This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [C, C++] fundamental type object representation
- From: Paul Schlie <schlie at comcast dot net>
- To: Gabriel Dos Reis <gdr at integrable-solutions dot net>
- Cc: Ian Lance Taylor <ian at airs dot com>, <gcc at gcc dot gnu dot org>
- Date: Sun, 11 Dec 2005 17:40:32 -0500
- Subject: Re: [C, C++] fundamental type object representation
Gabriel Dos Reis wrote
> Paul Schlie writes:
> [...]
> | > OK, thanks. I was rewriting part of valarray and I saw I made that
> | > assumption, and since it is bug chasing time I thought I should make
> | > sure my use of memset() isn't going to bomb out.
> |
> | - If of any help, the c4x's float format is: [exp:8|sign:1|fract:23]
> | where the exponent is encoded in 2's complement form thereby:
> |
> | (float)0 :: [0|0|0] :: 1.0*2^0 :: 1.0
> |
> | and further defined that an exponent of -128 regardless of the value
> | of it's sign and fraction fields is interpreted as 0.0, thereby:
> |
> | (int)0.0 :: (int)X*2^-128 :: may = (int)[-128|0|0] :: -2,147,483,648
> |
> | So if it's defined that non otherwise initialized static data fields
> | are initialized to all zero's, then correspondingly such floats will be
> | interpreted as being initialized to 1.0 as far as the c4x is concerned;
> | which may or may not be ok, but possibly reasonable if simply known to
> | the the case?
>
> That may probably be OK -- I haven't heard of a C4X user doing
> instensive computations with valarray. But I guess I'll stick to the
> explicit loop and avoid having to revisit the valarray mess again.
>
> However, Ian and you have intrigued me; do you have a (preferably
> online) reference to C4X model?
- General TI DSP Docs:
http://focus.ti.com/dsp/docs/dspsupporttechdocs.tsp?sectionId=3&tabId=409&t
echDoc=6&familyId=497&documentCategoryId=6
- c4x User Guide:
http://focus.ti.com/lit/ug/spru063c/spru063c.pdf