This fixes a bug where "make -j<N> check" may fail after a change to one or more header files, or where testsuite code doesn't get properly recompiled after internal changes.
Details:
- Disabled a sanity check in bli_pool_finalize() that was meant to alert
the user if a pool_t was being finalized while some blocks were still
checked out. However, this is exactly the situation that might happen
when a pool_t is re-initialized for a larger blocksize, and currently
bli_pool_reinit() is implemeneted as _finalize() followed by _init().
So, this sanity check is not universally appropriate. Thanks to
AMD-India for reporting this issue.
Details:
- Updated stale calls to the bli_membrk API within the 'gemmlike'
sandbox. This API is now called bli_pba (packed block allocator).
Ideally, this forgotten update would have been included as part of
21911d6, which is when the branch where the membrk->pba changes was
introduced was merged into 'master'.
- Comment updates.
-- In existing design memory pool supports buffers of only one
size, This size is determined at compile time to support buffer
needed for biggest block size and data type. However, the size
calculation is not generic as it considers sizes only for GEMM.
Also it assumes that the buffer will not be used for any other
purpose than packing operations.
-- If the new buffer is requested whose size is bigger than existing
size, The pool is re-initialized to contain buffers of new size,
however, this is done only if the pool is empty. If the pool is not
empty the execution is aborted. This is undesirable as in
mulithreaded scenarios it is possible that different threads needs
buffers of different sizes at the same time (i.e. while other thread
is still in middle of the operation).
-- This commit removes the restriction of single size buffer, when
new buffer is requested whose size if bigger than exiting one, no
re-init is done. New buffer of required size is allocated, added
to the pool and returned to the user.
Change-Id: I20acdb60eb06ab2e53366d51713aa83c4b2df0da
The added fields:
1. `pack_t schema`: storing the pack schema on the object allows the macrokernel to act accordingly without side-channel information from the rntm_t and cntx_t. The pack schema and "pack_[ab]" fields could be removed from those structs.
2. `void* user_data`: this field can be used to store any sort of additional information provided by the user. The pointer is propagated to submatrix objects and copies, but is otherwise ignored by the framework and the default implementations of the following three fields. User-specified pack, kernel, or ukr functions can do whatever they want with the data, and the user is 100% responsible for allocating, assigning, and freeing this buffer.
3. `obj_pack_fn_t pack`: the function called when a matrix is packed. This functions receives the expected arguments, as well as a mdim_t and mem_t* as memory must be allocated inside this function, and behavior may differ based on which matrix is being backed (i.e. transposition for B). This could also be achieved by passing a desired pack schema, but this would require additional information to travel down the control tree.
4. `obj_ker_fn_t ker`: the function called when we get to the "second loop", or the macro-kernel. Behavior may depend on the pack schemas of the input matrices. The default implementation would perform the inner two loops around the ukr, and then call either the default ukr or a user-supplied one (next field).
5. `obj_ukr_fn_t ukr`: the function called by the default macrokernel. This would replace the various current "virtual" microkernels, and could also be used to supply user-defined behavior. Users could supply both a custom kernel (above) and microkernel, although the user-specified kernel does **not** necessarily have to call the ukr function specified on the obj_t.
Note that no macros or functions for accessing these new fields have been defined yet. That is next once these are finalized. Addresses https://github.com/flame/blis/projects/1#card-62357687.
Improve DGEMM performance for smaller sizes. AOCL DYNAMIC is incorporated at blas interface to enable
calling bli_dgemm_small when optimum number of threads implied is 1 for (n and k < 10).
Improved smart threading logic for dgemm,
Additional conditions at the blas interface added to invoke bli_dgemm_small.
Removed N > 3 condition from bli_dgemm_small.
Change-Id: Id751528dfe9de37800b02ffaf765b6c82487093e
Details:
- Added a check condition in GEMMT native path to choose
update_triang routines based on whether the kernels are
row-preferred or column preferred.
- Moved the zen-specific SUP thresholds under BLIS_CONFIG_EPYC
macro. For non-zen architectures, it falls back to L3 SUP
thresholds.
- Modified SUP code path to always choose 2m variant. 1n variant
is not implemented for GEMMT.
Change-Id: Ifdd55815c588f645e337de80b5b9d1864f6b5dd3
Description:
1. Added fflush after fprintf in aocldtl so avoid mising any log
info when application crashes
2. Removed one redundent if def statement
Change-Id: I79a645060fa36bd8c7e5eaca4d9183dc944329ea
Description:
1. While processing reminder cases in bli_trsm_small algorithm
there were few loads and stores which were accessing
beyond the given matrix buffer because of vectorized instructions.
2. Modified 256bit vector loads at edges into 128bit or 64 bit loads/stores
such that no read/write happens beyond the matrix boundary.
AMD-Internal: [CPUPL-1759] [SWLCSG-819]
Change-Id: Iba51d0ed9bb28d1b0948a219755b8dbcc86a7fa9
Details:
- Fixed a compile-time error in bli_init.c when compiling with OSX's
clang. This error was introduced in 868b901, which introduced a
post-declaration struct assignment where the RHS was a struct
initialization expression (i.e. { ... }). This use of struct
initializer expressions apparently works with gcc despite it not
being strict C99. The fix included in this commit declares a temporary
variable for the purposes of being initialized to the desired value,
via the struct initializer, and then copies the temporary struct (via
'=' struct assignment) to the persistent struct. Thanks to Devin
Matthews for his help with this.
Details:
- Accept either 'clang' or 'LLVM' in vendor string when greping for
the version number (after determining that we're working with clang).
Thanks to Devin Matthews for this fix.
Details:
- Fixes a rather obvious bug that resulted in segmentation fault
whenever the calling application tried to re-initialize BLIS after
its first init/finalize cycle. The bug resulted from the fact that
the bli_init.c APIs made no effort to allow bli_init() to be called
subsequent times at all due to it, and bli_finalize(), being
implemented in terms of pthread_once(). This has been fixed by
resetting the pthread_once_t control variable for initialization
at the end of bli_finalize_apis(), and by resetting the control
variable for finalization at the end of bli_init_apis(). Thanks to
@lschork2 for reporting this issue (#525), and to Minh Quan Ho and
Devin Matthews for suggesting the chosen solution.
- CREDITS file update.
1. single instance case sup is enabled.
2. Env BLIS_SINGLE_INSTANCE should be set to 1 to enable single instance tuning.
AMD-Internal: [CPUPL-1743]
Change-Id: Iadb05a6e9313ac41271c0522da243fd47d80abec
Details :
- Added packing Of Y for incy >1 cases for dgemv_unf_var2.
- Added packing Of X for incx >1 cases for dgemv_unf_var1.
AMD-Internal: [SWLCSG-735]
Change-Id: Ib395f478ba984a85533e4f79b3521d0b2500c30c
1. Added the compiler flags for the clang-cl compiler to build blis multithreading using openmp library.
2. Updated format of presenting version string.
AMD Internal : [CPUPL-1630]
Change-Id: I979de541fa57415c08c20b0d5b684ae6bd242d19
Details:
- Added single-threaded and multithreaded performance results to
docs/Performance.md. These results were gathered on a Graviton2
Neoverse N1 server. Special thanks to Nicholai Tukanov for
collecting these results via the Arm-HPC/AWS hackaton.
- Corrected what was supposed to be a temporary tweak to the legend
labels in test/3/octave/plot_l3_perf.m.
Details:
- Added vzeroupper instruction to the end of all 'gemm' and 'gemmtrsm'
microkernels so as to avoid a performance penalty when mixing AVX
and SSE instructions. These vzeroupper instructions were once part
of the haswell kernels, but were inadvertently removed during a source
code shuffle some time ago when we were managing duplicate 'haswell'
and 'zen' kernel sets. Thanks to Devin Matthews for tracking this down
and re-inserting the missing instructions.
Details:
- Thanks to Chengguo Sun for submitting #515 (5ef7f68).
- Thanks to Andrew Wildman for submitting #519 (551c6b4).
- Whitespace update to configure (spaces to tabs).
