On 22.01.2010 17:01, Keith Whitwell wrote: > Christoph, > > The USAGE_CONSTANT flag is a bit confusing -- the intention is to state that > this buffer can be bound as a constant-buffer, not that it is itself > immutable. It should be perfectly legal to create a single constant buffer > and update it over time... > > I didn't interpret the USAGE_CONSTANT as meaning the buffer's contents should be constant at all, I just meant it should not be used for throw-away constant buffers. The same is done for vertex buffers created for glVertexPointer etc., they do NOT have PIPE_BUFFER_USAGE_VERTEX (they are user_buffer_create), but STILL get set as vertex buffers.
Alternatively, rather than not using USAGE_CONSTANT/VERTEX on those, we could finally add PIPE_BUFFER_USAGE_STATIC,STREAM,THROW_AWAY ... Of course, maybe we shouldn't take the buffer_create function literally and ask TTM to create a new buffer object every time it's called ... I admit that this whole thing is a pipe driver specific issue. > Keith > ________________________________________ > From: Christoph Bumiller [e0425...@student.tuwien.ac.at] > Sent: Thursday, January 21, 2010 3:48 PM > To: Michal Krol > Cc: mesa3d-dev; Keith Whitwell > Subject: Re: [Mesa3d-dev] [RFC] gallium-multiple-constant-buffers merge > > On 21.01.2010 20:20, michal wrote: > >> Hi, >> >> This simple feature branch adds support for two-dimensional constant >> buffers in TGSI. >> >> An example shader would look like this: >> >> FRAG >> >> DCL IN[0], COLOR, LINEAR >> DCL OUT[0], COLOR >> DCL CONST[1][1..2] >> >> MAD OUT[0], IN[0], CONST[1][2], CONST[1][1] >> >> END >> >> For this to work, one needs to bind a buffer to slot nr 1 containing at >> least 3 vectors. >> >> >> > I've been thinking about what to do with these multiple constant > buffers now, and I thought we might want to avoid creating those > buffers with PIPE_BUFFER_USAGE_CONSTANT that will be used > only once and then discarded, as the currently used uniform buffer > is. > Instead it should only be set for persistent ones that the user > created with glGenBuffers'd etc. (this would be similar to our usage of > PIPE_BUFFER_USAGE_VERTEX). > > Then I would be able to decided whether to put them in VRAM or > in user memory (since creation of VRAM buffers is a bit costly), > and then either copy/upload constants from the buffer if it's user, > or bind it's GPU virt. address to the program CB it if it's VRAM. > > Of course, I don't know if it would even be noticeable performance-wise > if I just kept around a bunch of buffers and pushed constants all the > time ... > > Any thoughts ? > Thanks, > Christoph > >> ------------------------------------------------------------------------------ >> Throughout its 18-year history, RSA Conference consistently attracts the >> world's best and brightest in the field, creating opportunities for >> Conference >> attendees to learn about information security's most important issues through >> interactions with peers, luminaries and emerging and established companies. >> http://p.sf.net/sfu/rsaconf-dev2dev >> _______________________________________________ >> Mesa3d-dev mailing list >> Mesa3d-dev@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/mesa3d-dev >> >> > ------------------------------------------------------------------------------ Throughout its 18-year history, RSA Conference consistently attracts the world's best and brightest in the field, creating opportunities for Conference attendees to learn about information security's most important issues through interactions with peers, luminaries and emerging and established companies. http://p.sf.net/sfu/rsaconf-dev2dev _______________________________________________ Mesa3d-dev mailing list Mesa3d-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
