<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW - Unify the format conversion code" href="https://bugs.freedesktop.org/show_bug.cgi?id=84566#c22">Comment # 22</a> on <a class="bz_bug_link bz_status_NEW " title="NEW - Unify the format conversion code" href="https://bugs.freedesktop.org/show_bug.cgi?id=84566">bug 84566</a> from <a class="email" href="mailto:itoral@igalia.com" title="Iago Toral <itoral@igalia.com>"> Iago Toral</a> <pre>(In reply to Jason Ekstrand from <a href="show_bug.cgi?id=84566#c20">comment #20</a>) > (In reply to Iago Toral from <a href="show_bug.cgi?id=84566#c18">comment #18</a>) > > (In reply to Jason Ekstrand from <a href="show_bug.cgi?id=84566#c17">comment #17</a>) > > > (In reply to Iago Toral from <a href="show_bug.cgi?id=84566#c16">comment #16</a>) > > > > We also need new mesa_format enums and pack/unpack functions for byte > > > > swapped variants of non-array types. > > > > > > Why? Is this for the byte_swapped flag on GL upload/downloads? If that's > > > the only reason, then I'd rather not bother adding mesa types. We can > > > easily do a byte swap with _mesa_swizzle_and_convert. While it's a little > > > bit of a performance hit, people shouldn't be using that attribute anyway. > > > > I think that won't help in this case. We can only use > > _mesa_swizzle_and_convert for array types, but as I say, this involves > > non-array types like GL_UNSIGNED_INT_10_10_10_2. The master converter will > > not use _mesa_swizzle_and_convert with these, it goes through things like > > _mesa_unpack_rgba_row and _mesa_pack_float_rgba_row for example (which rely > > on auto-generated pack/unpack functions). Do you think we should try expand > > these to consider byte-swapping? > > My inclination on byte-swapping it to keep it out of the master conversion > function. It's really an oddity of the old GL api and not a core format > conversion thing. No sanely written app should use byte swapping, so I'm > not too worried about making it fast. My inclination there is to keep going > byteswapping and the other format conversion ops in TexImage, GetTexImag, > ReadPixels, and DrawPixels. They can call _mesa_swizzle_and_convert to do > the byteswap operation befor calling into the actual format conversion > function. I'll try it this way then. > The downside here is that there are a few cases where we can do better. > However, that's pretty much limited to GL_UNSIGNED_INT_8_8_8_8[_REV] and I > don't think it's worth the effort. I am already handling GL_UNSIGNED_INT_8_8_8_8[_REV] with byte swapping when we generate the mesa_array_format from the gl format and type.(it is only a matter of reverting the component mapping in this case), so if we want we can have TexImage and cia handle byte-swapping for all formats but this one.</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the QA Contact for the bug.</li> </ul> </body> </html>