

1464c62d88
metal: implement DSv4 Lightning Indexer (#25893)
* metal: implement F16 Lightning Indexer
- Implement GGML_OP_LIGHTNING_INDEXER for 128-dimensional, 64-head inputs
with F32 queries and weights plus F16 keys and masks.
- Add tiled and tail kernels and test KV lengths around 8- and 64-element
boundaries.
llama-bench (--mmap 1, -fa 1, -p 512, -n 128; d=0/10k/20k/30k):
Before:
- pp512: 153.73 ± 0.87 t/s
- tg128: 8.91 ± 0.04 t/s
- pp512 @ d10000: 73.90 ± 0.39 t/s
- tg128 @ d10000: 8.66 ± 0.03 t/s
- pp512 @ d20000: 45.83 ± 0.18 t/s
- tg128 @ d20000: 8.26 ± 0.03 t/s
- pp512 @ d30000: 33.40 ± 0.21 t/s
- tg128 @ d30000: 7.94 ± 0.01 t/s
After:
- pp512: 155.19 ± 0.91 t/s
- tg128: 8.95 ± 0.04 t/s
- pp512 @ d10000: 86.95 ± 0.69 t/s
- tg128 @ d10000: 9.00 ± 0.05 t/s
- pp512 @ d20000: 62.01 ± 0.45 t/s
- tg128 @ d20000: 8.68 ± 0.04 t/s
- pp512 @ d30000: 49.18 ± 0.33 t/s
- tg128 @ d30000: 8.60 ± 0.02 t/s
Assisted-by: Codex
* metal: stage Lightning Indexer K tiles
- Stage and dequantize K in F16 threadgroup memory before simdgroup matrix loads.
- Zero-fill partial tiles and guard stores so all KV segments use the same numerical path.
- Support F32, F16, BF16, Q4_0, Q4_1, Q5_0, Q5_1, and Q8_0 K caches.
llama-bench (--mmap 1, -fa on, -p 512, -n 128; d=0/10k/20k):
- pp512: 160.38 +/- 1.01 t/s
- tg128: 9.08 +/- 0.03 t/s
- pp512 @ d10000: 88.37 +/- 0.46 t/s
- tg128 @ d10000: 9.07 +/- 0.04 t/s
- pp512 @ d20000: 62.53 +/- 0.46 t/s
- tg128 @ d20000: 8.84 +/- 0.03 t/s
Assisted-by: Codex
* dedup Lightning Indexer constants, fix flaky test
* cont : fix whitespace
---------
Co-authored-by: forforever73 <690105611@qq.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2026-08-03 07:33:37 +03:00
..
2026-07-17 11:39:33 +03:00
2026-04-30 17:37:13 +02:00
2026-07-31 01:14:16 +08:00
2026-07-31 11:15:44 +03:00
2026-07-10 12:38:34 +08:00
2026-07-25 15:47:29 -07:00
2026-07-28 07:13:48 +08:00
2026-08-03 07:33:37 +03:00
2026-03-22 11:05:51 +01:00
2026-08-02 11:44:00 -07:00
2026-07-21 22:43:11 +08:00
2026-07-29 15:04:30 +08:00
2026-08-02 15:10:32 +08:00
2026-05-10 17:32:41 +02:00
2026-08-01 14:18:07 -05:00
2026-08-02 08:28:31 +02:00
2026-04-30 17:37:13 +02:00
2026-07-31 09:40:52 +03:00
2026-07-22 17:26:40 +08:00
2026-05-25 12:38:01 +03:00
2026-01-29 12:33:21 -08:00
2026-01-29 12:33:21 -08:00
2026-04-09 16:42:19 +02:00
2026-07-17 00:33:33 +08:00
2026-07-10 12:38:34 +08:00
2026-07-27 14:54:46 +03:00
2026-07-07 12:05:47 -07:00
2026-04-16 17:21:28 +08:00
2026-04-08 17:40:15 +02:00
2026-07-07 12:05:47 -07:00
2026-07-07 12:05:47 -07:00
2024-11-14 18:04:35 +01:00
2024-12-12 19:02:49 +01:00
2026-07-28 16:23:24 +03:00
2025-06-01 13:43:57 +03:00
2026-07-25 14:32:37 +02:00