| |
| |
| |
|
Comments:
<0> http://modseven.de/pastebin.php?id=999
<0> the huffman codes are <~26 bits long so I can use the lower 5 bits for the length
<1> okay, i'll take a look
<1> hmm, does the code actually work? (more specifically, does it work if the value of a byte in src is over 0x7f)?
<1> that zero-extending in the first move doesn't really make much sense
<0> it works..
<0> I think the zero extending is done because you can't write "[r10+4*al]"
<1> yeah, but zero-extending basically means "interpret as signed number", and the byte value seems to be used as a table offset. so actually if the byte was unsigned 128, which is in signed -128, it would be extended to a huge number, causing the offset to be huge as well, don't you think?
<0> I don't know.. I don't understand ***embler.. yet
<1> well, if you feed it an input array which has values in the range 128U-255U, does it not segfault?
<0> no, it doesn't
<1> hmmok. ah, right. the offset is of course signed, so actually it is just then accessing r10+-512, which may not segfault, but still won't be what you wanted
<1> the huffman code table is an array of 256 32bit values, starting at tbl right?
<0> yes
<1> have you verified the output to be the same with a reference encoder?
<0> there is no reference encoder
<1> how do you check everything works properly then?
<0> md5sum.. for now
<1> xor r11, r11 before loop then and just mov r11b, byte [esi] would be correct, I think. (then it would access memory from tbl to tbl+1024, which is what I think you want)
<2> :( sse has no adc
<0> using shl on one register instead of shld really IS faster :)
<3> using a microsoft c compiler, how to write bytecode with __asm directive.
<3> want to leave a message for crackers :)
<3> found it, it's _emit pseudoinstruction
<4> hello, i've got a problem with my asm programme, it's a ridiculous basic problem but i've got an error : instruction is not allowed http://www.rafb.net/paste/results/ev0SqG95.html
<4> it's a ridiculous basic programme
<5> kutio: on which line?
<6> kutio: line 27 is never executed
<4> why is never executed edcba ?
<6> because of ret
<6> in the boucle 'proc'
<6> but which instruction is not allowed ?
<6> which line ?
<6> it seems syntaxically correct
<6> at least for masm
<6> or tasm
<6> maybe mov ax,data
<6> mov ax,@data ?
<4> edcba, ok i try to compile without the ret of the boucle
<5> kutio: on which line do you get the error?
<4> i don't know skuggi
<4> when i compile my program , i don't have any error
<4> but when i launch it
<4> the processor have found an instruction forbidden
<4> CS:f000 IP:e5f3 OP:0f 00 00 ff ff
<5> ah, yes, that's probably because you return from a procedure you haven't called..
<5> after line 12 the program will proceed directly in to the procedure. you will have to either call the procedure or not use ret
<4> i should to not use the ret into boucle procedure ?
<5> kutio: i think what you should do is this: move lines 27-28 up to just below line 12 and then put a line with "call boucle" just after line 12
<4> ok thanx skuggi i try!
<4> it's ok skuggi thx :)
<5> np
<7> o_O
<8> how much will cost to have a blog like this ???? http://freelance-job.blogspot.com
Return to
#asm
or
Go to some related
logs:
#math
#math
charbal
lmms 0.1.4 ubuntu
pirut Unable to import modules
#math
#web
mplayer-moz
pigeonhole-sort
#math
|
|