Short week, Labor day (2/10)
== Issues ==
* None
== Progress ==
* Launchpad bugs: (4/10)
o LP #1307197 : gcc-4.9 miscompiles linux kernel zlib for armv3
- This an LRA issue for architectures < armv4 withi movhi
- Investigation ongoing
o LP #1169164 : including signal.h exposes various PSR_MODE #defines
- The proposed patch fix the issue
- Still need some validation before upstream submission
* Test new backport workflow and Cbuild2 (2/10)
- Some troubles to access the build farm.
* Misc:
o Cbuildv1 baby-sitting (1/10)
o Various meetings (1/10).
== Next ==
- Two days off
- FSF branch merge + backports
== Issues ==
* None
== Progress ==
* May 1-2: Laybor day holiday
* Commit fcsel support patch for aarch64 @r209889.
* Tuning move-loop-invariants heuristics for -fira-loop-pressure:
- Record register pressure for each invariant other than a max for all.
- Take register pressure into account when selecting the best candidate.
== Plan ==
* Tuning move-loop-invariants heuristics
* Ping pending patches.
== Planned leaves ==
* June 2.
* -fno-sched-interblock
* loop-inv
> Hello,
> I have a problem building gcc-linaro-arm-linux-gnueabihf
> toolchain for my x86_64 system. I download crosstool-ng (v1.13) and
> toolchain source code (4.8-2014.03) from
> http://releases.linaro.org/14.03/components/toolchain/binaries. I use
> the example configuration linaro-arm-linux-gnueabihf (obtained from
> ct-ng list-samples) and modify some entries using menuconfig option,
> like the local tarballs directory (pointing to downloaded toolchains
> source code directory) and the prefix directory (custom directory output).
>
> This is the problem during the execution of "ct-ng build" command:
>
> =================================================================
>
> [INFO ] Installing final compiler
>
> [EXTRA] Configuring final compiler
>
> [EXTRA] Building final compiler
>
> [EXTRA] AFMOYA: EMPIEZA A EJECUTARSE make jobsflags all
>
> [EXTRA] -j5
>
> [EXTRA] /home/afmoya/Descargas/build/.build/arm-linux-gnueabihf/build/build-cc
>
> [ERROR] configure: error: Link tests are not allowed after GCC_NO_EXECUTABLES.
>
> [ERROR] make[2]: *** [configure-zlib] Error 1
>
> [ERROR] make[1]: *** [all] Error 2
>
> [ERROR]
>
> [ERROR] >>
>
> [ERROR] >> Error happened in: main[scripts/crosstool-NG.sh]
>
> [ERROR] >>
>
> [ERROR] >> For more info on this error, look at the file: 'build.log'
>
> [ERROR] >> There is a list of known issues, some with workarounds, in:
>
> [ERROR] >> '/opt/cross-linaro/share/doc/ct-ng-linaro-1.13.1-4.8-2014.03/B - Known issues.txt'
>
> [ERROR]
>
> [ERROR] Build failed in step 'Extracting and patching toolchain components'
>
> [ERROR]
>
> [ERROR] (elapsed: 11:18.57)
>
> make: *** [build] Error 2
>
>
> Highlighted in green some entries added for me in attemp to understand
> the execution secuence.
>
> The error is triggered in the "configure-zlib" target of
> "${CT_TOP_DIR}/.build/arm-linux-gnueabihf/build/build-cc/Makefile". The
> "configure.log" of the
> "${CT_TOP_DIR}/.build/arm-linux-gnueabihf/build/build-cc/zlib/"
> directory shows some errors (I don't know if the are significants):
>
> LINE 83: Thread model: posix
>
> LINE 84: gcc version 4.7.2 (Debian 4.7.2-5)
>
> LINE 85: configure:3234: $? = 0
>
> LINE 86: configure:3223: x86_64-build_unknown-linux-gnu-gcc -V >&5
>
> LINE 87: gcc: error: unrecognized command line option '-V'
>
> LINE 88: gcc: fatal error: no input files
>
> LINE 89: compilation terminated.
>
> LINE 90: configure:3234: $? = 4
>
> LINE 91: configure:3223: x86_64-build_unknown-linux-gnu-gcc -qversion >&5
>
> LINE 92: gcc: error: unrecognized command line option '-qversion'
>
> LINE 93: gcc: fatal error: no input files
>
> LINE 94: compilation terminated.
>
> ...
>
> LINE 188: configure:6661: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
>
> LINE 189: conftest.c:11:28: fatal error: ac_nonexistent.h: No such file or directory
>
> LINE 190: compilation terminated.
>
> ...
>
> LINE 207: configure:6720: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
>
> LINE 208: conftest.c:11:28: fatal error: ac_nonexistent.h: No such file or directory
>
> LINE 209: compilation terminated.
>
> ...
>
> LINE 318: configure:10852: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
>
> LINE 319: conftest.c:23:28: fatal error: ac_nonexistent.h: No such file or directory
>
> LINE 320: compilation terminated.
>
> Some idea of what's the problem?
>
> This is my system info:
> Linux afmoya-pc 3.2.0-4-amd64 #1 SMP Debian 3.2.57-3 x86_64 GNU/Linux
>
> and some tools version installed in my system:
> gcc (Debian 4.7.2-5) 4.7.2
> GNU Make 3.81
> GNU Awk 4.0.1
> ...
>
> Best regards:
> Alexis Fajardo Moya.
I have removed the "zlib" folder from "gcc-linaro-4.8-2014.03.tar.xz" tarball, the compiler is also using zlib-1.2.5 statically linked, and thats permit compile the Linaro toolchain.
Best regards
Lic. Reinier Millo Sánchez
Centro de Estudios de Informática
Universidad Central "Marta Abreu" de Las Villas
-----
"...hay locuras que son la locura
personales locuras de dos...
hay locuras sin nombre
sin fecha sin cura
que no vale la pena curar...
que una de ellas será mi morir."
Lic. Reinier Millo Sánchez
Centro de Estudios de Informática
Universidad Central "Marta Abreu" de Las Villas
Hello,
I have a problem building gcc-linaro-arm-linux-gnueabihf
toolchain for my x86_64 system. I download crosstool-ng (v1.13) and
toolchain source code (4.8-2014.03) from
http://releases.linaro.org/14.03/components/toolchain/binaries. I use
the example configuration linaro-arm-linux-gnueabihf (obtained from
ct-ng list-samples) and modify some entries using menuconfig option,
like the local tarballs directory (pointing to downloaded toolchains
source code directory) and the prefix directory (custom directory output).
This is the problem during the execution of "ct-ng build" command:
=================================================================
[INFO ] Installing final compiler
[EXTRA] Configuring final compiler
[EXTRA] Building final compiler
[EXTRA] AFMOYA: EMPIEZA A EJECUTARSE make jobsflags all
[EXTRA] -j5
[EXTRA] /home/afmoya/Descargas/build/.build/arm-linux-gnueabihf/build/build-cc
[ERROR] configure: error: Link tests are not allowed after GCC_NO_EXECUTABLES.
[ERROR] make[2]: *** [configure-zlib] Error 1
[ERROR] make[1]: *** [all] Error 2
[ERROR]
[ERROR] >>
[ERROR] >> Error happened in: main[scripts/crosstool-NG.sh]
[ERROR] >>
[ERROR] >> For more info on this error, look at the file: 'build.log'
[ERROR] >> There is a list of known issues, some with workarounds, in:
[ERROR] >> '/opt/cross-linaro/share/doc/ct-ng-linaro-1.13.1-4.8-2014.03/B - Known issues.txt'
[ERROR]
[ERROR] Build failed in step 'Extracting and patching toolchain components'
[ERROR]
[ERROR] (elapsed: 11:18.57)
make: *** [build] Error 2
Highlighted in green some entries added for me in attemp to understand
the execution secuence.
The error is triggered in the "configure-zlib" target of
"${CT_TOP_DIR}/.build/arm-linux-gnueabihf/build/build-cc/Makefile". The
"configure.log" of the
"${CT_TOP_DIR}/.build/arm-linux-gnueabihf/build/build-cc/zlib/"
directory shows some errors (I don't know if the are significants):
LINE 83: Thread model: posix
LINE 84: gcc version 4.7.2 (Debian 4.7.2-5)
LINE 85: configure:3234: $? = 0
LINE 86: configure:3223: x86_64-build_unknown-linux-gnu-gcc -V >&5
LINE 87: gcc: error: unrecognized command line option '-V'
LINE 88: gcc: fatal error: no input files
LINE 89: compilation terminated.
LINE 90: configure:3234: $? = 4
LINE 91: configure:3223: x86_64-build_unknown-linux-gnu-gcc -qversion >&5
LINE 92: gcc: error: unrecognized command line option '-qversion'
LINE 93: gcc: fatal error: no input files
LINE 94: compilation terminated.
...
LINE 188: configure:6661: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
LINE 189: conftest.c:11:28: fatal error: ac_nonexistent.h: No such file or directory
LINE 190: compilation terminated.
...
LINE 207: configure:6720: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
LINE 208: conftest.c:11:28: fatal error: ac_nonexistent.h: No such file or directory
LINE 209: compilation terminated.
...
LINE 318: configure:10852: x86_64-build_unknown-linux-gnu-gcc -E conftest.c
LINE 319: conftest.c:23:28: fatal error: ac_nonexistent.h: No such file or directory
LINE 320: compilation terminated.
Some idea of what's the problem?
This is my system info:
Linux afmoya-pc 3.2.0-4-amd64 #1 SMP Debian 3.2.57-3 x86_64 GNU/Linux
and some tools version installed in my system:
gcc (Debian 4.7.2-5) 4.7.2
GNU Make 3.81
GNU Awk 4.0.1
...
Best regards:
Alexis Fajardo Moya.
== Progress ==
* Prototype scripting for macro-benchmarks for malloc (4/10, TCWG-441)
* glibc single thread optimization work (3/10, TCWG-436)
* Built releases of eglibc, gdb and binutils (2/10)
* glibc patch review and followup (1/10)
* Misc admin (expenses, LCU travel)
== Issues ==
* None
== Plan ==
* Various bits of patch followup
* More malloc macro-benchmark work
* See how single thread optimization discussion goes
* glibc benchmark graphing?
--
Will Newton
Toolchain Working Group, Linaro
== Progress ==
* Named Register (CARD-1246 2/10)
- Changing type of intrinsics, rebasing patch
* Build & Benchmark (CARD-716 6/10)
- Adding Clang+LLVM builds to CBuild2
- Re-setting APM's testing environment after replacement
* Background (2/10)
- Code review, etc.
- EuroLLVM 2014 outcome analysis, plans for 2015
== Plan ==
* Get Clang+LLVM in CBuild2
* Get Named registers in, then try Clang's parser
* Try SPEC on APM
hi,
For building arm64 kernel, I tried all toolchain released from
http://releases.linaro.org/latest/components/toolchain/binaries/
But it complained about some instructions bad:
/home/work/linux/arch/arm64/include/asm/irqflags.h: Assembler
messages:
/home/work/linux/arch/arm64/include/asm/irqflags.h:49: Error: no
such instruction: `msr daifset,'
/home/work/linux/arch/arm64/include/asm/irqflags.h:68: Error: no
such instruction: `mrs %rax,daif//arch_local_save_flags'
/home/work/linux/arch/arm64/include/asm/irqflags.h:49: Error: no
such instruction: `msr daifset,'
/home/work/linux/arch/arm64/include/asm/irqflags.h:68: Error: no
such instruction: `mrs %rax,daif//arch_local_save_flags'
/home/work/linux/arch/arm64/include/asm/irqflags.h:40: Error: no
such instruction: `msr daifclr,'
/home/work/linux/arch/arm64/include/asm/irqflags.h:68: Error: no
such instruction: `mrs %rax,daif//arch_local_save_flags'
/home/work/linux/arch/arm64/include/asm/irqflags.h:40: Error: no
such instruction: `msr daifclr,'
Could anyone tell how to solve it? thanks!
== Progress==
TCWG-435 needless busy-wait in lowlevellock.c (3/10)
* Patches ready, testing a bit slow & fiddly
TCWG-156 cortex-strings memset (3/10)
* Worried about noisy benchmarks
* Learned to use Lava
* Found some helpful internal docs
* Managed to tweak the code slightly
== Misc ==
Meetings/mail/etc 2/10
(Public) holiday 2/10
== Plan ==
Stop worrying and follow the general curve of the noisy benchmarks
(But also try to convince Maxim's spec scripts + lava to generate some
less noisy results)
Another public holiday next Monday
== Progress ==
* GDB reverse debugging on aarch64
-- Complete decoding of aarch64 data processing immediate instructions.
[TCWG-399] [1/10]
-- Complete decoding of aarch64 data processing register instructions.
[TCWG-402] [1/10]
-- Completed decoding of aarch64 exception and system instructions.
[TCWG-400] [2/10]
-- Further progress on decoding of aarch64 load store instructions.
[TCWG-401] [2/10]
-- Started implementation of aarch64 syscall record/replay. [TCWG-409]
[2/10]
* Sick Day Off on Monday [2/10]
== Plan ==
* GDB reverse debugging on aarch64
-- Further progress on decoding of aarch64 load store instructions.
[TCWG-401]
-- Further progress on aarch64 syscall record/replay. [TCWG-409]
* Public Holiday on 1st May.
Short week, Easter Monday + child care (3/10)
== Issues ==
* Toolchain64 disk still full every 2 days
== Progress ==
* Linaro GCC 4.9 2014.04 release (3/10)
- Branch merge with FSF 4.9.1
- Release tarball available on releases.linaro.org
- Announcement will follow
* Launchpad bugs: (2/10)
o LP #1169164 : including signal.h exposes various PSR_MODE #defines
- Discussed and implemented a fix
- Validation on-going
* Misc:
o Cbuildv1 baby-sitting (1/10)
o Various meetings (1/10).
o Resolved cards :
- TCWG-343. Make LRA the default for the ARM backend
- TCWG-422. LP-bug 1296676 : ICE in assign_by_spills building linux
btrfs module
== Next ==
- One day off (Labor day)
- Submit patch for LP #1169164
- TCWG-345. Analyse performance of LRA for ARM
== Progress ==
PGO - AArch64 (TCWG-179) (4/10)
* Completed SPEC2006 runs ( -O3 + PGO) in chroot + qemu-arm64-saucy
with Linaro branch(4.9).
* Perlbench train run failed, issue with qemu.
* DealII train runs failed due to system libstdc++.so.6 not compatible
with GCC 4.9.
Changing the LD_LIBRARY_PATH to point to the libstdc++.so.6 that
gets built along with GCC 4.9 solves the problem.
* CPU2000 runs are completed my maxim so functional testing of PGO for
spec benchmarks completed.
* Wating on Arm64 hardware to benchmark SPEC2000 and SPEC2006 and
compare against the PGO runs for x86.
Misc (2/10)
* Setup arm64-trusty chroot and built GCC 4.9 compiler
* Maxim 1-1 discussions and set up opennx client
* Checked libssp patch upstream status
* Read about gcc debug counters.
* Upgrade laptop to Ubuntu 14.04 LTS
Short week (22nd and 23rd leave (4/10))
== Plan ==
* Bug fixing.
== Issues==
* Waiting on hardware to Benchmark SPEC2006 PGO runs in hardware.
== Issue ==
* None
== Progress ==
* More tests on shrink-wrap changes (TCWG-133, 6/10).
- Cortex-m3 tests exposes a data flow issue: New BB created after
dfinit does not have correct df_lr info.
- Collect Spec2k benchmarks on X86-64 and ARM.
* Linaro 4.9 binaries release (2/10).
- Update Linaro crosstool-ng config and samples to support 4.9.
- Disable parallel build for gcc manual (pdf & html).
- Prebuild binaries for aarch64:
http://cbuild.validation.linaro.org/binaries/4.9-prerelease-2014.04
* Ping aarch64 fcel patch. But still no comments.
* Fix a trunk build fail issue @r209556.
* One day off (2/10).
== Plans ==
* Send the shrink-wrap related patches for review.
* PING the pending patches.
* Investigate move-loop-invariants heuristics.
== Planed leaves ==
* May 1-3: Labour day holiday.
== Progress ==
* Got Jenkins working with matrix builds so we can utilize all LAVA
slaves. (TCWG 1387 - 3/10)
* Write script to take a list of revisions and then build and test
them all and diff the results. (TCWG 448 - 4/10)
* More experimenting with Kugan's benchmarking branch. (1/10)
* Meetings and Misc (2/10)
== Plan ==
* More work on matrix builds. (TCWG 1387)
* More work on regression test analysis and reporting. (TCWG 448)
== Progress ==
* Short week (21st and 25th are public holidays) (4/10)
* TCWG-447 (5/10)
* Implemented and tested fenv target hooks, necessary built-ins and md
patterns
* Posted RFC patches for review for both arm and aarch64
* http://gcc.gnu.org/ml/gcc-patches/2014-04/msg01743.html
* http://gcc.gnu.org/ml/gcc-patches/2014-04/msg01744.html
* TCWG-413 Spec2006 (1/10)
* Finished the set-up
* On hold for now
== Plan ==
* upstream zero/sign extension elimination activities
* start with literal pool merging
== Progress ==
* Holidays (3 days)
- Clearing emails/tasks backlog
- Some post-trip illness
* AArch64 vs. ARM64
- Comparing performance of both back-ends
* Named Register
- Re-implementing after code review
- http://reviews.llvm.org/D3261
* Time
- CARD-1246 4/10
- Others 6/10
== Plan ==
* Finish named register in LLVM, check Clang
* Continue testing and benchmarking ARM64 back-end
* Have a try at CBuildv2
== Progress ==
* Back to work part-time (50%) for the next 3 weeks
* Easter Monday (1/5)
* (No Jira card - 2/5)
Analyzed at GCC trunk validations and reported some regressions.
The current commit rate is very high since stage1, and our compute
farm isn't fast enough. Should be OK after the week-end.
Modified my scripts so that I move these validations out of Jenkins.
* (No Jira card - 2/5)
Continued converting Neon intrinsics tests for inclusion in GCC testsuite.
To make the review easier, I now believe that it would be better that
I convert most of the testsuite first, and rework the patches order.
* Meetings and conf calls(1/5)
== Next ==
* Continue to closely monitor GCC trunk validations
* Neon intrinsics tests
* Off Thursday/Friday next week
=== Previous ===
libvpx NEON intrinsics investigation [TCWG-429 14/10]
. if interested, you can read the report in card 429:
https://cards.linaro.org/browse/TCWG-429?focusedCommentId=22247
tested division optimisation patches against current trunk and pinged
on list [2/10]
good friday bank holiday [2/10]
=== Next ===
Resume looking at NEON scheduling TCWG-135, which ties into libvpx
Upgrade laptop to Ubuntu 14.04 LTS
== Progress ==
* GDB reverse debugging on aarch64
-- Scored through a64 instruction set details. [1/10]
-- Completed testing and implementation of all basic structures.
[TCWG-398] [1/10]
-- Started decoding aarch64 load store instructions for recording.
[TCWG-401] [2/10]
-- Completed decoding of aarch64 branch instructions. [TCWG-400] [2/10]
-- Scored through a32 instruction set and researched on a32 recording.
[2/10]
* Miscellaneous [2/10]
-- UK visa application process
-- Tools cauldron and TCWG sprint bookings etc.
-- Meetings
== Plan ==
* GDB reverse debugging on aarch64
-- Complete decoding of aarch64 load store instructions. [TCWG-401]
-- Completed decoding of aarch64 exception and system instructions.
[TCWG-400]
== Progress ==
PGO - AArch64 (TCWG-179) (4/10)
* Completed SPEC runs -O3 -mcpu=cortex-a57 in chroot + qemu
arm64-saucy with Linaro release source (4.8) march 2014.
* Built Linaro 4.9 GCC branch based tool chain under chroot + qemu arm64-saucy
* Created a PGO config file for aarch64 as a peak configure and use --tune=peak
SPEC gave warning that FDO cannot be used as base flags and will
ignore the flags.
Also setting basepeak = yes gave issues and ran base runs and
setting it to no does both base and peak runs
* Started SPEC2006 runs for PGO under chroot + qemu arm64-saucy
* Perlbench train run failed in qemu.
* Other benchmarks are progressing.
GLIBC Systemtap (2/10)
* Ran glibc make check under qemu. Seen some illegal instructions and
ntpl tests hung in qemu aarch64. Will has completed these tests in
hardware.
Misc (2/10)
* Tested cbuildv2 to build gcc natively under qemu-arm64-chroot-ubuntu-saucy.
git gave issues not able to clone and hangs when downloading
binutils and gdb repos.
* 17th was Public Holiday (2/10)
== Plan ==
* Go through the results for SPEC2006 PGO runs.
* Start functional PGO runs for SPEC2000
* Bug fixing.
== Issues ==
* Toolchain64 disk full
== Progress ==
* Linaro GCC 4.9 2014.04 (3/10)
- Created FSF Linaro 4.9 branch
- Testing FSF 4.9.0 RC
* Launchpad bugs: (3/10)
o LP #1169164 : including signal.h exposes various PSR_MODE #defines
- Two possible ways to fix the issue (to be discussed with maintainers)
* Misc:
o Cbuildv1 baby-sitting (1/10)
o Various meetings (1/10).
== Next ==
- Easter Monday off
- Continue on LP #1169164
- Linaro GCC 4.9 release (when FSF release will be made)
== Week of April 14th ==
- Benchmarked autoprefetcher patches. (TCWG-388, 2/10)
- Continued working on spec2xxx-utils scripts. These are now gaining support for AArch64 and SPEC2006. (TCWG-238, 4/10)
- Various discussions. (2/10)
- Short week due to Good Friday / Easter. (2/10)
== Week of April 21st ==
- Take Tyler B.'s LAVA test definition for SPEC benchmarking and adjust spec2xxx-utils scripts to match LAVA environment.
- Investigate regressions on gap, eon and wupwise from autoprefetcher patches.
-- SPEC2000's gap in particular is sensitive to 1st scheduling pass with register pressure, and, apparently has a 5% performance improvement potential (at -O2 Cortex-A15).
- Various discussions.
- Even shorter week due to Easter Monday and ANZAC Day on Friday. (4/10)
--
Maxim Kuvyrkov
www.linaro.org
== Progress ==
* TCWG-238 (4/10)
- Created scripts and spectools for spec2006 to work with cbuild2
* TCWG-413 Spec2006 (4/10)
- Worked out Spec2006 config and src.alt for v1.1 src we have to work
correctly
- Got it to work natively with Maxim's scripts
- Did a trail run in apm openembedded; Needs ccrypt, tar with xz
compression support and system binutils with -mabi=ilp64 support
* 18th is Public Holiday (2/10)
== Plan ==
Benchmarking and FENV support
== Progress ==
* Attempted to get qemu-aarch64 working, fails on saucy, works on
precise. Did native build, hung during 'make check'. Setup
qemu-aarch64 chroot on all TCWG x86_64 build farm
machines. (4/10)
* Tracked down problem that's kept TCWG Jenkins from
working. (1/10)
* Tried Kugan's benchmarking branch of cbuildv2. (1/10)
* More work on Neon intrinsics tests. (TCWG-322 - 1/10)
* Meetings and Misc (3/10)
- Worked on benchmarking doc with Ryan.
== Plan ==
* Get Jenkins working with matrix builds so we can utilize all LAVA
slaves.
* More experimenting with Kugan's benchmarking branch.
== Issue ==
* None.
== Progress ==
* multilib can not work with multiarch anymore, so change Linaro
crosstool-ng to make 2014.04 release (2/10).
- For 4.8 release, we will revert the change and keep it align with
previous releases.
* Linaro 4.9 toolchain binaries pre-releases (Cross & native) for
aarch64 (2/10):
http://cbuild.validation.linaro.org/binaries/4.9-prerelease-2014.04.
* Try to set up native spec2006 tests on Chromebook. But got "out of
memory" when building the tools (1/10).
* Clean up and test for shrink-wrap enhancement (TCWG-133: 2/10)
* Rebase the aarch64 fcsel patch and testing. (1/10)
* ARM internal work (2/10)
== Plans ==
* Send the shrink-wrap related patches for review.
* PING the pending patches.
== Planed leaves ==
* April 21: Team event.
* May 1-3: Labour day holiday.