explicit outputs, just as it would for a non-LTO clang compile output. Can you clarify the motivation?

Wed, Oct 23, 12:05 PM · Restricted Project, Restricted Project
tejohnson added a comment to D69331: [LegacyPM] Port CGProfile pass.

Thanks for looking at it, Teresa!

Here is a llvm-dev post, suggesting to use new PM by default:
http://lists.llvm.org/pipermail/llvm-dev/2017-October/118280.html

As you may have noticed it was two years ago and new PM is still "brand new" :)
Also, AFAIK, codegen passes still use old PM, so I guess it will be around for quite a while.
Until new PM is adopted as default by LLVM community, switching to it is not an option for us, unfortunately.
And I need just this pass (it does pretty good job optimizing PGO instrumented images) and don't have plans to port anything else.

Wed, Oct 23, 11:55 AM · Restricted Project
tejohnson added a comment to D69331: [LegacyPM] Port CGProfile pass.

FYI what I heard at the dev meeting yesterday is that we are very close to being able to switch to the new PM by default, after which the legacy PM will be removed (I'm sure not immediately, but the goal is to get down to one PM soon). Before you do a lot of work backporting some of the new passes like this, is switching to the new PM an option for you?

Wed, Oct 23, 7:46 AM · Restricted Project

Mon, Oct 21

tejohnson added a comment to D69107: [ThinLTO] Don't internalize during promotion.

@tejohnson Teresa, are you ok with it?

Mon, Oct 21, 11:05 AM · Restricted Project

Oct 15 2019

tejohnson accepted D68782: [ThinLTO] Import virtual method with single implementation in hybrid mode.

lgtm

Oct 15 2019, 8:19 AM · Restricted Project
tejohnson added inline comments to D68782: [ThinLTO] Import virtual method with single implementation in hybrid mode.
Oct 15 2019, 6:36 AM · Restricted Project

Oct 14 2019

tejohnson accepted D68924: CodeExtractor: NFC: Use Range based loop.

LGTM

Oct 14 2019, 1:06 PM · Restricted Project
tejohnson committed rG8408d95e3199: [ThinLTO] Fix printing of NoInline function summary flag (authored by tejohnson).
[ThinLTO] Fix printing of NoInline function summary flag
Oct 14 2019, 11:40 AM
tejohnson closed D68948: [ThinLTO] Fix printing of NoInline function summary flag.
Oct 14 2019, 11:39 AM · Restricted Project
tejohnson committed rL374802: [ThinLTO] Fix printing of NoInline function summary flag.
[ThinLTO] Fix printing of NoInline function summary flag
Oct 14 2019, 11:39 AM
tejohnson created D68948: [ThinLTO] Fix printing of NoInline function summary flag.
Oct 14 2019, 9:20 AM · Restricted Project

Oct 12 2019

tejohnson added inline comments to D68924: CodeExtractor: NFC: Use Range based loop.
Oct 12 2019, 8:48 PM · Restricted Project

Oct 10 2019

tejohnson added a comment to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

Is this the only place where bad things can happen? May be simply raise an error in addGlobalValueSummary when new summary type is different from that of SummaryList[0]?

Oct 10 2019, 12:37 PM · Restricted Project
tejohnson added a comment to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.

I wonder how such thing happened in the first place. Can linker produce working image even if we circumvent the crash?

Oct 10 2019, 9:18 AM · Restricted Project
tejohnson added inline comments to D67322: [LLD][ThinLTO] Handle GUID collision in import global processing.
Oct 10 2019, 8:52 AM · Restricted Project

Oct 4 2019

tejohnson added inline comments to D68063: Propeller: LLVM support for basic block sections.
Oct 4 2019, 11:38 AM · Restricted Project

Oct 2 2019

tejohnson accepted D68318: Fix: Actually erase remove the elements from AssumeHandles.

LGTM

Oct 2 2019, 9:47 AM · Restricted Project
tejohnson committed rG077cc3fcb01d: [ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary (authored by tejohnson).
[ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary
Oct 2 2019, 9:36 AM
tejohnson committed rL373485: [ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary.
[ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary
Oct 2 2019, 9:36 AM
tejohnson closed D68159: [ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary.
Oct 2 2019, 9:36 AM · Restricted Project
tejohnson added a comment to D68318: Fix: Actually erase remove the elements from AssumeHandles.

Relevant test case: llvm/test/Transforms/HotColdSplit/assumption-cache-invalidation.ll is already in trunk.

Oct 2 2019, 8:48 AM · Restricted Project
tejohnson accepted D68331: [CodeExtractor] NFC: Refactor sanity checks into isEligible.

LGTM with one suggestion

Oct 2 2019, 6:40 AM · Restricted Project

Oct 1 2019

tejohnson added inline comments to D68159: [ThinLTO/WPD] Ensure devirtualized targets use promoted symbol when necessary.
Oct 1 2019, 4:14 PM · Restricted Project
tejohnson committed rGdca5b94e798c: [ThinLTO] Enable index-only WPD from clang (authored by tejohnson).
[ThinLTO] Enable index-only WPD from clang
Oct 1 2019, 11:08 AM
tejohnson committed rL373370: [ThinLTO] Enable index-only WPD from clang.
[ThinLTO] Enable index-only WPD from clang
Oct 1 2019, 11:08 AM
tejohnson closed D68029: [ThinLTO] Enable index-only WPD from clang.
Oct 1 2019, 11:08 AM · Restricted Project, Restricted Project
tejohnson added a comment to D68029: [ThinLTO] Enable index-only WPD from clang.

Ping

Oct 1 2019, 9:50 AM · Restricted Project, Restricted Project

Sep 30 2019

tejohnson committed rG2024ea148cc0: Fix buildbot failure from r373217 (don't match metadata id exactly) (authored by tejohnson).
Fix buildbot failure from r373217 (don't match metadata id exactly)
Sep 30 2019, 10:29 AM
tejohnson added a comment to D67592: [Clang] Use -main-file-name for source filename if not set.

This is failing on my mac like so:

Sep 30 2019, 10:29 AM · Restricted Project, Restricted Project
tejohnson committed rL373237: Fix buildbot failure from r373217 (don't match metadata id exactly).
Fix buildbot failure from r373217 (don't match metadata id exactly)
Sep 30 2019, 10:24 AM
tejohnson committed rG57dbfe194cb5: [Clang] Use -main-file-name for source filename if not set (authored by tejohnson).
[Clang] Use -main-file-name for source filename if not set
Sep 30 2019, 8:05 AM
tejohnson committed rL373217: [Clang] Use -main-file-name for source filename if not set.
[Clang] Use -main-file-name for source filename if not set
Sep 30 2019, 8:05 AM
tejohnson closed D67592: [Clang] Use -main-file-name for source filename if not set.
Sep 30 2019, 8:05 AM · Restricted Project, Restricted Project
tejohnson added a comment to D67592: [Clang] Use -main-file-name for source filename if not set.

I don't have commit access so I'd need someone to push this, thanks.

Sep 30 2019, 7:14 AM · Restricted Project, Restricted Project

Sep 27 2019

tejohnson added inline comments to D59709: [ThinLTO] Auto-hide prevailing linkonce_odr only when all copies eligible.
Sep 27 2019, 4:53 PM · Restricted Project