How about this?
I'm using it locally for nv40 immediate mode vertex emission.
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -236,6 +236,8 @@ enum pipe_transfer_usage {
#define PIPE_BUFFER_USAGE_VERTEX(1 << 5)
#define PIPE_BUFFER_USAGE_INDEX (1 <
How about this?
I'm using it locally for nv40 immediate mode vertex emission.
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -236,6 +236,8 @@ enum pipe_transfer_usage {
#define PIPE_BUFFER_USAGE_VERTEX(1 << 5)
#define PIPE_BUFFER_USAGE_INDEX (1 <
Well, we'd certainly want to keep that performance, so fixing the "bug" should
involve adding an alternative technique for getting the same behaviour...
Keith
From: Christoph Bumiller [e0425...@student.tuwien.ac.at]
Sent: Friday, January 22, 2010 9:12 AM
T
On Fri, Jan 22, 2010 at 11:41 AM, Keith Whitwell wrote:
> Christoph,
>
> The vertex usage issue is a bug and can be fixed.
>
> It sounds like you're really asking for another flag that indicates the
> expected lifetime of the buffer. That sounds orthogonal to the ones we
> already have describi
On 22.01.2010 18:12, Christoph Bumiller wrote:
> On 22.01.2010 17:41, Keith Whitwell wrote:
>
>> Christoph,
>>
>> The vertex usage issue is a bug and can be fixed.
>>
>>
>>
> Oh ... I thought it was intentional, and relied on it to decide whether
> to use immediate mode or to copy the bu
On 22.01.2010 17:41, Keith Whitwell wrote:
> Christoph,
>
> The vertex usage issue is a bug and can be fixed.
>
>
Oh ... I thought it was intentional, and relied on it to decide whether
to use immediate mode or to copy the buffers to VRAM/GART first.
That bug helped increase nexuiz' FPS tenfold
Christoph,
The vertex usage issue is a bug and can be fixed.
It sounds like you're really asking for another flag that indicates the
expected lifetime of the buffer. That sounds orthogonal to the ones we already
have describing the possible binding points.
I suspect ttm is already reusing buf
