Conversation

Notices

1. þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:19 JST þernia

   • :ihavenomouth:
   @Inginsub ok ok ic.
   
   so, i'm thinking for the sum, could i just
   
   add resultDh with resultCl,
   inc for the carry,
   add resultBl to that,
   inc for carry again,
   mov that result into the lowest part of a 3 word buffer,
   add the inc to resultCh
   inc for carry again starting from 0
   add to that resultBh
   inc for carry
   add to that resultAl
   inc for carry again
   mov the result into the lower part of the 3 word buffer
   add inc to result Ah
   mov that result into the highest part of the 3 word buffer
   
   and then i would have the entire result in one buffer.
   
   and if not in one buffer, could i mov the results into separate buffers and then figure out how to print those out?
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:11 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint
     > You asked me!
     you went through one paper, 3 opinions on tools that do the same thing, and missed the whole point of the question once, only to repeat something someone else said 3 days ago. read the FUCKING THREAD NIGGER.
     
     gn
     New Janny in Town likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 02-Sep-2024 19:18:12 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint
     
     > i got as far as that
     
     That's as far as you need!
     
     > before you joined the thread
     
     You asked me!
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 02-Sep-2024 19:18:13 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint
     
     > when i have a different word for fractional and whole parts,
     
     No, that is your problem. It's fixed-point and you confuse fixed-point and floating-point. What you want to do is multiply 0xNNNN by 0xMMMM. You get 0xWWXXYYZZ, and the W is overflow, the Z is roundoff, your answer is 0xXXYY.
     
     > maybe i'm not being clear: i'm not allowed to.
     
     Well then why are you talking about it?
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:13 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint
     >You get 0xWWXXYYZZ
     i got as far as that before you joined the thread
     > Well then why are you talking about it?
     if i ever did talk about fpu's, it was to say i wasn't allowed to use them
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 02-Sep-2024 19:18:14 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint
     
     > lolno. "just multiply" is reiterating the problem
     
     Unless you want to implement multiplication yourself, you want to use the multiplication instructions. I don't know how to make that easier. mul, imul.
     
     > you think id be asking for help if i could?
     
     Yes, you can. I listed all of the PoC||GTFO issues that have boot-sector hacks, including complete guides. "Here is the thing you want to do, here is a list of places that tell you how to do it. All of these PDFs are also zip files that include the code."
     
     I type "x87 fpu" into a search engine and I see this immediately: https://wiki.osdev.org/FPU .
     
     MOV EDX, CR0 ; Start probe, get CR0
     AND EDX, (-1) - (CR0_TS + CR0_EM) ; clear TS and EM to force fpu access
     MOV CR0, EDX ; store control word
     FNINIT ; load defaults to FPU
     FNSTSW [.testword] ; store status word
     CMP word [.testword], 0 ; compare the written status with the expected FPU state
     JNE .nofpu ; jump if the FPU hasn't written anything (i.e. it's not there)
     JMP .hasfpu
     
     .testword: DW 0x55AA ; store garbage to be able to detect a change
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:14 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint
     >mul, imul
     no shit. when i have a different word for fractional and whole parts, or am doing ieee flaoting point numbers (because the format is something i had to figure out too), its a few more steps
     >Yes, you can
     maybe i'm not being clear: i'm not allowed to. or wasn't anyway, its done
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 02-Sep-2024 19:18:15 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint
     
     > im saying, how do i multiply two numbers with decimals
     
     I think you should do your own homework but when I wrote "Just do a 32-bit multiply, this yields a 64-bit result. You need to shift it right by 16 bits to correct the fractional part.", I already *did* your homework for you.
     
     > ingisub suggested i use ieee 754 numbers, probably the most sane approach at this point.
     
     You could work directly on the IEEE-754 numbers but "turn on the FPU" is probably way easier. If you wanna build it yourself from scratch, here, this is allegedly better:
     beating_floating_point--posit.pdf
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:15 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint
     > i already did your homework for you
     lolno. "just multiply" is reiterating the problem
     >turn on the fpu
     you think id be asking for help if i could?
     
     thanks for the cool read though
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:16 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint im saying, how do i multiply two numbers with decimals (eg. 0.5, 9999.9999, 1.0, etc), without an fpu coprocessor.
     
     ingisub suggested i use ieee 754 numbers, probably the most sane approach at this point.
     
     >pernis
     what
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:17 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @Inginsub @p @qint yall niggas got any ideas? i'm multiplying two floating point numbers, each number with a whole part and fractional part in their own word
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 02-Sep-2024 19:18:17 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint I don't know what ideas you want. Just do a 32-bit multiply, this yields a 64-bit result. You need to shift it right by 16 bits to correct the fractional part. There must be a recipe book for fixed-point math somewhere.
     
     > each number with a whole part and fractional part in their own word
     
     That would be fixed-point, not floating-point.
     
     > xor ax, ax
     > xor dx, dx
     > mov ax, whole0
     
     pernis
   • þernia (pernia@cum.salon)'s status on Monday, 02-Sep-2024 19:18:18 JST þernia

     in reply to

     • :ihavenomouth:

     @Inginsub vro i did it, although its kind of broken

     .data whole0 dw 5000 float0 dw 0 whole1 dw 0 float1 dw 5 producth dw 0 productl dw 0 productfloat dw 0 ; 9999,9999*9999,9999 = 99999998 ; actual output: 5F59ECC,2BB7 => 99983052.1707 ; 5000*0.5 = 2500 ; actual output: 0,61A8 (25000, but in the productfloat) carry dw 0 resultah dw 0 resultal dw 0 resultbh dw 0 resultbl dw 0 resultch dw 0 resultcl dw 0 resultdh dw 0 resultdl dw 0 .code check_carry proc near jnc no_increment inc carry no_increment: ret check_carry endp mul16 proc near ; input: word0h:word0l * word1h:word1l, each 16bit whole 16bit float ; output: producth:productl:productfloat, 32bit whole 16bit float mov ax, whole0 mul whole1 ; dx:ax mov resultah, dx mov resultal, ax xor ax, ax xor dx, dx mov ax, whole0 mul float1 mov resultbh, dx mov resultbl, ax xor ax, ax xor dx, dx mov ax, whole1 mul float0 mov resultch, dx mov resultcl, ax xor ax, ax xor dx, dx mov ax, float0 mul float1 mov resultdh, dx mov resultdl, ax xor ax, ax xor dx, dx mov ax, resultdh add ax, resultcl call check_carry clc ; CF = 0 add ax, resultbl call check_carry clc mov productfloat, ax xor ax, ax mov ax, carry mov carry, 0 add ax, resultch call check_carry clc add ax, resultbh call check_carry clc add ax, resultal call check_carry clc mov productl, ax xor ax, ax mov ax, carry add ax, resultah mov producth, ax ; clear variables xor ax, ax clc mov carry, 0 mov resultah, 0 mov resultal, 0 mov resultbh, 0 mov resultbl, 0 mov resultch, 0 mov resultcl, 0 mov resultdh, 0 mov resultdl, 0 ret mul16 endp start: mov ax, @data mov ds, ax call mul16 mov ax, 4c00h int 21h end start
   • þernia (pernia@cum.salon)'s status on Wednesday, 04-Sep-2024 04:55:33 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @Inginsub @qint @p what i noticed is that it works mostly when i do this (throwing out the last 16 bits of the 16.16.16.16 product), but if im doing for example 5000*0.5, i end up with
     
     0000.0000.25000.0000
     
     which is almost right, and i think this is where a shift comes in correct?
   • pistolero (p@fsebugoutzone.org)'s status on Wednesday, 04-Sep-2024 04:55:33 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint Dude, just use bochs and step through the instructions.
   • pistolero (p@fsebugoutzone.org)'s status on Wednesday, 04-Sep-2024 04:55:34 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @Inginsub @pernia @qint If he does a 16x16 multiplication, he'll just have to take the low half of one register and the high half of the other. But notionally, it's a shift, right, like that is the operation you do, you throw out the upper 8 as overflow and the lower 8 as underflow.
   • :ihavenomouth: (inginsub@clubcyberia.co)'s status on Wednesday, 04-Sep-2024 04:55:34 JST :ihavenomouth:

     in reply to

     • pistolero
     • .–.
     @p @qint @pernia he multiplies 32 bit numbers in 16.16 format. 16x16 comes from the 8086 limitation of only having 16 bit regs to work with, but in this case each product will have 0, 16, or 32 fractional bits.
   • :ihavenomouth: (inginsub@clubcyberia.co)'s status on Wednesday, 04-Sep-2024 04:55:35 JST :ihavenomouth:

     in reply to

     • pistolero
     • .–.
     @p @qint @pernia to be fair, with a 16 bit fractional part he doesn't need to shift anything, discarding the lowest word is enough
   • þernia (pernia@cum.salon)'s status on Wednesday, 04-Sep-2024 04:55:36 JST þernia

     in reply to

     • :ihavenomouth:
     • pistolero
     • .–.
     @p @Inginsub @qint what was wrong is i probably never shifted the result. ill definately revisit this proyect once i got time again
   • pistolero (p@fsebugoutzone.org)'s status on Wednesday, 04-Sep-2024 04:55:36 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint I told you to shift it, you said that you already figured that out!
   • pistolero (p@fsebugoutzone.org)'s status on Wednesday, 04-Sep-2024 04:55:37 JST pistolero

     in reply to

     • :ihavenomouth:
     • .–.
     @pernia @Inginsub @qint You sent me a page of assembly and asked what was wrong! I don't know, I didn't read it!