How to dissociate constants from text

How to dissociate constants from text

[Riadh Elloumi]

Hi all,

We are using GCC 3.3.3 cross compilation to ARM9e target. In our ARM9 
chip, we have TCM (tightly coupled memory) which is divided in 32K 
instrcutions (I-TCM) and 32K data (D-TCM). As this memory has a Harvard 
architecture, we get 1 wait state when loading a data from the 
instruction part (I-TCM), and zero wait state when fetching it from 
D-TCM. But arm-elf-gcc involves some constants in the end of functions 
and this makes the code slower when executed in I-TCM than in cache...

Is there any option in gcc, or a patch to make him separate these 
constants from the functions, and map them in, let's say, a .constants 
section that we can load in D-TCM.

Thank you in advance for your help.

Riadh.

Re: How to dissociate constants from text

[Riadh Elloumi]

Sorry for the confusion between emacs and gcc forums. Sure this is not 
the right place for my post.

Riadh Elloumi wrote:
> Hi all,
> 
> We are using GCC 3.3.3 cross compilation to ARM9e target. In our ARM9 
> chip, we have TCM (tightly coupled memory) which is divided in 32K 
> instrcutions (I-TCM) and 32K data (D-TCM). As this memory has a Harvard 
> architecture, we get 1 wait state when loading a data from the 
> instruction part (I-TCM), and zero wait state when fetching it from 
> D-TCM. But arm-elf-gcc involves some constants in the end of functions 
> and this makes the code slower when executed in I-TCM than in cache...
> 
> Is there any option in gcc, or a patch to make him separate these 
> constants from the functions, and map them in, let's say, a .constants 
> section that we can load in D-TCM.
> 
> Thank you in advance for your help.
> 
> Riadh.
> 


-- 
Riadh Elloumi
Software Architect
PARROT SA
174, quai de Jemmapes
75010 Paris, France

Tel +33 (0) 1 48 03 70 14
Fax +33 (0) 1 48 03 06 66
Email: riadh.elloumi@parrot.fr
http://www.parrot.biz

Re: How to dissociate constants from text

[Thien-Thi Nguyen]

Riadh Elloumi <riadh.elloumi@parrot.fr> writes:

> Sorry for the confusion between emacs and gcc forums. Sure this is not
> the right place for my post.

maybe it's the right place but the wrong time (someday, emacs will
support M-x compile-to-native for various HLLs, but not today... ;-).

thi