Index
> Windows > Rol and to convert num to text.Goto page 1, 2 Next |
| Author |
|
|
Overclick 02 Jul 2022, 12:07
You have to add 30h for each number of 1-9 to get ASCII code. But we have hexadecimal system so you definitely need to divide value by 10 to easily get lower number. And do that for each next one.
If you really want to do that without div or mul you have to compare each part and extract it from source. It is more complicated way to do that simple things. |
|||
02 Jul 2022, 12:07 |
|
|
macomics 02 Jul 2022, 12:57
Code: mov eax,4444 ; {00000000`00000000`00010001`01011100} = 4444 rol eax,4 ; {00000000`00000001`00010101`11000000} = 71104 mov ebx,eax bswap ebx ; {11000000`00010101`00000001`00000000} = 3222601984 and ebx,0xffff;0000 {00000000`00000000`00000001`00000000} = 256 ;xor ebx,0xffff;0000 add bh,'0' ; {00000000`00000000`00110001`00000000} = 12544 mov [txt],bh ;txt db 80 dup (0) Last edited by macomics on 02 Jul 2022, 12:58; edited 1 time in total |
|||
|
02 Jul 2022, 12:57 |
|
|
FlierMate1 02 Jul 2022, 12:58
I came across 'Binary number to decimal ASCII string conversion' ( https://board.flatassembler.net/topic.php?t=3924 ) the other day when doing utf8count.asm. Too advanced for average coders like me.
|
|||
|
02 Jul 2022, 12:58 |
|
|
Roman 02 Jul 2022, 13:41
I found int to hex text.
But i want int to decimal text numbers Code: ;convert int to hex text mov rsi,txt ;in data txt db 16 dup (0) mov eax,100 ;100=64h mov cl,8 @@rnd1: rol eax,4 mov ebx,eax and al,00001111b add al,'0' cmp al,'9' jbe @@rnd2 add al,07h @@rnd2: mov [rsi],al inc rsi mov eax,ebx dec cl jnz @@rnd1 |
|||
|
02 Jul 2022, 13:41 |
|
|
macomics 02 Jul 2022, 13:53
First then you will have to translate int into packed BCD
|
|||
|
02 Jul 2022, 13:53 |
|
|
Roman 02 Jul 2022, 14:19
First i found this
formula: A/10 = a*((1<<67)/10) >> 67 A/10 = a*0x0CCCCCCCCCCCCCCCD >> 67 Code: ;int64 to decimal string number ! mov r8, -3689348814741910323 ;magic = 0x0CCCCCCCCCCCCCCCD = ((1<<67)/10) mov rsi,txt+40 ;in data txt db 42 dup (0) mov rdi,4689348814741910323 ;19 digits ! .LBB1_1: ; =>This Inner Loop Header: Depth=1 mov rax, rdi mul r8 ; mul to rdx shr rdx, 3 ;>> 67 lea eax, [rdx + rdx] lea eax, [rax + 4*rax] mov ecx, edi sub ecx, eax or cl, '0' mov byte [rsi - 1], cl dec rsi ;what faster ? dec rsi or add rsi,-1 ? cmp rdi, 9 mov rdi, rdx ja .LBB1_1 invoke MessageBox,0,rsi,0,0 |
|||
|
02 Jul 2022, 14:19 |
|
|
Overclick 03 Jul 2022, 04:54
Code: .data result db 10 dup(0x30) db 0 simple dd 1000000000 dd 100000000 dd 10000000 dd 1000000 dd 100000 dd 10000 dd 1000 dd 100 dd 10 number dd ? .code xor rdx,rdx mov eax,[number] begin: cmp eax,[simple+rdx*4] jb @F sub eax,[simple+rdx*4] add byte[result+rdx],1 jmp begin @@: inc rdx cmp rdx,9 jne begin add byte[result+rdx],al You're welcome |
|||
|
03 Jul 2022, 04:54 |
|
|
I 03 Jul 2022, 10:38
Umm, some results. Maybe someone should check if they are right?
Roman Code: 1 Million Iterations @2GHz, Counter Frequency 10000000 Number Used Number Ret ms Clocks ================================================================ 9999999999999999999 48.028 96 9999999999999999999 47.998 96 9999999999999999999 47.933 96 999999999999999 36.403 73 999999999999999 36.410 73 999999999999999 36.431 73 99999999999 25.098 50 99999999999 25.115 50 99999999999 25.102 50 9999999 15.095 30 9999999 15.111 30 9999999 15.106 30 999 7.001 14 999 7.014 14 999 7.003 14 5555555555555555555 47.939 96 5555555555555555555 47.954 96 5555555555555555555 47.948 96 555555555555555 36.411 73 555555555555555 36.409 73 555555555555555 36.414 73 55555555555 25.107 50 55555555555 25.109 50 55555555555 25.109 50 5555555 15.096 30 5555555 15.095 30 5555555 15.095 30 555 6.064 12 555 6.457 13 555 6.065 12 0 2.251 5 0 2.250 5 0 2.250 5 1 2.251 5 1 2.251 5 1 2.250 5 18446744073709551615 50.859 102 18446744073709551615 50.856 102 18446744073709551615 50.865 102 17777777777777777777 50.892 102 17777777777777777777 50.879 102 17777777777777777777 50.873 102 Overclick adjusted to 64-bit number. Code: 1 Million Iterations @2GHz, Counter Frequency 10000000 Number Used Number Ret ms Clocks ================================================================ 9999999999999999999 09999999999999999999 235.319 471 9999999999999999999 09999999999999999999 236.097 472 9999999999999999999 09999999999999999999 235.538 471 999999999999999 00000999999999999999 194.617 389 999999999999999 00000999999999999999 194.554 389 999999999999999 00000999999999999999 194.545 389 99999999999 00000000099999999999 151.655 303 99999999999 00000000099999999999 151.676 303 99999999999 00000000099999999999 151.744 303 9999999 00000000000009999999 118.357 237 9999999 00000000000009999999 119.566 239 9999999 00000000000009999999 119.229 238 999 00000000000000000999 47.596 95 999 00000000000000000999 47.626 95 999 00000000000000000999 47.560 95 5555555555555555555 05555555555555555555 161.903 324 5555555555555555555 05555555555555555555 161.908 324 5555555555555555555 05555555555555555555 161.869 324 555555555555555 00000555555555555555 134.227 268 555555555555555 00000555555555555555 134.046 268 555555555555555 00000555555555555555 134.054 268 55555555555 00000000055555555555 112.019 224 55555555555 00000000055555555555 112.050 224 55555555555 00000000055555555555 111.968 224 5555555 00000000000005555555 83.302 167 5555555 00000000000005555555 83.572 167 5555555 00000000000005555555 83.536 167 555 00000000000000000555 48.116 96 555 00000000000000000555 48.183 96 555 00000000000000000555 47.984 96 0 00000000000000000000 37.765 76 0 00000000000000000000 37.758 76 0 00000000000000000000 37.753 76 1 00000000000000000001 37.840 76 1 00000000000000000001 37.905 76 1 00000000000000000001 37.917 76 18446744073709551615 18446744073709551615 227.419 455 18446744073709551615 18446744073709551615 228.378 457 18446744073709551615 18446744073709551615 229.771 460 17777777777777777777 17777777777777777777 221.867 444 17777777777777777777 17777777777777777777 221.654 443 17777777777777777777 17777777777777777777 221.658 443 5000000000000000000 05000000000000000000 45.061 90 5000000000000000000 05000000000000000000 45.171 90 5000000000000000000 05000000000000000000 45.275 91 5000000000000000 00005000000000000000 42.174 84 5000000000000000 00005000000000000000 41.875 84 5000000000000000 00005000000000000000 41.656 83 5000000000000 00000005000000000000 46.476 93 5000000000000 00000005000000000000 46.386 93 5000000000000 00000005000000000000 46.214 92 5000000000 00000000005000000000 46.548 93 5000000000 00000000005000000000 46.466 93 5000000000 00000000005000000000 46.606 93 5000000 00000000000005000000 41.679 83 5000000 00000000000005000000 34.602 69 5000000 00000000000005000000 34.600 69 5000 00000000000000005000 34.523 69 5000 00000000000000005000 34.519 69 5000 00000000000000005000 34.513 69 50 00000000000000000050 34.616 69 50 00000000000000000050 34.621 69 50 00000000000000000050 34.610 69 Both unsigned, one with leading zero's, one without. Last edited by I on 04 Jul 2022, 06:14; edited 1 time in total |
|||
|
03 Jul 2022, 10:38 |
|
|
macomics 03 Jul 2022, 10:56
So it was clear that Overclick gave a very slow garbage bust. There are up to 200 iterations of a cycle with a 20-digit 64-bit number.
Naturally, he will lose outright to the cycle that performs division and allocates a digit from a number. At the same time, no more than 20 iterations take place. |
|||
|
03 Jul 2022, 10:56 |
|
|
Overclick 03 Jul 2022, 11:15
Quote:
right? |
|||
|
03 Jul 2022, 11:15 |
|
|
macomics 03 Jul 2022, 11:46
Code: .data result db 20 dup(0x30) dd 0 ; align 8 number dq ? simple dq 10000000000000000000 dq 1000000000000000000 dq 100000000000000000 dq 10000000000000000 dq 1000000000000000 dq 100000000000000 dq 10000000000000 dq 1000000000000 dq 100000000000 dq 10000000000 dq 1000000000 dq 100000000 dq 10000000 dq 1000000 dq 100000 dq 10000 dq 1000 dq 100 dq 10 .code xor rdx,rdx mov rax,[number] begin: mov r9, [simple+rdx*8] cmp rax, r9 jb skip shl r9, 3 mov r8l, 8 digit: mov rcx, rax sub rcx, r9 jc @f mov rax, rcx add byte[result+rdx], r8l cmp rax, [simple+rdx*8] jc skip @@: shr r9, 1 shr r8l, 1 jnz digit skip: inc rdx cmp rdx,19 jl begin add byte[result+rdx],al Last edited by macomics on 03 Jul 2022, 15:46; edited 1 time in total |
|||
|
03 Jul 2022, 11:46 |
|
|
macomics 03 Jul 2022, 13:22
For my example, the worst number is 7, not 9. If you check, then check for a number consisting of all 7.
Code: number dq 17777777777777777777 |
|||
|
03 Jul 2022, 13:22 |
|
|
revolution 03 Jul 2022, 13:57
But shr reg,1 is also a division.
 |
|||
|
03 Jul 2022, 13:57 |
|
|
macomics 03 Jul 2022, 14:49
revolution wrote: But shr reg,1 is also a division. |
|||
|
03 Jul 2022, 14:49 |
|
|
Overclick 03 Jul 2022, 15:20
Quote:
You talk about optimization I just showed concept. Believe me there is a lot of optimizations can be done but it is Roman's headache, not ours. He want it without div mul for some reason, just give it to be. Also there you missing something with your rcx register at beginning |
|||
|
03 Jul 2022, 15:20 |
|
|
macomics 03 Jul 2022, 15:42
Overclick wrote: Also there you missing something with your rcx register at beginning Last edited by macomics on 03 Jul 2022, 15:47; edited 1 time in total |
|||
|
03 Jul 2022, 15:42 |
|
|
Overclick 03 Jul 2022, 15:47
It may jump as rcx is randomized
Code: cmp rax, rcx
jb skip
|
|||
|
03 Jul 2022, 15:47 |
|
|
Roman 03 Jul 2022, 19:01
Overclick
Thanks , interesting example. I modify your example: Code: align 8 simple dq 1000000000 dq 100000000 dq 10000000 dq 1000000 dq 100000 dq 10000 dq 1000 dq 100 dq 10 xor rdx,rdx mov eax,9999 ;[number] beg: sub bl,bl mov ecx,dword [simple+rdx*8] begin: cmp eax,ecx jb @F sub eax,ecx ;add byte[result+rdx],1 inc bl jmp begin @@: add [result+rdx],bl ;sub bl,bl inc rdx cmp rdx,9 jnz beg add byte[result+rdx],al PS: User I please test this variant. How its faster ? Last edited by Roman on 03 Jul 2022, 20:55; edited 1 time in total |
|||
|
03 Jul 2022, 19:01 |
|
|
Overclick 03 Jul 2022, 20:53
U can minimize memory access by move dword [simple+rdx*8] to register next to beg:
|
|||
|
03 Jul 2022, 20:53 |
|
| Goto page 1, 2 Next < Last Thread | Next Thread > |
Forum Rules:
|