mesa.git
13 years agoi965/gen4: Move the GS state to state streaming.
Eric Anholt [Mon, 25 Apr 2011 03:36:22 +0000 (20:36 -0700)]
i965/gen4: Move the GS state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen4: Move clip state to state streaming
Eric Anholt [Mon, 25 Apr 2011 03:20:16 +0000 (20:20 -0700)]
i965/gen4: Move clip state to state streaming

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Drop the now unused brw_cache_data() function.
Eric Anholt [Mon, 25 Apr 2011 03:05:02 +0000 (20:05 -0700)]
i965: Drop the now unused brw_cache_data() function.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Fix up state dumping for SF state.
Eric Anholt [Mon, 25 Apr 2011 02:57:33 +0000 (19:57 -0700)]
i965: Fix up state dumping for SF state.

It was moved to state streaming a while back and this was left over.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen4: Move VS state to state streaming.
Eric Anholt [Mon, 25 Apr 2011 02:47:37 +0000 (19:47 -0700)]
i965/gen4: Move VS state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agomesa: Add a bunch of documentation to the register allocator.
Eric Anholt [Sun, 24 Apr 2011 20:44:32 +0000 (13:44 -0700)]
mesa: Add a bunch of documentation to the register allocator.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Use the dynamic state base address to reduce relocations.
Eric Anholt [Fri, 22 Apr 2011 23:00:14 +0000 (16:00 -0700)]
i965/gen6: Use the dynamic state base address to reduce relocations.

Now that all the dynamic state is streamed through the top of the
batchbuffer, we can cut out many of our relocations to that state by
using the base address.

Improves 3DMMES taiji performance 3.3% +/- 0.4% (n=15).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move scissor state to state streaming.
Eric Anholt [Mon, 25 Apr 2011 03:02:38 +0000 (20:02 -0700)]
i965/gen6: Move scissor state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Move sampler state to state streaming.
Eric Anholt [Fri, 22 Apr 2011 22:30:42 +0000 (15:30 -0700)]
i965: Move sampler state to state streaming.

Overall, across this series since the last set of numbers, gen6 3DMMES
taiji performance has dropped 0.8% +/- 0.3% (n=15), probably due to
the increased reissuing of state from some of the state objects that
otherwise never changed, and increased occurrence of the per-batch
overhead as we've increased how much we put in the batch BO without
increasing the batch BO's size.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen4: Move WM state to state streaming.
Eric Anholt [Sat, 23 Apr 2011 07:33:43 +0000 (00:33 -0700)]
i965/gen4: Move WM state to state streaming.

The samplers are about to become streamed for gen6 performance, which
would cause this unit to blow out the state cache.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move the depth/stencil state to state streaming.
Eric Anholt [Fri, 22 Apr 2011 21:26:12 +0000 (14:26 -0700)]
i965/gen6: Move the depth/stencil state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move the blend state to state streaming.
Eric Anholt [Fri, 22 Apr 2011 21:07:30 +0000 (14:07 -0700)]
i965/gen6: Move the blend state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move the color calc state to state streaming.
Eric Anholt [Fri, 22 Apr 2011 20:54:10 +0000 (13:54 -0700)]
i965/gen6: Move the color calc state to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Move the CC VP to state streaming.
Eric Anholt [Fri, 22 Apr 2011 20:44:00 +0000 (13:44 -0700)]
i965: Move the CC VP to state streaming.

This is in a way a revert of f5bb775fd1f333d8e579d07a5cac1ded2bd54a2f.
The tiny win that had will be overwhelmed by the win of using the gen6
dynamic state base address.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move the SF VP to state streaming.
Eric Anholt [Fri, 22 Apr 2011 20:25:24 +0000 (13:25 -0700)]
i965/gen6: Move the SF VP to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Move the clip VP to state streaming.
Eric Anholt [Fri, 22 Apr 2011 20:20:08 +0000 (13:20 -0700)]
i965/gen6: Move the clip VP to state streaming.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Stream the VS push constants.
Eric Anholt [Fri, 22 Apr 2011 16:55:25 +0000 (09:55 -0700)]
i965/gen6: Stream the VS push constants.

Improves 3DMMES taiji demo performance by 10.1% +/- 0.9% (n=15).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/gen6: Stream the WM push constants.
Eric Anholt [Fri, 22 Apr 2011 02:03:18 +0000 (19:03 -0700)]
i965/gen6: Stream the WM push constants.

Improves 3DMMES taiji demo performance by 5.1% +/- 1.9% (n=15), by
reducing CPU time spent thrashing around those tiny little constant BOs.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Give the push constants and pull constants unique function names.
Eric Anholt [Fri, 22 Apr 2011 01:33:22 +0000 (18:33 -0700)]
i965: Give the push constants and pull constants unique function names.

This helps clarify profiling results.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Fix fragcoord_w on gen6 with 16-wide.
Eric Anholt [Fri, 29 Apr 2011 19:17:02 +0000 (12:17 -0700)]
i965: Fix fragcoord_w on gen6 with 16-wide.

The payload regs can go all the way up to register 60+, so just give
them 8 bits to be addressed by instead of 3-4 (which made source_w_reg
of 8 end up 0).  There's no reason to aggressively pack these fields,
as they are just used as compiler information, where being easier to
access is probably more important than shaving a byte or two off of
the structure.

Fixes piglit fragcoord_w.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36649

13 years agometa: Don't ask for floating point textures if not ARB_texture_float.
Eric Anholt [Tue, 26 Apr 2011 19:56:37 +0000 (12:56 -0700)]
meta: Don't ask for floating point textures if not ARB_texture_float.

I was promoting to float for ARB_color_buffer_float unclamped, which
failed when ARB_texture_float wasn't present.  Since the metaops don't
need results outside of [0,1] when not drawing to a floating point
destination, they can just use a fixed point texture when floating
point destinations are impossible.

Fixes regression in fdo23670-depth_test when --enable-texture-float is
not present.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36473

13 years agomesa: fix a few incorrect error messages
Brian Paul [Fri, 29 Apr 2011 18:03:28 +0000 (12:03 -0600)]
mesa: fix a few incorrect error messages

13 years agoegl/wayland: Fix possible lockup in drm initialization
Benjamin Franzke [Fri, 29 Apr 2011 15:04:11 +0000 (17:04 +0200)]
egl/wayland: Fix possible lockup in drm initialization

Lockup happens when wl_drm interface is not available.

13 years agoAdd git_sha1.h to .gitignore
Jon TURNEY [Tue, 26 Apr 2011 12:11:23 +0000 (13:11 +0100)]
Add git_sha1.h to .gitignore

Since commit de579a1 "Include GIT SHA1 in GL version string"

$ git status
 On branch master
 Your branch is ahead of 'origin/master' by 2 commits.

 Untracked files:
   (use "git add <file>..." to include in what will be committed)

       src/mesa/main/git_sha1.h
nothing added to commit but untracked files present (use "git add" to track)

Add git_sha1.h to .gitignore so git knows not to warn it is present but untracked

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
13 years agor600g: print opcodes names instead of numbers
Marek Olšák [Fri, 29 Apr 2011 11:16:24 +0000 (13:16 +0200)]
r600g: print opcodes names instead of numbers

13 years agodocs: update relnotes-7.11
Marek Olšák [Thu, 28 Apr 2011 15:33:05 +0000 (17:33 +0200)]
docs: update relnotes-7.11

13 years agodocs: update GL3 status
Marek Olšák [Thu, 28 Apr 2011 15:30:55 +0000 (17:30 +0200)]
docs: update GL3 status

13 years agomesa: remove unused variable bytesPerRow
Marek Olšák [Thu, 28 Apr 2011 17:34:56 +0000 (19:34 +0200)]
mesa: remove unused variable bytesPerRow

13 years agor600g: implement EXT_packed_float
Marek Olšák [Wed, 27 Apr 2011 14:30:28 +0000 (16:30 +0200)]
r600g: implement EXT_packed_float

13 years agost/mesa: support EXT_packed_float
Marek Olšák [Tue, 26 Apr 2011 00:28:33 +0000 (02:28 +0200)]
st/mesa: support EXT_packed_float

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agomesa: implement EXT_packed_float
Marek Olšák [Tue, 26 Apr 2011 00:27:25 +0000 (02:27 +0200)]
mesa: implement EXT_packed_float

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
13 years agoutil: implement R11G11B10_FLOAT pack/unpack functions
Marek Olšák [Wed, 27 Apr 2011 11:37:27 +0000 (13:37 +0200)]
util: implement R11G11B10_FLOAT pack/unpack functions

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agomesa, util: move RGB9E5 conversion functions to gallium/util
Marek Olšák [Wed, 27 Apr 2011 10:52:10 +0000 (12:52 +0200)]
mesa, util: move RGB9E5 conversion functions to gallium/util

Also use MAX3 and incorporate Ian's suggestion in texformat.c.

I don't think wrapping u_format_rgb9e5.h in another header and thus making it
more complicated is worth it.

13 years agor600g: trivially implement EXT_texture_shared_exponent
Marek Olšák [Tue, 26 Apr 2011 00:24:47 +0000 (02:24 +0200)]
r600g: trivially implement EXT_texture_shared_exponent

Nothing else needed.

13 years agoutil: implement R9G9B9E5 pack and unpack functions
Marek Olšák [Tue, 26 Apr 2011 00:29:00 +0000 (02:29 +0200)]
util: implement R9G9B9E5 pack and unpack functions

softpipe and llvmpipe support done (sampler only).

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agost/mesa: support EXT_texture_shared_exponent
Marek Olšák [Tue, 26 Apr 2011 00:23:33 +0000 (02:23 +0200)]
st/mesa: support EXT_texture_shared_exponent

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agomesa: implement EXT_texture_shared_exponent
Marek Olšák [Tue, 26 Apr 2011 00:18:24 +0000 (02:18 +0200)]
mesa: implement EXT_texture_shared_exponent

swrast support done.

There is no renderbuffer support in swrast, because it's not required
by the extension.

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agomesa: set reasonable defaults in update_wrapper
Marek Olšák [Mon, 25 Apr 2011 23:53:20 +0000 (01:53 +0200)]
mesa: set reasonable defaults in update_wrapper

I was wondering why I had been getting GL_RGBA for GL_RGB9_E5.

Instead of setting GL_RGBA and CHAN_TYPE for most types,
use the helper functions to obtain the info.

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agoi915g: Enable S3TC texture support
Christopher Egert [Fri, 29 Apr 2011 00:13:35 +0000 (02:13 +0200)]
i915g: Enable S3TC texture support

Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Christopher Egert <cme3000@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
13 years agoprog_print: Add support for printing the TXD opcode.
Kenneth Graunke [Tue, 26 Apr 2011 23:14:06 +0000 (16:14 -0700)]
prog_print: Add support for printing the TXD opcode.

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agollvmpipe: move active_query assignment
Brian Paul [Thu, 28 Apr 2011 18:41:45 +0000 (12:41 -0600)]
llvmpipe: move active_query assignment

Fixes piglit regression.

13 years agollvmpipe: move active_query assignment in lp_setup_begin_query()
Brian Paul [Thu, 28 Apr 2011 13:38:53 +0000 (07:38 -0600)]
llvmpipe: move active_query assignment in lp_setup_begin_query()

If we run out of bin memory and do an early return from
lp_setup_begin_query() we'd omit setting the setup->active_query
pointer.  Then, when lp_setup_end_query() was later called, the
assertion for setup->active_query == pq would fail.  Moving the
assigment in lp_setup_begin_query() avoids that.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
13 years agosgi: Fix MSVC build.
José Fonseca [Wed, 27 Apr 2011 18:05:29 +0000 (19:05 +0100)]
sgi: Fix MSVC build.

Including windows.h was ineffective on MSVC because we define the NOGDI macro,
which skips the wingdi.h include.

Unsetting NOGDI is also a bad idea because it causes all sort of symbol
clashes with SGI code.

The real problem is that WINGDAPI was not being defined, also due to NOGDI,
so simply define it to blank if not done already.  This seems to make
everybody happy.

13 years agost/mesa: raise shader MaxParameters if driver supports more
Brian Paul [Wed, 27 Apr 2011 15:31:30 +0000 (09:31 -0600)]
st/mesa: raise shader MaxParameters if driver supports more

The default value is 64 but drivers usually advertise more, like 4096.
Allows ARB vp/fp programs to use more parameters.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
13 years agomesa: emit more info in program parser error message
Brian Paul [Wed, 27 Apr 2011 15:23:16 +0000 (09:23 -0600)]
mesa: emit more info in program parser error message

13 years agosvga: emit user-defined clip plane state
Brian Paul [Mon, 25 Apr 2011 16:26:19 +0000 (10:26 -0600)]
svga: emit user-defined clip plane state

User-defined clip planes were a swtnl fallback before.

13 years agodocs: Note that ES2_compatibility is already done.
Eric Anholt [Tue, 26 Apr 2011 22:14:01 +0000 (15:14 -0700)]
docs: Note that ES2_compatibility is already done.

13 years agoRevert "intel: use throttle ioctl for throttling"
Eric Anholt [Mon, 25 Apr 2011 17:36:29 +0000 (10:36 -0700)]
Revert "intel: use throttle ioctl for throttling"

This reverts commit 50ade6ea697953bb17e3ca7210515fbd0411cd1e.

Fixes jerky rendering again on apps that don't block on the GPU per
frame and are GPU bound (e.g. 3DMMES on Ironlake).  The whole point of
this complicated throttle scheme is to wait on frame n-1 to have
started rendering before starting frame n's rendering.  Otherwise, the
GPU-bound app will race ahead and call the GL to draw many
nearly-identical frames, then >0ms later get stuck waiting for them
(all dispatched at about the same time) to retire, then render a new
batch of nearly-identical frames.

13 years agost/mesa: choose 3-component float formats before 4-component formats
Brian Paul [Wed, 27 Apr 2011 15:06:07 +0000 (09:06 -0600)]
st/mesa: choose 3-component float formats before 4-component formats

If GL_RGB16F or GL_RGB32F is specified let's try the 3-component float
texture formats before trying the 4-component ones.  Before this,
GL_RGB16/32F were treated the same as GL_RGBA16/32F.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
13 years agost/mesa: whitespace fixes
Brian Paul [Wed, 27 Apr 2011 15:05:30 +0000 (09:05 -0600)]
st/mesa: whitespace fixes

13 years agoutil: make macros MIN3, MAX3, MIN4, MAX4 little more efficient
Marek Olšák [Wed, 27 Apr 2011 10:54:24 +0000 (12:54 +0200)]
util: make macros MIN3, MAX3, MIN4, MAX4 little more efficient

13 years agomesa: add macros MIN3 and MAX3
Marek Olšák [Wed, 27 Apr 2011 10:53:58 +0000 (12:53 +0200)]
mesa: add macros MIN3 and MAX3

13 years agogallium/nouveau: fix printf warnings
Marek Olšák [Wed, 27 Apr 2011 11:08:25 +0000 (13:08 +0200)]
gallium/nouveau: fix printf warnings

13 years agogallivm: fix warning: ‘value’ may be used uninitialized in this function
Marek Olšák [Wed, 27 Apr 2011 11:05:53 +0000 (13:05 +0200)]
gallivm: fix warning: ‘value’ may be used uninitialized in this function

The path where it's uninitialized is guarded by an assert.

13 years agost/mesa: fix warning: ‘user_memory’ may be used uninitialized in this function
Marek Olšák [Wed, 27 Apr 2011 11:04:21 +0000 (13:04 +0200)]
st/mesa: fix warning: ‘user_memory’ may be used uninitialized in this function

It's initialized later in a conditional the condition of which is always true
the first time it's evaluated.

13 years agosvga: fix warning: ‘uc.ui’ may be used uninitialized in this function
Marek Olšák [Wed, 27 Apr 2011 11:02:32 +0000 (13:02 +0200)]
svga: fix warning: ‘uc.ui’ may be used uninitialized in this function

This is safe because it's initialized if buffers & PIPE_CLEAR_COLOR and
probably doesn't have any effect otherwise.

13 years agoos: simplify ifdef
nobled [Sun, 18 Jul 2010 16:08:19 +0000 (16:08 +0000)]
os: simplify ifdef

The actual code that needs this include is just using
"if defined (PIPE_OS_UNIX)", and the two conditions should match.

This should also make the file compile under Hurd.

13 years agodocs: fix typo, wrap lines in 7.11 release notes
Brian Paul [Tue, 26 Apr 2011 20:58:33 +0000 (14:58 -0600)]
docs: fix typo, wrap lines in 7.11 release notes

13 years agodocs: add GL_ARB_robustness to 7.11 release notes
Brian Paul [Tue, 26 Apr 2011 20:57:57 +0000 (14:57 -0600)]
docs: add GL_ARB_robustness to 7.11 release notes

13 years agoSquashed commit of the following:
Brian Paul [Tue, 26 Apr 2011 20:54:41 +0000 (14:54 -0600)]
Squashed commit of the following:

commit 864fe253b04105b7469e5f7b064dc37637b944f8
Author: Brian Paul <brianp@vmware.com>
Date:   Thu Apr 21 20:13:07 2011 -0600

    mesa: s/exec/disp/ in _mesa_init_histogram_dispatch()

    This function isn't normally compiled (FEATURE_histogram).

commit f4bf45e2b94b582cacd19cdca873c5be627e4250
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:58 2011 -0600

    mesa: hook up GL_ARB_robustness dispatch functions

    ...and advertise the extension.

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 2b89e38e5f572dc40cebc06381ae7c5d04386998
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:58 2011 -0600

    mesa: regenerated API files for GL_ARB_robustness

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 5d5ebfb7135cec9d833adef86cbf4d0f3d9beca8
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    glapi: add ARB_robustness xml

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 0159d1d6d99f4bbc18381dc2081c20d3aff17ac9
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: implement GL_ARB_robustness functions

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 938fd71f4c4742f274922d53492a7290ab8d9c9b
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: add context fields for GL_ARB_robustness

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 72075137bc79e65be03dac7e97b6dba93c3a86a4
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: standardize more bounds-checking error messages

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 32a3fc23746db49da903fbc08afa0135af3007d2
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: standardize some bounds-checking error messages

Signed-off-by: Brian Paul <brianp@vmware.com>
commit cecbf1f4d164207de373dec0cadee2e84e1f9656
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: add more bounds-checking support for client memory buffers

Signed-off-by: Brian Paul <brianp@vmware.com>
commit edc895b52383d5bd274422db56adead1d81daf5f
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: add bounds-checking support for client memory buffers

Signed-off-by: Brian Paul <brianp@vmware.com>
commit 3a96ef28a538f158a219b406cd090dee70470c85
Author: nobled <nobled@dreamwidth.org>
Date:   Thu Apr 21 07:53:57 2011 -0600

    mesa: use is_bufferobj() helper function

Signed-off-by: Brian Paul <brianp@vmware.com>
13 years agoi965/fs: Add support for compute-to-mrf in 16-wide mode.
Eric Anholt [Mon, 28 Mar 2011 23:54:39 +0000 (16:54 -0700)]
i965/fs: Add support for compute-to-mrf in 16-wide mode.

This is more painful than instruction scheduling, as we have to
compare two MRF writes to see if they coincide, and have to handle
partial GRF writes before that (for example, the result of a math
instruction written to color).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Typo fix a comment.
Eric Anholt [Mon, 28 Mar 2011 23:55:14 +0000 (16:55 -0700)]
i965/fs: Typo fix a comment.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Enable constant propagation in 16-wide.
Eric Anholt [Fri, 25 Mar 2011 18:19:53 +0000 (11:19 -0700)]
i965/fs: Enable constant propagation in 16-wide.

All that needed fixing was skipping the newly-possible
uncompressed/sechalf partial GRF constant writes.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Fix and enable the instruction scheduler for 16-wide.
Eric Anholt [Wed, 23 Mar 2011 20:53:26 +0000 (13:53 -0700)]
i965/fs: Fix and enable the instruction scheduler for 16-wide.

Most of the work of the scheduler is agnostic to wide dispatch.  It
operates on our virtual GRF file, which means instructions are
generally referring to 8 or 16 wide naturally.  For the MRF file
management we're trying to track the actual hardware MRF file, so we
need to watch if an instruction writes multiple MRFs.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for 16-wide dispatch with uniforms in use.
Eric Anholt [Wed, 23 Mar 2011 19:50:53 +0000 (12:50 -0700)]
i965/fs: Add support for 16-wide dispatch with uniforms in use.

This is glued in in a bit of an ugly way -- we rely on the uniforms
having been set up by 8-wide dispatch, and we just reuse them without
the ability to add new uniforms for any reason, since the 8-wide
compile is already completed.  Today, this all works out because our
optimization passes are effectively the same for both and even if they
weren't, we don't reduce the set of uniforms pushed after
optimization.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agohash_table: Add an iterator for doing things like cleanup of the HT.
Eric Anholt [Wed, 23 Mar 2011 19:05:56 +0000 (12:05 -0700)]
hash_table: Add an iterator for doing things like cleanup of the HT.

Without this, consumers often have to keep linked lists of the
entries, at additional malloc cost.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add a little whitespace between shader dumping debug.
Eric Anholt [Wed, 23 Mar 2011 19:36:52 +0000 (12:36 -0700)]
i965/fs: Add a little whitespace between shader dumping debug.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for compr4 MRF writes.
Eric Anholt [Mon, 28 Mar 2011 18:29:55 +0000 (11:29 -0700)]
i965/fs: Add support for compr4 MRF writes.

These reduce an emitted (not decoded) instruction per shader on
g4x/gen5, but may allow for additional register coalescing as well.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for 16-wide dispatch on gen5.
Eric Anholt [Mon, 14 Mar 2011 17:29:12 +0000 (10:29 -0700)]
i965/fs: Add support for 16-wide dispatch on gen5.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add initial support for 16-wide dispatch on gen6.
Eric Anholt [Sat, 12 Mar 2011 03:19:01 +0000 (19:19 -0800)]
i965/fs: Add initial support for 16-wide dispatch on gen6.

At this point it doesn't do uniforms, which have to be laid out the
same between 8 and 16.  Other than that, it supports everything but
flow control, which was the thing that forced us to choose 8-wide for
general GLSL support.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for discard instructions in 16-wide mode.
Eric Anholt [Thu, 24 Mar 2011 16:15:55 +0000 (09:15 -0700)]
i965/fs: Add support for discard instructions in 16-wide mode.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for math instructions in 16-wide mode.
Eric Anholt [Tue, 29 Mar 2011 01:03:37 +0000 (18:03 -0700)]
i965/fs: Add support for math instructions in 16-wide mode.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Fix interference calculation of pixel_[xy] in 16-wide.
Eric Anholt [Thu, 24 Mar 2011 16:08:58 +0000 (09:08 -0700)]
i965/fs: Fix interference calculation of pixel_[xy] in 16-wide.

Fixes glsl-fs-ceil in that mode, which produced the code in the comment.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Disable some optimization passes under 16-wide for now.
Eric Anholt [Wed, 23 Mar 2011 21:00:01 +0000 (14:00 -0700)]
i965/fs: Disable some optimization passes under 16-wide for now.

These are fixable for 16, but that can wait until after it's basically
working.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for 16-wide texturing on gen5+.
Eric Anholt [Sun, 13 Mar 2011 20:27:59 +0000 (13:27 -0700)]
i965/fs: Add support for 16-wide texturing on gen5+.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for computing pixel_[xy] in 16-wide.
Eric Anholt [Mon, 21 Mar 2011 22:54:18 +0000 (15:54 -0700)]
i965/fs: Add support for computing pixel_[xy] in 16-wide.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965/fs: Add support for 16-wide dispatch to the register allocator.
Eric Anholt [Sun, 13 Mar 2011 20:26:09 +0000 (13:26 -0700)]
i965/fs: Add support for 16-wide dispatch to the register allocator.

Note that the virtual grfs are in increments of the dispatch_width,
not hardware registers -- this makes the 16-wide emit and 8-wide emit
mostly the same.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agoi965: Move the destination reg setup for 8/16 wide to the emit code.
Eric Anholt [Mon, 14 Mar 2011 17:51:19 +0000 (10:51 -0700)]
i965: Move the destination reg setup for 8/16 wide to the emit code.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agointel: Use tiling for dri2AllocateBuffer implementation
Kristian Høgsberg [Mon, 25 Apr 2011 13:53:22 +0000 (09:53 -0400)]
intel: Use tiling for dri2AllocateBuffer implementation

13 years agointel: Set gen in intelInitScreen, just copy value in intelInitContext
Kristian Høgsberg [Mon, 25 Apr 2011 13:49:09 +0000 (09:49 -0400)]
intel: Set gen in intelInitScreen, just copy value in intelInitContext

13 years agointel: Use X tiling for DRM EGL Images
Kristian Høgsberg [Sat, 23 Apr 2011 01:05:40 +0000 (21:05 -0400)]
intel: Use X tiling for DRM EGL Images

13 years agomesa: Remove SWcontext::_FogMode, use gl_context::gl_fog_attrib::Mode everywhere
Ian Romanick [Fri, 22 Apr 2011 00:36:17 +0000 (17:36 -0700)]
mesa: Remove SWcontext::_FogMode, use gl_context::gl_fog_attrib::Mode everywhere

Reviewed-by: Brian Paul <brianp@vmware.com>
13 years agoegl: Use the right extension name in #ifdef
Kristian Høgsberg [Mon, 25 Apr 2011 19:34:04 +0000 (15:34 -0400)]
egl: Use the right extension name in #ifdef

13 years agowayland: Fix prototypes for EGL_WL_bind_wayland_display extension
Kristian Høgsberg [Mon, 25 Apr 2011 19:06:41 +0000 (15:06 -0400)]
wayland: Fix prototypes for EGL_WL_bind_wayland_display extension

13 years agomain: remove duplicated includes
Nicolas Kaiser [Mon, 25 Apr 2011 16:37:08 +0000 (18:37 +0200)]
main: remove duplicated includes

Remove duplicated includes of guarded headers.

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: Brian Paul <brianp@vmware.com>
13 years agowinsys/wayland: Fix typo in Makefile
Benjamin Franzke [Mon, 25 Apr 2011 16:16:35 +0000 (18:16 +0200)]
winsys/wayland: Fix typo in Makefile

Reported by dir1212 on irc.

13 years agost/egl: Add wayland shm softpipe support
Benjamin Franzke [Sat, 23 Apr 2011 12:21:05 +0000 (14:21 +0200)]
st/egl: Add wayland shm softpipe support

13 years agost/egl: Generalize wayland backend a bit
Benjamin Franzke [Sat, 23 Apr 2011 12:20:24 +0000 (14:20 +0200)]
st/egl: Generalize wayland backend a bit

13 years agowinsys: Add wayland shm sw winsys
Benjamin Franzke [Sat, 23 Apr 2011 12:14:29 +0000 (14:14 +0200)]
winsys: Add wayland shm sw winsys

13 years agor600g: Unify comment style somewhat.
Henri Verbeet [Mon, 25 Apr 2011 11:28:55 +0000 (13:28 +0200)]
r600g: Unify comment style somewhat.

Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
13 years agor600g: Cleanup the big endian support a bit.
Henri Verbeet [Mon, 25 Apr 2011 11:28:55 +0000 (13:28 +0200)]
r600g: Cleanup the big endian support a bit.

In particular, make sure the code is at least compiled on little endian
systems.

Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
13 years agor600g: Use EG constants in EG r600_colorformat_endian_swap().
Henri Verbeet [Mon, 25 Apr 2011 11:28:55 +0000 (13:28 +0200)]
r600g: Use EG constants in EG r600_colorformat_endian_swap().

This would actually fail to compile when PIPE_ARCH_BIG_ENDIAN is defined.

Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
13 years agoconfigure.ac: require LLVM for r300g on x86 and x86_64
Marek Olšák [Thu, 21 Apr 2011 11:33:29 +0000 (13:33 +0200)]
configure.ac: require LLVM for r300g on x86 and x86_64

13 years agoconfigure.ac: enable LLVM by default on x86 and x86_64
Marek Olšák [Thu, 21 Apr 2011 11:27:55 +0000 (13:27 +0200)]
configure.ac: enable LLVM by default on x86 and x86_64

13 years agomesa: add stricter checks for float formats in the texstore memcpy path
Marek Olšák [Fri, 22 Apr 2011 17:29:55 +0000 (19:29 +0200)]
mesa: add stricter checks for float formats in the texstore memcpy path

E.g. when the internal format was RGBA16F and the source was RG, it would use
memcpy.

13 years agomesa: implement generate-mipmap fallback for RGB10_A2
Marek Olšák [Fri, 22 Apr 2011 16:22:51 +0000 (18:22 +0200)]
mesa: implement generate-mipmap fallback for RGB10_A2

I hit this when testing RV350, which lacks RGB10_A2 render target
support. It had been missed when implementing the format and probably
unused by anything else too.

Not applicable to 7.10.

Reviewed-by: Eric Anholt <eric@anholt.net>
13 years agor600g: Remove r600_helper.c from SConscript.
Vinson Lee [Mon, 25 Apr 2011 06:58:39 +0000 (23:58 -0700)]
r600g: Remove r600_helper.c from SConscript.

This is a follow-up to commit d737857ed2ff4313fd6046dcd80018c6308a53c5.

13 years agodocs: update GL3 status
Marek Olšák [Sun, 24 Apr 2011 23:52:33 +0000 (01:52 +0200)]
docs: update GL3 status

13 years agor600g: remove some pointless and unused functions
Marek Olšák [Thu, 21 Apr 2011 15:24:50 +0000 (17:24 +0200)]
r600g: remove some pointless and unused functions

13 years agor600g: do not reset device to 0 when doing unrelated operations
Marek Olšák [Thu, 21 Apr 2011 15:20:27 +0000 (17:20 +0200)]
r600g: do not reset device to 0 when doing unrelated operations

Seems to be a copy-paste bug.