Closed 299792458m closed 5 years ago
Thanks, great help, I'll look at it soon. Probably this inline-asm reverse engineering and elimination was the ugliest part of the update project, all help is welcomed, seems that I could not make it bugfree despite a lot of before-after tests.
I hope my wasted time helps you. reverse engineering from assembly with little comments make us crazy...
Right. As a blind test, I've checked it independently and found the same differences.
thanks for new release. but, still "nxtnnf" not updated. is it? (or need not....?)
Thx, needed. Check v1.0.14 pls.
thanks! the result was matched with 1.05.
I used tritical's TIVTC(1.05) with slow=2 option, and the result is diffrent from yours. and I noticed that in TFM.cpp, compareFieldsSlow2 with field!=0 section something is wrong. (just compared with asm block.) --arround line 1720 the address of nxtnnf and prvnnf does not updated. --at line 1685, may "(eax & 7)" instead of "56"? --at line 1708, may "(diff_n_c > 42)" instead of "2" ? --at line 1709, may "(eax & 2)" instead of "16"?