OSchip
/
OpenBLAS
mirror of https://github.com/xianyi/OpenBLAS.git
0
0
Fork 2
Code Issues Packages Projects Releases Wiki Activity
10,714 Commits 58 Branches 70 Tags 121 MiB
Commit Graph

147 Commits

Author SHA1 Message Date
Martin Kroeker 75b93b4465
Fix missing declaration of info 2026-08-07 13:56:43 +02:00
Martin Kroeker 761ebe91ab
Fix return without code and buffer leak on other error path 2026-08-07 13:51:01 +02:00
Martin Kroeker ed2ba87203
Add error return on buffer assignment failure 2026-08-07 12:28:01 +02:00
Martin Kroeker b080247a21
Revert "Mark temporaries as volatile to avoid SIGBUS in multithreaded builds with gcc15+ on 32bit FreeBSD/x86" 2026-07-26 22:51:58 +02:00
Keno Fischer 7a07c8e1f4 Add an asynchronous cancellation API for in-flight operations 
Long-running BLAS calls (a large gemm can run for minutes) cannot
currently be interrupted: callers embedding OpenBLAS (e.g. the Julia
runtime responding to a user's ^C) can only wait for completion or kill
the process. Add a minimal cooperative cancellation protocol:

Every thread owns a pointer-sized generation slot in thread-local
storage, whose stable address is returned by openblas_cancel_token().
Instrumented compute drivers advance the slot to a fresh even
generation at operation entry on the issuing thread (forwarding the
slot and generation to worker threads through blas_arg_t) and poll it
at block granularity. openblas_cancel(token, loaded_token) - callable
from any thread - sets the cancel bit (bit 0) iff the slot still holds
loaded_token, so a canceller that loaded the value while an operation
was in flight stops exactly that operation, while stale or racing
requests either miss or dirty an already-dead generation, both
harmless. There is no object lifecycle: nothing to allocate, bind,
reset, or free.

A cancelled operation returns quickly, leaving its output buffer in an
unspecified partially-updated state that the caller must discard; every
synchronization point in the threaded driver is still executed, so
sibling threads never stall and the library remains consistent for
subsequent calls. Coverage: the level-3 gemm/symm/hemm drivers
(level3.c and level3_thread.c).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012nCkyKUguncLLJrH9K5o7m
 2026-07-19 04:20:28 +00:00
Martin Kroeker 15351019ea
Mark subbuffer as volatile to fix SIGBUS in multithreading with gcc15+ on 32bit FreeBSD/x86 2026-07-16 18:35:14 +02:00
Kumar Aditya a2261f9e52 fix thread safety of openmp backend 2026-07-02 15:39:09 +05:30
Kumar Aditya 298d53c061 fix data races in blas_server and level3_thread 2026-06-23 18:42:25 +05:30
Nathan Goldbaum 9363452742 Fix corruption due to lock sharding issues by centralizing locking 2026-06-14 15:32:51 -06:00
Martin Kroeker 54b939e4e3
Undefine GEMM_PREFERRED_SIZE before redefining to silence warning 2026-04-22 16:30:12 +02:00
Martin Kroeker c9185e91ad
Make GEMM_DIVIDE_RATE and GEMM_PREFERRED_SIZE available in DYNAMIC_ARCH builds 2026-03-20 15:34:04 +01:00
Martin Kroeker b7601ea92f
Retrieve cpu-specific GEMM_DIVIDE_LIMIT if DYNAMIC_ARCH 2026-03-19 08:29:15 +01:00
lujiaweics 1f3b81e562 Serialize accesses to parallelized syrk functions from multiple callers, like it was already done for GEMM in level3_thread.c and GEMM3M in level3_gemm3m_thread.c 2026-01-20 21:31:35 +08:00
Martin Kroeker b2b9abc20b
Revert "Enhancing Core Utilization in BLAS Calls: A Scalable Architecture" 2025-10-22 15:44:43 +02:00
Martin Kroeker 5b640b1cbc
add bgemm_thread_xx 2025-10-16 10:03:04 -07:00
Masato Nakagawa 7e29f11396 Multi-thread GEMM Performance Improvement on NeoverseV1 (DIVIDE_RATE=1) 2025-07-29 18:54:36 +09:00
youcai 41f9701ebc Fix cmake building with cblas_bgemm 2025-07-23 22:10:53 +08:00
Chris Sidebottom 48394384ef Use correct constants for per-target BGEMM/SBGEMM 
This fixes the build and tests on `NEOVERSEV1` target, which was failing
with specific constants for `SBGEMM`

Co-authored-by: Ye Tao <ye.tao@arm.com>
 2025-07-08 16:23:27 +01:00
Chris Sidebottom f95e7b0e32 Add infrastructure for BGEMM 
Setting up all the infrastructure for BGEMM support in OpenBLAS, hopefully I found all the right places.

Derived mostly from the previous work done in https://github.com/OpenMathLib/OpenBLAS/pull/5287

Co-authored-by: Ye Tao <ye.tao@arm.com>
 2025-07-08 16:22:41 +01:00
Chris Sidebottom 7a97c4ca97 Rename HALF -> BFLOAT16 in some more places 2025-07-07 10:13:39 +00:00
Masato Nakagawa 5253c8f165 Multi-thread Performance Improvement of GEMM with DIVIDE_RATE=1 for 
A64FX.
 2025-06-30 21:35:16 +09:00
Srangrang 9f13b2c6ac style: modify HALF to BFLOAT16 in benchmark folder 2025-06-15 20:57:05 +08:00
gkdddd 670ec6f757 Added shgemm_kernel_8x8 for RISCV64_ZVL128B and shgemm_kernel_16x8 for RISCV64_ZVL256B 
Added HFLOAT16 support for RISCV64
Added shgemm_kernel_8x8 for RISCV64_ZVL128B and shgemm_kernel_16x8 for RISCV64_ZVL256B based on HFLOAT16
The instruction sets used are ZVFH and ZFH, which need to be supported by RVV1.0

Related to issue #5279
Co-authored-by Linjin Li <linjin_li@163.com>
 2025-06-03 20:14:30 +08:00
Martin Kroeker 20f2ba0141
Move declaration of i for pre-C99 compilers 2025-05-21 23:44:17 +02:00
Masato Nakagawa 2351a98005 Update 2D thread-partitioned GEMM for M << N case. 2025-05-21 21:21:52 +09:00
Ruiyang Wu 02fd1df10b CMake: Pass `OpenMP` compiler and linker flags through CMake targets 
Using `OpenMP::OpenMP_LANG` targets for CMake is less error-prone than
passing the compiler and linker flags manually. Furthermore, it allows
the user to customize those flags by setting `OpenMP_LANG_FLAGS`,
`OpenMP_LANG_LIB_NAMES`, and `OpenMP_omp_LIBRARY`.
 2025-03-26 23:09:54 -04:00
Masato Nakagawa 80d3c2ad95 Add Improving Load Imbalance in Thread-Parallel GEMM 2025-03-11 20:18:20 +09:00
Martin Kroeker 77c638db67
Revert "Fix potential inaccuracy in multithreaded level3 related to SWITCH_RATIO" 2025-02-15 20:37:48 +01:00
John Hein 6cd9bbe531 fix signedness of pointer to integer type passed to blas_lock() 2025-02-01 17:22:57 -07:00
Martin Kroeker 8a1710dd0d
don't apply switch_ratio to tail of loop 2024-10-06 20:03:32 +02:00
shivammonaka 9e22d70957 Dynamic locking in Pthread Backend to allow multiple BLAS calls to be executed parallelly 2024-06-07 08:40:17 +05:30
Martin Kroeker db070a9223
add gemm_batch drivers 2024-05-31 18:29:27 +02:00
Martin Kroeker d0794f88dc
add gemm_batch driver 2024-05-29 15:49:20 +02:00
yamazaki-mitsufumi 51ab1903e7 Expanding the scop of 2D thread distribution 2024-04-18 18:20:25 +09:00
shivammonaka d49ebc54e1 Merge branch 'shivam-develop' into shivam-Locks 2024-02-29 11:58:14 +05:30
shivammonaka bc191015e3 Using OpenMP locks with NUM_PARALLEL 2024-02-29 11:47:05 +05:30
Martin Kroeker c4bd4a2e5d
fix improper function prototypes (empty parentheses) 2023-09-30 12:49:24 +02:00
Chris Sidebottom 32f2fafde7 Propagate SWITCH_RATIO to DYNAMIC_ARCH builds 
Previously dynamic builds were either using the default SWITCH_RATIO
or one from the higher level architecture; this patch ensures the
dynamic builds can use this parameter as well.
 2023-04-17 15:34:12 +01:00
Honglin Zhu 4989e039a5 Define SBGEMM_ALIGN_K for DYNAMIC_ARCH build 2022-10-27 14:10:26 +08:00
Honglin Zhu b00d5b9746 New sbgemm implementation for Neoverse N2 
1. Use UZP instructions but not gather load and scatter store instructions to get lower latency.
    2. Padding k to a power of 4.
 2022-10-26 15:09:41 +08:00
Wangyang Guo 3dc6052c7e initial support for Sapphire Rapids platform 2021-10-12 01:30:40 -07:00
Martin Kroeker 2f8220d757
Add sbgemm 2021-09-14 16:14:43 +02:00
Martin Kroeker 307c4c0786
Fix typo 2021-06-16 13:41:16 +02:00
Martin Kroeker e83df93975
Work around another recent macro name collision with winnt.h 2021-06-16 12:32:34 +02:00
Martin Kroeker a554712439
remove extra/intermediate size step for min_jj introduced in PR747 2020-12-08 21:01:36 +01:00
Martin Kroeker 5d26223f4a
remove extra/intermediate size step of min_jj from PR747 2020-12-08 20:59:56 +01:00
Martin Kroeker d3ff1f889f
Convert ifndefs to ifneq 2020-11-22 16:27:17 +01:00
Rajalakshmi Srinivasaraghavan b5d30b390d Fix build issues with bfloat16 
This patch fixes compilation errors due to recent renaming from SH to SB
with BUILD_BFLOAT16.
 2020-10-13 11:00:22 -05:00
Martin Kroeker 006c7f6671
Change "HALF" and "sh" to "BFLOAT16" and "sb" 2020-10-12 00:06:06 +02:00
Martin Kroeker 886a8e3190
Adapt for supporting only a subset of variable types 2020-10-11 14:57:32 +02:00