code size latest snapshop versus 200206

[Till Straumann]

Chris Johns wrote:
> Till Straumann wrote:
> 
>> These are the numbers for hello.exe on a PowerPC (SVGM, derived from
>> shared) BSP:
>>
>> ss-20020301:
>>
>>     text       data        bss        dec        hex    filename
>>  138568       4236      18673     161477      276c5    samples/hello.exe
>>
>> ss-20030128:
>>
>>    text       data        bss        dec        hex    filename
>>  148820       5332      19205     173357      2a52d samples/hello.exe
> 
> 
> Made running some variation of the following :
> 
>   $ powerpc-rtems-nm --print-size --size-sort samples/hello.exe
> 
> will help show what has increased.
> 

Well - I compiled some statistics (see attachments).
The file 'hello.symdiffs' (generated using the [quickndirty]
diffcalc script)
lists the names of all symbols who's size changed from
ss-20020301 to ss-20030128 (I find readelf better suited than
nm --size-sort).
It's not clear who is really responsible for the net increase
as some things got bigger and some new symbols appear but other
symbols shrunk or disappeared.
There is also noticable fluctuation among small items.

_vfiprintf_r/_vfprintf_r and the two instances of __udivmoddi4
are candidates, though...

-- Till
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hello.symdiffs.gz
Type: application/x-gzip
Size: 4391 bytes
Desc: not available
URL: <http://lists.rtems.org/pipermail/users/attachments/20030226/3536d8fc/attachment.bin>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: diffcalc
URL: <http://lists.rtems.org/pipermail/users/attachments/20030226/3536d8fc/attachment.ksh>