Hi,
I am looking at best approach for
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43721 - Failure to optimise
(a/b) and (a%b) into single __aeabi_idivmod call in ARM architecture
In sumary, the following c code results in __aeabi_idivmod() call and
one __aeabi_idiv() call even though the former already calculates the
quotient.
int q = a / b;
int r = a % b;
return q + r;
My question is what would be the best way to handle it. As I see there
are few options with some issues.
1. Handling in gimple level, try to reduce the operations to equivalent
of this. We should do this for the targets without integer divide.
{q, r} = a % b;
Gimple assign stmts have only one lhs operation (?). Therefore, lhs has
to be made 64bit to signify return values of R0 and R1 returned
together. I am not too sure of any implications on other architectures here.
2. Handling in expand_divmod. Here, when we see a div or mod operation,
we will have to do a linear search to see if there is a valid equivalent
operation to combine. If we find one, we can generate __aeabi_idivmod()
and cache the result for the equivalent operation. As I see, this can
get messy and might not be acceptable.
3. An RTL pass to process and combine these library calls. Possibly
using cse. I am still looking at this.
4. Ramana tried a prototype to do the same using target pattens. He has
ruled this out. (if you want more info, please refer to at
https://code.launchpad.net/~ramana/gcc-linaro/divmodsi4-experiments)
Any suggestion for best way to handle this?
Thanks,
Kugan
Hi all,
I am facing build error, when I try to bootstrap GCC trunk for native
aarch64-unknown-linux-gnu configuration in openemedded/V8 model.
Linker error occurs while building stage 1 GCC.
(Snip)
/usr/lib/gcc/aarch64-oe-linux/4.7.3/../../../../aarch64-oe-linux/bin/ld:
gcov: hidden symbol `__deregister_frame_info' in
/usr/lib/gcc/aarch64-oe-linux/4.7.3/../../../aarch64-oe-linux/4.7.3/libgcc_eh.a(unwind-dw2-fde-dip.o)
is referenced by DSO
/usr/lib/gcc/aarch64-oe-linux/4.7.3/../../../../aarch64-oe-linux/bin/ld:
final link failed: Bad value
collect2: error: ld returned 1 exit status
(Snip)
The steps to reproduce the issue is attached.
Need some help to solve this.
regards,
Venkat.
== Progress ==
* Neon intrisincs: compiled my testsuite with GCC/trunk and filed
bugzilla 57431 for ICE.
* Merges for linaro-4.8-2013.06: started actual merges
- fixed a cbuild reporting problem
- faced calxeda and e2c problems, fixed by Matt.
* Jira: a few updates
* Libsanitizer: patched upstream, to be backported in GCC/trunk, then
in gcc-linaro-4.8.
== Next ==
* Merges for linaro-4.8-2013.06: complete them.
* Disable-peeling: resume
* Look at Kugan question about trunk regression
* Libsanitizer/aarch64: resume
* PGO/LTO/python bug: resume
* Neon intrinsics/vzup/vero: resume
== Progress ==
* Investigated attach to process and threads failures
(TCWG-95<http://cards.linaro.org/browse/TCWG-95>).
Almost all problems were fixed with any patch by making an os configuration
change on chromebook.
* Ran GDB test suites after OS configuration change on chromebook and
updated test results.
* Investigated inline-break.exp test suite failures on arm. GDB seems to be
missing one inlined instance on two different functions. Obtained debug
info and dumps of obj file without any luck to find the possible cause and
fix.
* Started work on integration of different testing scripts. Initially added
commandline options in bash script to test let user choose target/host,
native-none/native-gdbserver/remote, and testsuite/testcase configurations.
== Plan ==
* Figure out a reason/solution for inline-break.exp failures on arm.
* Investigate and Fix more ARM bugs shown in gdb 7.6 testsuite results.
* Chromebook os update on a faster sd card.
* Side activity work on automation of testing gdb in different
configurations and uploading comparison result.
== Issues ==
* None.
== Progress ==
* LRA on ARM and AArch64:
- Tried to workaround the issue with new insns (reload_outdi,...).
- But it doesn't seem to be handle by LRA.
- Debug still ongoing.
* Internal meetings
== Plan ==
* Releases merge reviews
* LRA
== Progress ==
Very short week (monday and wednesday off).
* AARCH64 testing
Got boot strap failure with GCC 4.9 trunk on open embedded image with
glibc changes. Retired with latest openembedded image on V8 model
Noted down the steps to reproduce boot strap failure in the model for
broadcasting.
* libssp support for Aarch64
Read documentation to understand and evaluate the work.
== Plan ==
* Continue bootstrap testing and push patches to GCC
* Implement Libssp GCC back end hooks.
* Linaro connect travel prep, book tickets hotel and apply visa.
== Progress ==
* VRP based zero/sign extension
- Tested and posted the latest patch
* Better end of loop counter optimisation
- Tree level optimization are optimized in mainline
- Christophe noted a slight change in asm generated from earlier
version
- tracked down the patch causing this and communicated this.
* Generate a single call to divmod
- Looked at expand_divmod to understand how __aeabi_idiv and
__aeabi_idivmod are generated.
== Plan ==
* Better end of loop counter optimisation
- Change the pattern to remove this additional instruction if
necessary.
* Generate a single call to divmod
- Come up with a solution
== Issues ==
* None
== Progress ==
* Shrink wrapping improvement (TCWG-133)
- Call copyprop to optimized the parameter register move instructions.
- Test is ongoing.
* Update aarch64-none-elf toolchain newlib version to 2.0~20130530 (lp:1185711).
== Plan ==
* Collect performance data for TCWG-133.
* Continue on conditional compare.
Best Regards!
-Zhenqiang
== Progress ==
* Monday holiday here too, so went mountain climbing
* Finally fixed Lava token for cbuild
* Dug into EC2 issues, still confused
* Cloned infrastructure files from gcc.gnu.org, which turned out to be
out of date, so updated them to the current releases. That then
required fixing a few minor configure bugs in libppl. Files in
~/cbuild/var/snapshots/infrastructure. Infrastructure libs are
statically linked so they're consistent across platforms.
* Worked on cbuildv2, it now downloads a tarball or does a
bzr|git|svn checkout, configures, compiles, and installs them in
a sysroot as other packages depend on them. Also added a function
to download, build, and install all the toolchain infrastructure
libraries in a sysroot for binutils & GCC to use at build time
https://git.linaro.org/gitweb?p=people/rsavoye/cbuild2;a=summary
* Automated import of all historical test results till running.
Currently
the DB has 137,687,996 test results, 1909 separate test runs of GCC,
and covering 544 versions.
== Plan ==
* Add remote SSH support to DejaGnu for remote testing on
Chromebook. (carried over from this week)
* Figure out the EC2 mess and document it better
* Clone gcc_release.sh script but enhance so a similar process can be
applied to releasing Eglibc, Newlib, and the Binutils.
* Start gearing up for the releases
* Finish cross build and testing support in Cbuildv2, which would
basically eliminate the need for crosstool-ng.
== Progress ==
* 4 day week due to bank holiday.
* Merged binutils arm ifunc changes onto stable branch.
* Worked on getting some git mirrors setup for various toolchain deliverables.
* A certain amount of planning and JIRA card work in preparation for
the next iteration.
* Fixed aarch64_be-* and aarch64-elf testsuite issues with AArch64 ifunc patch.
* Investigated glibc malloc and alternatives.
== Issues ==
* None.
== Plan ==
* Submit new AArch64 ifunc patch after testing complete.
* Finish getting git mirrors setup.
* Look at some gdb testsuite failures.
--
Will Newton
Toolchain Working Group, Linaro
