Monday 17th June 2013
This month's meetings
<< < 2013 / 1 > >>
- Matthew Gretton-Dann y
- James Elliott
- Christophe Lyon
- Yvan Roux
- Venkat Kumar
- Kugan Vivekarandarajah
- Zhenqiang Chen
LLVM + Libs & Dev Tools + Support & Maintenance:-
- Renato Golin
- Will Newton
- Mans Rullgard
- Omair Javaid
- Peter Maydell (absent)
- Rob Savoye
- Richard Earnshaw (absent)
- Ramana Radhakrishnan (absent)
- Mathias y
- Review activity last week
- Discuss plans for coming week
- Raise any issues holding up progress
Action Items from this Meeting
Action Items from Previous Meeting
- Merges for linaro-4.8-2013.06:
- - investigated why the cross-validation lack libpthread and libdl. - fixed cbuild to make these libs available in the same dir as ld-linux.so - not sure why it wasn't necessary several weeks ago when I first spawn libsanitizer tests: we changed the prebuilt tarball since then, but the libs are still at the same place; the binutils version changed though.
- Re-spawned a merge request for libsanitizer, using the updated cbuild (libpthread/libdl support, qemu flags updated) and the updated gcc-4.8 (with my patch to cope with qemu's output of /proc/self/maps)
- Merges for linaro-4.8-2013.07: updated list
- Aarch64 frame growing downward: started.
- Design spec needs to be written, or implementation specified in Card.
- Disable-peeling: restarting looking at vectorizer.
- Merges for linaro-4.-2013.07: update list
- disable peeling: see how to make the vectorizer less aggressive
- Aarch64 frame growing downward.
- Neon intrinsics/vuzip/veor: resume work
- Book hotel/flight for Connect
- Releases merge reviews:
- - Finished reviews for releases.
- Launchpad bug #1187247:
- - Closed as invalid since the issue was on an attempt to access at an invalid address.
- Cbuild Babysitting:
- - Fixed 4.8 repository update on toolchain64. - Recreate ancestors tarballs and re-spawned jobs - Re-started Calxedas...
- LRA on ARM and AArch64:
- - Resumed the task. - Investigation of AArch64 failure on-going.
- Mainly LRA
- Calxeda instability
- libssp support for AArch64
Libssp support needed AArch64 frame to grow downward. But it is currently defined as growing upward. This work is currently in hold. Needed frame layout changes before restarting. Waiting for feedback from Matt.
- AAarch64 testing
Drilling down Boot strap failure with GCC 4.9 trunk on open embedded image with V8 model. Fails at time xg++ is linking. Not much progress. Set up Cross build and test in V8 model for GCC testsuites. Also looking at increasing timeout for individual tests.
- AArch64 LTO and PGO support.
Started on understanding PGO and LTO runs with AArch64.
- Continue bootstrap testing and push patches to GCC
- Continue LTO and PGO runs for AArch64
- VRP based zero/sign extension
- - Pinged the patch.
- Generate a single call to divmod
- - Builtin based implementation bootstrapped and passes regression. - posted patch and initiated discussion
- - long long is not handled now
- Better end of loop counter optimisation
- - Dropped md patch - Verified that Bin Cheng's RTL change noop_move_p fixes this.
- Respond to patch review
- Set-up X86 benchmarking locally for Spec2k
- Monday, 10 June Public Holiday (4 day week)
- Investigate conditional compare RTL representation.
- - Expand conditional compare to cmp_and/cmp_ior. - Test is ongoing
- Investigate lp: 1189445. Patch is in testing.
- Identify the root cause for lp: 1189448. FSF4.8 does not have the buggy code. Linaro 2013.06 and trunk have fixed it.
- Continue on conditional compare.
- internal release - validation, 2days.
Progress / Announcements / Updates / Comments to team
- Pandaboard - Using GNU ld didn't help the problem, and made compilation time double
- - Reverting to Gold and will keep an eye on it
- Release 3.3 finished - Tested and released, just waiting the final email
- Linaro+ARM sync meeting - Setting priorities, requesting approval for some work items
- Phoronix - Script to run batch mode ready at:
(slp vect) cannot be tested
- - No way yet to share this data internally, working with IT to find a solution
- CBuild - Merged Clang changes, running job again - Autodetecting ARMv7
- LLVM administrativia - many patches to review this week - vectorizer also being enabled by default on -O2 and -Os - testing vectorizer with -Os on chromebook, found no big differences
- Phoronix - Store the base runs somewhere (people.linaro?) and have the script install them to compare with any run on the board.
- Think of a way to put bootstrap & phoronix in LAVA/CBuild (whatever is easier), so I can use the calxeda nodes without worry if they'll be up or down
- Pandaboard - Use GCC 4.8 on linaro-panda-02 and hope to solve the problem
- CBuild - Finish LLVM+Clang change (merge pending) and try to run some benchmarks with it
- PerfDB - Install website locally, try to cook some analysis
LIBS & DEV TOOLS:-
Eglibc, newlibc, binutils - located in releases.linaro.org
- Branched and released binutils, newlib and eglibc.
- Wrote up some release process docs on the wiki.
- Fixes and further investigation of AArch64 ifunc issue.
- Submitted memcpy to upstream gerrit and ran more benchmarks.
- Submitted new fix for gdb frexpl issue.
- Fix AArch64 ifunc issue.
- Fix gdb.thread issue.
- feedback from Qemu patches.
- new revised version of patches this week with mistakes corrected.
- Investigated gdb.dwarf2 unsupported issues wrote two new patches for gdb.dwarf2 testsuite.
- Read some background on dwarf2 to translate test written in x86 assembly into arm assembly.
- Got refusal of visa from Irish embassy. Ran after visa consultants for issues related to visa application.
- Calls with Irish consulate officers for appeal process and causes of refusal.
- Submit ready patches for dwarf2 untested/unsupported problems.
- Translate dwarf2 tests written in x86 to arm asm.
- Chromebook os update on a faster sd card. (Still pending)
- Add JIRA cards for gdb features missing on arm. (Still pending)
- Prepare documentation for future visa applications.
- * made qemu-linaro 2013.06 release
- * remaining VIRT-4 work handed over to other people for now
- * Huawei's AArch64 TCG patches now committed upstream
- * working through mach-virt and virtio patches to see what
- still needs to be done here and in what order
- continue mach-virt/virtio work
- set up a new qemu-linaro tree/branch as our CI/LAVA input [to keep it
- separate from our "we release this" tree]
- restart work on upstreaming omap3 patches as part of my generic qemu
- maintenance work (will reduce our maintenance burden in the long term)
[very short week; two days]
- Did QEMU testing for the release by running several images on a few platforms.
- Did the GCC 4.7-2013.06 release.
- Did the GCC 4.8-2013.06 release.
- Did a spin of the GCC 4.7-2013.06 release.
- Updated benchmark parsing scripts, downloaded historical benchmark data and data from the release testing.
- Updated SQL query and gnuplot scripts to work with new data.
- Produced some benchmark graphs.
- Finally figured out how to spawn remote Cbuildv1 tests and
- Focus on performance and regression test graphs.
- Do builds of FSF GCC branches to compare with Linaro branches.
- Continue improving Cbuildv2 as use it to do all these builds and
Progress / Announcements / Updates / Comments to team
WorkingGroups/ToolChain/Meetings/Archive/2013-06-17 (last modified 2013-08-30 11:47:31)