0.3.32

New issue

Closed

Due by March 8, 2026

•Closed Mar 24, 2026

released March 23, 2026

100% complete

Fix too small DGEMM_R for some Loongson LA464 cpus
Unread
#5611by was merged Jan 18, 2026
Fix ARMV9SME/VORTEXM4 GEMM compilation with SMP disabled
Unread
#5613by was merged Mar 8, 2026
Avoid integer overflow in dynamic_riscv64.c
Unread
#5609by was merged Jan 18, 2026
1
fix: don't use _Float16 type on GCC 12
Unread
#5606by was merged Jan 19, 2026
Fix building without multithreading or LAPACK
Unread
#5615by was merged Jan 18, 2026
Fix Intel OpenMP flag
Unread
#5616by was merged Jan 19, 2026
1
arm64: fix clang ICE on Windows for thunderx2t kernels
Unread
#5618by was merged Jan 20, 2026
9
Provide optimized ?SUM kernels for NeoverseN1 and related
Unread
#5621by was merged Jan 21, 2026
Serialize accesses to parallelized syrk functions from multiple calle…
Unread
#5619by was merged Feb 6, 2026
3
Rename the DllMain copy used in static linking to OpenBLASDllMain
Unread
#5623by was merged Jan 23, 2026
Fix CMake/LLVM compilation issues seen under Windows-on-Arm
Unread
#5626by was closed Jan 29, 2026
1
Fix stack address of flag parameter in (pre-POWER6) POWER ?SCAL kernels
Unread
#5628by was merged Jan 29, 2026
Add Intel Panther Lake CPUID
Unread
#5630by was merged Jan 30, 2026
Fix CMake/LLVM compilation issues seen under Windows-on-Arm
Unread
#5631by was merged Jan 30, 2026
Fix: Remove invalid parentheses after endif
Unread
#5634by was merged Feb 5, 2026
1
Add CPUID identification for Intel Arrow Lake H/U
Unread
#5636by was merged Feb 8, 2026
build: fix Makefile build with LLVM flang on macOS
Unread
#5633by was merged Feb 8, 2026
1
Fix SGEMM returning wrong results in multithreading on NeoverseV2
Unread
#5642by was closed Feb 11, 2026
Work around llvm failing to compile the AVX512 sgemm kernel
Unread
#5644by was merged Feb 12, 2026
Add CPU autodetection for Arm Cortex X925/A725
Unread
#5645by was merged Feb 14, 2026
Fix gmake build with only a subset of precision types
Unread
#5651by was merged Feb 20, 2026
Fix passing of C/ZDOTC results in C-converted LAPACK on non-Windows systems
Unread
#5652by was merged Feb 19, 2026
Use generic SCAL kernels for PPC970 running FreeBSD
Unread
#5654by was merged Feb 20, 2026
Added ability to accumulate in FP16. Convert BF16 to FP32. For FP16 and BF16 GEMM in RISC-V (BF16 now works for pre-RVA23)
Unread
#5640by was merged Feb 20, 2026
10
Add feature-based fallbacks for unknown/future Intel CPUIDs
Unread
#5655by was merged Feb 20, 2026

Navigation Menu

0.3.32

Fix too small DGEMM_R for some Loongson LA464 cpus

Fix ARMV9SME/VORTEXM4 GEMM compilation with SMP disabled

Avoid integer overflow in dynamic_riscv64.c

fix: don't use `_Float16` type on GCC 12

Fix building without multithreading or LAPACK

Fix Intel OpenMP flag

arm64: fix clang ICE on Windows for thunderx2t kernels

Provide optimized ?SUM kernels for NeoverseN1 and related

Serialize accesses to parallelized syrk functions from multiple calle…

Rename the DllMain copy used in static linking to OpenBLASDllMain

Fix CMake/LLVM compilation issues seen under Windows-on-Arm

Fix stack address of flag parameter in (pre-POWER6) POWER ?SCAL kernels

Add Intel Panther Lake CPUID

Fix CMake/LLVM compilation issues seen under Windows-on-Arm

Fix: Remove invalid parentheses after endif

Add CPUID identification for Intel Arrow Lake H/U

build: fix Makefile build with LLVM flang on macOS

Fix SGEMM returning wrong results in multithreading on NeoverseV2

Work around llvm failing to compile the AVX512 sgemm kernel

Add CPU autodetection for Arm Cortex X925/A725

Fix gmake build with only a subset of precision types

Fix passing of C/ZDOTC results in C-converted LAPACK on non-Windows systems

Use generic SCAL kernels for PPC970 running FreeBSD

Added ability to accumulate in FP16. Convert BF16 to FP32. For FP16 and BF16 GEMM in RISC-V (BF16 now works for pre-RVA23)

Add feature-based fallbacks for unknown/future Intel CPUIDs

Issues