head	1.1;
access;
symbols;
locks; strict;
comment	@// @;


1.1
date	2026.11.04.17.07.04;	author wiz;	state Exp;
branches;
next	;
commitid	EB0QPj3NthnIla2H;


desc
@@


1.1
log
@ggml: update to 0.29.0.

ggml v0.29.0 adds an OCP FP8 E4M3 tensor type, a new
ggml_flash_attn_ext_rows op where each query slice attends to its
own list of K/V rows, a batched tensor copy API, and initial support
for multi-node tensor parallelism through a new multi-process
communicator interface. Core improvements include bounding tensor
views by the bytes their strides reach, zero-copy GGUF export of
host-buffer tensors, and vm_allocate-based allocations on Apple
platforms. Performance work spans all major backends: new SVE
kernels in the CPU backend, DPAS flash attention for quantized KV
in SYCL, fused MoE MMQ in CUDA, faster MXFP4 mat-vec in Metal,
split-K tuning in Vulkan, and a large expansion of quantized
GEMM/GEMV coverage in OpenCL. The Hexagon backend gains support
for several new ops, including the lightning indexer, top-k mask,
conv2d and upscale.
@
text
@$NetBSD$

error: 'isnan' was not declared in this scope; did you mean 'std::isnan'?
error: 'signbit' was not declared in this scope; did you mean 'std::signbit'?
https://github.com/ggml-org/ggml/issues/1681

--- tests/test-quantize-fns.cpp.orig	2026-11-04 17:03:17.251240863 +0000
+++ tests/test-quantize-fns.cpp
@@@@ -177,10 +177,10 @@@@ static uint8_t fp32_to_f8_e4m3_ref(float value) {
 };
 
 static uint8_t fp32_to_f8_e4m3_ref(float value) {
-    if (isnan(value)) {
+    if (std::isnan(value)) {
         return 0x7F;
     }
-    const uint8_t sign = signbit(value) ? 0x80 : 0;
+    const uint8_t sign = std::signbit(value) ? 0x80 : 0;
     const double magnitude = fabs((double) value);
     if (magnitude >= F8_E4M3_VALUES[126]) {
         return sign | 126;
@@@@ -222,8 +222,8 @@@@ static int test_f8_e4m3(bool verbose) {
     for (int i = 0; i < 256; ++i) {
         const bool is_nan = (i & 0x7F) == 0x7F;
         const float expected = is_nan ? NAN : copysignf(F8_E4M3_VALUES[i & 0x7F], i & 0x80 ? -1.0f : 1.0f);
-        const bool decode_failed = is_nan ? !isnan(decoded[i])
-            : decoded[i] != expected || (expected == 0.0f && signbit(decoded[i]) != signbit(expected));
+        const bool decode_failed = is_nan ? !std::isnan(decoded[i])
+            : decoded[i] != expected || (expected == 0.0f && std::signbit(decoded[i]) != std::signbit(expected));
         const bool encode_failed = !is_nan && roundtrip[i] != encoded[i];
         if (decode_failed || encode_failed) {
             num_failed++;
@@@@ -293,7 +293,7 @@@@ static int test_f8_e4m3(bool verbose) {
         x[0] = x.back() = code;
         float actual;
         traits_cpu->vec_dot(x.size(), &actual, 0, x.data(), 0, y.data(), 0, 1);
-        num_failed += !isnan(actual);
+        num_failed += !std::isnan(actual);
     }
 
     if (verbose || num_failed) {
@