Stormphrax

a work-in-progress UCI chess and chess960 engine, with NNUE evaluation trained from zero knowledge starting with random weights

this project is a continuation of my HCE engine Polaris

Strength

Version	CCRL 40/15	CCRL Blitz	CCRL 40/2 FRC	SPCC	MCERL
2.0.0	-	-	3639	-	-
1.0.0	3254	3364	3513	3344	3342

Features

standard PVS with quiescence search and iterative deepening
- aspiration windows
- check extensions
- countermoves
- futility pruning
- history
  - capture history
  - countermove history (1-ply continuation history)
  - follow-up history (2-ply continuation history)
- internal iterative reduction
- killers (1 per ply)
- late move reductions
- mate distance pruning
- multicut
- nullmove pruning
- reverse futility pruning
- SEE move ordering and pruning
- singular extensions
- Syzygy tablebase support
NNUE
- (768->384)x2->1 architecture
- trained from zero knowledge with reinforcement learning from a randomly-initialised network
BMI2 attacks in the bmi2 build, otherwise fancy black magic
- pext/pdep for rooks
- pext for bishops
lazy SMP
static contempt

To-do

tune search constants
make it stronger uwu

UCI options

Name	Type	Default value	Valid values	Description
Hash	integer	64	[1, 131072]	Memory allocated to the transposition table (in MB).
Clear Hash	button	N/A	N/A	Clears the transposition table.
Threads	integer	1	[1, 2048]	Number of threads used to search.
UCI_Chess960	check	`false`	`false`, `true`	Whether Stormphrax plays Chess960 instead of standard chess.
UCI_ShowWDL	check	`true`	`false`, `true`	Whether Stormphrax displays predicted win/draw/loss probabilities in UCI output.
Move Overhead	integer	10	[0, 50000]	Amount of time Stormphrax assumes to be lost to overhead when making a move (in ms).
SyzygyPath	string	`<empty>`	any path, or `<empty>`	Location of Syzygy tablebases to probe during search.
SyzygyProbeDepth	spin	1	[1, 255]	Minimum depth to probe Syzygy tablebases at.
SyzygyProbeLimit	spin	7	[0, 7]	Maximum number of pieces on the board to probe Syzygy tablebases with.
EvalFile	string	`<internal>`	any path, or `<internal>`	NNUE file to use for evaluation.

Builds

avx512: requires AVX-512 (Zen 4, Skylake-X)
avx2-bmi2: requires BMI2 and AVX2 and assumes fast pext and pdep (i.e. no Zen 1 and 2)
avx2: requires BMI and AVX2 - primarily useful for pre-Zen 3 AMD CPUs back to Excavator
sse41-popcnt: needs SSE 4.1 and popcnt - for older x64 CPUs

Alternatively, build the CMake target stormphrax-native for a binary tuned for your specific CPU (see below)
(note that this does not automatically disable pext and pdep for pre-Zen 3 AMD CPUs that implement them in microcode)

Note:

If you have an AMD Zen 1 (Ryzen x 1xxx) or 2 (Ryzen x 2xxx) CPU, use the avx2 build even though your CPU supports BMI2. These CPUs implement the BMI2 instructions pext and pdep in microcode, which makes them unusably slow for Stormphrax's purposes.

Building

The makefile is not intended for building by users. It exists purely for OpenBench compliance.
Requires CMake and a competent C++20 compiler. Currently, GCC is known to vectorise Stormphrax's NNUE code very poorly compared to Clang.

> cmake -DCMAKE_BUILD_TYPE=Release -S . -B build/
> cmake --build build/ --target stormphrax-<TARGET>

(replace <TARGET> with your preferred target - native/avx512/avx2-bmi2/avx2/sse41-popcnt)

If you have a pre-Zen 3 AMD Ryzen CPU (see the notes in Builds above) and want to build the native target, use these commands instead (the second is unchanged):

> cmake -DCMAKE_BUILD_TYPE=Release -DSP_FAST_PEXT=OFF -S . -B build/
> cmake --build build/ --target stormphrax-native

Disabling the CMake option SP_FAST_PEXT builds the non-BMI2 attack getters.

Credit

Stormphrax uses Fathom for tablebase probing, licensed under the MIT license, and a slightly modified version of incbin for embedding neural network files, under the Unlicense.

The name "Stormphrax" is a reference to the excellent Edge Chronicles :)

jhonnold/Stormphrax