edax-reversi-AVX

Automatically exported from code.google.com/p/okuharaandroid-edax-reversi

Edax is a strong othello program. Its main features are:

fast bitboard based & multithreaded engine.
accurate midgame-evaluation function.
opening book learning capability.
text based rich interface.
multi-protocol support to connect to graphical interfaces or play on Internet (GGS).
multi-OS support to run under MS-Windows, Linux and Mac OS X.

This is SSE/AVX optimized version of Edax 4.4.0. Functionally equivalent to the parent project, provided no bugs are introduced.

Thanks to AVX2, x64-modern build solves fforum-40-59.obf 60% faster than official edax-4.4 on Haswell, and runs level 30 autoplay 80% faster.

See http://www.amy.hi-ho.ne.jp/okuhara/bitboard.htm and http://www.amy.hi-ho.ne.jp/okuhara/edaxopt.htm for optimization details in Japanese.

1. Mobility (board_sse.c, board_mmx.c)

1.1 new SSE2 version of get_moves

Diagonals are SIMD'd using vertical mirroring by bswap.

Athlon -get_moves_sse
problem\fforum-20-39.obf: 111349635 nodes in 0:07.998 (13922185 nodes/s).
mobility: 81.10 < 81.28 +/- 0.17 < 82.03
Athlon +get_moves_sse
problem\fforum-20-39.obf: 111349635 nodes in 0:07.889 (14114544 nodes/s).
mobility: 71.08 < 71.72 +/- 0.34 < 73.53
Core2 -get_moves_sse
problem/fforum-20-39.obf: 111349635 nodes in 0:10.180 (10938078 nodes/s).
mobility: 78.06 < 78.18 +/- 0.08 < 78.41
Core2 +get_moves_sse
problem/fforum-20-39.obf: 111349635 nodes in 0:09.978 (11159514 nodes/s).
mobility: 60.84 < 61.19 +/- 0.13 < 61.47

1.2 can_move

Now calls SIMD'd get_moves for x86/x64 build.

2. Stability (board.c, board_sse.c, board_mmx.c)

2.1 get_full_lines_h, get_full_lines_v

get_full_lines for horizontal and vertical are simplified. The latter is compiled into rotation instrunction.

2.2 rearranged loop

The last while loop is rearranged not to call bit_count in case stable == 0.

2.3 new SSE2 version with bswap and pcmpeqb

Athlon -get_stability_sse
stability: 90.10 < 90.28 +/- 0.24 < 91.20
Athlon +get_stability_sse
stability: 81.59 < 81.93 +/- 0.73 < 86.25
Core2 -get_stability_sse
stability: 79.24 < 79.39 +/- 0.15 < 79.93
Core2 +get_stability_sse
stability: 71.80 < 71.85 +/- 0.06 < 72.07

2.4 get_corner_stability

Kindergarten version eliminates bit_count call.

2.5 find_edge_stable

Loop optimization and flip using carry propagation. One time execution but affect total solving time.

3. eval.c (4.4.5)

Eval feature calculation using SSE2 / AVX2 (now in eval_sse.c) improves midgame by 15-30% and endgame by 8-12%. Restoring eval from backup instead of rewinding. eval_open (one time execution) is also optimized.

4. hash.c

I think hash->data.move[0] on line 677 should be hash->data.move[1].

5. board_symetry, board_unique (board.c, board_sse.c)

SSE optimization and mirroring reduction. (Not used in solving game)

6. endgame_sse.c (4.4.7)

Keep more variables in SSE registers. SSE optimized count_last_flip. Parity sort by shuffle.

7. board_get_hash_code (4.5.0)

Changed to use CRC32c. This enables hardware acceleration on modern build.

8. AVX2 versions (x64-modern build only)

In many cases AVX2 version is simplest, thanks to variable shift instructions (although they are 3 micro-op instructions).

Benchmarks are on Core i5-4260U (Haswell) 1.4GHz (TB 2.7GHz) single thread.

4.4.0 original x64-modern clang
problem/fforum-20-39.obf: 111349635 nodes in 0:05.726 (19446321 nodes/s).
+optimizations 1-5 above, no-avx2
problem/fforum-20-39.obf: 111349635 nodes in 0:05.342 (20844185 nodes/s).
+get_moves (board_sse.c)
problem/fforum-20-39.obf: 111349635 nodes in 0:05.142 (21654927 nodes/s).
+flip_avx.c
problem/fforum-20-39.obf: 111349635 nodes in 0:04.946 (22513068 nodes/s).
+count_last_flip_sse.c
problem/fforum-20-39.obf: 111349635 nodes in 0:04.906 (22696624 nodes/s).

9. makefile

gcc-old, x86 build should be -m32, not -m64. Some flags and defines added for optimization.