Re: [jamboost] mipspro-tools.jam file

From: Rene Rivera (grafik666_at_[hidden])
Date: 2002-06-11 15:33:10

On 2002-06-11 at 03:14 PM, gohmandj_at_[hidden] (Dan Gohman) wrote:

>On Mon, Jun 10, 2002 at 02:20:03PM -0500, Rene Rivera wrote:
>> On 2002-06-10 at 01:54 PM, gohmandj_at_[hidden] (Dan Gohman) wrote:
>> >I'm thinking about suggesting something like this:
>> >
>> > feature integer-model : unspecified ilp32 lp64 ; # add others as
needed
>> 
>> Good name! It's specific enough that other options that also affect the
ABI
>> on other platforms don't collide :-)
>> 
>> Question, though... What does "ilp" and "lp" stand for?
>
>i, l, and p stand for int, long, and pointer. It really isn't that useful
>to be able to specify the size of `int' or `long' (as long as
><boost/cstdint.hpp> works :-)), but being able to specify the pointer
>size is very useful.
>
>Maybe the feature would be more appropriately be named `pointer-size'.
>
> # This feature specifies the pointer size in bits.
> feature pointer-size : unspecified 16 32 64 ;
>
>Optionally, `default' might look better than `unspecified' in everyone's
>build directories. How does that sound?

I'm on my way back to confusion...

What is it that the n64/32 specify for the mips compiler? pointer, integer,
register, or float size?

Do we need a feature for each of those? That is can they be specified
independantly to the compiler(s)?

Or is it safe to enforce a pointer=integer=register=float size of 16, 32, or
64?

Or is it that speficying ilp, specifies the size of integer=long=pointer?
And therefore lp is size of integer=long.

-- grafik - Don't Assume Anything
-- rrivera_at_[hidden] - grafik_at_[hidden]
-- 102708583_at_icq - Grafik666_at_AIM - Grafik_at_[hidden]