Find a file
Tom de Vries ab3f4b27ab [omp, ftracer] Don't duplicate blocks in SIMT region
When running the libgomp testsuite on x86_64-linux with nvptx accelerator on
the test-case included in this patch, we run into:
...
FAIL: libgomp.fortran/pr95654.f90 -O3 -fomit-frame-pointer -funroll-loops \
  -fpeel-loops -ftracer -finline-functions  execution test
...

The test-case is a minimal version of this FAIL:
...
FAIL: libgomp.fortran/pr66199-5.f90 -O3 -fomit-frame-pointer -funroll-loops \
  -fpeel-loops -ftracer -finline-functions  execution test
...
but that one has stopped failing at commit c2ebf4f10d "openmp: Add support
for non-rect simd and improve collapsed simd support".

The problem is that ftracer duplicates a block containing GOMP_SIMT_VOTE_ANY.

That is, before ftracer we have (dropping the GOMP_SIMT_ prefix):
...
bb4(ENTER_ALLOC)
*----------+
|           \
|            \
|             v
|             *
v             bb8
*<------------*
bb5(VOTE_ANY)
*-------------+
|             |
|             |
|             |
|             |
|             v
|             *
v             bb7(XCHG_IDX)
*<------------*
bb6(EXIT)
...

The XCHG_IDX internal-fn does inter-SIMT-lane communication, which for nvptx
maps onto shfl, an operator which has the requirement that the warp executing
the operator is convergent.  The warp diverges at bb4, and
reconverges at bb5, and does not diverge by going to bb7, so the shfl is
indeed executed by a convergent warp.

After ftracer, we have:
...
bb4(ENTER_ALLOC)
*----------+
|           \
|            \
|             \
|              \
v               v
*               *
bb5(VOTE_ANY)   bb8(VOTE_ANY)
*               *
|\             /|
| \  +--------+ |
|  \/           |
|  /\           |
| /  +----------v
|/              *
v               bb7(XCHG_IDX)
*<--------------*
bb6(EXIT)
...

The warp diverges again at bb5, but does not reconverge again before bb6, so
the shfl is executed by a divergent warp, which causes the FAIL.

Fix this by making ftracer ignore blocks containing ENTER_ALLOC, VOTE_ANY and
EXIT, effectively treating the SIMT region conservatively.

An argument can be made that the test needs to be added in a more
generic place, like gimple_can_duplicate_bb_p or some such, and that ftracer
then needs to use the generic test.  But that's a discussion with a much
broader scope, so I'm leaving that for another patch.

Bootstrapped and reg-tested on x86_64-linux.

Build on x86_64-linux with nvptx accelerator, tested with libgomp.

gcc/ChangeLog:

	PR fortran/95654
	* tracer.c (ignore_bb_p): Ignore GOMP_SIMT_ENTER_ALLOC,
	GOMP_SIMT_VOTE_ANY and GOMP_SIMT_EXIT.

libgomp/ChangeLog:

2020-10-05  Tom de Vries  <tdevries@suse.de>

	PR fortran/95654
	* testsuite/libgomp.fortran/pr95654.f90: New test.
2020-10-05 08:53:11 +02:00
config Fixup config/ChangeLog. 2020-09-10 10:17:51 +02:00
contrib Daily bump. 2020-09-22 00:16:31 +00:00
fixincludes Daily bump. 2020-10-04 00:16:21 +00:00
gcc [omp, ftracer] Don't duplicate blocks in SIMT region 2020-10-05 08:53:11 +02:00
gnattools
gotools libgo: update to Go1.14beta1 2020-01-21 23:53:22 -08:00
include Daily bump. 2020-09-25 00:16:27 +00:00
INSTALL
intl intl: Unbreak intl build with bison 3 when no regeneration is needed [PR92008] 2020-04-16 11:55:00 +02:00
libada Add `--with-toolexeclibdir=' configuration option 2020-01-24 11:24:25 +00:00
libatomic Daily bump. 2020-09-28 00:16:21 +00:00
libbacktrace Daily bump. 2020-09-29 00:16:30 +00:00
libcc1 Daily bump. 2020-10-03 00:16:25 +00:00
libcpp Daily bump. 2020-09-27 00:16:24 +00:00
libdecnumber Daily bump. 2020-07-31 00:16:26 +00:00
libffi Daily bump. 2020-09-25 00:16:27 +00:00
libgcc Daily bump. 2020-10-02 00:16:27 +00:00
libgfortran Daily bump. 2020-09-29 00:16:30 +00:00
libgo compiler: set varargs correctly for type of method expression 2020-10-01 16:10:17 -07:00
libgomp [omp, ftracer] Don't duplicate blocks in SIMT region 2020-10-05 08:53:11 +02:00
libhsail-rt Add `--with-toolexeclibdir=' configuration option 2020-01-24 11:24:25 +00:00
libiberty Daily bump. 2020-09-25 00:16:27 +00:00
libitm Daily bump. 2020-10-02 00:16:27 +00:00
libobjc Daily bump. 2020-05-30 00:16:27 +00:00
liboffloadmic Add missing ChangeLog entries. 2020-05-05 16:10:13 +02:00
libphobos Daily bump. 2020-09-11 00:16:28 +00:00
libquadmath Daily bump. 2020-05-30 00:16:27 +00:00
libsanitizer Daily bump. 2020-09-22 00:16:31 +00:00
libssp Daily bump. 2020-05-30 00:16:27 +00:00
libstdc++-v3 Daily bump. 2020-10-03 00:16:25 +00:00
libvtv Daily bump. 2020-05-30 00:16:27 +00:00
lto-plugin Daily bump. 2020-09-11 00:16:28 +00:00
maintainer-scripts Daily bump. 2020-05-30 00:16:27 +00:00
zlib Daily bump. 2020-05-30 00:16:27 +00:00
.dir-locals.el
.gitattributes Add *.md diff=md. 2020-01-15 14:29:53 +01:00
.gitignore Add .clang-tidy to git ignored. 2020-06-12 16:03:02 +02:00
ABOUT-NLS
ar-lib
ChangeLog Daily bump. 2020-09-19 00:16:29 +00:00
ChangeLog.jit
ChangeLog.tree-ssa
compile
config-ml.in MSP430: Add -fno-exceptions multilib 2019-12-11 19:19:50 +00:00
config.guess Update config.sub and config.guess. 2019-09-09 11:14:32 +02:00
config.rpath
config.sub Update config.sub and config.guess. 2019-09-09 11:14:32 +02:00
configure configure: Also check C++11 (flags) for ${build} compiler not only for ${host} 2020-08-20 21:59:00 +02:00
configure.ac configure: Also check C++11 (flags) for ${build} compiler not only for ${host} 2020-08-20 21:59:00 +02:00
COPYING
COPYING.LIB
COPYING.RUNTIME
COPYING3
COPYING3.LIB
depcomp
install-sh
libtool-ldflags
libtool.m4 [ARM/FDPIC v6 02/24] [ARM] FDPIC: Handle arm*-*-uclinuxfdpiceabi in configure scripts 2019-09-10 09:37:00 +02:00
ltgcc.m4
ltmain.sh Do not use HAVE_DOS_BASED_FILE_SYSTEM for Cygwin. 2020-04-17 09:22:51 +02:00
ltoptions.m4
ltsugar.m4
ltversion.m4
lt~obsolete.m4
MAINTAINERS MAINTAINERS: Add myself for write after approval. 2020-09-18 15:31:46 +01:00
Makefile.def Merge top-level configury changes from gdb 2020-04-09 06:52:55 -06:00
Makefile.in Merge top-level configury changes from gdb 2020-04-09 06:52:55 -06:00
Makefile.tpl Makefile.tpl (HOST_EXPORTS): Add CXX_FOR_BUILD. 2019-08-23 15:37:22 -06:00
missing
mkdep
mkinstalldirs
move-if-change
multilib.am
README
symlink-tree
test-driver
ylwrap

This directory contains the GNU Compiler Collection (GCC).

The GNU Compiler Collection is free software.  See the files whose
names start with COPYING for copying permission.  The manuals, and
some of the runtime libraries, are under different terms; see the
individual source files for details.

The directory INSTALL contains copies of the installation information
as HTML and plain text.  The source of this information is
gcc/doc/install.texi.  The installation information includes details
of what is included in the GCC sources and what files GCC installs.

See the file gcc/doc/gcc.texi (together with other files that it
includes) for usage and porting information.  An online readable
version of the manual is in the files gcc/doc/gcc.info*.

See http://gcc.gnu.org/bugs/ for how to report bugs usefully.

Copyright years on GCC source files may be listed using range
notation, e.g., 1987-2012, indicating that every year in the range,
inclusive, is a copyrightable year that could otherwise be listed
individually.