Question to Crafty's search output

[Joerg Oster]

Running Crafty in a terminal, 4 Threads, start position, I get the following output:

Code: Select all

               21    17.84   0.22   1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. d4
                                    exd4 5. Nxd4 Bb4 6. Nxc6 bxc6 7. Bd3
                                    d5 8. exd5 Qe7+ 9. Be3 Bxc3+ 10. bxc3
                                    Nxd5 11. Qf3 O-O
               21->  22.15   0.22   1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. d4
                                    exd4 5. Nxd4 Bb4 6. Nxc6 bxc6 7. Bd3
                                    d5 8. exd5 Qe7+ 9. Be3 Bxc3+ 10. bxc3
                                    Nxd5 11. Qf3 O-O (s=2)
              time=22.15  mat=0  n=198722622  fh=91%  nps=9.0M
              extensions=1.3M qchecks=4.5M reduced=20.3M pruned=43.7M
              predicted=0  evals=88.0M  50move=0  EGTBprobes=0  hits=0
              SMP->  splits=2315  aborts=299  data=12/1024  elap=22.15

SMP info is most interesting. Yet I have some questions.

splits=2315 Is this the number of splits only for the last iteration or summed up since search depth 1?

aborts= 299 Are these breaks from smp-search due to fail-highs?

data=12/1024 what's this?

[bob]

Running Crafty in a terminal, 4 Threads, start position, I get the following output:

Code: Select all

               21    17.84   0.22   1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. d4
                                    exd4 5. Nxd4 Bb4 6. Nxc6 bxc6 7. Bd3
                                    d5 8. exd5 Qe7+ 9. Be3 Bxc3+ 10. bxc3
                                    Nxd5 11. Qf3 O-O
               21->  22.15   0.22   1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. d4
                                    exd4 5. Nxd4 Bb4 6. Nxc6 bxc6 7. Bd3
                                    d5 8. exd5 Qe7+ 9. Be3 Bxc3+ 10. bxc3
                                    Nxd5 11. Qf3 O-O (s=2)
              time=22.15  mat=0  n=198722622  fh=91%  nps=9.0M
              extensions=1.3M qchecks=4.5M reduced=20.3M pruned=43.7M
              predicted=0  evals=88.0M  50move=0  EGTBprobes=0  hits=0
              SMP->  splits=2315  aborts=299  data=12/1024  elap=22.15

SMP info is most interesting. Yet I have some questions.

splits=2315 Is this the number of splits only for the last iteration or summed up since search depth 1?

Total splits done for that specific search, which includes all iterations.

aborts= 299 Are these breaks from smp-search due to fail-highs?

Yes.

data=12/1024 what's this?

Tells how many split blocks were used, and allocated. Only point of concern is when the first begins to approach the latter. I added this simply for info. I'd suspect you have 4 cpus, which won't use a ton of split blocks, and that apparently the version you are running was compiled with -DCPUS=16. You could reduce that and make the program smaller.

If Crafty runs out of split blocks, it won't crash, but it will be unable to split for a bit leaving one or more processors idle.

[Joerg Oster]

Thank you for explainig. Very appreciated.