9393dce1ff
Take v0.19.3 of perf libs which improves sigverify perf 2x ( #11894 )
2020-08-28 12:34:25 -07:00
1733f6c9e9
Update perf libs to 0.19.1 ( #11301 )
...
* fix for cuda sigverify check for small order pubkey
2020-07-31 09:19:43 -07:00
0de6c444d6
Simd poh ( #10604 )
...
* Simd poh
* Fix poh verify bench
2020-06-16 23:03:26 -07:00
0d38257dbb
Fix perf-libs version detection ( #10571 )
...
automerge
2020-06-14 12:35:06 -07:00
d5c889d6b0
Re-enable gpu sigverify ( #9870 )
...
Add sigverify fuzz and scalar test
2020-05-06 15:44:55 -07:00
dc91698b3a
Cache solana-perf.tgz to speed up CI ( #9360 )
...
automerge
2020-04-07 13:13:45 -07:00
b5125479ec
Bump perf libs to v0.18.0 for CUDA 10.2 support
2020-01-24 21:39:49 -07:00
ecdea54203
Bump perf-libs version to v0.16.2 to get cudaMalloc opt ( #7478 )
2019-12-13 17:20:06 -08:00
9ea398416e
Sign shreds on the GPU ( #6595 )
...
* sign gpu shreds
* wip
* checks
* tests build
* test
* tests
* test
* nits
* sign cpu test
* write out the sigs in parallel
* clippy
* cpu test
* prepare secret for gpu
* woot!
* update
* bump perf libs
2019-11-02 06:23:14 -07:00
b4da83a3ab
Remove CUDA feature ( #6094 )
2019-09-26 13:36:51 -07:00
ee0c570d54
Rework solana-validator-cuda to automatically prepare the perf-libs env ( #5849 )
...
automerge
2019-09-08 21:20:08 -07:00
fbea9d8621
Page-pin packet memory for cuda ( #4250 )
...
* Page-pin packet memory for cuda
Bring back recyclers and pin offset buffers
* Add packet recycler to streamer
* Add set_pinnable to sigverify vecs to pin them
* Add packets reset test
* Add test for recycler and reduce the gc lock critical section
* Add comments/tests to cuda_runtime
* Add recycler to recv_blobs path.
* Add trace/names for debug and PacketsRecycler to bench-streamer
* Predict realloc and unpin beforehand.
* Add helper to reserve and pin
* Cap buffered packets length
* Call cuda wrapper functions
2019-06-27 09:32:32 +02:00
36aa876833
Avoid linking with CUDA directly
2019-06-21 15:26:22 -07:00
8bd1c57448
Defer CUDA selection to env.sh, also always create env.sh
2019-06-19 08:47:27 -07:00
6bc0d2a0cb
exit with success even if no CUDA version detected
2019-06-18 21:18:13 -07:00
5683282c94
Update to solana-perf-libs v0.14.0, with support for both CUDA 10.0 and 10.1
2019-06-18 10:41:03 -07:00
fe04fb4cd3
Refetch perf-libs when the release version is changed ( #4706 )
...
automerge
2019-06-17 14:31:41 -07:00
0c4cb76acf
Add GPU based PoH verification ( #4524 )
...
* Add GPU poh verify
* Switch to single PoH verify function
* Add EntrySlice verify tests with hashes and txs
* Add poh-verify benchmarks
2019-06-08 10:21:43 -06:00
c2dcbee6af
cd within the subshell
2019-05-24 18:10:25 -07:00
f56955a17c
Use absolute path to env.sh
2019-05-24 12:27:12 -07:00
45642c4da1
Add path to env.sh
2019-05-24 09:56:07 -07:00
8eac199e8b
Include perf-libs in release tarball ( #4422 )
2019-05-24 09:28:52 -07:00
1d7a758c97
Bump perf libs version to fix out buffer sizing ( #4385 )
2019-05-22 13:00:03 -07:00
408bdbce7a
Add non_default_stream parameter to cuda_verify ( #4079 )
2019-04-30 13:34:46 -07:00
3543a9a49f
Add check for missing signature with fee'ed transaction
...
And update fetch-perf-libs version
2018-11-15 16:23:13 -08:00
10e1e0c125
Switch to perf-libs v0.11.0 for CUDA 10 support
2018-11-12 20:58:52 -08:00
d0ebee5e3b
Correct path to solana-perf-CUDA_HOME.txt
2018-11-12 19:17:54 -08:00
aa7c741ec0
Switch to perf-libs v0.10.6
2018-11-12 19:17:54 -08:00
9e7b9487b0
perf-libs now drives setting CUDA_HOME
2018-11-12 18:49:15 -08:00
def7d156f6
codemod --extensions sh '#!/usr/bin/env bash -e' '#!/usr/bin/env bash\nset -e'
2018-11-11 16:24:36 -08:00
33aab094ef
codemod --extensions sh '#!/bin/bash' '#!/usr/bin/env bash'
2018-11-11 16:24:36 -08:00
cf6f344ccc
Add CUDA_HOME env var to permit overriding the CUDA install location
2018-11-11 16:24:18 -08:00
c9c1564d26
Fetch v0.10.5 perf libs ( #1727 )
...
- includes SGX enclave for signing
2018-11-06 14:20:22 -08:00
7cb5c0708b
Fetch v0.10.4 which has v100 binary compiled in
...
This may or may not fix high latencies seen on the snap build on v100.
GPU driver will not have to JIT the device code for V100 though which
is an improvement.
2018-10-30 18:06:16 -07:00
9a6a399a29
Bump version number to pick up fixed cuda library
...
Has fix for unaligned memory access in chacha_encrypt_many_sample
function.
2018-10-26 14:57:14 -07:00
1e53f4266a
Fetch perf-libs with configurable packet size
...
sig verify library uses passed in size directly
to get packet size, so rust side can be modified
without changing cuda library.
2018-10-25 08:26:35 -07:00
ead7f4287a
Storage mining fixups...
...
* Use IV to make unique identies
* Use hex! macro for hex literal and not string converted to u8 slice
* fix sha sampling to control init/end of sha state
2018-10-24 09:58:41 -07:00
37a0b7b132
Initial validator code for rust side hooks for chacha cuda parallel encrypt
2018-10-18 13:50:19 -07:00
aa2a3fe201
Add chacha module to encrypt ledger files
2018-09-28 14:23:37 -07:00
e050160ce5
Use tagged perf-libs to enable controlled updates
2018-09-18 09:21:44 -07:00
7aeef27b99
not quite banishing build.rs, but better
2018-08-16 22:33:31 -07:00
771478bc68
Add simple CUDA version check, warn on mismatch
2018-06-21 13:42:06 -07:00
89fc0ad7a9
Add convenience script to download performance libraries
2018-06-20 16:48:32 -07:00
