pkgsrc-Changes archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

CVS commit: pkgsrc/math/ggml



Module Name:    pkgsrc
Committed By:   wiz
Date:           Sat Sep 26 09:45:32 UTC 2026

Modified Files:
        pkgsrc/math/ggml: Makefile distinfo

Log Message:
ggml: update to 0.25.3.

0.25.3

A small maintenance release: the core graph size calculation
(ggml_graph_nbytes) no longer triggers a UBSan "non-zero offset to
null pointer" diagnostic, and the CI setup moves CUDA builds to
the hf-jobs runner and adds a new AMD Vulkan build job.

0.25.2

A small point release focused on backend improvements: a new CUDA
convolution kernel with an implicit-GEMM fast path, Vulkan fixes
and tuning (misalignment handling in convolution shaders and
cooperative-matrix support for Adreno GPUs), an optimized DP4A
binary kernel for OpenCL Q6_K GEMM, and a precision guard in the
Hexagon backend.

0.25.1

A hotfix release focused on the CUDA backend: sparse flash attention,
which had been disabled due to a batch-dependent gate, is re-enabled
for long-context prefill (up to ~1.4x prefill speedup at 131k
context), with the supporting sparse mask scan kernel significantly
sped up. Additionally, Metal gains the missing f32 × bf16 matmul
kernels (fixing depthwise convolutions with bf16 weights) and its
flash-attention tuning tables are now keyed by GPU family instead
of individual SKU, while Vulkan adds MMQ/MMV matmul kernels for
the IQ4_XS quantization type.

0.25.0

This release expands hyper-connection, flash-attention, and fused
MoE/SSM support across CPU, GPU, and accelerator backends. It also
improves backend robustness, quantization, data-layout handling,
and RPC/meta buffer management. Numerous correctness fixes and
performance tuning land across all major backends, including new
kernels, fusions, and op coverage.

0.24.0

This release focuses on expanding backend coverage and robustness,
with a new precision-control API, major Vulkan/SYCL/Hexagon/OpenCL
work, and numerous correctness and performance fixes across CPU,
CUDA, Metal, and other backends.


To generate a diff of this commit:
cvs rdiff -u -r1.13 -r1.14 pkgsrc/math/ggml/Makefile \
    pkgsrc/math/ggml/distinfo

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.

Modified files:

Index: pkgsrc/math/ggml/Makefile
diff -u pkgsrc/math/ggml/Makefile:1.13 pkgsrc/math/ggml/Makefile:1.14
--- pkgsrc/math/ggml/Makefile:1.13      Sat Sep 12 20:27:32 2026
+++ pkgsrc/math/ggml/Makefile   Sat Sep 26 09:45:32 2026
@@ -1,6 +1,6 @@
-# $NetBSD: Makefile,v 1.13 2026/09/12 20:27:32 wiz Exp $
+# $NetBSD: Makefile,v 1.14 2026/09/26 09:45:32 wiz Exp $
 
-DISTNAME=      ggml-0.23.0
+DISTNAME=      ggml-0.25.3
 CATEGORIES=    math
 MASTER_SITES=  ${MASTER_SITE_GITHUB:=ggml-org/}
 GITHUB_TAG=    v${PKGVERSION_NOREV}
Index: pkgsrc/math/ggml/distinfo
diff -u pkgsrc/math/ggml/distinfo:1.13 pkgsrc/math/ggml/distinfo:1.14
--- pkgsrc/math/ggml/distinfo:1.13      Sat Sep 12 20:27:32 2026
+++ pkgsrc/math/ggml/distinfo   Sat Sep 26 09:45:32 2026
@@ -1,5 +1,5 @@
-$NetBSD: distinfo,v 1.13 2026/09/12 20:27:32 wiz Exp $
+$NetBSD: distinfo,v 1.14 2026/09/26 09:45:32 wiz Exp $
 
-BLAKE2s (ggml-0.23.0.tar.gz) = 5d0c4805ea8df62b546455efb40e0bd0175e4f4a80d930512827fd0d75b03b92
-SHA512 (ggml-0.23.0.tar.gz) = 5bab13fbe931e4a78759449e91eadad979d1efaf427439c674ebb22639fde32d1468791488b28ae3dda4952dd8c0514b2c4fee9ecc46033f7aff16de442dffc2
-Size (ggml-0.23.0.tar.gz) = 3931297 bytes
+BLAKE2s (ggml-0.25.3.tar.gz) = a00d4ec2eb165d2b0077924d8111e5fb5760186ddd42b21976c1c97a1d1b481a
+SHA512 (ggml-0.25.3.tar.gz) = 2344cf53dadeb1f1b2fd392aa8ae1850992fa127740a0a78b085be11708fe3c6423bfdd94009d3409c4a99fac7f525d8dfca0ba955526e3ae9b1024a5e73bcbe
+Size (ggml-0.25.3.tar.gz) = 4103121 bytes



Home | Main Index | Thread Index | Old Index