Skip to content

feat: Backend-specific prefill chunk sizes #39

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

Open
Skyrion9 wants to merge 2 commits into
base: main
Choose a base branch
from
Open

feat: Backend-specific prefill chunk sizes #39

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c3ffaf0
feat: Backend-specific prefill chunk sizes
Skyrion9 Jul 15, 2026
42e1a93
fix(prefill): use runtime backend detection and fix unreachable CPU c…
Skyrion9 Jul 15, 2026
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
47 changes: 39 additions & 8 deletions src/s2_model.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -815,14 +815,45 @@ bool SlowARModel::prefill(const std::vector<int32_t> & flat_tokens, int32_t n_to
backend_requires_single_token_semantic_prefill(backend_gpu_)) {
chunk_tokens = 1;
} else {

const int32_t adaptive_gpu_chunk = std::clamp(
128 / std::max(1, n_gpu_layers_),
8,
64
);
if (n_tokens > adaptive_gpu_chunk) {
chunk_tokens = adaptive_gpu_chunk;
// Backend-specific prefill chunk sizes for naive attention path.
// Vulkan: Limited by ggml_soft_max shader workgroup/shared memory limits
// CUDA/Metal: Highly optimized soft_max kernels handle large batches
// CPU: Memory bandwidth bound, scale with available parallelism
int32_t adaptive_chunk = std::min<int32_t>(256, std::max<int32_t>(64, resolve_n_threads(n_threads) * 16));

if (backend_gpu_ != nullptr && n_gpu_layers_ > 0) {
const char * backend_name = ggml_backend_name(backend_gpu_);
if (backend_name != nullptr) {
std::string name(backend_name);
if (name.find("CUDA") != std::string::npos ||
name.find("Metal") != std::string::npos) {
adaptive_chunk = 512;
} else if (name.find("Vulkan") != std::string::npos) {
adaptive_chunk = 64;
}
}

int32_t semantic_prompt_tokens = 0;
for (int32_t t = 0; t < n_tokens; ++t) {
const int32_t semantic = flat_tokens[static_cast<size_t>(t) * codebook_dim];
if (semantic >= hparams_.semantic_begin_id &&
semantic <= hparams_.semantic_end_id) {
semantic_prompt_tokens++;
if (semantic_prompt_tokens > 1) {
break;
}
}
}

if (semantic_prompt_tokens > 1 &&
backend_requires_single_token_semantic_prefill(backend_gpu_)) {
adaptive_chunk = 1;
}
}

int32_t chunk_tokens = n_tokens;
if (n_tokens > adaptive_chunk) {
chunk_tokens = adaptive_chunk;
}
}
}
Expand Down