[GEM-dev] [Fwd: PD-cvs Digest, Vol 20, Issue 22]
Hans-Christoph Steiner
hans at eds.org
Fri Oct 27 05:39:36 CEST 2006
On Oct 24, 2006, at 12:42 PM, james tittle wrote:
> On Oct 24, 2006, at 12:10 AM, Hans-Christoph Steiner wrote:
>> FYI, This is the gcc version:
>>
>> gcc (GCC) 3.3 20030304 (Apple Computer, Inc. build 1671)
>>
>> .hc
>>
>> On Oct 22, 2006, at 1:03 PM, IOhannes m zmoelnig wrote:
>>
>>> anybody knows, why alitvec fails to compile on os-x 10.3?
>>> is this an issue of an outdated gcc?
>
> ...10.3 is only supported up to a certain Xcode (I think 2.1?), so
> do we know what's installed on the compile farm?
The 10.3 box has the latest Xcode and the gcc update.
.hc
>>> <20061022081040.E4495FF114 at macosx103-powerpc.idmi.poly.edu>
>>> ...
>>> g++ -c -g -O2 -fPIC -freg-struct-return -O3 -falign-loops=16
>>> -funroll-loops -ffast-math -faltivec -I..
>>> -I/Users/pd/auto-build/pd-extended/pd/src
>>> GemPixConvertAltivec.cpp -o
>>> GemPixConvertAltivec.o
>
> ...snip...maybe we should add -maltivec to this compile? It's
> really for something else, and guarantees crashes if you try to run
> the code on a G3, but...?
>
>>> GemPixConvertAltivec.cpp: In function `void YUV422_to_YV12_altivec
>>> (short
>>> int*,
>>> short int*, short int*, short int*, unsigned char*, int, int)':
>>> GemPixConvertAltivec.cpp:636: error: no instance of overloaded
>>> builtin
>>> function
>>> `vec_sl' matches the parameter list
>
> ...snip...hmm, I had just made all the vectors "vector signed
> short", but on close re-examination of the altivec programmer's
> interface manual, the shift should be a 'vector unsigned short':
> I'll commit this and see if it goes away...
>
>>> GemPixConvertAltivec.cpp: In function `void YUV422_to_BGRA_altivec
>>> (unsigned
>>> char*, long unsigned int, unsigned char*)':
>>> GemPixConvertAltivec.cpp:772: error: cannot convert `vector int'
>>> to `vector
>>> unsigned int' in assignment
>
> ...snip...sure enough, similar problem here: the vec_add() is
> trying to do "vector unsigned int" = "vector signed int" + "vector
> signed int", but they should all be the same type...again, I'll try
> to change some types around, but I'll have to do a bit more testing
> to make sure it still does what I think it should...
>
> ...ultimate question, tho, is why don't I see these errors on my
> 10.4/gcc4 compiles?
>
> jamie
------------------------------------------------------------------------
http://at.or.at/hans/
More information about the GEM-dev
mailing list