Page MenuHomePhabricator
Feed Advanced Search

Tue, May 26

aganea added a comment to D69778: Make -fmodules-codegen and -fmodules-debuginfo work also with precompiled headers.

@aganea Have you tried how this version of the patch affects PR44953? If not, could you please try?

Tue, May 26, 8:38 AM · Restricted Project

Fri, May 22

aganea accepted D80454: [Clang][test] fix tests when using external assembler.

Are you planning on adding another patch for the change that you've sent initially? Ideally, I'd like to relax a bit that constraint in Driver.cpp, as in D74447, but that requires first to disable -disable-free to ensure we exit cleanly, at least when runing the tests (a bit like the LLD_IN_TEST env var).

Fri, May 22, 2:31 PM · Restricted Project
aganea updated subscribers of D19908: [X86] Support the "ms-hotpatch" attribute..

Hello! It seems this patch and another related patch were reverted and @cdavis5x has been inactive since.
Any objection if I tried to reland this? Is there a policy for abandonned patches?

Fri, May 22, 1:58 PM · Restricted Project
aganea added a comment to D80454: [Clang][test] fix tests when using external assembler.

Could you please attach a test to demonstrate the issue?

Fri, May 22, 1:24 PM · Restricted Project
aganea added inline comments to D80433: [clang-tblgen][CommandLine][ManagedStatic] Fix build errors with clang-tblgen in Debug mode using MSVC 2019 v16.6.
Fri, May 22, 8:33 AM · Restricted Project, Restricted Project

Wed, May 20

aganea accepted D79847: [Support] On Windows, handle interrupt signals without crash message.

LGTM.

Wed, May 20, 9:16 AM · Restricted Project

Tue, May 19

aganea added a comment to D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link().

I read back through some of the history, but I still don't feel like I have all the context. That said, this change looks simple and straightforward to me.

Would it be feasible and useful to create a test that links two trivial programs with a single instantiation of the lld process? It seems like it would detect regressions and catch any introduction of a new global that needs to be reset.

Tue, May 19, 4:00 PM · lld, Restricted Project
aganea updated the diff for D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link().

Taking over the patch, after discussing with Erik.

Tue, May 19, 1:44 PM · lld, Restricted Project
aganea commandeered D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link().
Tue, May 19, 1:43 PM · lld, Restricted Project

Fri, May 15

aganea added a comment to D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link().

Seems good, could please remove the unrelated clang-format changes?
Do you have commit access?

Fri, May 15, 11:24 AM · lld, Restricted Project
aganea committed rG76c5f277f25e: Re-land [Debug][CodeView] Emit fully qualified names for globals (authored by aganea).
Re-land [Debug][CodeView] Emit fully qualified names for globals
Fri, May 15, 8:07 AM
aganea closed D79447: [Debug][CodeView] Emit fully qualified names for globals.
Fri, May 15, 8:06 AM · Restricted Project, Restricted Project
aganea added inline comments to D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link().
Fri, May 15, 6:28 AM · lld, Restricted Project

Thu, May 14

aganea added a comment to D68820: win: Move Parallel.h off concrt to cross-platform code.

BTW ThreadPoolExecutor has undefined behavior since there's a detached thread touching the standard library when the program exits violating [basic.start.term]/6. Detached threads are almost never safe.

This was fixed later by D70447.

Thu, May 14, 6:31 PM · Restricted Project
aganea added inline comments to D68820: win: Move Parallel.h off concrt to cross-platform code.
Thu, May 14, 2:44 PM · Restricted Project
aganea abandoned D79512: [DebugInfo][CodeView] Fix lowering of UDT.

I've merged this back with D79447, I think it's better if it landed in one piece.

Thu, May 14, 11:57 AM · Restricted Project
aganea abandoned D59226: [LLD][COFF] DebugTypes prototype -- NOT FOR SUBMIT.

Commited by @rnk as rG54a335a2f60b0f7bb85d01780bb6bbf653b1f399.

Thu, May 14, 11:23 AM · Restricted Project
aganea added inline comments to D79672: [COFF] Move type merging to TpiSource::mergeDebugT virtual method.
Thu, May 14, 10:16 AM · Restricted Project
aganea accepted D79672: [COFF] Move type merging to TpiSource::mergeDebugT virtual method.

LGTM, thanks again! (minor comment below)

Thu, May 14, 8:38 AM · Restricted Project
aganea added a reviewer for D70378: [LLD][COFF] Fix missing cache cleanup in COFF::link(): aganea.
Thu, May 14, 7:32 AM · lld, Restricted Project

Wed, May 13

aganea added a comment to D79672: [COFF] Move type merging to TpiSource::mergeDebugT virtual method.

Thanks a lot for rebasing this!

Wed, May 13, 9:12 AM · Restricted Project

Tue, May 12

aganea requested review of D79447: [Debug][CodeView] Emit fully qualified names for globals.
Tue, May 12, 9:07 AM · Restricted Project, Restricted Project
aganea reopened D79447: [Debug][CodeView] Emit fully qualified names for globals.

Could you please take another look? Thank you in advance!

Tue, May 12, 9:07 AM · Restricted Project, Restricted Project
aganea updated the diff for D79447: [Debug][CodeView] Emit fully qualified names for globals.

Reopening because I had to revert.

  • Fix lldb's variables.test
  • Fix type lowering, as described here: D79512
  • Added assert in emitDebugInfoForUDTs to ensure no lowering can occur during UDT emission, as suggested by @rnk
Tue, May 12, 9:07 AM · Restricted Project, Restricted Project

Sat, May 9

aganea added inline comments to D55585: RFC: [LLD][COFF] Parallel GHASH generation at link-time -- NOT FOR SUBMIT.
Sat, May 9, 8:29 AM · Restricted Project
aganea updated subscribers of D79672: [COFF] Move type merging to TpiSource::mergeDebugT virtual method.
Sat, May 9, 8:29 AM · Restricted Project

Fri, May 8

aganea added a comment to D79467: [PDB] Optimize public symbol processing.
In D79467#2027736, @rnk wrote:

Speaking of which, @aganea, since you are using clang to compile now, you should try adding -Xclang -debug-info-kind=constructor if you haven't already. It greatly reduces the amount of duplicate type info that clang emits.

We were discussing about that today. Is there any drawback for doing so?

Fri, May 8, 3:36 PM · Restricted Project
aganea added a comment to D79467: [PDB] Optimize public symbol processing.

Yes, I wanted to get back to that GHash parallelization at some point, but I'm swamped in deploying Clang on production and shipping some of our games. I was planning to back to that eventually: D55585 -- However the plan was to first move all Type-related code from PDB.cpp to DebugTypes.cpp, to ease things a bit: D59226 -- this still needs to be completed (steps 5-7).

Fri, May 8, 12:20 PM · Restricted Project
aganea updated the summary of D59226: [LLD][COFF] DebugTypes prototype -- NOT FOR SUBMIT.
Fri, May 8, 11:14 AM · Restricted Project
aganea accepted D79467: [PDB] Optimize public symbol processing.

Looks good, thanks!

Fri, May 8, 8:00 AM · Restricted Project

Thu, May 7

aganea added a comment to D79467: [PDB] Optimize public symbol processing.

Seems good to me, just a few things:

Thu, May 7, 1:02 PM · Restricted Project

Wed, May 6

aganea committed rG6adc45d3fd67: [LLD][COFF] Move debug info for thread-local variables into PDB global stream (authored by aganea).
[LLD][COFF] Move debug info for thread-local variables into PDB global stream
Wed, May 6, 12:27 PM
aganea committed rGf78b674de453: Revert "[Debug][CodeView] Emit fully qualified names for globals" (authored by aganea).
Revert "[Debug][CodeView] Emit fully qualified names for globals"
Wed, May 6, 12:27 PM
aganea added a reverting change for rG06591b6d191a: [Debug][CodeView] Emit fully qualified names for globals: rGf78b674de453: Revert "[Debug][CodeView] Emit fully qualified names for globals".
Wed, May 6, 12:27 PM
aganea closed D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.
Wed, May 6, 12:27 PM · Restricted Project
aganea updated the summary of D79512: [DebugInfo][CodeView] Fix lowering of UDT.
Wed, May 6, 12:25 PM · Restricted Project
aganea added inline comments to D79512: [DebugInfo][CodeView] Fix lowering of UDT.
Wed, May 6, 12:25 PM · Restricted Project
aganea created D79512: [DebugInfo][CodeView] Fix lowering of UDT.
Wed, May 6, 12:25 PM · Restricted Project
aganea added a comment to D79467: [PDB] Optimize public symbol processing.

Can you diff the full output of microsoft-pdb\cvdump\cvdump.exe your_exe.pdb and llvm-pdbutil dump -all your_exe.pdb before and after this patch? (with any large exe). This can be a bit tedious because the text dump is very large, but you can at least validate that things are still right (you're probably already doing this).

Wed, May 6, 9:07 AM · Restricted Project
aganea committed rG06591b6d191a: [Debug][CodeView] Emit fully qualified names for globals (authored by aganea).
[Debug][CodeView] Emit fully qualified names for globals
Wed, May 6, 6:25 AM
aganea committed rGc503d97d1934: [Support] Silence warning: comparison of integers of different signs: 'const… (authored by aganea).
[Support] Silence warning: comparison of integers of different signs: 'const…
Wed, May 6, 6:25 AM
aganea committed rGdb817d15d121: [InstrProf] Silence warnings when targeting x86 with VS2019 16.5.4 (authored by aganea).
[InstrProf] Silence warnings when targeting x86 with VS2019 16.5.4
Wed, May 6, 6:25 AM
aganea closed D79447: [Debug][CodeView] Emit fully qualified names for globals.
Wed, May 6, 6:25 AM · Restricted Project, Restricted Project
aganea committed rG3483cdc8344d: [Sema] Silence warnings when targeting x86 with VS2019 16.5.4 (authored by aganea).
[Sema] Silence warnings when targeting x86 with VS2019 16.5.4
Wed, May 6, 6:25 AM
aganea closed D79337: Silence warnings when compiling x86 with latest MSVC.
Wed, May 6, 6:25 AM · Restricted Project, Restricted Project
aganea added a comment to D79447: [Debug][CodeView] Emit fully qualified names for globals.

I have a backlog of issues I'm noting along the way, I will only fix them if they have an impact in regards to debugging in Visual Studio.

Wed, May 6, 5:52 AM · Restricted Project, Restricted Project

Tue, May 5

aganea accepted D79390: [Support] Sink LLD's parallel algorithm wrappers to support.
Tue, May 5, 3:08 PM · Restricted Project
aganea added a comment to D79390: [Support] Sink LLD's parallel algorithm wrappers to support.

LGTM as well, thank you!

Tue, May 5, 3:08 PM · Restricted Project
aganea created D79447: [Debug][CodeView] Emit fully qualified names for globals.
Tue, May 5, 2:03 PM · Restricted Project, Restricted Project
aganea added a comment to D79390: [Support] Sink LLD's parallel algorithm wrappers to support.

If the goal is to replace Parallel.h with pSTL sometime in the future, then maybe it's better to keep the STL-like API and the policy (but I wouldn't favor that ATM)
I find it pleasant to read the way it is -- if you want to protect from someone doing using llvm::parallel, why not keep the LLD nomenclature? llvm::parallelSort(...). It will not be ambiguous, even if you do using llvm; parallelSort(...);.

Tue, May 5, 12:26 PM · Restricted Project
aganea added a comment to D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.

Ping! Is there anything else to be done here?

Tue, May 5, 6:25 AM · Restricted Project
aganea added a comment to D79390: [Support] Sink LLD's parallel algorithm wrappers to support.

This is a prerequisite for parallel public symbol processing in the PDB
library, which is in LLVM.

Tue, May 5, 6:25 AM · Restricted Project

Mon, May 4

aganea added a comment to D79337: Silence warnings when compiling x86 with latest MSVC.

Will do, thank you for your time!

Mon, May 4, 2:32 PM · Restricted Project, Restricted Project
aganea committed rG721ea5b380b8: [DebugInfo][CodeView] Include namespace into emitted globals (authored by aganea).
[DebugInfo][CodeView] Include namespace into emitted globals
Mon, May 4, 11:16 AM
aganea closed D79028: [DebugInfo][CodeView] Include namespace into emitted globals.
Mon, May 4, 11:16 AM · Restricted Project
aganea added a comment to D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Thank you!

Mon, May 4, 10:09 AM · Restricted Project
aganea created D79337: Silence warnings when compiling x86 with latest MSVC.
Mon, May 4, 10:09 AM · Restricted Project, Restricted Project

Fri, May 1

aganea added inline comments to D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.
Fri, May 1, 5:12 PM · Restricted Project
aganea updated the diff for D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.
Fri, May 1, 5:12 PM · Restricted Project
aganea requested review of D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Thanks @amccarth ! Would you mind taking a second look please? The stream now matches MSVC behavior.

Fri, May 1, 5:11 PM · Restricted Project
aganea updated the diff for D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Fix static thread_local, ie. S_LTHREAD32.

Fri, May 1, 5:11 PM · Restricted Project

Thu, Apr 30

aganea added a comment to D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

I'm noting by the way that extern thread_local int A; does not generate a debug symbol :( I haven't checked why yet.

Thu, Apr 30, 12:53 PM · Restricted Project
aganea added inline comments to D79028: [DebugInfo][CodeView] Include namespace into emitted globals.
Thu, Apr 30, 12:53 PM · Restricted Project
aganea updated the diff for D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Fix local thread_local which shouldn't generate a scoped name in S_LTHREAD32.

Thu, Apr 30, 12:53 PM · Restricted Project
aganea added a comment to D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.

Is bitness relevant to the problem being solved? Do we know thread-local variables are in the globals stream regardless of bitness?

In this case, the globals stream is the same for 32-bit and 64-bit builds (except maybe for file offsets).

Thu, Apr 30, 12:21 PM · Restricted Project
aganea updated the diff for D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.
  • Generate yaml with the 32-bit toolchain to reduce variability.
  • Added local thread_local to cover S_LTHREAD32.
Thu, Apr 30, 12:21 PM · Restricted Project
aganea retitled D79028: [DebugInfo][CodeView] Include namespace into emitted globals from [DebugInfo][COFF] Include namespace into emitted globals to [DebugInfo][CodeView] Include namespace into emitted globals.
Thu, Apr 30, 6:21 AM · Restricted Project

Wed, Apr 29

aganea added inline comments to D78903: [Driver] Add option -fproc-stat-report.
Wed, Apr 29, 10:43 AM · Restricted Project
aganea added inline comments to D78903: [Driver] Add option -fproc-stat-report.
Wed, Apr 29, 7:28 AM · Restricted Project
aganea added a comment to D79043: [Driver] Skip validation of system sanitizer blacklists files if -fno-sanitizer-blacklist was specified.

clang -fsanitize=cfi -fno-sanitize-blacklist -fsanitize-blacklist=/tmp/blacklist.txt foo.c

Wed, Apr 29, 6:24 AM · Restricted Project, Restricted Project

Tue, Apr 28

aganea updated the diff for D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Fix tests.

Tue, Apr 28, 12:25 PM · Restricted Project
aganea updated the diff for D79028: [DebugInfo][CodeView] Include namespace into emitted globals.

Simplify.

Tue, Apr 28, 11:50 AM · Restricted Project
aganea created D79028: [DebugInfo][CodeView] Include namespace into emitted globals.
Tue, Apr 28, 11:50 AM · Restricted Project
aganea added a comment to D78845: [COFF] Add a fastpath for /INCLUDE: in .drective sections.
In D78845#2008164, @rnk wrote:

I think early in LLVM project history, developers did a lot of micro-optimization focusing on reducing heap allocations (see prevalence (and overuse!) of SmallVector), and a lot of that has gone by the wayside as generic containers proliferate in new code.

Tue, Apr 28, 11:17 AM · Restricted Project
aganea added a comment to D74846: fix -fcodegen-modules code when used with PCH (PR44953).

Once you do that, maybe you could send another message to the cfe-dev mailing list to ask for reviewers who have interest into optimizing PCH? I do, but I don't have the deep knowledge of this code.

Tue, Apr 28, 8:01 AM · Restricted Project
aganea added a comment to D74846: fix -fcodegen-modules code when used with PCH (PR44953).

To ease reviewers comprehension you have to propose patches that diff against the master. People sometimes apply the patches locally before accepting them.
This current patch is a subtraction after D69778 is applied locally on master. It seems a bit complicated.

Tue, Apr 28, 8:01 AM · Restricted Project
aganea created D79005: [LLD][COFF] Move debug info for thread-local variables into PDB global stream.
Tue, Apr 28, 7:29 AM · Restricted Project

Apr 28 2020

aganea added a comment to D74846: fix -fcodegen-modules code when used with PCH (PR44953).

My point is that all the changes in this patch are already in the upstream. If you want to re-land D69778, you should rather re-open and update that patch instead? (and close this current one)

Apr 28 2020, 6:24 AM · Restricted Project

Apr 27 2020

aganea added a comment to D74846: fix -fcodegen-modules code when used with PCH (PR44953).

@llunak : It seems @hans already reverted this in rG7ea9a6e0220da36ff2fd1fbc29c2755be23e5166, could you please ensure this patch is still relevant?

Apr 27 2020, 2:00 PM · Restricted Project
aganea added a comment to D74846: fix -fcodegen-modules code when used with PCH (PR44953).

If I understand correctly, this patch reverts the behavior to what it was before your changes, to only handle modules, but not PCH? Is that correct?

Apr 27 2020, 1:27 PM · Restricted Project
aganea added inline comments to D78903: [Driver] Add option -fproc-stat-report.
Apr 27 2020, 11:17 AM · Restricted Project
aganea added a reviewer for D78901: [Support] Get process statistics in ExecuteAndWait and Wait: aganea.
Apr 27 2020, 11:17 AM · Restricted Project
aganea added inline comments to D78901: [Support] Get process statistics in ExecuteAndWait and Wait.
Apr 27 2020, 11:17 AM · Restricted Project

Apr 26 2020

aganea committed rGfd773e8a51b8: Re-land [MC] Fix quadratic behavior in addPendingLabel (authored by aganea).
Re-land [MC] Fix quadratic behavior in addPendingLabel
Apr 26 2020, 7:58 AM
aganea closed D78775: [MC] Fix quadratic behavior in addPendingLabel.
Apr 26 2020, 7:57 AM · Restricted Project
aganea updated the summary of D78775: [MC] Fix quadratic behavior in addPendingLabel.
Apr 26 2020, 7:57 AM · Restricted Project

Apr 25 2020

aganea committed rG72b6fcbe781b: [Support] Fix fragile ThreadPool test (authored by aganea).
[Support] Fix fragile ThreadPool test
Apr 25 2020, 12:12 PM
aganea closed D78832: [Support] Fix fragile ThreadPool test.
Apr 25 2020, 12:12 PM · Restricted Project
aganea added inline comments to D78856: [Support] Simplify and optimize ThreadPool.
Apr 25 2020, 11:08 AM · Restricted Project
aganea added inline comments to D78856: [Support] Simplify and optimize ThreadPool.
Apr 25 2020, 10:04 AM · Restricted Project
aganea added a comment to D78845: [COFF] Add a fastpath for /INCLUDE: in .drective sections.

As for LLVMOptions, what prevents a BumpAllocator + placement new on the Arg(s)? Or is the perf. wasted somewhere else?

Apr 25 2020, 10:04 AM · Restricted Project
aganea reopened D78775: [MC] Fix quadratic behavior in addPendingLabel.
Apr 25 2020, 8:28 AM · Restricted Project
aganea requested review of D78775: [MC] Fix quadratic behavior in addPendingLabel.
Apr 25 2020, 8:28 AM · Restricted Project
aganea updated the diff for D78775: [MC] Fix quadratic behavior in addPendingLabel.

As suggested by @efriedma, PTAL.

Apr 25 2020, 8:28 AM · Restricted Project
aganea updated the diff for D78832: [Support] Fix fragile ThreadPool test.

As suggested by @mehdi_amini

Apr 25 2020, 7:56 AM · Restricted Project

Apr 24 2020

aganea created D78832: [Support] Fix fragile ThreadPool test.
Apr 24 2020, 3:11 PM · Restricted Project
aganea committed rG65fe71be48d6: Revert "[MC] Fix quadratic behavior in addPendingLabel()" (authored by aganea).
Revert "[MC] Fix quadratic behavior in addPendingLabel()"
Apr 24 2020, 2:07 PM
aganea added a reverting change for rGe98f73a62907: [MC] Fix quadratic behavior in addPendingLabel(): rG65fe71be48d6: Revert "[MC] Fix quadratic behavior in addPendingLabel()".
Apr 24 2020, 2:07 PM
aganea committed rG0e13a0331fb9: [llvm-cov] Prevent llvm-cov from using too many threads (authored by aganea).
[llvm-cov] Prevent llvm-cov from using too many threads
Apr 24 2020, 1:00 PM
aganea closed D78408: [llvm-cov] Prevent llvm-cov from using too many threads.
Apr 24 2020, 1:00 PM · Restricted Project