Test not only for -march support, but also the actual
presence of the intrinsic routines. Not sure why, but gcc
4.8.5 passes the first but not the second.
Fixes: http://tracker.ceph.com/issues/19386
Signed-off-by: Dan Mick <dan.mick@redhat.com>
ARMv8 defines PMULL crypto instruction.
This patch optimizes crc32c calculate with the instruction when
available rather than original linear crc instructions.
ceph crc32c performance unit test shows that the optimization get
~ x3.90 speedups on ThunderX ARM Core@2.0GHz (Cavium)
~ x1.45 speedups on ARM Cortex-A57@2.1GHz (Huaiwei)
~ x1.16 speedups on ARM Cortex-A57@2.0GHz (Softiron)
Jira: ENTLLT-358
Change-Id: I657422cd20c9ca78237cd060210a5383f4122575
Signed-off-by: wei xiao <wei.xiao@linaro.org>
The used compiler may or may not be recent enough to recognize the
crc32 extended cpu type. However, it does not really have to know about
them either, since all we do is pass inline assembly instructions to
the assembler.
This patch moves the crc cpu extension detection from compiler based
to assembler based, so that we can build optimized code even when the
compiler does not know about the cpu type yet.
Fixes: http://tracker.ceph.com/issues/17516
Signed-off-by: Alexander Graf <agraf@suse.de>
Resurrected jerasure_generic, jerasure_sse3, jerasure_sse4, jerasure_neon,
shec_generic, shec_sse3, shec_sse4 and shec_neon. These all are exact
copies of the new jerasure and shec plugins that support SIMD detection.
Moved EC preload code in ceph-mon and ceph-osd to a central location, added
warning when preloading legacy plugins.
OSMonitor::get_erasure_code and OSDMonitor:normalize_profile will now check
if legacy EC plugins are used and log a warning.
Added tests to check that warnings make it to the log.
Signed-off-by: Bassam Tabbara <bassam.tabbara@quantum.com>
By switching to a new gf-complete with SIMD runtime detection, we can now remove all the different flavors of jerasure and shec. This simplifies deployment and configuration of erasure coding, enables hetergenous OSDs, and enables us to take advantage of new performance improvements in jerasure without config/build changes.
This commit removes flavors from cmake, removes ErasureCodePluginSelect___, and fixes unit tests. There is now a single plugin for jerasure and a single plugin for shec.
SIMDExt.cmake was changed so that its a little more generic, and is not polluted with gf-complete specific CFLAG defines. The #define for SIMD instruction were based on gf-complete.
I also added a small init helper for jerasure that has code that was common between jerasure and shec.
Signed-off-by: Bassam Tabbara <bassam.tabbara@quantum.com>
* in addition to the -march=foobar, we also need to pass -DFOOBAR to
gf-complete sources' cflags, so that the SIMD instructions can be
generated as expected.
* also extract the compiler support for instruction extensions detections
into SIMDExt.cmake.
* and only check the compiler support if CMAKE_SYSTEM_PROCESSOR matches
with the instruction-set/arch to be checked.
Signed-off-by: tone.zhang <tone.zhang@linaro.org>
Signed-off-by: Kefu Chai <kchai@redhat.com>
Dan Mick
wei xiao
Alexander Graf
Bassam Tabbara
Kefu Chai