Index
> Main > How I can to declare of abnormal float-numbers as constants? |
| Author |
|
|
Plue 27 Feb 2007, 20:20
infinity dd 2139095040
|
|||
27 Feb 2007, 20:20 |
|
|
Garthower 27 Feb 2007, 22:00
Plue: Do you can post constants for negative infinity, QNAN and other and for all float types (single, double, extended)? Or, if you can, give me some urls for info about it?
|
|||
|
27 Feb 2007, 22:00 |
|
|
Reverend 27 Feb 2007, 22:22
Code: PRECISION = 32 if PRECISION = 32 PLUS = 0 shl 31 MINUS = 1 shl 31 DENORMALIZED = 1 shl 22 INFINITY = ((1 shl 8) - 1) shl 23 NAN = DENORMALIZED or INFINITY else if PRECISION = 64 PLUS = 0 shl 63 MINUS = 1 shl 63 DENORMALIZED = 1 shl 51 INFINITY = ((1 shl 11) - 1) shl 52 NAN = DENORMALIZED or INFINITY ; else if PRECISION = 80 ; PLUS = 0 shl 79 ; MINUS = 1 shl 79 ; DENORMALIZED = 1 shl 65 ; INFINITY = ((1 shl 15) - 1) shl 66 ; NAN = DENORMALIZED or INFINITY end if denormalized dd DENORMALIZED infinity dd INFINITY or MINUS ; -infinity nan dd NAN  . Tomasz disallows using anything but floats with 'dt'. It only accepts 1.0, 3.14, etc. The 80 bits block is just put here, because maybe you'll dig into macros to enable its usage - it is possible, I'm just too lazy to create it  .
EDIT: Oh I forgot fasm uses 64-bit numbers so no chance for compiler-time 80 bits INFINITY ot NAN, this one can be only saved during runitme :/. |
|||
|
27 Feb 2007, 22:22 |
|
|
Borsuc 28 Feb 2007, 14:14
Quote: EDIT: Oh I forgot fasm uses 64-bit numbers so no chance for compiler-time 80 bits INFINITY ot NAN, this one can be only saved during runitme :/. |
|||
|
28 Feb 2007, 14:14 |
|
|
vid 28 Feb 2007, 14:32
Code: macro dt [arg] { else equ 1 match =NAN, arg \{ dd ... dd ... dw ... restore else else equ 0 \} match =-=NAN, arg \{ dd ... dd ... dw ... restore else else equ 0 \} ... etc ... match =1, else \{ dt arg \} restore else } that's the idea. Additionaly you can add checking for "+NAN" etc.. |
|||
|
28 Feb 2007, 14:32 |
|
|
LocoDelAssembly 28 Feb 2007, 14:41
Don't forget to define "struc dt ..." also making use of that macro.
|
|||
|
28 Feb 2007, 14:41 |
|
|
Reverend 28 Feb 2007, 17:07
OK, I underestimated fasm's power
. But from now it's Garthower's task to make the macro work if he needs 80 bits NANS and INFINITIES - I'm too lazy, and besides I don't need it at all ;] |
|||
|
28 Feb 2007, 17:07 |
|
|
Garthower 01 Mar 2007, 08:40
Thanks for info. Using data from Reverend I create constants with db directive. Here they are:
Code: @FcInfinity_Ext: db 7Fh,0FFh,80h,00h,00h,00h,00h,00h,00h,00h ;1.0/0.0 @FcNan_Ext: db 0FFh,0FFh,0C0h,00h,00h,00h,00h,00h,00h,00h ;0.0/0.0 @FcnInfinity_Ext: db 0FFh,0FFh,80h,00h,00h,00h,00h,00h,00h,00h ;-1.0/0.0 @FcInfinity_Double: db 7Fh,0F0h,00h,00h,00h,00h,00h,00h ;1.0/0.0 @FcNan_Double: db 0FFh,0F8h,00h,00h,00h,00h,00h,00h ;0.0/0.0 @FcnInfinity_Double: db 0FFh,0F0h,00h,00h,00h,00h,00h,00h ;-1.0/0.0 @FcInfinity_Single: db 7Fh,80h,00h,00h ;1.0/0.0 @FcNan_Single: db 0FFh,0C0h,00h,00h ;0.0/0.0 @FcnInfinity_Single: db 0FFh,080h,00h,00h ;-1.0/0.0 I hope in future FASM's builds Tomasz will add an opportunity of record of similar numbers without any cunnings. |
|||
|
01 Mar 2007, 08:40 |
|
|
vid 01 Mar 2007, 10:56
Garthower: slightly better version:
Code: label @FcInfinity_Ext tbyte db 7Fh,0FFh,80h,00h,00h,00h,00h,00h,00h,00h ;1.0/0.0 label @FcNan_Ext tbyte db 0FFh,0FFh,0C0h,00h,00h,00h,00h,00h,00h,00h ;0.0/0.0 ... label @FcInfinity_Double qword db 7Fh,0F0h,00h,00h,00h,00h,00h,00h ;1.0/0.0 ... label @FcInfinity_Single dword db 7Fh,80h,00h,00h ;1.0/0.0 ... |
|||
|
01 Mar 2007, 10:56 |
|
|
asmfan 01 Mar 2007, 11:04
By the way? is this correct? Little endian or big?
|
|||
|
01 Mar 2007, 11:04 |
|
|
Garthower 01 Mar 2007, 11:04
Yes, you are right vid. So it will not be need in manual to specify type of data in commands. Thanks!
|
|||
|
01 Mar 2007, 11:04 |
|
|
Garthower 01 Mar 2007, 11:09
asmfan wrote: By the way? is this correct? Little endian or big? Sorry, my english is not so well, but what do you mean in word "endian"? |
|||
|
01 Mar 2007, 11:09 |
|
|
vid 01 Mar 2007, 11:13
endianness means which byte comes first. For example, is 1234ABCDh this:
Code: db 12h, 34h, 0ABh, 0CDh or Code: db 0CDh, 0ABh, 34h, 12h for more info consult wikipedia |
|||
|
01 Mar 2007, 11:13 |
|
|
Garthower 01 Mar 2007, 11:17
Thanks vid again, I undertood
2asmfan: Yes, it's a correct constants with currect endian  |
|||
|
01 Mar 2007, 11:17 |
|
|
asmfan 01 Mar 2007, 11:35
Guys, your constants need to be byteswapped to be correct - that's all i meant.
|
|||
|
01 Mar 2007, 11:35 |
|
|
Garthower 01 Mar 2007, 12:30
asmfan wrote: Guys, your constants need to be byteswapped to be correct - that's all i meant. These constants are in Intel format, i.e. in little endian. And you should solve, as to you they should be rearranged. |
|||
|
01 Mar 2007, 12:30 |
|
|
asmfan 01 Mar 2007, 12:40
Les's say no words and run Olly. try to load your constants as if they little endian and you'll see what is correct.
|
|||
|
01 Mar 2007, 12:40 |
|
|
Garthower 01 Mar 2007, 13:14
2asmfan:: Oh, you are right
 For normal using it must be rearranged. I'm so sorry for post incorect values.
Here are tested constants with vid recomendation: Code:
label @FcInfinity_Ext tbyte
db 00h,00h,00h,00h,00h,00h,00h,80h,0FFh,7Fh ;1.0/0.0
label @FcNan_Ext tbyte
db 00h,00h,00h,00h,00h,00h,00h,0C0h,0FFh,0FFh ;0.0/0.0
label @FcnInfinity_Ext tbyte
db 00h,00h,00h,00h,00h,00h,00h,80h,0FFh,0FFh ;-1.0/0.0
label @FcInfinity_Double qword
db 00h,00h,00h,00h,00h,00h,0F0h,7Fh ;1.0/0.0
label @FcNan_Double qword
db 00h,00h,00h,00h,00h,00h,0F8h,0FFh ;0.0/0.0
label @FcnInfinity_Double qword
db 00h,00h,00h,00h,00h,00h,0F0h,0FFh ;-1.0/0.0
label @FcInfinity_Single dword
db 00h,00h,80h,7Fh ;1.0/0.0
label @FcNan_Single dword
db 00h,00h,0C0h,0FFh ;0.0/0.0
label @FcnInfinity_Single dword
db 00h,00h,080h,0FFh ;-1.0/0.0
|
|||
|
01 Mar 2007, 13:14 |
|
< Last Thread | Next Thread > |
Forum Rules:
|