Dword to Bin$

[jj2007]

Ultimately, I often read posts (not here) by "experts" who claim the C/C++ are faster than assembly because, y'know, those compiler developers are real geniuses.

The geniuses have so far no decent algo to offer for translating a number to a binary string. However, you can you use invoke crt__itoa, DDvalue, ADDR _rv_bin_string_, 2 to achieve that (credits to MichaelW).
Does it work? Yes
Is it faster than Assembly? Not really

Code Select Expand

AMD Athlon Gold 3150U with Radeon Graphics      (SSE4)

6449    cycles for 100 * bin$ (CRT)
991     cycles for 100 * Bin$ (MasmBasic)
627     cycles for 100 * new Bin$

6537    cycles for 100 * bin$ (CRT)
993     cycles for 100 * Bin$ (MasmBasic)
629     cycles for 100 * new Bin$

6420    cycles for 100 * bin$ (CRT)
975     cycles for 100 * Bin$ (MasmBasic)
630     cycles for 100 * new Bin$

6471    cycles for 100 * bin$ (CRT)
981     cycles for 100 * Bin$ (MasmBasic)
618     cycles for 100 * new Bin$

6427    cycles for 100 * bin$ (CRT)
988     cycles for 100 * Bin$ (MasmBasic)
627     cycles for 100 * new Bin$

Averages:
6449    cycles for bin$ (CRT)
987     cycles for Bin$ (MasmBasic)
628     cycles for new Bin$

bin$ (CRT)                              11101010101010101010000111000111
Bin$ (MasmBasic)                        11101010101010101010000111000111
new Bin$                                11101010101010101010000111000111
Usually, we are already happy if we beat the CRT by a factor 2

[jj2007]

Version 2

Code Select Expand

AMD Athlon Gold 3150U with Radeon Graphics      (SSE4)
loop overhead is approx. 106/100 cycles

240    cycles for 100 * bin$ (CRT)
975    cycles for 100 * Bin$ (MasmBasic)
553    cycles for 100 * new Bin$ P2 (4*glob/al, ah)
629    cycles for 100 * new Bin$ P3 (edi/al, ah)
570    cycles for 100 * new Bin$ P4 (4*glob/shr)
626    cycles for 100 * new Bin$ P5 (edi/shr)

234    cycles for 100 * bin$ (CRT)
986    cycles for 100 * Bin$ (MasmBasic)
552    cycles for 100 * new Bin$ P2 (4*glob/al, ah)
636    cycles for 100 * new Bin$ P3 (edi/al, ah)
552    cycles for 100 * new Bin$ P4 (4*glob/shr)
627    cycles for 100 * new Bin$ P5 (edi/shr)

254    cycles for 100 * bin$ (CRT)
975    cycles for 100 * Bin$ (MasmBasic)
560    cycles for 100 * new Bin$ P2 (4*glob/al, ah)
630    cycles for 100 * new Bin$ P3 (edi/al, ah)
552    cycles for 100 * new Bin$ P4 (4*glob/shr)
633    cycles for 100 * new Bin$ P5 (edi/shr)

230    cycles for 100 * bin$ (CRT)
977    cycles for 100 * Bin$ (MasmBasic)
550    cycles for 100 * new Bin$ P2 (4*glob/al, ah)
628    cycles for 100 * new Bin$ P3 (edi/al, ah)
559    cycles for 100 * new Bin$ P4 (4*glob/shr)
624    cycles for 100 * new Bin$ P5 (edi/shr)

241    cycles for 100 * bin$ (CRT)
986    cycles for 100 * Bin$ (MasmBasic)
552    cycles for 100 * new Bin$ P2 (4*glob/al, ah)
632    cycles for 100 * new Bin$ P3 (edi/al, ah)
553    cycles for 100 * new Bin$ P4 (4*glob/shr)
638    cycles for 100 * new Bin$ P5 (edi/shr)

Averages:
238    cycles for bin$ (CRT)
979    cycles for Bin$ (MasmBasic)
552    cycles for new Bin$ P2 (4*glob/al, ah)
630    cycles for new Bin$ P3 (edi/al, ah)
555    cycles for new Bin$ P4 (4*glob/shr)
629    cycles for new Bin$ P5 (edi/shr)

bin$ (CRT)                              11101010101010101010000111000111
Bin$ (MasmBasic)                        11101010101010101010000111000111
new Bin$ P2 (4*glob/al, ah)            11101010101010101010000111000111
new Bin$ P3 (edi/al, ah)                11101010101010101010000111000111
new Bin$ P4 (4*glob/shr)                11101010101010101010000111000111
new Bin$ P5 (edi/shr)                  11101010101010101010000111000111
Note that bin$ (CRT) runs with one tenth of the iterations.

Now it would be nice to see some Intel timings

[jj2007]

Code Select Expand

Intel(R) Core(TM) i5-2450M CPU @ 2.50GHz (SSE4)
Averages:
288     cycles for bin$ (CRT)
1112    cycles for Bin$ (MasmBasic)
720     cycles for new Bin$ P2 (4*glob/al, ah)
714     cycles for new Bin$ P3 (edi/al, ah)
637     cycles for new Bin$ P4 (4*glob/shr)
725     cycles for new Bin$ P5 (edi/shr)

[sinsi]

Code Select Expand

Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz (SSE4)
loop overhead is approx. 79/100 cycles

Averages:
337     cycles for bin$ (CRT)
825     cycles for Bin$ (MasmBasic)
434     cycles for new Bin$ P2 (4*glob/al, ah)
495     cycles for new Bin$ P3 (edi/al, ah)
426     cycles for new Bin$ P4 (4*glob/shr)
486     cycles for new Bin$ P5 (edi/shr)


[jj2007]

Thanks, Sinsi - so P4 wins. The original version dates from 2008, 16 years ago. and in my source I noted "credits go to Sinsi" ;-)