Details:
- Fixed a bug that manifested as improperly-computed 1-norm for vectors
and matrices. This is one of the few operations in BLIS that does not
have its own test module within the testsuite, hence why it went
undetected for so long. The bad 1-norms were being used to normalize
matrices in the testsuite after initialization, which led to some
matrices containing a combination of "large" and "small" values. This
tended to push the residuals computed after each test away from zero.
In some cases, they were off *just* enough to the testsuite to label
it a "failure". Many thanks to Jeff Hammond for reporting this bug.
(Wonky details: the bug was due to improperly-defined level-0 scalar
macros for abval2, an operation that computes the absolute square,
or complex magnitude/modulus. Certain complex domain instances of
abval2 were being incorrectly defined in terms of real-only solutions,
leading to bad results. This level-0 operation forms the basis of
norm1v/norm1m. absq2 was also affected, but almost nothing uses
this operation.)
Details:
- Disabled testsuite tests of all level-3 implementations based on 3m
and 4m. This will improve testing runtime on Travis CI as well as for
anyone manually running the testsuite using default test parameters.
Thanks to Devin Matthews for suggesting this change.
we want to be able to run BLIS KNL binaries on non-KNL machines via SDE.
although it is possible to install hbwmalloc implementation on such
systems, it is easier not to, since obviously the performance of SDE
execution is not representative so there is no reason to emulate HBW
allocation.
Details:
- Fixed a bug introduced in 1c732d3 that affected trsm1m_r. The result
was nondeterministic behavior (usually segmentation faults) for certain
problem sizes beyond the 1m instance of kc (e.g. 128 on haswell). The
cause of the bug was my commenting out lines in bli_gemm1m_ukr_ref.c
which explicitly directed the virtual gemm micro-kernel to use temporary
space if the storage preference of the [real domain] gemm ukernel did
not match the storage of the output matrix C. In the context of gemm,
this handling is not needed because agreement between the storage pref
and the matrix is guaranteed by a high-level optimization in BLIS.
However, this optimization is not applied to trsm because the storage
of C is not necessarily the same as the storage of the micro-panels of
B--both of which are updated by the micro-kernel during a trsm
operation. Thus, the guarantee of storage/preference agreement is not
in place for trsm, which means we must handle that case within the
virtual gemm micro-kernel.
- Comment updates and a minor macro change to bli_trsm*_cntx_init() for
3m1, 4m1a, and 1m.
Details:
- Commented out code in frame/ind/oapi/bli_l3_3m4m1m_oapi.c that was
specifically inserted to facilitate the benchmarking of 1m block-panel
and panel-block algorithms.
- Updates to test/3m4m/Makefile, runme.sh script, and test_gemm.c to
reflect changes used/needed during benchmarking.
Details:
- Defined bli_gemmbp_cntl_create(), bli_gemmpb_cntl_create(), with the
body of bli_gemm_cntl_create() replaced with a call to the former.
- Defined bli_cntl_free_w_thrinfo(), bli_cntl_free_wo_thrinfo(). Now,
bli_cntl_free() can check if the thread parameter is NULL, and if so,
call the latter, and otherwise call the former.
- Defined bli_gemm1mbp_cntx_init(), bli_gemm1mpb_cntx_init(), both in
terms of bli_gemm1mxx_cntx_init(), which behaves the same as
bli_gemm1m_cntx_init() did before, except that an extra bool parameter
(is_pb) is used to support both bp and pb algorithms (including to
support the anti-preference field described below).
- Added support for "anti-preference" in context. The anti_pref field,
when true, will toggle the boolean return value of routines such as
bli_cntx_l3_ukr_eff_prefers_storage_of(), which has the net effect of
causing BLIS to transpose the operation to achieve disagreement (rather
than agreement) between the storage of C and the micro-kernel output
preference. This disagreement is needed for panel-block implementations,
since they induce a transposition of the suboperation immediately before
the macro-kernel is called, which changes the apparent storage of C. For
now, anti-preference is used only with the pb algorithm for 1m (and not
with any other non-1m implementation).
- Defined new functions,
bli_cntx_l3_ukr_eff_prefers_storage_of()
bli_cntx_l3_ukr_eff_dislikes_storage_of()
bli_cntx_l3_nat_ukr_eff_prefers_storage_of()
bli_cntx_l3_nat_ukr_eff_dislikes_storage_of()
which are identical to their non-"eff" (effectively) counterparts except
that they take the anti-preference field of the context into account.
- Explicitly initialize the anti-pref field to FALSE in
bli_gks_cntx_set_l3_nat_ukr_prefs().
- Added bli_gemm_ker_var1.c, which implements a panel-block macro-kernel
in terms of the existing block-panel macro-kernel _ker_var2(). This
technique requires inducing transposes on all operands and swapping
the A and B.
- Changed bli_obj_induce_trans() macro so that pack-related fields are
also changed to reflect the induced transposition.
- Added a temporary hack to bli_l3_3m4m1m_oapi.c that allows us to easily
specify the 1m algorithm (block-panel or panel-block).
- Renamed the following cntx_t-related macros:
bli_cntx_get_pack_schema_a() -> bli_cntx_get_pack_schema_a_block()
bli_cntx_get_pack_schema_b() -> bli_cntx_get_pack_schema_b_panel()
bli_cntx_get_pack_schema_c() -> bli_cntx_get_pack_schema_c_panel()
and updated all instantiations. Also updated the field names in the
cntx_t struct.
- Comment updates.
Details:
- Implemented the 1m method for inducing complex domain matrix
multiplication. 1m support has been added to all level-3 operations,
including trsm, and is now the default induced method when native
complex domain gemm microkernels are omitted from the configuration.
- Updated _cntx_init() operations to take a datatype parameter. This was
needed for the corresponding function for 1m (because 1m requires us
to choose between column-oriented or row-oriented execution, which
requires us to query the context for the storage preference of the
gemm microkernel, which requires knowing the datatype) but I decided
that it made sense for consistency to add the parameter to all other
cntx initialization functions as well, even though those functions
don't use the parameter.
- Updated bli_cntx_set_blkszs() and bli_gks_cntx_set_blkszs() to take
a second scalar for each blocksize entry. The semantic meaning of the
two scalars now is that the first will scale the default blocksize
while the second will scale the maximum blocksize. This allows scaling
the two independently, and was needed to support 1m, which requires
scaling for a register blocksize but not the register storage
blocksize (ie: "packdim") analogue.
- Deprecated bli_blksz_reduce_dt_to() and defined two new functions,
bli_blksz_reduce_def_to() and bli_blksz_reduce_max_to(), for reducing
default and maximum blocksizes to some desired blocksize multiple.
These functions are needed in the updated definitions of
bli_cntx_set_blkszs() and bli_gks_cntx_set_blkszs().
- Added support for the 1e and 1r packing schemas to packm, including
1e/1r packing kernels.
- Added a minor optimization to bli_gemm_ker_var2() that allows, under
certain circumstances (specifically, real domain beta and row- or
column-stored matrix C), the real domain macrokernel and microkernel
to be called directly, rather than using the virtual microkernel
via the complex domain macrokernel, which carries a slight additional
amount of overhead.
- Added 1m support to the testsuite.
- Added 1m support to Makefile and runme.sh in test/3m4m. Also simplified
some code in test_gemm.c driver.
Intel compilers include a highly optimized math library (libimf) that
should be used instead of GNU libm.
yes, this change is for ALL targets, including those that are not
supported by the Intel compiler. there is no harm in doing this, and it
is future-proof in the event that the Intel compilers support other
architectures.
Details:
- Removed the vast majority of directories named "old", which contained
deprecated code that I wasn't quite ready to jettison from the source
tree.
Details:
- Changed the interface of bli_getopt() to take a new argument, a getopt_t
struct, that stores the values of optarg, optind, opterr, and optopt,
and updated the implementation accordingly. (Previously, these
variables were assumed to be global.)
- Added a function for initializing a getopt_t struct.
- Changed test_libblis.c--currently the only consumer of bli_getopt()--to
utilize the new getopt_t state object.
Details:
- Reimplemented several sets of get/set-style preprocessor macros with
static functions, including those in the following frame/base headers:
auxinfo, cntl, mbool, mem, membrk, opid, and pool. A few headers in
frame/thread were touched as well: mutex_*, thrcomm, and thrinfo.
Details:
- Expanded checking of the arch_t id in bli_gks.c--either passed in from
the caller or as returned from bli_arch_query_id()--against the expected
range of id values. Thanks to Devangi Parikh for suggesting these
additional sanity checks.
Details:
- Defined a new 'uninstall-old-headers' target that allows users of BLIS to
uninstall no-longer-needed headers left over from previous installations.
- Fixed the 'uninstall-old' target so that it will install both .a and .so
libraries.
- Renamed 'uninstall-old' to 'uninstall-old-libs'.
- Added 'uninstall-old' target (different from previous 'uninstall-old'
target) that combines 'uninstall-old-libs' and 'uninstall-old-headers'.
Details:
- When CBLAS is enabled at configure-time, BLIS now creates a monolithic
cblas.h using the same flatten-header.sh script that was recently
introduced for creating monolithic blis.h header files. The top-level
Makefile will also install this cblas.h file into the install prefix
alongside blis.h when the 'install' target is invoked. The two header
files are compatible with one another. Regardless whether the user's
source #includes cblas.h, both blis.h and cblas.h, or just blis.h,
the user will get the CBLAS function prototypes and enums, as expected.
Details:
- Fixed a subtle bug in bli_cntx_get_[un]packm_ker_dt() in which the
function fails to return NULL when passed a kernel id argument that is
equal to or beyond BLIS_NUM_[UN]PACKM_KERS. Instead, the function was
attempting to index into the cntx_t's packm kernel array, which resulted
in undefined behvaior. Thanks to Devangi Parikh for finding this bug.
Details:
- Rewrote monolithify-header.sh (and renamed to flatten-header.sh) so that
headers are inserted recursively. This improves performance by a factor
of 3-4x.
- Modified configure to create an 'include/<configname>' directory in which
make can create a monolithic header.
- Modified the top-level Makefile so that a monolithic header is generated
unconditionally prior to compilation (stored in include/<configname>) and
so that the single header is installed instead of the 450 or so header
files that reside throughout the framework source tree.
- Added "include/*/*.h" to .gitignore file.
- Removed some pnacl/emscripten leftovers that I intended to include in
a1caeba (mostly in testsuite/Makefile).
- Trivial comment changes to frame/include/bli_f2c.h.
Details:
- Added support for the x86_64 configuration family to bli_arch.c and
bli_arch_config.h. Thanks to Johannes Dieterich for reporting this
issue.
- Bumped the default value for BLIS_SIMD_NUM_REGISTERS from 16 to 32 and
the default value for BLIS_SIMD_SIZE from 32 to 64. This will support
configuration families that include Skylake and newer processors without
any supported needed in the bli_family_*.h file. The semantics of these
values have always been "maximum" and not exact values; comments in
bli_kernel_macro_defs.h and the github wiki have been adjusted
accordingly.
Details:
- Erroneously placed the "don't overwrite existing blocksize" logic in
bli_blksz_init*() rather than in bli_cntx_set_blkszs(). It belongs in
the latter because that function copies blocksizes as-is from the
blksz_t function argument to the appropriate field in the cntx_t. If
the blksz_t was previously initialized selectively, based on the sign
of the blocksize value passed into bli_blksz_init*(), that just leaves
some fields possibly uninitialized (with garbage values), which
definitely will not work.
- The aforementioned logic has been moved to bli_cntx_set_blkszs() via
a new function bli_blksz_copy_if_pos(), which selectively copies only
the blocksizes that are greater than zero.
Details:
- Employ the new semantics of bli_blksz_init*() in e31f0b3 in various
sub-configurations' bli_cntx_init_*() functions by passing in 0 for
register and cache blocksizes that correpond to gemm microkernel
datatypes that were not registered, allowing the default values
set by the bli_cntx_init_*_ref() function call to remain.
Details:
- Updated the semantics of bli_blksz_init() and bli_blksz_init_ed() so
that non-positive blocksize values are ignored entirely. This provides
an easy way to indicate that certain existing values should not be
touched by the update. Thanks to Devangi Parikh for feedback that led
to these changes.
Details:
- Added missing x86_64 configuration directory, which was intended to be
part of b7ca580.
- Added -Wfatal-errors compiler warning flag to all configurations so that
compilation stops after the first error.
- Changed the vectorization flags for intel64 configuration to be compatible
with 'penryn', the oldest sub-config included in that family.
- Changed the vectorization flags for penryn to target the 'core2'
microarchitecture and ssse3.
