64bit versions of engines?

[Bryan Hofmann]

Thanks.
This means the 64 bit executable in these two cases is 8 and 15% slower than the public 32 bit version (which may be a very fast compilation though, certainly not compiled the same way as the 64 bit version, and thus not directly comparable).

Nevertheless this suggests that there is no significant 64bit speed-up to be expected at least for Fruit/Toga, if any at all.

Robert

There are no optimization in fruit/toga for 64bit. The above results do not suprise me in that this was done awhile back on the CCC by Slater Wold who found that most of the programs either showed no improvement or were slower.

Since I do not have access to a 64 bit system I can not run it thru the PGO compile which would show some improvements in speed. I did do this with the 32 bit version of Fruit when it came out and it was faster than the one on Leo's site using the Microsoft VS 2005 compiler on a AMD platform.

[Robert Allgeuer]

Then possibly two "normal" non-PGO compilations of Toga with identical flags, one with 32 bit, one with 64 bit would be the best way to compare which relative speed-up (or slow-down) the non-bitboard Toga receives from the move to 64bit.

Robert

[Fabien Letouzey]

There are no optimization in fruit/toga for 64bit. The above results do not suprise me in that this was done awhile back on the CCC by Slater Wold who found that most of the programs either showed no improvement or were slower.

Hi Bryan,

In fact Slater first reported a slow down. I think it was Fruit 1.0 and I can't remember by how much (something like 20%).

The slow down went away later, probably after I got rid of most small integers (< 32 bits). The speed was the same in 32 bits and 64 bits after that point.

Fabien.