head	1.15;
access;
symbols
	pkgsrc-2026Q3:1.13.0.2
	pkgsrc-2026Q3-base:1.13;
locks; strict;
comment	@# @;


1.15
date	2026.10.06.07.46.08;	author wiz;	state Exp;
branches;
next	1.14;
commitid	adsdnjcUhto6boYG;

1.14
date	2026.09.26.09.45.32;	author wiz;	state Exp;
branches;
next	1.13;
commitid	503fdom2PDeI97XG;

1.13
date	2026.09.12.20.27.32;	author wiz;	state Exp;
branches;
next	1.12;
commitid	ibVyPeNRK3N7anVG;

1.12
date	2026.08.26.10.27.31;	author wiz;	state Exp;
branches;
next	1.11;
commitid	3qeofor0gXtao8TG;

1.11
date	2026.08.23.07.44.35;	author wiz;	state Exp;
branches;
next	1.10;
commitid	mCs7bcys72WeAJSG;

1.10
date	2026.08.19.09.18.48;	author wiz;	state Exp;
branches;
next	1.9;
commitid	ckcSl6Xwwl8xeeSG;

1.9
date	2026.08.19.08.25.29;	author wiz;	state Exp;
branches;
next	1.8;
commitid	x5j8JEyJeltfWdSG;

1.8
date	2026.08.15.08.04.29;	author wiz;	state Exp;
branches;
next	1.7;
commitid	6ob0AIG6GRYUWHRG;

1.7
date	2026.08.08.07.14.51;	author wiz;	state Exp;
branches;
next	1.6;
commitid	2x7bUKr9yCpVTNQG;

1.6
date	2026.08.05.05.35.43;	author wiz;	state Exp;
branches;
next	1.5;
commitid	3zcBYgQTS9IUrpQG;

1.5
date	2026.08.03.17.14.33;	author wiz;	state Exp;
branches;
next	1.4;
commitid	73wrtuiTtQaDndQG;

1.4
date	2026.07.19.07.50.48;	author wiz;	state Exp;
branches;
next	1.3;
commitid	CI59o2EqVtd7KeOG;

1.3
date	2026.07.13.14.35.20;	author wiz;	state Exp;
branches;
next	1.2;
commitid	l67bRCoJluAMavNG;

1.2
date	2026.07.13.10.16.36;	author wiz;	state Exp;
branches;
next	1.1;
commitid	aKYNGOWwcy4BJtNG;

1.1
date	2026.07.02.10.58.21;	author wiz;	state Exp;
branches;
next	;
commitid	cgGQYP2SmaYmk4MG;


desc
@@


1.15
log
@ggml: update to 0.26.0.

ggml v0.26.0 brings a new alloc_buffer_n/get_alloc_size_n buffer
allocation API, sparse flash attention kernels on SYCL, Vulkan and
Metal, and major lightning indexer improvements (halved score
memory, tiling, MUSA support). The CPU backend gains BF16 ops and
a tiled k-quant mul_mat, CUDA gains a model-driven W4A4 (NVFP4/MXFP4)
mul_mat path plus MMVQ shared-expert fusion, and the Hexagon backend
adds a sampler and more quant types. Model loading is faster with
stricter GGUF size validation, Windows ARM64 MSVC builds are enabled,
and WebGPU/OpenVINO/OpenCL/SYCL pick up numerous new kernels, ops
and fixes.
@
text
@# $NetBSD: Makefile,v 1.14 2026/09/26 09:45:32 wiz Exp $

DISTNAME=	ggml-0.26.0
CATEGORIES=	math
MASTER_SITES=	${MASTER_SITE_GITHUB:=ggml-org/}
GITHUB_TAG=	v${PKGVERSION_NOREV}

MAINTAINER=	pkgsrc-users@@NetBSD.org
HOMEPAGE=	https://github.com/ggml-org/ggml/
COMMENT=	Tensor library for machine learning
LICENSE=	mit

USE_TOOLS+=	pkg-config
USE_LANGUAGES=	c c++

USE_CC_FEATURES=	c11
USE_CXX_FEATURES=	c++17

PKGCONFIG_OVERRIDE+=	ggml.pc.in

CMAKE_CONFIGURE_ARGS+=	-DGGML_BLAS=ON
CMAKE_CONFIGURE_ARGS+=	-DGGML_BLAS_VENDOR=OpenBLAS
CMAKE_CONFIGURE_ARGS+=	-DBLAS_LIBRARIES=${CBLAS_LIBS:Q}

BLAS_INDEX64=		yes
BLAS_ACCEPTED=		openblas_pthread #openblas_openmp
BLAS_C_INTERFACE=	yes

SUBST_CLASSES+=		findblas
SUBST_STAGE.findblas=	pre-configure
SUBST_MESSAGE.findblas=	Fixing libpci soname
SUBST_FILES.findblas+=	src/ggml-blas/CMakeLists.txt
SUBST_SED.findblas+=	-e 's,DepBLAS openblas64,DepBLAS ${BLAS_PC},'

.include "options.mk"
.include "../../devel/cmake/build.mk"
.include "../../mk/blas.buildlink3.mk"
.include "../../mk/bsd.pkg.mk"
@


1.14
log
@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.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.13 2026/09/12 20:27:32 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.25.3
@


1.13
log
@ggml: update to 0.23.0.

Version 0.23.0 adds new operators and backend scheduling APIs for
sparse attention, asynchronous execution, and allocation-dependency
tracking. It includes broad correctness and performance work across
CPU, GPU, and accelerator backends, plus an expanded RPC stack with
Apple RDMA support.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.12 2026/08/26 10:27:31 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.23.0
@


1.12
log
@ggml: update to 0.22.0.

This release adds tensor-split support to the multi-backend (meta)
backend with improved split-state propagation, reworks the Metal
kernels into per-op sources with parallel compilation, and fixes
ggml_clamp to be a proper non-in-place op. It also brings new ops
(POOL_1D, PAD_REFLECT_1D), Q2_K SYCL kernels, MoE bias fusion on
OpenCL, and assorted fixes across the CUDA, Metal, SYCL, Vulkan,
OpenCL and WebGPU backends.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.11 2026/08/23 07:44:35 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.22.0
@


1.11
log
@ggml: update to 0.21.0.

Changes not found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.10 2026/08/19 09:18:48 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.21.0
@


1.10
log
@ggml: update to 0.20.2.

Changes not found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.9 2026/08/19 08:25:29 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.20.2
@


1.9
log
@ggml: update to 0.20.1.

No changelog found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.8 2026/08/15 08:04:29 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.20.1
@


1.8
log
@ggml: update to 0.20.0

no summary available
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.7 2026/08/08 07:14:51 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.20.0
@


1.7
log
@ggml: update to 0.19.0.

No summary found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.6 2026/08/05 05:35:43 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.19.0
@


1.6
log
@ggml: update to 0.18.1.

Summary not found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.5 2026/08/03 17:14:33 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.18.1
@


1.5
log
@ggml: update to 0.18.0.

Another two weeks of updates.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.4 2026/07/19 07:50:48 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.18.0
@


1.4
log
@ggml: update to 0.17.0.

No changes summary found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.3 2026/07/13 14:35:20 wiz Exp $
d3 1
a3 1
DISTNAME=	ggml-0.17.0
@


1.3
log
@ggml: remove patch for stable-diffusion.cpp

stable-diffusion.cpp uses local copy anyway.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.2 2026/07/13 10:16:36 wiz Exp $
d3 1
a3 2
DISTNAME=	ggml-0.16.0
PKGREVISION=	1
@


1.2
log
@ggml: update to 0.16.0.

No summary of changes found.
@
text
@d1 1
a1 1
# $NetBSD: Makefile,v 1.1 2026/07/02 10:58:21 wiz Exp $
d4 1
@


1.1
log
@math/ggml: import ggml-0.15.3

Tensor library for machine learning

## Features

- Low-level cross-platform implementation
- Integer quantization support
- Broad hardware support
- Automatic differentiation
- ADAM and L-BFGS optimizers
- No third-party dependencies
- Zero memory allocations during runtime
@
text
@d1 1
a1 1
# $NetBSD$
d3 1
a3 1
DISTNAME=	ggml-0.15.3
@