[VSDK] Framework prebinding

[Totte Alm]

2003-12-03 kl. 21.17 skrev Eric Forget:

> Hi Totte,
>
>> I would like to add some stuff I learned on the
>> Apple Optimization Kitchen/Workshop I attended just two weeks ago.
>
> Did Apple provided some HTML/PDF doc available for those who didn't  
> attend
> it?
Nope. Was just some presentations, and mostly coding and sharking.

If you wanna check code, use Shark from CHUD tools, its Awesome,  
absolutly awesome.

Free with Panther.

>
>> In the Build Style panel, do the following settings:
>>
>> OTHER_CFLAGS = "-falign-loops=16 -funroll-loops"
>
> I believe that -falign-loops=32 is better for G5:

Yes, it is.
>
> <http://developer.apple.com/technotes/tn/tn2087.html>
>
>> GCC_ALTIVEC_EXTENSIONS = YES
>
> Won't that prevent for running on G3? All iBooks, all old iMacs, etc...
Optimize G4 may also utilize non G3 opcodes.

>
>> COPY_PHASE_STRIP = YES
>> GCC_OPTIMIZATION_LEVEL = 3
>
>> From what I understand:
>
>     GCC_OPTIMIZATION_LEVEL          Optimization Level
>
>             3                           Fastest
>             s                           Fastest, smallest
>
> So you are sure it is not better to use "s" (or Fastest, smallest).  
> From
> apple doc:
>
> <http://developer.apple.com/documentation/Performance/Conceptual/ 
> CodeFootpri
> nt/Concepts/CompilerOptions.html>
>
> It says:
>
> Fastest
> -------
> Performs all optimizations in an attempt to improve the speed of the
> generated code. This option can increase the size of generated code as  
> the
> compiler performs aggressive inlining of functions.
>
> This option is generally not recommended. See “Avoid Excessive Function
> Inlining” for more information.
>
> Fastest, smallest
> -----------------
> Performs all optimizations that do not typically increase code size.  
> This is
> the preferred option for shipping code as it gives your executable a  
> smaller
> memory footprint.
>
> Just wondering what Apple said about that at the Kitchen?

Thay did mention both. Sometimes faster IS faster than  
fastest/smallest, depends on things like
"should very large loops be unrolled or not", smallest will prevent  
that, speed will benefit.

>
>> For G4 also use:
>> GCC_MODEL_CPU = G4
>> GCC_MODEL_TUNING = G4
>>
>> for G5, also use;
>> GCC_MODEL_CPU = G5
>> GCC_MODEL_TUNING = G5
>
> I think that for customer app, right now, we should use G4. For Pro,  
> G5.
> Maybe, the server (V.2) should use G5, but for the current kernel, G4?

maybe have G4 AND G5 versions of VSDK, so your custom installer for  
your app takes the right thing. G5 on G5 IS very good.

>
>> GCC_MODEL_PPC64 = YES
> That will make it none executable on G4, don’t it?

Yupp, but VERY fast on G5 with 64bit integer math

// Totte
>
> Éric
>
> ___________________________________________________________________
>
>  Eric Forget                       Cafederic
>  ForgetE at cafederic.com             <http://www.cafederic.com/>
>
>  Fingerprint <86D5 38F5 E1FD 5D9C 71C3  BAA3 797E 70A4 6210 C684>
>
>
> _______________________________________________
> Valentina mailing list
> Valentina at lists.macserve.net
> http://lists.macserve.net/mailman/listinfo/valentina
>
>
------------------------------------------------------------------------ 
--------------------------
It's the bean counters on the east coast that use windows.
We creative people on the west coast use macs. / American market  
research guru.