From: Pascal <redacted>
Date: Wed, 5 Aug 2026 22:47:59 +0000 (+0200)
Subject: mtmd/ggml: add ggml_build_forward_order (llama/26649)
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=60894f1e3766ef0224b57ad8840a83a66e05cb6f;p=pkg%2Fggml%2Fsources%2Fwhisper.cpp

mtmd/ggml: add ggml_build_forward_order (llama/26649)

* ggml: add ggml_build_forward_order

ggml_build_forward_expand marks the tensor and all its ancestors for
compute, so using it as a pure ordering hint (keeping q, k and v
together) defeats ggml_build_forward_select: the unselected branch is
forced to run with inputs that were never uploaded. In the mtmd audio
graph this makes GEN_WAV calls execute the GEN_CODE branch with a
stale inp_code0, hitting the get_rows bound assert on CPU.

Add ggml_build_forward_order, which inserts nodes without the compute
flag; the flag is restored when the branch is actually selected.
Switch the q/k/v hints in clip_graph::build_attn to it.

* nit: reduce comments (AGENTS.md)
---

diff --git a/ggml/include/ggml.h b/ggml/include/ggml.h
index 35f0c44e..5cb49d0e 100644
--- a/ggml/include/ggml.h
+++ b/ggml/include/ggml.h
@@ -2788,6 +2788,12 @@ extern "C" {
             struct ggml_cgraph * cgraph,
             struct ggml_tensor * tensor);
 
+    // add the tensor and its parents to the graph without marking them for compute
+    // the flag is set later, when the tensor is reached from a node that computes
+    GGML_API void ggml_build_forward_order(
+            struct ggml_cgraph * cgraph,
+            struct ggml_tensor * tensor);
+
     GGML_API void ggml_build_backward_expand(
         struct ggml_context *  ctx,        // context for gradient computation
         struct ggml_cgraph  *  cgraph,
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index 59191c66..da7f3a5f 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -7200,6 +7200,10 @@ void ggml_build_forward_expand(struct ggml_cgraph * cgraph, struct ggml_tensor *
     ggml_build_forward_impl(cgraph, tensor, true, true);
 }
 
+void ggml_build_forward_order(struct ggml_cgraph * cgraph, struct ggml_tensor * tensor) {
+    ggml_build_forward_impl(cgraph, tensor, true, false);
+}
+
 void ggml_build_backward_expand(
         struct ggml_context *  ctx,
         struct ggml_cgraph  *  cgraph,