Profile data request

[mcostalba]

Glaurung (all functions with at least 1.0%):

Code: Select all

do_move():                               5.0%

Have you tried with memcpy to populate the undo struct? I have started but gived up, not enough inertia.

Code: Select all

TranspositionTable::Retrieve():          4.4%

This is a bit strange I got some less here. I will check better.

Code: Select all

pop_1st_bit():                           2.6%

I would guess you are on a 64 bit box here.


Marco

[Tord Romstad]

Glaurung (all functions with at least 1.0%):

Code: Select all

do_move():                               5.0%

Have you tried with memcpy to populate the undo struct? I have started but gived up, not enough inertia.

I don't think I have done any low-level optimization anywhere in the program, apart from trying to write the code in a simple and straightforward way and not doing too much unnecessary work. I doubt that a memcpy would be faster, though. The compiler should be able to optimize the assignments and a memcopy to the same code anyway.

Code: Select all

TranspositionTable::Retrieve():          4.4%

This is a bit strange I got some less here. I will check better.

It might depend on the position. I did a "./glaurung bench 128 1" from the command line, which runs 15 benchmark positions for 60 seconds each, with 128 MB TT and one search thread.

Code: Select all

pop_1st_bit():                           2.6%

I would guess you are on a 64 bit box here.

Yes, this was on a Core 2 Duo.

Tord