| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
ideally this should be set once in the beginning of CS but there's
no way to change values there while in the middle of rendering.
For now reemitting SQ setup seems to work probably due to
r700WaitForIdleClean after each render
currently does not to try to decrease values once increased
fixes hangs in glsl-vs-vec4-indexing-temp-src-in-nested-loop-combined
glsl-vs-vec4-indexing-temp-dst-in-nested-loop-combined for my rv740
maybe more for other chips
|
|
When --enable-shared-glapi is specified, libGL will share libglapi with
OpenGL ES instead of defining its own copy of glapi. This makes sure an
app will get only one copy of glapi in its address space.
The new option is disabled by default. When enabled, libGL and libglapi
must be built from the same source tree and distributed together. This
requirement comes from the fact that the dispatch offsets used by these
libraries are re-assigned whenever GLAPI XMLs are changed.
For GLX, indirect rendering for has_different_protocol() functions is
tricky. A has_different_protocol() function is assigned only one
dispatch offset, yet each entry point needs a different protocol opcode.
It cannot be supported by the shared glapi. The fix to this is to make
glXGetProcAddress handle such functions specially before calling
_glapi_get_proc_address.
Note that these files are automatically generated/re-generated
src/glx/indirect.c
src/glx/indirect.h
src/mapi/glapi/glapi_mapi_tmp.h
|
|
|
|
I don't have evidence for this amounting to any improvement,
but it does codify a bit more what we understand so far about
the pipeline.
|
|
|
|
This fixes a bunch of unnecessary barriers due to the scheduler not
knowing what that arbitrary register description refers to when trying
to reason about its dependencies.
The result is rescheduling in the convolution kernel shader in
Lightsmark, which results in avoiding register spilling and increasing
the performance of the first scene from 6-7 fps midway through the
panning to 11fps. The register spilling was a regression from Mesa
7.9 to Mesa 7.10.
|
|
Improves performance of my GLSL demo by 5.1% (+/- 1.4%, n=7). It also
reschedules the giant multiply tree at the end of
glsl-fs-convolution-1 so that we end up not spilling registers,
producing the expected level of performance.
|
|
|
|
This is a candidate for 7.9 and 7.10
|
|
|
|
(really not sure why I'm doing this).
This is a candidate for 7.9 and 7.10 branches.
|
|
Can happen during swrast fallbacks if a buffer is somehow bound as
a render target and a texture.
Fixes gnome-shell on nv20, and gets it mostly working on nv10.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
sw clears were being used and not getting the correct offsets in the span
code.
also not emitting correct offsets for CB draws to texture levels.
(I've no idea why I'm playing with r100).
This is a candidate for 7.9 and 7.10
|
|
|
|
Fixes piglit glsl-fs-texture2d-branching. I couldn't come up with a
testcase that didn't involve dead code, but it's still worthwhile to
fix I think.
|
|
Hopefully better than previous - this passes more mipgen tests
|
|
border color is RGBA for samples - this passes texenv tests
|
|
use introduced STATE_FB_WPOS_Y_TRANSFORM variable (thanks Marek)
this gets coords also right when using fbo
|
|
Fixes texrect-many regression with ff_fragment_shader -- as we added
refs to the subsequent texcoord scaling paramters, the array got
realloced to a new address while our params[] still pointed at the old
location.
|
|
We just choose the texture format depending on the srgb decode bit
for the sRGB formats.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
|
|
We don't have all of the features of this extension hooked up yet, but
the consensus yesterday was that since those features are things that
we should also be supporting in our ES2 implementation, claiming ES2
here too doesn't make anything worse and will make incremental
improvement through piglit easier.
|
|
This code should never have been triggered, but I often did anyway
when I disabled optimization passes during debugging, then spent my
time debugging that this code doesn't work.
|
|
No effect, since it was called before live intervals were calculated.
|
|
|
|
Until we get the EXT_framebuffer_sRGB extension we should bind the sRGB
formats for FBO as linear.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
OES_standard_derivatives must be manually disabled for i915 because Mesa
enables it by default.
|
|
|
|
Without this, X doesn't start with UMS on r300g.
NOTE: This is a candidate for the 7.9 and 7.10 branches.
Signed-off-by: Paulo Zanoni <pzanoni@mandriva.com>
Signed-off-by: Brian Paul <brianp@vmware.com>
|
|
We were trying to interpolate, which would end up doing unnecessary
math, and doing so on undefined values. Fixes glsl-fs-flat-color.
|
|
With the change to not reset baselevel, this GL_LINEAR filtering was
resulting in generating mipmaps off of the base level instead of the
next higher detail level. Fixes fbo-generatemipmap-filtering.
Reported by: Neil Roberts <neil@linux.intel.com>
|
|
The ad-hoc placement of recalculation somewhere between when they got
invalidated and when they were next needed was confusing. This should
clarify what's going on here.
|
|
We were returning the negative absolute value, instead of the absolute
value. Fixes glsl-vs-abs-neg.
|
|
We were returning the negative absolute value, instead of the absolute
value. Fixes glsl-fs-abs-neg.
|
|
This greatly improves codegen for programs with flow control by
allowing coalescing for all instructions at the top level, not just
ones that follow the last flow control in the program.
|
|
The _Enabled field is the thing that takes into account whether
there's a stencil buffer. Tested with piglit glx-visuals-stencil.
|
|
Signed-off-by: Alberto Milone <alberto.milone@canonical.com>
|
|
When attaching a small mipmap level to an FBO, the original gen4
didn't have the bits to support rendering to it. Instead of falling
back, just blit it to a new little miptree just for it, and let it get
revalidated into the stack later just like any other new teximage.
Bug #30365.
|
|
It's been replaced by just setting texObj->mt to image->mt at TexImage
time.
|
|
This avoids relayouts in the common case of glGenerateMipmap() or
people doing similar things.
Bug #30366.
|
|
If we hit this path, we're level 1+ and the base level got allocated
as a single level instead of a full tree (so we don't match
intelObj->mt). This tries to recover from that so that we end up with
2 allocations and 1 validation blit (old -> new) instead of
allocations equal to number of levels and levels - 1 blits.
|
|
We don't need to worry about levels other than MaxLevel because we're
minifying -- the lower levels (higher detail) won't contribute to the
result. By changing BaseLevel, we forced hardware that doesn't
support BaseLevel != 0 to relayout the texture object.
|
|
This reverts commit 7ce6517f3ac41bf770ab39aba4509d4f535ef663.
This reverts commit d60145d06d999c5c76000499e6fa9351e11d17fa.
I was wrong about which generations supported baselevel adjustment --
it's just gen4, nothing earlier. This meant that i915 would have
never used the mag filter when baselevel != 0. Not a severe bug, but
not an intentional regression. I think we can fix the performance
issue another way.
|
|
|
|
|
|
Again, this makes it match the documentation.
|
|
Most _3DSTATE defines contain the command type, sub-type, opcode, and
sub-opcode (i.e. 0x7905). These, however, contain only the sub-opcode
(i.e. 0x05). Since they are inconsistent with the rest of the code and
nothing uses them, simply delete them.
The _3DOP and _3DCONTROL defines seemed similar, and were also unused.
|
|
|
