Corey - Do you know of any documentation anywhere that discusses doing this quickly? I can think of a method which is basically like allocating your own stack of aligned variables... since I only ever use about 5 temps, and it's unlikely I'll be using more than a few threads, I could just allocate 20 __m128 variables aligned, and let them be requested in the functions. I can't help but be slightly disenchanted with the idea, but if this is the sort of thing that needs to be done... Thanks Corey, Brian On Wed, 16 Feb 2005 10:46:54 -0600, corey taylor <corey.taylor@xxxxxxxxx> wrote: > Implementation's I've used and worked on always do aligned allocations > manually. Typically the hidden and real sizes of the allocation are > put into the memory allocation itself and the returned pointer is > incremented a few bytes. The downside to this is that you must be > strict in using the aligned free routine also. > > corey > > > On Wed, 16 Feb 2005 10:09:27 -0600, Eljay Love-Jensen <eljay@xxxxxxxxx> wrote: > > Hi Brian, > > > > >Wow, that's terrible news if true. > > > > Take it with a grain of salt. I'm just speculating. > > > > You might want to try looking at the assembly output of the compiler, using > > the -S switch. It might shed some light as to what's going on under the > > covers, especially pertaining to alignment. > > > > >But surely thousands of people are writing sse code... how do they make > > it work? > > > > I presume by taking measures to assure the SSE structs are properly aligned. > > > > >Do I need to switch to the intel compiler/linker? > > > > I do not know. > > > > >Or is there a way to tell ld to allow larger alignment sizes? > > > > I'm not sure -- perhaps it is an option when ld is configure'd and make'd. > > > > >Are there other linkers available for (free) use with gcc? > > > > I do not know. > > > > HTH, > > --Eljay > > > > >
