Commit graph

4115 commits

Author SHA1 Message Date
Zonghang Li
e543fcd55f fix buf_compute_meta size 2025-01-04 19:16:38 +04:00
Zonghang Li
b2f806a572 increase default max nodes from 8192 to 16384 2025-01-04 11:08:35 +04:00
Zonghang Li
b712ad3044 fix flops test error 2024-12-31 09:37:46 +04:00
Zonghang Li
0a5450487c fix serialize error 2024-12-31 08:24:31 +04:00
Zonghang Li
e7c75d6b4a fix get device os error on linux 2024-12-30 21:03:51 +04:00
Lizonghang
fa31ca8e35 add os detect 2024-12-30 09:13:12 +04:00
Lizonghang
d9beb030ee add EPS in device_compute_delay 2024-12-29 22:31:45 +04:00
Lizonghang
fa210d2034 remove duplicate calls 2024-12-29 22:00:17 +04:00
Lizonghang
a7ec685eda add memcpy speed test 2024-12-29 16:19:08 +04:00
Lizonghang
5b46c4e848 correct GB to GiB 2024-12-28 10:09:43 +04:00
Lizonghang
b2bc836423 reformat 2024-12-28 08:36:24 +04:00
Lizonghang
70811d85b3 remove OOM warning 2024-12-22 11:01:45 +04:00
Lizonghang
b6fd762fa8 fix log 2024-12-21 19:01:00 +04:00
Lizonghang
aa6c352aa2 reverse 128MiB free mem in termux 2024-12-12 21:27:36 +04:00
Lizonghang
1b169f8f37 remove anonpages for repeat count 2024-12-12 20:28:59 +04:00
Zonghang Li
6ea692d974 use small tensors for test on cpu 2024-12-12 15:52:27 +04:00
Lizonghang
b642d70188 fix swappable mem in termux 2024-12-12 15:15:16 +04:00
Lizonghang
fa05a482f8 fix swapout and mmapin delay 2024-12-12 09:58:10 +04:00
Lizonghang
8c491c0fb1 consider swappable mem and swap space in termux 2024-12-11 21:24:37 +04:00
Lizonghang
28ebc6478a use disk.read_rnd.bw in termux 2024-12-11 14:56:08 +04:00
Lizonghang
dfb37671d3 fix memory count and use sequantial disk read speed on linux 2024-12-11 09:30:02 +04:00
Lizonghang
bef2662d42 fix infinite loop in fio test 2024-12-10 20:58:17 +04:00
Lizonghang
8e9ab45458 fix model bytes counter 2024-12-10 14:57:48 +04:00
Zonghang Li
2d79554694 fix device_physical_memory 2024-12-09 14:42:49 +04:00
Zonghang Li
4104eb5b7a fix std::max val type 2024-12-09 13:36:53 +04:00
Zonghang Li
a11e0dd0fa fix device_cgroup_physical_memory in docker container 2024-12-09 13:34:01 +04:00
Lizonghang
d78fa427e7 add memory copy speed test 2024-12-09 10:07:42 +04:00
Lizonghang
1aee5bd6da hide metal log if not debug 2024-12-09 09:17:49 +04:00
Lizonghang
cead0f2e5a device_disk_access_delay: adapt for linux (on termux) 2024-12-08 22:58:08 +04:00
Lizonghang
2c2171cebf fix display 2024-12-08 22:57:12 +04:00
Lizonghang
aad604ec8b add disk access delay while swapping 2024-12-08 20:07:30 +04:00
Zonghang Li
f5243571d1 use sequential vram read test 2024-12-08 10:14:52 +04:00
Zonghang Li
d2dc8fd3b8 set CUDA_READ_BLOCK_SIZE to 512 2024-12-08 10:14:20 +04:00
Zonghang Li
df813675d0 fix flops count and ram/vram speed test 2024-12-08 10:14:05 +04:00
Lizonghang
26c2ffb5b7 set fio read-job startdelay 1.5s 2024-12-06 17:03:42 +04:00
Lizonghang
a79ab3dde3 device_disk_access_delay: fix delay calculation when physical memory is busy (all mapped tensors will be reloaded) 2024-12-06 16:41:55 +04:00
Lizonghang
cd823546dd llama_profile_device: add arg n_predict 2024-12-06 16:37:25 +04:00
Lizonghang
a46d56cc60 llama_model_n_flops: remove ctxs 2024-12-06 11:31:53 +04:00
Lizonghang
f1c1d1b929 add support for Q5_K and fix byte count for Q6_K 2024-12-06 07:59:45 +04:00
Lizonghang
e6f4c009ab device_disk_access_delay: add delay for loading one row from lookup table 2024-12-05 20:38:28 +04:00
Lizonghang
4fd3b6679e fix n_gpu_layers 2024-12-05 20:28:19 +04:00
Lizonghang
3c8dbd11fc device_compute_delay: compatible with the shared memory in Metal 2024-12-05 20:02:13 +04:00
Lizonghang
e74967c488 throw error when using async upload, will support later 2024-12-05 14:18:39 +04:00
Zonghang Li
5a99c5894d update test model, enable warm-up and sched 2024-12-05 09:52:16 +04:00
Lizonghang
b9538f048a add sched in device_flops but disabled by default 2024-12-04 23:54:57 +04:00
Zonghang Li
7521e532c4 device_memory_bw: simulate cache-friendly block access and multi-threading 2024-12-04 15:36:59 +04:00
Zonghang Li
44b4718c8b recover device_memory_access_delay 2024-12-04 12:29:24 +04:00
Zonghang Li
89aad7880e fix sysctlbyname on Linux and Win 2024-12-04 12:24:07 +04:00
Lizonghang
585864b05a add inactive_memory in Mac UMA 2024-12-04 11:56:09 +04:00
Lizonghang
99d48157a0 compatible with Mac ARM64 and x86_64 arch, and ignore memory access delay 2024-12-04 11:33:46 +04:00