Re: [Nasm-devel] Optimizer check needed

[Ed B. <be...@mi...>]

H. Peter Anvin wrote:
> John Coffman wrote:
> 
>>
>> "Very tricky" it is, and "very tricky" it must stay.  The optimization 
>> was rather difficult to put in on top of an assembler that was not 
>> designed for it from the beginning.
>>
>> BTW:  The "passX" variable sequences are documented in the source 
>> code.  I don't know how much more explicit I can make them.
> 
> A big chunk of the problem is that it talks about sequencing, but not 
> what the variables actually control.  This makes it hard, not to say 
> impossible, to debug things.

I'd agree.  For instance, why isn't there a single "pass" variable which 
contains the actual pass number?  I still don't know and I haven't a 
clue as to why we need pass, pass_max, pass_cnt, pass0, pass1, and 
pass2, all of which are variable names that appear at least once just in 
nasm.c.

To be clear, I'm not blaming you, John, for this -- I'm just saying that 
it's one of the things that is an impediment to progress because I 
strongly suspect that nobody understands it.  I sure don't!  If you do, 
explaining it to the rest of us would be a truly valuable service, IMHO.

Ed