kernel: validate cuBLAS dead-end (sm_80 fallback) + W4A16 Marlin impl plan

Decisive DGX experiment: rebuilt with -DGGML_CUDA_FORCE_CUBLAS (it's a compile
#ifdef, not the runtime env we'd been setting - so prior 'cuBLAS no-op' tests
never engaged it). Real result: cuBLAS is SLOWER than MMQ for dense Q4 (pp2048
690 vs 750) and runs an Ampere cutlass_80_tensorop kernel - CUDA-13 has no sm_121
GEMM, falls back to sm_80. So both MMQ and cuBLAS sit at ~46 TFLOP/s; no library
shortcut to the 213 ceiling on GB10. Confirms a hand-tuned sm_120a kernel is
required. Added the phased W4A16 Marlin-style implementation plan (P0 harness ->
P5 enable) as the committed multi-week build; corrected the cuBLAS note.

Assisted-by: Claude:opus-4.8 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto
2026-06-20 21:16:13 +00:00
parent 76cc0b6abc
commit 13e6ee89c7
2 changed files with 63 additions and 0 deletions

View File

@@ -101,3 +101,5 @@ GB10 peaks (measured): forums.developer.nvidia.com/t/351993, /360142, /373618. M
arxiv 2408.11743, developers.redhat.com Marlin/Machete. MMQ untuned: llama.cpp docs/build.md, discussions/16578,
DandinPower/llama.cpp_bench. FP4 landing/sm121: llama.cpp PR #17906/#20644, issues #19662/#18331. Roofline:
vllm.ai/blog/2026-06-01-vllm-dgx-spark, lmsys.org DGX Spark.
> **Correction (measured):** the earlier `GGML_CUDA_FORCE_CUBLAS` env test was a no-op because it's a *compile-time* `#ifdef`, not a runtime flag — cuBLAS never engaged. A real rebuild with `-DGGML_CUDA_FORCE_CUBLAS=ON` shows cuBLAS is **slower** than MMQ for dense Q4 (pp2048 690 vs 750) and runs an **Ampere `cutlass_80_tensorop` FP16 kernel** — cuBLAS-13.0 has no sm_121-tuned GEMM and falls back to sm_80. So *both* MMQ and cuBLAS sit at ~46 TFLOP/s (~21% of the 213 BF16 peak); there is **no library shortcut** to the ceiling on GB10 — a hand-tuned sm_120a kernel (Marlin-style) is required.

View File

@@ -0,0 +1,61 @@
# W4A16 Marlin-style GEMM for ggml-cuda on Blackwell (sm_120/121) — implementation plan
The committed multi-week kernel. Goal: get 4-bit-weight dense matmul to the GB10 **BF16 ceiling (~213
TFLOP/s ≈ ~3,300 t/s prefill on Qwen3-32B)**, ~4.3× over today's 765. This is the *match-vLLM* path; vLLM's
own GB10 dense throughput runs on W4A16 Marlin (its FP4 path is broken on sm_121).
## Why a custom kernel (validated, not assumed)
On GB10 (sm_121), measured: **both** llama-MMQ (int8, Ampere-tuned) **and** cuBLAS-FP16 sit at ~46 TFLOP/s
(~21% of peak). cuBLAS falls back to an Ampere `cutlass_80_tensorop` kernel (CUDA-13 has no sm_121 GEMM for
these shapes); rebuilt with `-DGGML_CUDA_FORCE_CUBLAS=ON` it's *slower* than MMQ (690 vs 750). **No library
path reaches the ceiling on consumer Blackwell** — a hand-tuned sm_120a kernel is required. `mmapeak` measures
the 213 BF16 peak as reachable, and vLLM's Marlin hits it, so the ceiling is real; the work is reaching it.
## What Marlin does (the design we mirror)
Weights stored 4-bit, **dequantized in-register/shared-mem** in-flight; GEMM math on **FP16/BF16 tensor
cores** (`mma.sync m16n8k16`). Speed comes from: `cp.async` global→shared with a **multi-stage double-buffered
pipeline**, **offline weight reshuffle** into the MMA-friendly layout, activations kept resident in registers,
and **Stream-K** partitioning. Sources: IST-DASLab/marlin, arXiv 2408.11743, vLLM machete (Hopper successor).
## Phases (each ends with: numerical parity vs MMQ + a prefill benchmark)
### P0 — Harness + baseline (do first)
- Add a `test-backend-ops` MUL_MAT case for Q4_K/Q4_0 at prefill shapes (M=512/2048) — gives a numerical
reference and a microbench. Confirm baseline ~46 TFLOP/s.
- Model-level gate: token-identical greedy generation (Qwen3) before/after, like the paged Gate 0.
- Deliverable: a red/green parity check the kernel must pass at every phase.
### P1 — Dispatch seam (no behavior change)
- New `ggml/src/ggml-cuda/marlin-w4a16.cu` + a gated hook in `ggml_cuda_mul_mat` (dense, non-ids path),
behind `GGML_CUDA_W4A16` + sm_120/121 + type∈{Q4_0,Q4_K}. Initially returns false → falls back to MMQ.
(Mirror of the `fp4-grouped-moe.cu` scaffold seam.) Builds byte-identical by default.
### P2 — Correctness-first kernel (slow OK)
- Dequant Q4→BF16 (reuse ggml's `dequantize_block_q4_K`) into shared mem, naive `mma.sync m16n8k16` BF16
accumulate, small tiles. Goal: **bit-parity vs MMQ** (within fp tol) on the toy + the real model. Establishes
the data plumbing + the harness pass. Not expected to beat MMQ yet.
### P3 — The Marlin pipeline (the speedup)
- `cp.async` double/triple-buffered global→shared; offline weight reshuffle (a one-time repack of the Q4
tensor into the mma+pipeline layout — likely a load-time transform or a new tensor variant); register-
resident activation tiles; Stream-K split for the prefill M. Target: ≥150 TFLOP/s (≥~2,300 t/s), then ~213.
### P4 — Tune
- Tile (mmq_x/y analogues), warps, pipeline depth, occupancy. We have nsys (throughput) but **not ncu** on the
DGX — tuning is empirical (sweep configs, measure t/s). Note ncu would need sudo/driver perms we lack.
### P5 — Enable
- Default on for sm_120/121 + Q4_0/Q4_K dense when parity holds + faster; keep the flag as an escape hatch.
Ship as a LocalAI llama.cpp patch (the patches/ series) and/or upstream (ggml has no Marlin-equivalent —
issue #1519 — so it's net-new upstream value; float it with maintainers first).
## Risks / notes
- **Multi-week, expert-CUDA, DGX-only** (GB10 is the only sm_121). The session's network flakiness +
`llama-cli` hang make `llama-bench`/`test-backend-ops` the reliable verification tools (both work).
- Quantization correctness: Q4_K's superblock structure (256-elem, 6-bit scales) is more complex to dequant
in-kernel than Q4_0; consider landing Q4_0 first, then Q4_K.
- **Beat-path follow-on:** the FP4-MMA path (`mul_mat_q<MXFP4>`, ~5% of FP4 peak) tuned/fixed on sm_121 reaches
~6,600 (2× BF16). Separate track; this W4A16 kernel is the match-path foundation.
- Reuse ggml's `mma.cuh` tile abstractions (MMQ already uses them) rather than raw PTX where possible.