Details:
- Reran all existing single-threaded performance experiments comparing
BLIS sup to other implementations (including the conventional code
path within BLIS), using the latest versions (where appropriate).
- Added multithreaded results for the three existing hardware types
showcased in docs/PerformanceSmall.md: Kaby Lake, Haswell, and Epyc
(Zen1).
- Various minor updates to the text in docs/PerformanceSmall.md.
- Updates to the octave scripts in test/sup/octave, test/supmt/octave.
Details:
- Updated test/sup/Makefile and test/supmt/Makefile to allow specifying
different problem size ranges for the drivers where one, two, or three
matrix dimensions is large. This will facilitate the generation of
more meaningful graphs, particularly when two dimensions are tiny.
Details:
- Optimized scripts in test/sup/octave and test/supmt/octave for use
with octave 5.2.0 on Ubuntu 18.04.
- Fixed stray 'end' keywords in gen_opsupnames.m and plot_l3sup_perf.m,
which were not only unnecessary but also causing issues with versions
5.x.
Details:
- Renamed two bli_thread_*() APIs:
bli_thread_obarrier() -> bli_thread_barrier()
bli_thread_obroadcast() -> bli_thread_broadcast()
The 'o' was a leftover from when thrcomm_t objects tracked both
"inner" and "outer" communicators. They have long since been
simplified to only support the latter, and thus the 'o' is
superfluous.
Details:
- Return early from bli_thrinfo_sup_grow() if the thrinfo_t object
address is equal to either &BLIS_GEMM_SINGLE_THREADED or
&BLIS_PACKM_SINGLE_THREADED.
- Added preprocessor logic to bli_l3_sup_thread_decorator() in
bli_l3_sup_decor_single.c that (by default) disables code that
creates and frees the thrinfo_t tree and instead passes
&BLIS_GEMM_SINGLE_THREADED as the thrinfo_t pointer into the
sup implementation.
- The net effect of the above changes is that a small amount of
thrinfo_t overhead is avoided when running small/skinny dgemm
problems when BLIS is compiled with multithreading disabled.
Details:
- Fixed a syntax bug in bli_l3_sup_decor_single.c as a result of
changing function interface for the thread entry point function
(of type l3supint_t).
- Unfortunately, fixing the interface was not enough, as it caused
a memory leak in the sba at bli_finalize() time. It turns out that,
due to the new multithreading-capable variant code useing thrinfo_t
objects--specifically, their calling of bli_thrinfo_grow()--we
have to pass in a real thrinfo_t object rather than the global
objects &BLIS_PACKM_SINGLE_THREADED or &BLIS_GEMM_SINGLE_THREADED.
Thus, I inserted the appropriate logic from the OpenMP and pthreads
versions so that single-threaded execution would work as intended
with the newly upgraded variants.
Details:
- Added multithreading support to the sup framework (via either OpenMP
or pthreads). Both variants 1n and 2m now have the appropriate
threading infrastructure, including data partitioning logic, to
parallelize computation. This support handles all four combinations
of packing on matrices A and B (neither, A only, B only, or both).
This implementation tries to be a little smarter when automatic
threading is requested (e.g. via BLIS_NUM_THREADS) in that it will
recalculate the factorization in units of micropanels (rather than
using the raw dimensions) in bli_l3_sup_int.c, when the final
problem shape is known and after threads have already been spawned.
- Implemented bli_?packm_sup_var2(), which packs to conventional row-
or column-stored matrices. (This is used for the rrc and crc storage
cases.) Previously, copym was used, but that would no longer suffice
because it could not be parallelized.
- Minor reorganization of packing-related sup functions. Specifically,
bli_packm_sup_init_mem_[ab]() are called from within packm_sup_[ab]()
instead of from the variant functions. This has the effect of making
the variant functions more readable.
- Added additional bli_thrinfo_set_*() static functions to bli_thrinfo.h
and inserted usage of these functions within bli_thrinfo_init(), which
previously was accessing thrinfo_t fields via the -> operator.
- Renamed bli_partition_2x2() to bli_thread_partition_2x2().
- Added an auto_factor field to the rntm_t struct in order to track
whether automatic thread factorization was originally requested.
- Added new test drivers in test/supmt that perform multithreaded sup
tests, as well as appropriate octave/matlab scripts to plot the
resulting output files.
- Added additional language to docs/Multithreading.md to make it clear
that specifying any BLIS_*_NT variable, even if it is set to 1, will
be considered manual specification for the purposes of determining
whether to auto-factorize via BLIS_NUM_THREADS.
- Minor comment updates.
Details:
- Fixed an error that manifests only when using C++ (specifically,
modern versions of g++) to compile drivers in 'test' (and likely most
other application code that #includes blis.h. Thanks to Ajay Panyala
for reporting this issue (#374).
Details:
- Removed a line of code in common.mk that passed LDFLAGS through the
sort function. The purpose was not to sort the contents, but rather
to remove duplicates. However, there is valid syntax in a string of
linker flags that, when sorted, yields different/broken behavior.
So I've removed the line in common.mk that sorts LDFLAGS. Also, for
future use, I've added a new function, rm-dupls, that removes
duplicates without sorting. (This function was based on code from a
stackoverflow thread that is linked to in the comments for that
code.) Thanks to Isuru Fernando for reporting this issue (#373).
Details:
- Finally removed 'attic/windows' and its contents. This directory once
contained "proto" Windows support for BLIS, but we've since moved on
to (thanks to Isuru Fernando) providing Windows DLL support via
AppVeyor's build artifacts. Furthermore, since 'windows' was the only
subdirectory within 'attic', the directory path would show up in
GitHub's listing at https://github.com/flame/blis, which probably led
to someone being confused about how BLIS provides Windows support. I
assume (but don't know for sure) that nobody is using these files, so
this is admittedly a case of shoot first and ask questions later.
* Fix parsing in vpu_count on workstation SKX
* Document Skylake-X as Haswell for single FMA
* Update vpu_count for Skylake and Cascade Lake models
* Support printing the configuration selected, controlled by the environment
Intended particularly for diagnosing mis-selection of SKX through
unknown, or incorrect, number of VPUs.
* Move bli_log outside the cpp condition, and use it where intended
* Add Fixme comment (Skylake D)
* Mostly superficial edits to commits towards #351.
Details:
- Moved architecture/sub-config logging-related code from bli_cpuid.c
to bli_arch.c, tweaked names, and added more set/get layering.
- Tweaked log messages output from bli_cpuid_is_skx() in bli_cpuid.c.
- Content, whitespace changes to new bullet in HardwareSupport.md that
relates to single-VPU Skylake-Xs.
* Fix comment typos
Co-authored-by: Field G. Van Zee <field@cs.utexas.edu>
the comment contains an incorrect link, which is trivially fixed here.
@fgvanzee I hope you don't mind that I committed directly to master but this cannot break anything.
For the kernel of size 4x8, cs_b is used instead of cs_a to calculate address of diagonal elements of matrix A.
Correcting the mistake.
Change-Id: Ie74e0f6a397fcd32fefb5804cd00f1e90bfe5523
Details:
- Fixed a bug in sdsdot_sub() that redundantly added the "alpha" scalar,
named 'sb'. This value was already being added by the underlying
sdsdot_() function. Thus, we no longer add 'sb' within sdsdot_sub().
Thanks to Simon Lukas Märtens for reporting this bug via #367.
- Fixed a second bug in order of typecasting intermediate products in
sdsdot_(). Previously, the "alpha" scalar was being added after the
"outer" typecast to float. However, the operation is supposed to first
add the dot product to the (promoted) scalar and THEN downcast the sum
to float. Thanks to Devin Matthews for catching this bug.
Details:
- Added BLIS_EXPORT_BLIS annotation to function prototypes for
bli_thrcomm_bcast()
bli_thrcomm_barrier()
bli_thread_range_sub()
so that these functions are exported to shared libraries by default.
This (hopefully) fixes issue #366. Thanks to Kyungmin Lee for
reporting this bug.
- CREDITS file update.
Interchanged some loops to favour column-major storage.
Added check condiion to identify last column and load it using a 'for' loop to avoid memory accesses out of buffer
Change-Id: Id5d2e16c65017a7f4b641d33228d23903efd09ac
Updates and fixes to power9 subconfig.
Details:
- Register s,c,z reference gemm and trsm ukernels that assume elements
of B have been broadcast.
- Added prototypes for level-3 ukernels that assume elements of B have
been broadcast. Also added prototype for an spackm function that
employs a duplication/broadcast factor of 4.
- Register virtual gemmtrsm ukernels that work with broadcasting of B.
- Disable right-side hemm, symm, trmm, and trmm3 in bli_family_power9.h.
- Thanks to Nicholai Tukanov for providing these updates.
For matrix sizes which are not multiples of 4, trsm_small kernels access memory outside the allocated buffers which causes segmentation fault.
This is fixed by handling each of the corner cases separately.
Change-Id: Ia7cfad5d65339a209a7376cc1654382593c933af
amd64 family supports all the architectures before zen.
Assigned (BLIS_ARCH_GENERIC+1) to BLIS_NUM_ARCHS in order to avoid update for every new architecture.
Change-Id: I8241e643f6dfd0ebe272e053ca8b6a9c1463d9dc
