Skip to content

fix(hip): skip the fused-path BF16-only marks when GGML_CUDA_DISABLE_FUSION is set #834

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
davide221 merged 1 commit into from
Oct 10, 2026
Merged

fix(hip): skip the fused-path BF16-only marks when GGML_CUDA_DISABLE_FUSION is set #834

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions server/deps/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6282,7 +6282,10 @@ static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t backend,
}
return false;
};
for (int i = 0; i < cgraph->n_nodes; ++i) {
// The xn, gate-mix and routed GLU marks assume the fused kernels consume or produce those tensors; with fusion
// disabled the unfused ops would read or write them and compute_forward aborts, so skip those marks.
static const bool fusion_off = getenv("GGML_CUDA_DISABLE_FUSION") != nullptr;
for (int i = 0; i < cgraph->n_nodes && !fusion_off; ++i) {
const ggml_tensor * op = cgraph->nodes[i];
if (op->op != GGML_OP_HC_COMBINE_NORM) continue;
const size_t xn_off = (size_t) op->ne[0] * op->ne[1] * op->ne[2] * sizeof(float);
Expand Down Expand Up @@ -6311,7 +6314,7 @@ static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t backend,
}

// Only the gate-mix path with a quantized w_up writes bf16 in place; the reduce fallback writes f32.
for (int i = 0; i < cgraph->n_nodes; ++i) {
for (int i = 0; i < cgraph->n_nodes && !fusion_off; ++i) {
ggml_cuda_hc_mix_args ma;
const int cnt = ggml_cuda_hc_mix_closed(cgraph, i, ma);
if (cnt <= 0 || !ma.dst || i < 1 || ggml_nrows(ma.dst) < 512) continue;
Expand Down Expand Up @@ -6339,7 +6342,6 @@ static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t backend,
// The fused routed GLU (gate MMID, up MMID, SWIGLU) writes bf16 in place when every consumer is an MMB
// MUL_MAT_ID, so its F32 form is never stored. Mirror the fusion's own conditions exactly: a marked GLU that ran
// unfused would leave F32 where the consumer reads bf16 (compute_forward aborts on that).
static const bool fusion_off = getenv("GGML_CUDA_DISABLE_FUSION") != nullptr;
for (int i = 2; i < cgraph->n_nodes && !fusion_off; ++i) {
ggml_tensor * glu = cgraph->nodes[i];
if (glu->op != GGML_OP_GLU || (glu->flags & GGML_TENSOR_FLAG_OUTPUT)) continue;
Expand Down
Loading