Note1: working for Linaro every other week
=== Issues ===
* IT issues made me loose 1 day of work
=== Work done during this LLVM+GCC week ===
* LLVM PR34170 (crashes with inline asm with double output operand in GPR):
internal review
+ improved patch to also support case with several output operands
-> had to choose to relax test a bit to not put expectation on
register allocator
* Support arithmetic on FileCheck regex variables: in progress
+ an attempt to be able to express that 2 registers are related by some
arithmetic expression (eg. vmov dY, rX, rX+1), seems doable
Upstream work on Arm GCC side:
cmse_nonsecure_caller return wrong answer: backported to GCC 7
* __builtin_arm_set_fpscr crashes with literal values input: committed to
trunk, backport oked
+ do more testing and prepare backport to GCC 6 & 7
* PR85344 (GCC fails constraint check on valid code): in progress
+ 2 prototype patches done but inconclusive result
+ pondering what should be the expect behaviour
+ investigating the extent of the problem -> both immediate and register
constraints can be surprising
=== Plan for next week (week 12) ===
* Investigate stack-protector failure with Linaro GCC 7 toolchains
* DSGHACK-25 (Support arithmetic on FileCheck regex variable): finish patch
=== Planned leave ===
* 28th April -> 29th May Sabbatical (Trip to France, then China, then
France again)
Progress:
* VIRT-65 [QEMU upstream maintainership]
+ sent patch fixing problem where some aarch64 binaries would segfault
under linux-user (changes for SVE accidentally caused them to get
a too-small signal frame)
+ wrote an RFC kernel patch which sanitizes the ESR value reported to
userspace for faults on kernel addresses, rather than dropping it
entirely
+ wrote a patch to fix a reported QEMU compile failure for Arm
guests on MIPS hosts
+ code review:
- various bug fixes for 2.12
- another round of the patchset to implement the PMU properly
- DS1338 RTC device emulation overhaul patches
- QEMU side of changes to allow more than 123 CPUs with KVM
(mostly involving allowing split GIC redistributor regions)
- linux-user FDPIC ABI support
Hopefully next week I'll be able to get back to non-release
related work...
thanks
-- PMM
* Short week: off Mon/Tue/Wed
== Progress ==
* GCC
- FDPIC
- binutils: patch series approved, waiting for clarification on target name
- GCC: cleaning patch series, fixing lto problems
* GCC upstream validation:
* Infrastructure:
- Linux kernel + lava failing: still failing, it seems the fix
hasn't been merged yet
- TCWG-1335: GCC validation results now sent to gcc-testresults
- looked at some random problems (?) in our GCC validation
* misc (conf-calls, meetings, emails, ....)
== Next ==
* FDPIC:
- Handle feedback on binutils and QEMU patches
- GCC continue cleanup
* GCC upstream validation
Hi Team,
Could you please let me know which version of toolchain released from Linaro starts to support LTO? Thanks
David
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.
[TCWG-1375] Make sure MCSubtargetInfo is passed through to
ARMAsmBackend functions
Spent most of the week working on this and follow up patches. Have
submitted a patch for the most important functions. Nearly have one
ready for the correct Nop encoding, this is more complicated as Nops
can be used as padding.
[Plans]
In office Monday Tuesday only.
Progress [TCWG-1375]
Planned Absences:
- At ACCU conference Wednesday - Friday [Holiday]
- At Euro LLVM following Monday Tuesday
[VIRT-198 # QEMU: SVE Emulation Support ]
Cleaning up patches and fixing RISU failures.
Most of these have thankfully been errors in
last week's first pass reordering of the
sve_access_check. Typically wrt when we find
the address of a register vs the location of
the check vs assert_fp_access_checked.
[Upstream]
Examined a bug affecting aa64 guest on ppc64 host,
but only when using gcc 4.8.5 and not gcc 7.2.
Fixed; in the end a minor typo on the code,
but highly annoying to track down.
r~
o 1 day off
== Progress ==
o LLVM
* Buildbots babysitting:
- Investigate LNT failure due to virtualenv and pip versions (Now fixed).
* Working on LLVM benchmarking job
o Misc
* Various meetings and discussions.
Progress: (short week, 2 days)
* VIRT-65 [QEMU upstream maintainership]
+ email/code review/etc catchup following holiday
+ the usual run of admin work for the release process for rc2
+ investigated a problem where in 'icount' mode we weren't
firing timer interrupts in a timely manner; with assistance
from Paolo, tracked down the cause and sent a patch
(https://bugs.launchpad.net/qemu/+bug/1754038)
+ investigated a problem with v7M emulation reported by the LITE team;
this turned out to be a combination of a known limitation combined
with a badly worded warning message; sent QEMU patch which rewords
the message so it's clearer that it's telling you that you're running
into the limitation.
+ tracked down why aarch64 guest binaries run in icount mode on
a 32-bit host misbehave; I have a working but slightly ugly patch
which I'll send out next week unless somebody has a neater approach
thanks
-- PMM
