jirayu_sri/stockfish
1
0
Fork 0
mirror of https://github.com/official-stockfish/Stockfish.git synced 2026-07-22 12:47:08 +00:00
Code Issues Packages Projects Releases Wiki Activity
5,870 Commits 6 Branches 60 Tags
040dfedb3457ca6971d98c754362cde4dc767aff
Commit Graph
4 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
AndrovTandStéphane Nicolet a6d9a302b8 Implement AffineTransformSparseInput for armv8 
Implements AffineTransformSparseInput layer for the NNUE evaluation
for the armv8 and armv8-dotprod architectures. We measured some nice
speed improvements via 10 runs of our benchmark:

armv8, Cortex-X1                  :   18.5% speed-up
armv8, Cortex-A76                 :   13.2% speed-up
armv8-dotprod, Cortex-X1          :   27.1% speed-up
armv8-dotprod, Cortex-A76         :   12.1% speed-up
armv8, Cortex-A72, Raspberry Pi 4 :    8.2% speed-up (thanks Torom!)

closes https://github.com/official-stockfish/Stockfish/pull/4719

No functional change
 2023-08-06 21:22:37 +02:00
AndrovTandJoost VandeVondele a42ab95e1f remove large input specialization 
Removes unused large input specialization for dense affine transform. It has been obsolete since #4612 was merged.

closes https://github.com/official-stockfish/Stockfish/pull/4684

No functional change
 2023-07-16 15:12:21 +02:00
AndrovTandJoost VandeVondele 32d3284df5 Permute master net weights to increase sparsity 
Activation data collection using https://github.com/AndrovT/Stockfish/commit/ac468039ab544b03ad9a22c859a4217729c10a77 run as

bench 16 1 13 varied_1000.epd depth NNUE log.bin

on FENs from https://gist.github.com/AndrovT/7eae6918eb50764227e2bafe7938953c.

Permutation found using https://gist.github.com/AndrovT/359c831b7223c637e9156b01eb96949e.
Uses a greedy algorithm that goes sequentially through the output positions and
chooses a neuron for that position such that the number of nonzero quartets is the smallest.

Net weights permuted using https://gist.github.com/AndrovT/9e3fbaebb7082734dc84d27e02094cb3.

Benchmark:

Result of 100 runs of 'bench 16 1 13 default depth NNUE'
========================================================
base (...kfish-master) =     885869  +/- 7395
test (./stockfish    ) =     895885  +/- 7368
diff                   =     +10016  +/- 2984

speedup        = +0.0113
P(speedup > 0) =  1.0000

Passed STC:
https://tests.stockfishchess.org/tests/view/648866c4713491385c804728
LLR: 2.94 (-2.94,2.94) <0.00,2.00>
Total: 126784 W: 34003 L: 33586 D: 59195
Ptnml(0-2): 283, 13001, 36437, 13358, 313

closes https://github.com/official-stockfish/Stockfish/pull/4620

No functional change.
 2023-06-14 18:36:39 +02:00
AndrovTandJoost VandeVondele 38e61663d8 Use block sparse input for the first layer. 
Use block sparse input for the first fully connected layer on architectures with at least SSSE3.

Depending on the CPU architecture, this yields a speedup of up to 10%, e.g.

```
Result of 100 runs of 'bench 16 1 13 default depth NNUE'

base (...ockfish-base) =     959345  +/- 7477
test (...ckfish-patch) =    1054340  +/- 9640
diff                   =     +94995  +/- 3999

speedup        = +0.0990
P(speedup > 0) =  1.0000

CPU: 8 x AMD Ryzen 7 5700U with Radeon Graphics
Hyperthreading: on
```

Passed STC:
https://tests.stockfishchess.org/tests/view/6485aa0965ffe077ca12409c
LLR: 2.93 (-2.94,2.94) <0.00,2.00>
Total: 8864 W: 2479 L: 2223 D: 4162
Ptnml(0-2): 13, 829, 2504, 1061, 25

This commit includes a net with reordered weights, to increase the likelihood of block sparse inputs,
but otherwise equivalent to the previous master net (nn-ea57bea57e32.nnue).

Activation data collected with https://github.com/AndrovT/Stockfish/tree/log-activations, running bench 16 1 13 varied_1000.epd depth NNUE on this data. Net parameters permuted with https://gist.github.com/AndrovT/9e3fbaebb7082734dc84d27e02094cb3.

closes https://github.com/official-stockfish/Stockfish/pull/4612

No functional change
 2023-06-12 20:41:27 +02:00