Re: [9fans] 8c dying with out of fixed registers

Re: [9fans] 8c dying with out of fixed registers

["Nils O. Selåsdal"]

andrey mirtchovski wrote:
> 8c has a 32K char limit for each line of code it processes.  i had a
> copy somewhere with that limit increased to 128K, but that's just a
> partial solution.
That isn't the problem here though, as far as I can see.
The snippet I mailed is a total of 569 bytes :-)
( the original code was also way below 32K.)

Re: [9fans] 8c dying with out of fixed registers

[andrey mirtchovski]

> andrey mirtchovski wrote:
>> 8c has a 32K char limit for each line of code it processes.  i had a
>> copy somewhere with that limit increased to 128K, but that's just a
>> partial solution.
> That isn't the problem here though, as far as I can see.
> The snippet I mailed is a total of 569 bytes :-)
> ( the original code was also way below 32K.)

perhaps you should run cpp on the code to find out what it devolves to.

Re: [9fans] 8c dying with out of fixed registers

[andrey mirtchovski]

> perhaps you should run cpp on the code to find out what it devolves to.

here's a one-liner in ffmpeg that went through a couple of macro
expansions (ultimately making it > 32K chars):

http://pages.cpsc.ucalgary.ca/~mirtchov/screenshots/ffmpeg.gif

Re: [9fans] 8c dying with out of fixed registers

["Nils O. Selåsdal"]

andrey mirtchovski wrote:
>>It only uses u.h, and I see nothing funny, thats even near a 32K limit, even
>>after cpp.
>>
>>Or did I just not understand your statement correctly. (*confused*)
> 
> 
> without an lstk() of the broken 8c binary everything is just
> speculation.  i didn't know any context, so i just suggested one bug
> that i know of as a possibility.  it wasn't even an educated guess
> (hence i omitted 9fans from the recipients list)

You can compile the snippet and lstk the broken 8c. That's why I tried to
massage it down to a single little sweet function, so others could try it
if they're up to the task ;-)
(Or is it only for me the piece originally posted suicides ?)

[9fans] 8c dying with out of fixed registers

["Nils O. Selåsdal"]

Hello,
I've been trying to compile some crypto code, and it seems 8c
isn't always too happy about it. I eventually got the monster ting
reduced to the below (now very-nonsense code) snippet which dies with;
a.c:22 out of fixed registers
8c 2655: suicide: sys: trap: fault read addr=0x0 pc=0x00036d93

So, for anyone that's bored, and wants to fix the compiler:
#include <u.h>

static uvlong T7[256];

void f(unsigned char *key)
{
	uvlong K1;
	int r;
	uvlong *S = T7;

	for (r = 1; r < 16; r++) {
		K1 =
		    T7[S[(int)(K1 >> 56)       ] & 0xff] ^
		    T7[(int)S[(int)(K1 >> 48) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1 >> 40) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1 >> 32) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1 >> 24) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1 >> 16) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1 >>  8) & 0xff] & 0xff] ^
		    T7[(int)S[(int)(K1      ) & 0xff] & 0xff];
	}
}


In the mean time, I'll rewrite the piece of code...

--
Nils O. Selåsdal