Kawrakow ef16135920 Bitnet: trying an alternative iq1_bn grid ...

Faster on CUDA. The scalar version is faster too.
The issue with CUDA is that now I see wild performance
fluctuations. Running llama-bench I can get 220 t/s
for TG-128 one time, and 190 t/s another time, with
uncertaintiers of 1-2 t/s. Same for PP, results are
jumping back-and-fort between ~9500 t/s and ~8900 t/s.
So, basically no reliable measurement at this point,
but for sure faster than the previous version, which was
at around 170-180 t/s.

2024-06-25 11:32:48 +03:00

..

template-instances

CUDA: refactor mmq, dmmv, mmvq (#7716)

2024-06-05 16:53:00 +02:00

acc.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

acc.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

arange.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

arange.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

argsort.cu

CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921)

2024-06-14 18:41:49 +02:00

argsort.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

binbcast.cu

bitnet(scale in a separate tensor): mul -> scale on CUDA

2024-06-22 12:02:52 +03:00

binbcast.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

clamp.cu

Introduction of CUDA Graphs to LLama.cpp (#6766)

2024-05-08 22:55:49 +02:00

clamp.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

common.cuh

Bitnet(1.75 bpw): higher precision fp8 scale

2024-06-22 12:02:52 +03:00

concat.cu

cuda : non-cont concat support (#7610)

2024-05-29 15:38:26 +03:00

concat.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

convert.cu

Bitnet: trying an alternative iq1_bn grid

2024-06-25 11:32:48 +03:00

convert.cuh

llama : add Command R Plus support (#6491)

2024-04-09 11:16:13 +03:00

cpy.cu

Introduction of CUDA Graphs to LLama.cpp (#6766)

2024-05-08 22:55:49 +02:00

cpy.cuh

Introduction of CUDA Graphs to LLama.cpp (#6766)

2024-05-08 22:55:49 +02:00

dequantize.cuh

llama : add Command R Plus support (#6491)

2024-04-09 11:16:13 +03:00

diagmask.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

diagmask.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

dmmv.cu

CUDA: refactor mmq, dmmv, mmvq (#7716)

2024-06-05 16:53:00 +02:00

dmmv.cuh

sync : ggml (#6351)

2024-03-29 17:45:46 +02:00

fattn-common.cuh

CUDA: use tensor cores for MMQ (#7676)

2024-06-10 11:45:13 +02:00

fattn-tile-f16.cu

CUDA: use tensor cores for MMQ (#7676)

2024-06-10 11:45:13 +02:00

fattn-tile-f16.cuh

CUDA: faster large batch FA without tensor cores (#7314)

2024-05-17 18:54:52 +02:00

fattn-tile-f32.cu

CUDA: fix Pascal FA, deq. KV to FP16 for batch > 8 (#7681)

2024-06-01 15:47:04 +02:00

fattn-tile-f32.cuh

CUDA: faster large batch FA without tensor cores (#7314)

2024-05-17 18:54:52 +02:00

fattn-vec-f16.cuh

CUDA: use tensor cores for MMQ (#7676)

2024-06-10 11:45:13 +02:00

fattn-vec-f32.cuh

CUDA: fix broken oob check for FA vec f32 kernel (#7904)

2024-06-12 17:41:51 +02:00

fattn-wmma-f16.cuh

CUDA: use tensor cores for MMQ (#7676)

2024-06-10 11:45:13 +02:00

fattn.cu

CUDA: fix Pascal FA, deq. KV to FP16 for batch > 8 (#7681)

2024-06-01 15:47:04 +02:00

fattn.cuh

ggml : add Flash Attention (#5021)

2024-04-30 12:16:08 +03:00

getrows.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

getrows.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

im2col.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

im2col.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

mma.cuh

CUDA: int8 tensor cores for MMQ (q4_K, q5_K, q6_K) (#7860)

2024-06-11 08:26:07 +02:00

mmq.cu

CUDA: stream-k decomposition for MMQ (#8018)

2024-06-20 14:39:21 +02:00

mmq.cuh

CUDA: stream-k decomposition for MMQ (#8018)

2024-06-20 14:39:21 +02:00

mmvq.cu

Bitnet(2.25 bpw): CUDA

2024-06-22 12:02:52 +03:00

mmvq.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

norm.cu

ggml : fix YARN + add tests + add asserts (#7617)

2024-05-29 20:17:31 +03:00

norm.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

pad.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

pad.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

pool2d.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

pool2d.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

quantize.cu

CUDA: revise q8_1 data layout for mul_mat_q (#7824)

2024-06-09 09:42:25 +02:00

quantize.cuh

CUDA: revise q8_1 data layout for mul_mat_q (#7824)

2024-06-09 09:42:25 +02:00

rope.cu

ggml : refactor rope norm/neox (#7634)

2024-06-05 11:29:20 +03:00

rope.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

scale.cu

Introduction of CUDA Graphs to LLama.cpp (#6766)

2024-05-08 22:55:49 +02:00

scale.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

softmax.cu

CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921)

2024-06-14 18:41:49 +02:00

softmax.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

sumrows.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

sumrows.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

tsembd.cu

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

tsembd.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

unary.cu

Add support for sqrt on CUDA (#7953)

2024-06-17 00:23:04 +02:00

unary.cuh

Add support for sqrt on CUDA (#7953)

2024-06-17 00:23:04 +02:00

upscale.cu

ggml : add ggml_upscale_ext (ggml/814)

2024-05-15 13:23:33 +03:00

upscale.cuh

cuda : refactor into multiple files (#6269)

2024-03-25 13:50:23 +01:00

vecdotq.cuh

Bitnet: trying an alternative iq1_bn grid

2024-06-25 11:32:48 +03:00