llama.cpp: update to 0.4.0.

[Thomas Klausner (via pkgsrc-wip) <commit-notify%pkgsrc.org@localhost>]

Module Name:	pkgsrc-wip
Committed By:	Thomas Klausner <wiz%NetBSD.org@localhost>
Pushed By:	wiz
Date:		Sun Sep 13 00:36:03 2026 +0200
Changeset:	d366103f34ecb6fc95a97042487b6e37c653bfe0

Modified Files:
	TODO
	llama.cpp/Makefile
	llama.cpp/Makefile.common
	llama.cpp/PLIST
	llama.cpp/distinfo

Log Message:
llama.cpp: update to 0.4.0.

llama.cpp 0.4.0 adds initial Qwen3.8-Flash-Next and Nemotron-3-Puzzle
support, on-demand tensor reading, per-slot server context limits,
video input options, and a ggml update to 0.23.0 with major sparse
flash attention and RDMA work.

To see a diff of this commit:
https://wip.pkgsrc.org/cgi-bin/gitweb.cgi?p=pkgsrc-wip.git;a=commitdiff;h=d366103f34ecb6fc95a97042487b6e37c653bfe0

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

diffstat:
 TODO                      |  1 -
 llama.cpp/Makefile        |  1 +
 llama.cpp/Makefile.common |  6 +++---
 llama.cpp/PLIST           |  6 +++---
 llama.cpp/distinfo        | 12 ++++++------
 5 files changed, 13 insertions(+), 13 deletions(-)

diffs:
diff --git a/TODO b/TODO
index 6a286a9486..8f82878ccc 100644
--- a/TODO
+++ b/TODO
@@ -88,7 +88,6 @@ Suggested package updates
 	o libressl-2.3.1
 	o lightdm-1.32
 	o lives-2.8.2
-	o llama.cpp-0.4.0
 	o lucene-4.5.0
 	o man-1.6f
 	o mcwm-20130209.2
diff --git a/llama.cpp/Makefile b/llama.cpp/Makefile
index de22635c74..cd27d90aa5 100644
--- a/llama.cpp/Makefile
+++ b/llama.cpp/Makefile
@@ -26,6 +26,7 @@ post-extract:
 	${LN} -s ${WRKDIR}/llama-b${LLAMA_BUILD_NUMBER} ${WRKSRC}/tools/ui/dist
 
 .include "../../www/curl/buildlink3.mk"
+BUILDLINK_API_DEPENDS.ggml+=	ggml>=0.23.0
 .include "../../math/ggml/buildlink3.mk"
 .include "../../devel/cmake/build.mk"
 .include "../../mk/bsd.pkg.mk"
diff --git a/llama.cpp/Makefile.common b/llama.cpp/Makefile.common
index 71ac38ccd2..8a2b1beac2 100644
--- a/llama.cpp/Makefile.common
+++ b/llama.cpp/Makefile.common
@@ -9,6 +9,6 @@ GITHUB_PROJECT=	llama.cpp
 GITHUB_TAG=	v${PKGVERSION_NOREV}
 
 # embedded in version string in binary, and used for downloading UI assets
-LLAMA_VERSION=		0.3.0
-LLAMA_BUILD_NUMBER=	10621
-LLAMA_BUILD_COMMIT=	c1d0e7a004015f23bc0233470b747b596f29b264
+LLAMA_VERSION=		0.4.0
+LLAMA_BUILD_NUMBER=	10809
+LLAMA_BUILD_COMMIT=	5266f24da75dc449bd56cbed7addb9c8e4a6a73e
diff --git a/llama.cpp/PLIST b/llama.cpp/PLIST
index c74089ea92..5fe7d987fe 100644
--- a/llama.cpp/PLIST
+++ b/llama.cpp/PLIST
@@ -51,7 +51,7 @@ lib/libllama-bench-impl.so
 lib/libllama-cli-impl.so
 lib/libllama-common.so
 lib/libllama-common.so.0
-lib/libllama-common.so.0.3.0
+lib/libllama-common.so.${PKGVERSION}
 lib/libllama-completion-impl.so
 lib/libllama-fit-params-impl.so
 lib/libllama-perplexity-impl.so
@@ -59,8 +59,8 @@ lib/libllama-quantize-impl.so
 lib/libllama-server-impl.so
 lib/libllama.so
 lib/libllama.so.0
-lib/libllama.so.0.3.0
+lib/libllama.so.${PKGVERSION}
 lib/libmtmd.so
 lib/libmtmd.so.0
-lib/libmtmd.so.0.3.0
+lib/libmtmd.so.${PKGVERSION}
 lib/pkgconfig/llama.pc
diff --git a/llama.cpp/distinfo b/llama.cpp/distinfo
index 5ba0c028ea..3ae037dd4b 100644
--- a/llama.cpp/distinfo
+++ b/llama.cpp/distinfo
@@ -1,8 +1,8 @@
 $NetBSD$
 
-BLAKE2s (llama-b10621-ui.tar.gz) = ba8de37235d666547806ca02786587e5277e868aa12e1ec8a1d97d2efe6bc134
-SHA512 (llama-b10621-ui.tar.gz) = 76d1c53d84a2588055ce2066fee5a8f2ea8df37384a1d16456d90a09f38a290df860b3a18910b1b512ea6fdba5842ce0b2513bbc3cb6de70cb856584fabfe209
-Size (llama-b10621-ui.tar.gz) = 3082196 bytes
-BLAKE2s (llama.cpp-0.3.0.tar.gz) = ba7d8af96ec1c85e1b98eb970546e7828375aa94699038a7566c3c0990f31e28
-SHA512 (llama.cpp-0.3.0.tar.gz) = c92095a186920110cf9f542d00b004236d92f1ab3f22267bb6ed8fc4b65dc1c32b2705efab202849756d464ed2213c6ee4d5613482f56dad26ae6eb3ba16a91d
-Size (llama.cpp-0.3.0.tar.gz) = 36951021 bytes
+BLAKE2s (llama-b10809-ui.tar.gz) = 17183f2378478e33f029f0cfc11b4fade5e1867bd7a903968f5b87be4a877c4d
+SHA512 (llama-b10809-ui.tar.gz) = 485b4331f547bcf53fa40691e43ca97de90301f656b8176c2ae5091bb2c6d0dcee57439f12a02a06fa5a397f5aba2481e8614c2a06da1396fa8efa58b704c973
+Size (llama-b10809-ui.tar.gz) = 3084118 bytes
+BLAKE2s (llama.cpp-0.4.0.tar.gz) = ff45d6675a4970ac54f2496d31b984f8032e2dad28dc08a9bf5340af5af66701
+SHA512 (llama.cpp-0.4.0.tar.gz) = 12592516c0ac70ed53d64fc0c9f82bbb156defa507c7364f1a753eb29f3498e0e13a282e139c2eff7d2eb00921b4ef107b2fa90a5b66f8a3c45409d4fcb0834f
+Size (llama.cpp-0.4.0.tar.gz) = 37288249 bytes