== Progress ==
* Zero/sign extension elimination with widening types (TCWG-546 - 10/10)
- Fixed regression failures
- Fixed bootstrapping issues for ARM and AArch64
- Re-factored and added some comments
- x86-64 Bootstrapped and regression tested for all languages with
forced promotion. There is 6 differences in scanning for certain
instructions. All the execution tests are passing. Needs further
investigation.
== Plan ==
* Continue with Zero/sign extension pass.
- Benchmarking
- Get patch ready for upstream discussion
* Improve block memory operations by GCC (TCWG-142)
- Start work on this
== Progress ==
* GCC trunk/4.9 cross-validation (1/10)
- submitted a couple of patches to clean testsuite cases
* Neon intrinsic tests (1/10)
- submitted patch to avoid running the tests on ARM targets w/o Neon
- started adding new tests
- created 2 PR about intrinsic tests failing on AArch64_be
(1 assigned to Venkat, 1 to me)
* AArch64 sanitizer (1/10)
- submitted a patch upstream to allow supporting both older and newer kernels
No feedback so far.
* GCC 4.8 and 4.9 releases (3/10)
- preparing both releases including ARM's latest fixes for the A53 erratum
- had to respin mid-week after new fix was committed
- LINK_SPEC patch not committed yet in 4.8, and committed in 4.9
after I made the branch merge.
- now checking results with references. Several FAIL appear. TBC.
* cbuild2 schroot and master branches comparison (1/10)
- re-ran schroot branch after cleaning spurious "-static" flag left
in dejagnu configuration
- better results, faster
* Misc (3/10)
- calls, meetings
== Next ==
* GCC 4.8 and 4.9 releases: hopefully, after branch merge review
* AArch64 sanitizer
* Neon intrinsics tests update
* cbuild2:
- analyze previous results
- look at backport-test and tcwgweb scripts + logs
== Progress ==
* US LLVM Dev Mtg (4/6)
* Buildbots (TCWG-76 1/6)
- Fixing last bugs of the libcxx bot
- One last failure being looked at
* Background (1/6)
- Code review, meetings, discussions, etc.
- Meeting with Google/ARM/Qualcomm about Android+LLVM
* Two days off
== Plan ==
* Investigate last libcxx bug
* Move lab/office
Thanks for the reply, Will Newton.
Then can we expect the fix to be included for the official Android
toolchain 2014.11?
And, thanks for explaining the culprit of this bug, Jongsung Kim! :)
On Wed, Oct 29, 2014 at 9:00 PM, <linaro-toolchain-request(a)lists.linaro.org>
wrote:
> Send linaro-toolchain mailing list submissions to
> linaro-toolchain(a)lists.linaro.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://lists.linaro.org/mailman/listinfo/linaro-toolchain
> or, via email, send a message with subject or body 'help' to
> linaro-toolchain-request(a)lists.linaro.org
>
> You can reach the person managing the list at
> linaro-toolchain-owner(a)lists.linaro.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of linaro-toolchain digest..."
>
>
> Today's Topics:
>
> 1. RE: Enabling back linker plugin for Linaro Android toolchain
> (Jongsung Kim)
> 2. Re: Enabling back linker plugin for Linaro Android toolchain
> (Will Newton)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Wed, 29 Oct 2014 11:26:30 +0900
> From: "Jongsung Kim" <neidhard.kim(a)lge.com>
> To: '???' <qkrwngud825(a)gmail.com>, <linaro-android(a)lists.linaro.org>,
> <linaro-toolchain(a)lists.linaro.org>
> Subject: RE: Enabling back linker plugin for Linaro Android toolchain
> Message-ID: <012f01cff31f$c01cea90$4056bfb0$(a)lge.com>
> Content-Type: text/plain; charset="UTF-8"
>
> The version-string from binutils-linaro looks to be blamed. It once was:
>
> GNU ld (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC
> 2013.04) 2.23.1
>
> and the linker plugin works with this version of Linaro prebuilt
> toolchain. Now it is:
>
> GNU ld (crosstool-NG linaro-1.13.1-4.8-2014.04 - Linaro GCC 4.8-2014.04)
> 2.24.0.20140311 Linaro 2014.03
> GNU ld (crosstool-NG linaro-1.13.1-4.9-2014.08 - Linaro GCC 4.9-2014.08)
> 2.24.0.20140801 Linaro 2014.08
>
> and the linker plugin is not supported:
>
> $ arm-linux-gnueabihf-gcc -flto -fuse-linker-plugin -o hello hello.c
> arm-linux-gnueabihf-gcc: error: -fuse-linker-plugin is not supported in
> this configuration
>
> Look into gcc/configure script. It uses the version of ld to determine
> whether ld supports linker plugin. It extracts the version by doing
> something like:
>
> $ arm-linux-gnueabihf-ld --version | sed 1q | sed -n -e 's,^.*[
> ]\([0-9][0-9]*\.[0-9][0-9]*.*\)$,\1,p'
>
> and it will extract the last 2014.03 or 2014.08. By using proper
> substitution expression like 's,^GNU ld (.*) \([0-9][.0-9]*\).*$,\1,p', the
> script may enable linker plugin.
>
> However, patching the script looks like a bad idea, because it doesn?t
> help handling the version of gold:
>
> GNU gold (GNU Binutils for Ubuntu 2.24) 1.11
> GNU gold (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC
> 2013.04 2.23.1) 1.11
> GNU gold (crosstool-NG linaro-1.13.1-4.8-2014.04 - Linaro GCC 4.8-2014.04
> 2.24.0.20140311 Linaro 2014.03) 1.11
> GNU gold (crosstool-NG linaro-1.13.1-4.9-2014.08 - Linaro GCC 4.9-2014.08
> 2.24.0.20140801 Linaro 2014.08) 1.11
>
> I couldn?t find a reasonable general expression to extract the version.
>
>
> From: linaro-toolchain-bounces(a)lists.linaro.org [mailto:
> linaro-toolchain-bounces(a)lists.linaro.org] On Behalf Of ???
> Sent: Monday, October 27, 2014 10:15 PM
> To: linaro-android(a)lists.linaro.org; linaro-toolchain(a)lists.linaro.org
> Subject: Enabling back linker plugin for Linaro Android toolchain
>
> I'm using Linaro Android toolchain's arm-eabi- for compiling my Android
> Linux kernel with LTO.
>
> The main benefits of my kernel is that it uses
> LTO(Link-Time-Optimizations).
> (Patches found here: https://github.com/andikleen/linux-misc)
>
> But now, it's broken with Linaro Android toolchains from 2014.09~
>
> Building Linux kernel with LTO requires Linker plugin with the toolchain.
>
>
> But for some reason, linker plugin is disabled with 2014.09 and 2014.10
> (Which I used from here :
> https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…
> https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…
> )
>
> LTO build works flawlessly with 2014.08.
>
> With 2014.09 and 2014.10, I get the following error :
> cc1: error: -fno-fat-lto-objects are supported only with linker plugin
>
> If I explicitly remove " -fno-fat-lto-objects " from the Makefile, the
> linker fails to link all of the object files.
>
>
> I would like to ask Linaro to enable back the Linker plugin support :)
>
> Thanks in advance..
>
>
>
>
> ------------------------------
>
> Message: 2
> Date: Wed, 29 Oct 2014 09:12:53 +0000
> From: Will Newton <will.newton(a)linaro.org>
> To: Jongsung Kim <neidhard.kim(a)lge.com>
> Cc: ??? <qkrwngud825(a)gmail.com>, linaro-android(a)lists.linaro.org,
> Linaro Toolchain <linaro-toolchain(a)lists.linaro.org>
> Subject: Re: Enabling back linker plugin for Linaro Android toolchain
> Message-ID:
> <CANu=DmhPdCGYw0=hCs9tmkbCU6hMLLMJnYS-u_JGz=
> p17Bgonw(a)mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> On 29 October 2014 02:26, Jongsung Kim <neidhard.kim(a)lge.com> wrote:
> > The version-string from binutils-linaro looks to be blamed. It once was:
> >
> > GNU ld (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC
> 2013.04) 2.23.1
> >
> > and the linker plugin works with this version of Linaro prebuilt
> toolchain. Now it is:
> >
> > GNU ld (crosstool-NG linaro-1.13.1-4.8-2014.04 - Linaro GCC 4.8-2014.04)
> 2.24.0.20140311 Linaro 2014.03
> > GNU ld (crosstool-NG linaro-1.13.1-4.9-2014.08 - Linaro GCC 4.9-2014.08)
> 2.24.0.20140801 Linaro 2014.08
> >
> > and the linker plugin is not supported:
> >
> > $ arm-linux-gnueabihf-gcc -flto -fuse-linker-plugin -o hello hello.c
> > arm-linux-gnueabihf-gcc: error: -fuse-linker-plugin is not supported in
> this configuration
> >
> > Look into gcc/configure script. It uses the version of ld to determine
> whether ld supports linker plugin. It extracts the version by doing
> something like:
> >
> > $ arm-linux-gnueabihf-ld --version | sed 1q | sed -n -e 's,^.*[
> ]\([0-9][0-9]*\.[0-9][0-9]*.*\)$,\1,p'
> >
> > and it will extract the last 2014.03 or 2014.08. By using proper
> substitution expression like 's,^GNU ld (.*) \([0-9][.0-9]*\).*$,\1,p', the
> script may enable linker plugin.
> >
> > However, patching the script looks like a bad idea, because it doesn?t
> help handling the version of gold:
> >
> > GNU gold (GNU Binutils for Ubuntu 2.24) 1.11
> > GNU gold (crosstool-NG linaro-1.13.1-4.7-2013.04-20130415 - Linaro GCC
> 2013.04 2.23.1) 1.11
> > GNU gold (crosstool-NG linaro-1.13.1-4.8-2014.04 - Linaro GCC
> 4.8-2014.04 2.24.0.20140311 Linaro 2014.03) 1.11
> > GNU gold (crosstool-NG linaro-1.13.1-4.9-2014.08 - Linaro GCC
> 4.9-2014.08 2.24.0.20140801 Linaro 2014.08) 1.11
> >
> > I couldn?t find a reasonable general expression to extract the version.
>
> This should be fixed in binutils-linaro-2.24-2014.11.
>
> > From: linaro-toolchain-bounces(a)lists.linaro.org [mailto:
> linaro-toolchain-bounces(a)lists.linaro.org] On Behalf Of ???
> > Sent: Monday, October 27, 2014 10:15 PM
> > To: linaro-android(a)lists.linaro.org; linaro-toolchain(a)lists.linaro.org
> > Subject: Enabling back linker plugin for Linaro Android toolchain
> >
> > I'm using Linaro Android toolchain's arm-eabi- for compiling my Android
> Linux kernel with LTO.
> >
> > The main benefits of my kernel is that it uses
> LTO(Link-Time-Optimizations).
> > (Patches found here: https://github.com/andikleen/linux-misc)
> >
> > But now, it's broken with Linaro Android toolchains from 2014.09~
> >
> > Building Linux kernel with LTO requires Linker plugin with the toolchain.
> >
> >
> > But for some reason, linker plugin is disabled with 2014.09 and 2014.10
> > (Which I used from here :
> https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…
> https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…
> )
> >
> > LTO build works flawlessly with 2014.08.
> >
> > With 2014.09 and 2014.10, I get the following error :
> > cc1: error: -fno-fat-lto-objects are supported only with linker plugin
> >
> > If I explicitly remove " -fno-fat-lto-objects " from the Makefile, the
> linker fails to link all of the object files.
> >
> >
> > I would like to ask Linaro to enable back the Linker plugin support :)
> >
> > Thanks in advance..
> >
> >
> > _______________________________________________
> > linaro-toolchain mailing list
> > linaro-toolchain(a)lists.linaro.org
> > http://lists.linaro.org/mailman/listinfo/linaro-toolchain
>
>
>
> --
> Will Newton
> Toolchain Working Group, Linaro
>
>
>
> ------------------------------
>
> _______________________________________________
> linaro-toolchain mailing list
> linaro-toolchain(a)lists.linaro.org
> http://lists.linaro.org/mailman/listinfo/linaro-toolchain
>
>
> End of linaro-toolchain Digest, Vol 52, Issue 16
> ************************************************
>
I'm using Linaro Android toolchain's arm-eabi- for compiling my Android
Linux kernel with LTO.
The main benefits of my kernel is that it uses LTO(Link-Time-Optimizations).
(Patches found here: https://github.com/andikleen/linux-misc)
But now, it's broken with Linaro Android toolchains from 2014.09~
Building Linux kernel with LTO requires Linker plugin with the toolchain.
But for some reason, linker plugin is disabled with 2014.09 and 2014.10
(Which I used from here :
https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…https://android-build.linaro.org/builds/~linaro-android/toolchain-4.9-2014.…
)
LTO build works flawlessly with 2014.08.
With 2014.09 and 2014.10, I get the following error :
cc1: error: -fno-fat-lto-objects are supported only with linker plugin
If I explicitly remove " -fno-fat-lto-objects " from the Makefile, the
linker fails to link all of the object files.
I would like to ask Linaro to enable back the Linker plugin support :)
Thanks in advance..
Hi All,
There will be lab downtime in the month of November, and rather than
jeopardize the November quarterly binary toolchain release we've decided to
take some steps to ensure that it takes place on time.
The November binary toolchain quarterly release will contain some ARM64
erratum fixes. These are already in the Linaro GCC 4.9 2014.10 source
package release.
The November binary toolchain quarterly release will contain these fixes,
and will be delivered as planned, but will be based on the Linaro GCC 4.9
2014.10 package. We will not have a November Linaro GCC 4.9 2014.11 source
package release.
The December release will be the next Linaro GCC 4.9 source package release.
--
Ryan S. Arnold
Linaro Toolchain Working Group - Engineering Manager
www.linaro.org
cbuild2 benchmarking - TCWG-360 [2/10]
* Figured out how to cook my own OE images
* Started remembering how to build spec
libm exercising - CARD-1693 [2/10]
* Borrowed a usable Juno
* Found that lapack tests segfault on AArch64
* Ran linpack hpl, didn't observe it exercising libm much
** Haven't ruled out Bernie-error yet, though
Upstream - CARD-341 [1/10]
* Respun lowlevellock.h comments
Meetings/mail/etc [5/10]
* A lot of time (>3/10) in performance review and other annual ARM admin
* Some back and forth about difficulties of userspace access to ID registers
=Plan=
Get spec2006 working via my scripts
Find a good HPL libm exerciser
Spend a lot less time on ARM admin
== Progress ==
Further progress on GDB Tracepoints/Fast Tracepoints support on arm
[TCWG-480] [6/10]
-- Debugging and trying out some code enabling tracepoints in gdbserver
-- Research on past tracepoint patches and x86 vs arm architecture comparison
ARM/AArch64 GDB testsuite failures investigation [2/10]
Miscellaneous [2/10]
-- Meetings, Emails etc
-- Fix SSH configurations
-- Patch Scrolling
-- Hong Kong visa process
-- Multiple Dr visits due to continued sickness
== Plan ==
Further progress on GDB Tracepoints/Fast Tracepoints support on arm [TCWG-480]
Fix internet latency issues from lab hardware.
= Progress ==
* Short week was in vacation (22-24 October)
* Updated perf profile numbers for instruction count and documented my
analysis for Linaro compiler's O3 + LTO comparison on Aarch64 vs X86_64
(TCWG-544) (2/10)
* Misc [2/10]
emails, AMD meetings and 1-1 with inline manger.
1-1 with christophe.
== Plan ==
* Coremark benchmark and profiling.
Investigate trunk degradation for O3+ PGO + LTO and report.
== Progress ==
* Zero/sign extension elimination with widening types (TCWG-546 - 10/10)
- Re-wrote the pass from the results of experiments so far
- Fixed most of the regression failures
- 5 tests are still failing from C/C++/Fortran regression suite.
== Plan ==
* Continue with Zero/sign extension pass.
- Get bootstrapping for ARM and AArch64 working
- Fix remaining regression failures
- Add detail dumps
- Remove unnecessary copies (it is now being removed dead code
elimination pass)
- Get patch ready for upstream discussion
