Needs to track this ourself since because we get into a race condition with
the dri_util.c code on make current when rendering to the front buffer.
This is what happens:
Old context is rendering to the front buffer.
App calls MakeCurrent with a new context. dri_util.c sets
drawable->driContextPriv to the new context and then calls the driver make
current. st/dri make current flushes the old context, which calls back into
st/dri via the flush frontbuffer hook. st/dri calls dri loader flush
frontbuffer, which calls invalidate buffer on the drawable into st/dri.
This is where things gets wrong. st/dri grabs the context from the dri
drawable (which now points to the new context) and calls invalidate
framebuffer to the new context which has not yet set the new drawable as its
framebuffers since we have not called make current yet, it asserts.
When clearing a GL_LUMINANCE_ALPHA buffer, for example, we need to convert
the clear color (R,G,B,A) to (R,R,R,A). We were doing this for texture border
colors but not renderbuffers. Move the translation function to st_format.c
and share it.
This fixes the piglit fbo-clear-formats test.
NOTE: This is a candidate for the 7.9 and 7.10 branches.
If finalizing a non-POW mipmapped texture with an odd-sized base texture
image we were allocating the wrong size of gallium texture (off by one).
Need to be more careful about computing the base texture image size.
This fixes https://bugs.freedesktop.org/show_bug.cgi?id=34463
I was being overly miserly and gave the offset of the buffer into the bo
insufficient bits, distracted by the adjacency of the buffer[4096].
Ref: https://bugs.freedesktop.org/show_bug.cgi?id=34541
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reducing the number of relocations has lots of nice knock-on effects,
not least including reducing batch buffer size, auxilliary array sizes
(vmalloced and copied into the kernel), processing of uncached
relocations etc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Rather than waiting on the first batch after the last swapbuffers to be
retired, call into the kernel to wait upon the retirement of any request
less than 20ms old. This has the twofold advantage of (a) not blocking
any other clients from utilizing the device whilst we wait and (b) we
attain higher throughput without overloading the system.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the next vertex arrays are a (discontiguous) continuation of the
current arrays, such that the new vertices are simply offset from the
start of the current vertex buffer definitions we can reuse those
defintions and avoid the overhead of relocations and invalidations.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Using a temporary buffer for large discontiguous uploads into the common
buffer and a single buffered upload is faster than performing the
discontiguous copies through a mapping into the GTT.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
