Is this fast? (Crafty-22.2)

[zullil]

Code: Select all

Crafty v22.2 (2 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 37348357
Raw nodes per second: 3636646.251217
Total elapsed time: 10.270000

On my 2 GHz Intel Core 2 Duo MacBook with 4 GB of RAM.

[krazyken]

Looks reasonably good to me. You don't mention how much hash you are using though.

on on a 2.1Ghz iMac core 2 duo:

Code: Select all

hash table memory =  192M bytes.
White(1): bench
Running benchmark. . .
......
Total nodes: 38981440
Raw nodes per second: 3925623.363545
Total elapsed time: 9.930000

hash table memory =   96M bytes.
White(1): bench
Running benchmark. . .
......
Total nodes: 36901295
Raw nodes per second: 3913180.805938
Total elapsed time: 9.430000

2 threads on my power mac 2 x 2.8Ghz Q6600:

Code: Select all

hash table memory =  192M bytes.
White(1): mt 2
max threads set to 2
White(1): bench
Running benchmark. . .
......
Total nodes: 39222975
Raw nodes per second: 4431974.576271
Total elapsed time: 8.850000

[zullil]

Looks reasonably good to me. You don't mention how much hash you are using though.

The amount of hash seems to have almost no effect on the bench speed. For example, this is my "best" result:

Code: Select all

hash table memory = 1536K bytes.
pawn hash table memory =    1M bytes.

Crafty v22.2 (2 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 39315709
Raw nodes per second: 3723078.503788
Total elapsed time: 10.560000

I wonder how little becomes too little.

[bob]

Code: Select all

Crafty v22.2 (2 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 37348357
Raw nodes per second: 3636646.251217
Total elapsed time: 10.270000

On my 2 GHz Intel Core 2 Duo MacBook with 4 GB of RAM.

No, this is fast:

Code: Select all

White(1): bench
Running benchmark. . .
......
Total nodes: 50340338
Raw nodes per second: 11287071.300448
Total elapsed time: 4.460000

It would be faster but the tests are now too short for modern hardware. I will fix the bench command to run a bit longer for each position.

The above is on a 2.3ghz dual quad-core box we used for the last ACCA event where we averaged about 18M nodes per second over all searches for the tournament, but frequently saw numbers well above 20M

[zullil]

Code: Select all

Crafty v22.2 (2 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 37348357
Raw nodes per second: 3636646.251217
Total elapsed time: 10.270000

On my 2 GHz Intel Core 2 Duo MacBook with 4 GB of RAM.

No, this is fast:

Code: Select all

White(1): bench
Running benchmark. . .
......
Total nodes: 50340338
Raw nodes per second: 11287071.300448
Total elapsed time: 4.460000

It would be faster but the tests are now too short for modern hardware. I will fix the bench command to run a bit longer for each position.

The above is on a 2.3ghz dual quad-core box we used for the last ACCA event where we averaged about 18M nodes per second over all searches for the tournament, but frequently saw numbers well above 20M

My result represents 3637 nps/$. How much does the 2.3ghz dual quad-core box cost? Actually, don't tell me---since Apple products are pricey I'll probably lose again.

[bob]

Code: Select all

Crafty v22.2 (2 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 37348357
Raw nodes per second: 3636646.251217
Total elapsed time: 10.270000

On my 2 GHz Intel Core 2 Duo MacBook with 4 GB of RAM.

No, this is fast:

Code: Select all

White(1): bench
Running benchmark. . .
......
Total nodes: 50340338
Raw nodes per second: 11287071.300448
Total elapsed time: 4.460000

It would be faster but the tests are now too short for modern hardware. I will fix the bench command to run a bit longer for each position.

The above is on a 2.3ghz dual quad-core box we used for the last ACCA event where we averaged about 18M nodes per second over all searches for the tournament, but frequently saw numbers well above 20M

My result represents 3637 nps/$. How much does the 2.3ghz dual quad-core box cost? Actually, don't tell me---since Apple products are pricey I'll probably lose again.

Actually these boxes are cheap... They are 1U rack-mount systems to make them compact...

[krazyken]

For comparison's sake, what do you get with 2 threads?

[zullil]

For comparison's sake, what do you get with 2 threads?

You're asking Bob, right?

[krazyken]

For comparison's sake, what do you get with 2 threads?

You're asking Bob, right?

yeah.

[bob]

max threads set to 2
White(1): bench
Running benchmark. . .
......
Total nodes: 27119992
Raw nodes per second: 4497511.111111
Total elapsed time: 6.030000