[Code size investigation]
Results (clang 2% larger than gcc) replicated on cortex-m0 and
cortex-m4 on Zephyr.
- Clang optimisation to use BLX rather than BL when same function
called multiple times is a pessimisation on Zephyr, especially on M0.
- GCC register allocation seems to result in fewer spills
TODO: Get an estimate of how much code-size difference is down to
different inlining decisions.
On CMSIS DSP cortex-m4f clang appears to be producing smaller than
GCC, not measured averages yet.
[LLD]
- Quite a few upstream reviews, PRs and investigations surrounding them.
- Likely that LLD will be converting to the new variable naming convention.
- Received a request to add cortex-a8 erratum fix for Google Android team.
[Linaro Connect]
Registered and contacted travel.
Drafted a submission for presentation, will submit next week.
Planned Absences:
On holiday Wednesday, Thursday, Friday next week
Progress:
* VIRT-65 [QEMU upstream maintainership]
+ catching up with email and code review -- managed to get some
series reviewed in time for softfreeze on Tuesday, notably the
'sbsa-ref' reference platform model that Hongbo Zhang was working on
+ a lot of release-herding, working through the huge pile of pull
requests that need merging
+ fixed a silly bug in recent VFP refactoring, spotted by coverity
+ fixed a memory leak that broke our CI sanitizer build (not a new
piece of code, but we currently only sanitize the x86-64 targets
and a recent change meant this old code is now used on x86-64 for
the ATI PCI display device model)
* Misc:
+ first KVM Forum Programme Committee meeting (and attendant
review of all the submitted abstracts; bumper crop this year)
thanks
-- PMM
== Progress ==
* GCC:
- FDPIC: No progress, still waiting for feedback
- noinit attribute: reviewers asked to make it a generic attribute,
rather than target-specific. New patch sent.
* GNU-583 (Fix Linux kernel built for Thumb-2 with GCC using LTO)
- no progress this week
* GCC upstream validation:
- reported a few regressions.
* Binutils:
- PR24709 (linker crash and assertion failure with CMSE). Use case not
considered worth the headache of supporting correctly CSME+long-branch
stubs (tricky to get right). Replaced the linker crash with a user
error message.
- Non-contiguous memory regions support in the BFD linker: Received a
good summary of the consensus reached in 2017.
* misc:
- infra fixes / troubleshooting
== Next ==
GCC:
- handle feedback on FDPIC and noinit patches
- binutils/linker support for non-contiguous memory regions
- GNU-583
- GCC upstream validation: Add a config for cortex-m33 (v8-m)
== Holidays ==
July 13-27
Aug 2-11
== This Week ==
* PR88833
- Fixed pending issues with x86 and committed fix to trunk
* PR90723
- Issue seems to be infinite recursion overflowing the stack, investigating.
* Misc
- Meetings
== Next Week ==
- PR90723
- Add testsuite comparison to tcwg_gnu
== Progress ==
* LLVM SPEC2k6 Performance Analysis [LLVM-134]
- Still working around perf version mismatches, going to investigate
if we can use a newer version of perf to collect data
- Had a look at the assembly for sphinx from gcc-6, clang-3.9.1 and
clang-8.0.0, but going to wait for better perf before I rush to
conclusions
* IR SVE Reviews [ LLVM-545]
- Had a look at the clang patches, gave some feedback for one of
them; the other ones are very subtle and best left to the clang
maintainers
* [GlobalISel] Refactor CallLowering [LLVM-568]
- Fixed a little ASAN snag, should be done for good now
* Trying to get up an ABI fuzzer for GlobalISel
== Plan ==
* Update benchmarking infrastructure (in support of LLVM-134)
* Deprioritize GlobalISel
* Upcoming vacation: 6 - 13 August
Short week; 3 days.
[VIRT-263 # ARMv8.1-VHE Virtual Host Extensions ]
Fixed 3 bugs:
* SVE length calculation,
* PNX bit while in EL2&0 regime,
* Interrupt routing w/ TGE bit.
A bare 5.2 kernel boots to root file system not found.
A 4.19 kernel hangs during boot somewhere.
Now doing a fedora30 install, which I believe has a 5.x kernel...
[VIRT-327 # Richard's upstream QEMU work ]
Patch review:
* GSoC x86 risugen,
* arm semihost cleanups.
r~
== This Week ==
* PR88833 (6/10)
- Patch approved by Richard.
- Regresses pr88152.C due to a possibly latent issue with combine.
* PR90722 (3/10)
- Investigating the issue
* Misc (1/10)
- Meetings
o LLVM:
* 8.0.1-rc3: Built ARM Binaries, AArch64 on-going.
* Machine Outliner:
- Uncovered issues in stack fixup handling, working on a fix.
o Misc
* Various meetings and discussions.
[VIRT-263 # ARMv8.1-VHE Virtual Host Extensions ]
Rebased on master. Now trying to remember the incantation that produced a
minimum number of insns before the kernel actually tries to use this. At
present things seem to be crashing before I even get that far, as if I've
misconfigured something.
[VIRT-327 # Richard's upstream QEMU work ]
Fixed a couple of bugs in my tcg/ppc host vector patch set.
Reviewed qemu kvm sve patches.
Reviewed target/ppc altivec optimization patches.
Reviewed GSoC risugen x86 patches.
Other misc upstream review.
r~
QEMU Tooling ([VIRT-252])
=========================
QEMU plugin support ([VIRT-280])
- more work on [v4 branch] but cutting it too fine for 4.1
- reworking the memory tracing to track mmu_idx
- hope to have v4 posted on Monday if I can squash the bugs
- the plugin call isn't getting the full TCGMemopidx (maybe only
TCGmemop?)
- some interest on list from HW manufacturers
[VIRT-280] https://projects.linaro.org/browse/VIRT-280
[v4 branch] https://github.com/stsquad/qemu/tree/plugins/plugins-v4
GSoC Mentoring ([VIRT-348])
- reviewed {Qemu-devel} {PATCH v2 0/4} dumping hot TBs Message-Id:
<20190624055442.2973-1-vandersonmr2(a)gmail.com>
- worked up some [suggestions for HMP interface and refactoring]
- first evaluation period work
[VIRT-348] https://projects.linaro.org/browse/VIRT-384
[suggestions for HMP interface and refactoring]
https://github.com/stsquad/qemu/tree/review/hotblocks-v2-tweaks
Upstream Work ([VIRT-109])
==========================
- posted {PULL 00/19} testing/next (tests/vm, Travis and hyperv build
fix) Message-Id: <20190624134337.10532-1-alex.bennee(a)linaro.org>
- bit late, will wait until pm215 returns
[VIRT-109] https://projects.linaro.org/browse/VIRT-109
Other
=====
- booking flights for Connect
- submitted a talk for KVM Forum (plugins)
Completed Reviews [2/2]
=======================
{Qemu-devel} {PATCH v2 0/4} dumping hot TBs
Message-Id: <20190624055442.2973-1-vandersonmr2(a)gmail.com>
- CLOSING NOTE [2019-06-27 Thu 12:39]
Made a bunch of notes and [tweaks]
[tweaks]
https://github.com/stsquad/qemu/tree/review/hotblocks-v2-tweaks%0A
{PATCH} Makefile: Rename the 'vm-test' target as 'vm-help'
Message-Id: <20190531064341.29730-1-philmd(a)redhat.com>
- CLOSING NOTE [2019-06-27 Thu 12:40]
Queued to my tree
Absences
========
- June 21st
Current Review Queue
====================
* {Qemu-devel} {PATCH 0/4} Introduce the microvm machine type
Message-Id: <20190628115349.60293-1-slp(a)redhat.com>
* {PATCH 0/3} tests/acceptance: Add tests for the Leon3 board
Message-Id: <20190627115331.2373-1-f4bug(a)amsat.org>
* {PATCH 0/5} tests/acceptance: Add bFLT loader linux-user test
Message-Id: <20190625101524.13447-1-philmd(a)redhat.com>
* {PATCH v2 0/9} KVM: arm/arm64: vgic: ITS translation cache
Message-Id: <20190611170336.121706-1-marc.zyngier(a)arm.com>
* {Qemu-devel} {RFC PATCH 0/7} Proof of concept for Meson integration
Message-Id: <1560165301-39026-1-git-send-email-pbonzini(a)redhat.com>
* {PATCH 00/59} KVM: arm64: ARMv8.3 Nested Virtualization support
Message-Id: <20190621093843.220980-1-marc.zyngier(a)arm.com>
--
Alex Bennée
== Progress ==
* GCC:
- FDPIC: No progress, still waiting for feedback
- noinit attribute: no feedback yet
* GNU-583 (Fix Linux kernel built for Thumb-2 with GCC using LTO)
- managed to reproduce similar error messages
* GCC upstream validation:
- reported a few regressions.
* Binutils:
- PR24709 (linker crash and assertion failure with CMSE). CMSE stubs
do not support long branches. Tried two approaches, but didn't find a
solution yet
- resurrected a thread about non-contiguous memory regions support in
the BFD linker
* misc:
- reported regressions in qemu, found with the GCC testsuite
- infra fixes / troubleshooting
== Next ==
GCC:
- handle feedback on FDPIC and noinit patches
- binutils/PR24709
- GNU-583
== Progress ==
* [GlobalISel] Refactor CallLowering [LLVM-568]
- Committed upstream
* [ARM GlobalISel] Add support for integers > 32 bits wide [LLVM-310]
- In progress
* LLVM SPEC2k6 Performance Analysis [LLVM-134]
- Got some results with clang-3.9.1 and clang-8.0.0, trying to work
around a perf annotate issue so I can investigate
== Plan ==
* Continue with these
* Friday off
== Progress ==
* FDPIC
- GCC: No progress, still waiting for feedback
* GCC upstream validation:
- reported a few regressions.
* GCC:
- noinit attribute: no feedback yet
* Binutils:
- started looking at PR24709 (linker crash and assertion failure with CMSE)
* misc:
- ABE: pushed fix to glibc-2.29+ builds, tcwg-backport job now works again
- forwarded GCC/Linaro bug #5314 to upstream; quickly fixed by richi
- looked at a couple of old Jira cards. Not sure how to resume work on
GNU-583 (Fix Linux kernel built for Thumb-2 with GCC using LTO)
== Next ==
GCC:
- handle feedback on FDPIC and noinit patches
- binutils/PR24709
o 4 days week (1 training day)
o LLVM:
* 8.0.1-rc2: ARM and AArch64 binaries built and uploaded.
* Buildbots babysitting: couple of issues reported.
* Machine Outliner:
- Adding testcases before upstream submission.
o Misc
* Various meetings and discussions.
== Progress ==
* Out of office on Friday (bank holiday)
* [GlobalISel] Refactor CallLowering [LLVM-568]
- Patches upstream
* [ARM GlobalISel] Add support for integers > 32 bits wide [LLVM-310]
- Started looking into call lowering for 64-bit types
* LLVM SPEC2k6 Performance Analysis [LLVM-134]
- Trying to reproduce results from Connect, hit a little snag with Jenkins
== Plan ==
* Continue with these
[LLVM-542] Build Zephyr with clang
- Spent quite a bit of time working out why a Clang built zephy hello
world wouldn't boot, tracked down to a missing clobber list in an
inline assembly block
- Wrote some scripts to collect code size information on the samples.
Some initial figures on mostly cortex-m3 put llvm -Oz trunk about 2%
larger than arm-none-eabi-gcc (9.1.1) -Os with frame pointers
disabled. The samples are making very little use of the library
(newlib built with arm-none-eabi-gcc).
- Working out which samples will build with cortex-m0.
- Investigated latest version of bloaty mc bloat face a code size tool
from Google. Has some interesting features including an inline
detection feature that can map a portion of a function's code size to
inlined functions.
Misc:
LLD reviews and mailing list comments.
Planned Absences:
Likely to take some holiday around 13th July
Week ending 23 June is very short.
* Patch review
- target/ppc vsr cleanup
- target/tricore translator loop conversion
- target/arm vfp decodetree cleanup
- cortex-strings strrchr fix
- continuing on the plugin api
* Xilinx meeting
* Fix qemu assert for clyon
- Found two other bugs in the process.
* Debugging my own USHR/SSHR patch vs aa32.
r~
Progress: (very short week, 2 days)
* VIRT-65 [QEMU upstream maintainership]
+ found and sent patches for a handful of minor M-profile bugs
* VIRT-268 [QEMU support for dual-core Cortex-M Musca board]
+ the final patches for this have now gone upstream and I have
marked the JIRA issue as resolved. There may still be minor
bug fixes but we can handle those under the usual 'upstream
maintainership' JIRA
thanks
-- PMM
Hello folks,
I got a few people asking me to do this in the last Connect, so I've
proposed a beginner session that explores gcc under the hood. The
tentative plan I have for the talk is:
1. A high level view of how the source code is laid out
2. Front end, middle end, backend. This includes a high level
introduction of GIMPLE and machine descriptions
3. A walkthrough of one or two simple programs and usage of diagnostic
flags like -fdump-*
Additional suggestions are most welcome. Also, I was thinking maybe
it would be good to have a llvm under the hood talk along similar
lines. Thoughts?
Siddhesh
[LLVM-571] Build GNU rmprofile toolchain with Linaro scripts (abe)
The existing build-system was only set up to build the A-profile bare
metal toolchain. Managed to find right combination of flags and
modifications to get a toolchain that zephyr can use.
[LLVM-158] Buildbot maintenance
An interesting failure introduced in LLD, but causing segfaults in
2-stage build, now fixed.
[LLVM-542] Zephyr code size investigation
- Rebased modifications to Zephyr
- Wrote script to build all the examples with GCC and Clang
- Fixed problems with modifications found by building all the examples
- Clang built helloworld no longer booting, need to investigate
- Found some areas for more investigation:
-- llvm-objcopy missing --gap-fill (used by one of the sample programs)
-- lld missing --print-memory-usage, while I'm using gcc for the main
link, zephyr build system seems to be feature testing using clang bare
metal default linker (lld)
-- clang always generates .note.GNU.stack, gcc embedded does not,
leading to lots of orphan section warnings. Probably best solved by
linker script modification.
Progress:
* VIRT-65 [QEMU upstream maintainership]
+ some work on getting Sphinx to generate manpages, with a conversion
of the qemu-ga manpage from texinfo as the demonstration case
+ set up to run pre-merge tests on our packet.net qemu-test machine
rather than the gcc compile farm one (as the latter is running an
ancient ubuntu whose python is now too old to build QEMU with)
+ sent some follow-on cleanup patches now that VFP decodetree is in master
* VIRT-268 [QEMU support for dual-core Cortex-M Musca board]
+ sent out patches which correctly make the Cortex-M33 (and the -M4)
implement single-precision-only floating point, so the double-precision
instructions UNDEF as they should
+ once those and the other on-list patches have made their way through
code review and into master this epic will be complete
thanks
-- PMM
