Document and refine MoE hot-cache PP optimization path#1
Merged
Conversation
Add dedicated documentation for the prompt-processing hot-cache work, including a PP architecture explainer with class and flow diagrams and a PP optimization journey page with benchmark results. Refine the PP worklist path by replacing per-layer expert-major offset vectors with fixed local arrays, reducing allocator activity in the prompt-processing hot path without changing behavior. Record the full Qwen3.6 PP benchmark history, including kept optimizations and removed experiments such as hot branch scatter-add and compact worklist field flags, with rationale for each decision. Also update the MoE hot-cache documentation navigation so the new PP pages are linked from the existing docs. Updated docs Updated llama-bench with necessary arguments, to verify experiments feat: speed up PP
fix: window linker problem fix: other linkers fix: next round
adrianhoehne
pushed a commit
that referenced
this pull request
Jun 30, 2026
* Sycl tp stage1 (#1) * SYCL: tensor parallelism (--split-mode tensor) for dual-GPU Adds the comm_init/comm_free/comm_allreduce_tensor trio that the meta-backend queries via get_proc_address to enable backend-specific all-reduce, mirroring the pattern used by ggml-cuda.cu. For N=2 (the common dual-GPU case) implements a degenerate ring all-reduce with two size-branched paths: * Small (nelem < 32768): FP32 direct memcpy + per-device ADD kernel chained via depends_on(memcpy_event). 4 SYCL submissions/call. * Large (nelem >= 32768): BF16-compressed. Each device compresses FP32 -> BF16 in a local outbox, cross-device memcpys to the peer's inbox (HALF the PCIe bytes), then decompresses + adds into the local FP32 partial. 6 SYCL submissions/call but PCIe bytes halved -- wins for any tensor where PCIe dominates kernel time. Threshold and BF16 path pattern mirror the CUDA NCCL allreduce. Storage: ONE persistent uint8_t buffer per device, 4 * nelem bytes (matches both path layouts: FP32 nelem floats; BF16 outbox+inbox = 2 * nelem uint16_t each). Single alloc+free per device keeps the SYCL pool's strict-LIFO invariant trivial. Initial impl handles N=2 FP32 contiguous tensors. Other cases return false, causing the meta-backend to use its generic butterfly fallback. Per-call sync is intentionally omitted. SYCL in-order queue semantics ensure that the meta-backend's next compute on the same per-device queue waits for our final ADD, and the next allreduce's first op on the same persistent buffer waits via the same queue. Only comm_free does an explicit final wait. OneCCL is NOT used: OneCCL 2021.17 hardcodes single-device-per-process in communicator_impl.hpp:47 (condition devices.size() == 1), which is incompatible with llama.cpp's single-process multi-GPU model. Measured on dual Intel Arc Pro B70 (NEO 26.05.x, oneAPI 2025.3 + DPC++ nightly): Llama-3.3-70B Q4_K_M, -sm tensor -fa 1 -ctk f16 -ctv f16: pp512 = 377.08 t/s (vs 313.65 layer mode = +20.2%) tg128 = 17.40 t/s (vs 9.74 layer mode = +78.6%) Qwen3-Coder-Next-80B-A3B Q3_K_M (MoE): pp512 = 216.56 t/s (vs 156.58 meta-backend butterfly = +38.3%) tg128 = 17.60 t/s (vs 14.31 meta-backend butterfly = +23.0%) Qwen3-4B Q4_K_M: pp64 = 984.51 t/s, tg16 = 49.29 t/s Llama-3.3-70B in SYCL TP now comfortably beats production layer mode on both prefill and decode. Coder-Next-80B-A3B (MoE) also wins on both — the BF16 path is what unlocks the many-medium-allreduces prefill pattern. Build/CMake: no changes. No new dependencies. ~210 lines added across ggml-sycl.h and ggml-sycl.cpp. * Fix comments * documentation update to address PR feedback * Bring over my device-to-device memcpy chagnes * move the dev2dev_memcpy calls to the upstream 7-parameter variety * Fix a typo and remove a trailing whitespace
adrianhoehne
pushed a commit
that referenced
this pull request
Jul 5, 2026
* Sycl tp stage1 (#1) * SYCL: tensor parallelism (--split-mode tensor) for dual-GPU Adds the comm_init/comm_free/comm_allreduce_tensor trio that the meta-backend queries via get_proc_address to enable backend-specific all-reduce, mirroring the pattern used by ggml-cuda.cu. For N=2 (the common dual-GPU case) implements a degenerate ring all-reduce with two size-branched paths: * Small (nelem < 32768): FP32 direct memcpy + per-device ADD kernel chained via depends_on(memcpy_event). 4 SYCL submissions/call. * Large (nelem >= 32768): BF16-compressed. Each device compresses FP32 -> BF16 in a local outbox, cross-device memcpys to the peer's inbox (HALF the PCIe bytes), then decompresses + adds into the local FP32 partial. 6 SYCL submissions/call but PCIe bytes halved -- wins for any tensor where PCIe dominates kernel time. Threshold and BF16 path pattern mirror the CUDA NCCL allreduce. Storage: ONE persistent uint8_t buffer per device, 4 * nelem bytes (matches both path layouts: FP32 nelem floats; BF16 outbox+inbox = 2 * nelem uint16_t each). Single alloc+free per device keeps the SYCL pool's strict-LIFO invariant trivial. Initial impl handles N=2 FP32 contiguous tensors. Other cases return false, causing the meta-backend to use its generic butterfly fallback. Per-call sync is intentionally omitted. SYCL in-order queue semantics ensure that the meta-backend's next compute on the same per-device queue waits for our final ADD, and the next allreduce's first op on the same persistent buffer waits via the same queue. Only comm_free does an explicit final wait. OneCCL is NOT used: OneCCL 2021.17 hardcodes single-device-per-process in communicator_impl.hpp:47 (condition devices.size() == 1), which is incompatible with llama.cpp's single-process multi-GPU model. Measured on dual Intel Arc Pro B70 (NEO 26.05.x, oneAPI 2025.3 + DPC++ nightly): Llama-3.3-70B Q4_K_M, -sm tensor -fa 1 -ctk f16 -ctv f16: pp512 = 377.08 t/s (vs 313.65 layer mode = +20.2%) tg128 = 17.40 t/s (vs 9.74 layer mode = +78.6%) Qwen3-Coder-Next-80B-A3B Q3_K_M (MoE): pp512 = 216.56 t/s (vs 156.58 meta-backend butterfly = +38.3%) tg128 = 17.60 t/s (vs 14.31 meta-backend butterfly = +23.0%) Qwen3-4B Q4_K_M: pp64 = 984.51 t/s, tg16 = 49.29 t/s Llama-3.3-70B in SYCL TP now comfortably beats production layer mode on both prefill and decode. Coder-Next-80B-A3B (MoE) also wins on both — the BF16 path is what unlocks the many-medium-allreduces prefill pattern. Build/CMake: no changes. No new dependencies. ~210 lines added across ggml-sycl.h and ggml-sycl.cpp. * Fix comments * documentation update to address PR feedback * Bring over my device-to-device memcpy chagnes * move the dev2dev_memcpy calls to the upstream 7-parameter variety * Fix a typo and remove a trailing whitespace
adrianhoehne
pushed a commit
that referenced
this pull request
Jul 5, 2026
* oai moe * compat with new checkpoint * add attn sink impl * add rope scaling yarn * logits match with latest transformers code * wip chat template * rm trailing space * use ggml_scale_bias * rm redundant is_swa_all * convert interleaved gate_up * graph : fix activation function to match reference (#7) * vocab : handle o200k_harmony special tokens * ggml : add attention sinks support (#1) * llama : add attn sinks * ggml : add attn sinks * cuda : add attn sinks * vulkan : add support for sinks in softmax remove unnecessary return * ggml : add fused swiglu_oai op (#11) * ggml : add fused swiglu_oai op * Update ggml/src/ggml-cpu/ops.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update CUDA impl * cont : metal impl * add vulkan impl * test-backend-ops : more test cases, clean up * llama : remove unfused impl * remove extra lines --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> * repack mxfp4 upon conversion * clean up a bit * enable thinking * add quick hack to render only some special tokens * fix bf16 conversion * remove vocab hack * webui ok * support chat parsing for gpt-oss * fix webui * direct mapping mxfp4, FINALLY * force using mxfp4 * properly use lazy tensor * ggml : add mxfp4 ggml : use e8m0 conversion instead of powf Co-authored-by: Diego Devesa <slarengh@gmail.com> change kvalues_mxfp4 table to match e2m1 (#6) metal : remove quantization for now (not used) cuda : fix disabled CUDA graphs due to ffn moe bias vulkan : add support for mxfp4 cont : add cm2 dequant * ggml : add ggml_add_id (ggml-org#13) * ggml : add ggml_add_id * add cuda impl * llama : add weight support check for add_id * perf opt * add vulkan impl * rename cuda files * add metal impl * allow in-place ggml_add_id * llama : keep biases on CPU with --cpu-moe * llama : fix compile error ggml-ci * cuda : add fallback for __nv_cvt_e8m0_to_bf16raw ggml-ci * cleanup ggml-ci * sycl : fix supports_op for MXFP4 ggml-ci * fix Unknown reasoning format * ggml-cpu : fix AVX build ggml-ci * fix hip build ggml-ci * cuda : add mxfp4 dequantization support for cuBLAS ggml-ci * ggml-cpu : fix mxfp4 fallback definitions for some architectures ggml-ci * cuda : fix version required for __nv_cvt_e8m0_to_bf16raw --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> Co-authored-by: slaren <slarengh@gmail.com>
adrianhoehne
pushed a commit
that referenced
this pull request
Jul 5, 2026
…gml-org#16038) Initalizing RESERVED_NAME in is_reserved_name() is not thread safe and leads to corrupted memory when used from multiple threads as can be seen in the asan trace below. This fixes the initialization to make it thread-safe. #0 0x000100abd018 in std::__1::pair<std::__1::__hash_iterator<std::__1::__hash_node<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, void*>*>, bool> std::__1::__hash_table<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>>::__emplace_unique_key_args<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&>(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) __hash_table:1565 #1 0x000100ab0320 in SchemaConverter::visit(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) json-schema-to-grammar.cpp:802 #2 0x000100aafc48 in std::__1::__function::__func<build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&)::$_2, std::__1::allocator<build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&)::$_2>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> (std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>::operator()(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&) function.h:319 #3 0x000100a2c938 in std::__1::__function::__func<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0::operator()(common_grammar_builder const&) const::'lambda'(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&), std::__1::allocator<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0::operator()(common_grammar_builder const&) const::'lambda'(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>, void (nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>::operator()(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&) function.h:319 #4 0x000100a139f8 in foreach_function(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&, std::__1::function<void (nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)> const&) chat.cpp:762 #5 0x000100a2a7f4 in std::__1::__function::__func<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0, std::__1::allocator<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0>, void (common_grammar_builder const&)>::operator()(common_grammar_builder const&) function.h:319 #6 0x000100aa98f4 in build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&) json-schema-to-grammar.cpp:982 #7 0x0001009c9314 in common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool) chat.cpp:1110 #8 0x0001009b8afc in common_chat_templates_apply_jinja(common_chat_templates const*, common_chat_templates_inputs const&) chat.cpp:1992 #9 0x0001009b533c in common_chat_templates_apply(common_chat_templates const*, common_chat_templates_inputs const&) chat.cpp:2074 #10 0x000100810120 in llamacpp_apply_chat_template+0x724 (predict_oai-98384e17fb94e863:arm64+0x100090120) ... ==45482==Register values: x[0] = 0x00006020004147f8 x[1] = 0x00006080000013c8 x[2] = 0x0000000000000000 x[3] = 0x0000604006289738 x[4] = 0x0000000000000002 x[5] = 0x0000000000000001 x[6] = 0x04034000004b4000 x[7] = 0x0000000000000001 x[8] = 0xbebebebebebebebe x[9] = 0x17d7d7d7d7d7d7d7 x[10] = 0x00000c04000828ff x[11] = 0x0000000000000001 x[12] = 0x000000002018d383 x[13] = 0x0000000000000000 x[14] = 0xfa0000000000fafa x[15] = 0x000010700001ffff x[16] = 0x000000019dc012c0 x[17] = 0x00000001021284f8 x[18] = 0x0000000000000000 x[19] = 0x00000001700acdc0 x[20] = 0x0000000000000002 x[21] = 0x000000002018d384 x[22] = 0x16dd16fd2e731151 x[23] = 0x0000007000020000 x[24] = 0x0000000100c69c08 x[25] = 0x0000000100c69c20 x[26] = 0x00006080000013c7 x[27] = 0x0000000100c69c00 x[28] = 0x00000001700acd60 fp = 0x00000001700aceb0 lr = 0x0000000100abce30 sp = 0x00000001700acd60 AddressSanitizer can not provide additional info. SUMMARY: AddressSanitizer: SEGV __hash_table:1565 in std::__1::pair<std::__1::__hash_iterator<std::__1::__hash_node<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, void*>*>, bool> std::__1::__hash_table<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>>::__emplace_unique_key_args<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&>(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) Thread T5 created by T0 here: #0 0x0001020b99d4 in pthread_create+0x5c (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x359d4) #1 0x000100873910 in std::sys::pal::unix::thread::Thread::new::h77254fdd87a28e05+0x118 (predict_oai-98384e17fb94e863:arm64+0x1000f3910) #2 0x0001007c7a1c in test::run_test::haeb3c2bcd5ed6cf6+0x76c (predict_oai-98384e17fb94e863:arm64+0x100047a1c) #3 0x0001007aedb0 in test::console::run_tests_console::he9d142d704f3a986+0x149c (predict_oai-98384e17fb94e863:arm64+0x10002edb0) #4 0x0001007c5758 in test::test_main::hf86a5e20735245b9+0x118 (predict_oai-98384e17fb94e863:arm64+0x100045758) #5 0x0001007c5da0 in test::test_main_static::h61ee9c8fd30abca0+0x54 (predict_oai-98384e17fb94e863:arm64+0x100045da0) ... ==45482==ABORTING
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add dedicated documentation for the prompt-processing hot-cache work, including
a PP architecture explainer with class and flow diagrams and a PP optimization
journey page with benchmark results.
Refine the PP worklist path by replacing per-layer expert-major offset vectors
with fixed local arrays, reducing allocator activity in the prompt-processing
hot path without changing behavior.
Record the full Qwen3.6 PP benchmark history, including kept optimizations and
removed experiments such as hot branch scatter-add and compact worklist field
flags, with rationale for each decision.
Also update the MoE hot-cache documentation navigation so the new PP pages are
linked from the existing docs.
Updated docs
Updated llama-bench with necessary arguments, to verify experiments
feat: speed up PP