mesa.git
16 years ago[965] Convert VS unit to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 23:03:32 +0000 (15:03 -0800)]
[965] Convert VS unit to use a cache key instead of brw_cache_data.

16 years ago[965] Convert SF unit to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 22:47:49 +0000 (14:47 -0800)]
[965] Convert SF unit to use a cache key instead of brw_cache_data.

16 years ago[965] Convert GS unit to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 22:09:06 +0000 (14:09 -0800)]
[965] Convert GS unit to use a cache key instead of brw_cache_data.

16 years ago[965] Convert clip unit to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 21:53:32 +0000 (13:53 -0800)]
[965] Convert clip unit to use a cache key instead of brw_cache_data.

16 years ago[965] Convert CC unit to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 21:20:00 +0000 (13:20 -0800)]
[965] Convert CC unit to use a cache key instead of brw_cache_data.

16 years ago[965] Convert surface state to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 20:07:00 +0000 (12:07 -0800)]
[965] Convert surface state to use a cache key instead of brw_cache_data.

16 years ago[965] Convert sampler state to use a cache key instead of brw_cache_data.
Eric Anholt [Wed, 2 Jan 2008 19:33:55 +0000 (11:33 -0800)]
[965] Convert sampler state to use a cache key instead of brw_cache_data.

16 years agoRevert "[intel] Use the memory type mask containing the caching flags."
Eric Anholt [Wed, 2 Jan 2008 23:51:27 +0000 (15:51 -0800)]
Revert "[intel] Use the memory type mask containing the caching flags."

This reverts commit 8bb9ae3693362a302206255c61f512d942df9bbf.

Validating our kernel buffers with the caching off in flags but on in mask
means that the kernel migrates the buffer to be uncached, which is undesired.

16 years ago[intel] Use the memory type mask containing the caching flags.
Eric Anholt [Wed, 2 Jan 2008 17:45:16 +0000 (09:45 -0800)]
[intel] Use the memory type mask containing the caching flags.

16 years agoSet correct flags mask when validating buffers.
Keith Packard [Mon, 17 Dec 2007 06:25:33 +0000 (22:25 -0800)]
Set correct flags mask when validating buffers.

The 'mask' value used in the validation operation specifies which of the
'flags' bits are being modified. Buffer validation wants to pass the memory
type and access mode (rwx) to the kernel so that the buffer will be placed
correctly, and so that the right kind of fence will be created (read vs
write). That means we actually want a constant mask for these operations,
and not something computed from the bits coming in. The constant we want is
DRM_BO_MASK_MEM | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_EXE.

16 years ago[965] Improve performance by including reloc target buffer pointers in keys.
Eric Anholt [Tue, 1 Jan 2008 10:16:29 +0000 (02:16 -0800)]
[965] Improve performance by including reloc target buffer pointers in keys.

Without this, the WM binding tables would all collide, for example.  Improves
openarena performance by around 2%.

16 years agoadditional stub functions
Brian [Wed, 2 Jan 2008 14:24:07 +0000 (07:24 -0700)]
additional stub functions

16 years agoadditional GL_COLOR_ATTACHMENTx_EXT cases (bug 13767)
Brian [Wed, 2 Jan 2008 14:18:46 +0000 (07:18 -0700)]
additional GL_COLOR_ATTACHMENTx_EXT cases (bug 13767)

16 years agoi915: Needn't adjust pixel centers. fix #12944
Xiang, Haihao [Wed, 2 Jan 2008 03:38:51 +0000 (11:38 +0800)]
i915: Needn't adjust pixel centers. fix #12944

16 years agoRevert "r300: fix bug with maniadrive rendering"
Dave Airlie [Tue, 1 Jan 2008 20:19:47 +0000 (06:19 +1000)]
Revert "r300: fix bug with maniadrive rendering"

this is correct, there is another issue with sw fallbacks

This reverts commit cc50edbca2fd3111f9987d4117fa6656599d79dc.

16 years agoremove unneeded conditional
Brian [Tue, 1 Jan 2008 17:20:21 +0000 (10:20 -0700)]
remove unneeded conditional

16 years agofix vbo display list memleak upon context destruction
Brian [Fri, 28 Dec 2007 21:51:52 +0000 (14:51 -0700)]
fix vbo display list memleak upon context destruction

16 years agofix glUniform4f typo
Brian [Tue, 25 Dec 2007 00:05:39 +0000 (17:05 -0700)]
fix glUniform4f typo

16 years agoadditional GL_COLOR_ATTACHMENTx_EXT cases (bug 13767)
Brian [Tue, 1 Jan 2008 17:10:04 +0000 (10:10 -0700)]
additional GL_COLOR_ATTACHMENTx_EXT cases (bug 13767)

16 years agoConvert to 0/1 when setting boolean uniforms
Bruce Merry [Fri, 21 Dec 2007 21:18:40 +0000 (23:18 +0200)]
Convert to 0/1 when setting boolean uniforms

Also add some extra tests to the shader_api regression tests

16 years agoMake use of count in _mesa_uniform_matrix
Bruce Merry [Fri, 21 Dec 2007 14:04:43 +0000 (16:04 +0200)]
Make use of count in _mesa_uniform_matrix

16 years agoAdd a test program to test for assorted bugs in shader_api.c
Bruce Merry [Fri, 21 Dec 2007 13:23:39 +0000 (15:23 +0200)]
Add a test program to test for assorted bugs in shader_api.c

16 years agoMore fixes to shader_api
Bruce Merry [Fri, 21 Dec 2007 13:20:17 +0000 (15:20 +0200)]
More fixes to shader_api

- return GL_INVALID_OPERATION instead of GL_INVALID_VALUE if location is bad
- correct the type-checking of uniforms from my previous commit
- accept location of -1 in _mesa_uniform_matrix

16 years agoFix several bugs relating to uniforms and attributes in GLSL API
Bruce Merry [Fri, 21 Dec 2007 12:41:45 +0000 (14:41 +0200)]
Fix several bugs relating to uniforms and attributes in GLSL API

- fix sizes for GL_FLOAT_MAT2x3 and GL_FLOAT_MAT4x3 in sizeof_glsl_type
- fix size returns in _mesa_get_active_attrib
- fix out-of-bounds array access to vec_types in _mesa_get_active_attrib
- fix queries of matrix uniforms in _mesa_get_uniformfv
- fix _mesa_get_uniformfv to only return one base, even from an array
- allow location == -1 in _mesa_uniform
- validate types in _mesa_uniform
- allow array overruns in _mesa_uniform

16 years agoadded 'get' info for framebuffer object tokens
Brian [Tue, 1 Jan 2008 16:26:15 +0000 (09:26 -0700)]
added 'get' info for framebuffer object tokens

16 years agoadd 'Get' info for MAX_3D_TEXTURE_SIZE (for bug 13811)
Brian [Tue, 1 Jan 2008 16:02:36 +0000 (09:02 -0700)]
add 'Get' info for MAX_3D_TEXTURE_SIZE (for bug 13811)

16 years agot_vp_build: fix temporary register allocation to minimise the allocations
Hans de Goede [Tue, 1 Jan 2008 15:32:01 +0000 (01:32 +1000)]
t_vp_build: fix temporary register allocation to minimise the allocations

16 years agorx00: fix off by one error in tempreg check
Hans de Goede [Tue, 1 Jan 2008 15:29:51 +0000 (01:29 +1000)]
rx00: fix off by one error in tempreg check

16 years agor300: fix bug with maniadrive rendering
Dave Airlie [Tue, 1 Jan 2008 15:19:23 +0000 (01:19 +1000)]
r300: fix bug with maniadrive rendering

I've no idea why I added this so I'll have to spend time tracking it down

16 years agoautoconf: Remind developers to bump version numbers in configure.ac
Dan Nicholson [Sun, 30 Dec 2007 16:48:10 +0000 (08:48 -0800)]
autoconf: Remind developers to bump version numbers in configure.ac

Just like configs/default, the version number in configure.ac needs to
be bumped for releases. Maybe later we can figure out how to scrape the
version from configs/default into configure.ac.

16 years agofix fd.o bug #13847
Zou Nan hai [Sat, 29 Dec 2007 07:28:51 +0000 (15:28 +0800)]
fix fd.o bug #13847

16 years agoBug #13839: Fix 3D texture offset miscalculation with pixels versus bytes.
Roland Scheidegger [Sat, 29 Dec 2007 02:03:39 +0000 (18:03 -0800)]
Bug #13839: Fix 3D texture offset miscalculation with pixels versus bytes.

16 years agoMerge branch 'autoconf2'
Dan Nicholson [Fri, 28 Dec 2007 23:14:32 +0000 (15:14 -0800)]
Merge branch 'autoconf2'

16 years agoi915: reset swrast state after calling swrast DrawPixels.
Xiang, Haihao [Thu, 27 Dec 2007 02:43:43 +0000 (10:43 +0800)]
i915: reset swrast state after calling swrast DrawPixels.
In order to optimize DrawPixels, the i915 texenv program isn't
applied to swrast DrawPixels in the i915 driver. This causes this
program isn't applied to any following swrast functions. Resetting
the swrast state fixes this issue.  Fix #13614

16 years agoMerge branch 'master' into autoconf2
Dan Nicholson [Wed, 26 Dec 2007 21:41:24 +0000 (15:41 -0600)]
Merge branch 'master' into autoconf2

16 years agoautoconf: Helper options for adding GCC 32/64 bit flags
Dan Nicholson [Wed, 26 Dec 2007 17:12:29 +0000 (11:12 -0600)]
autoconf: Helper options for adding GCC 32/64 bit flags

Two new configure options to add -m32 or -m64 to the CFLAGS and CXXFLAGS
when GCC is in use. By default, the user supplied options are
environment variables are respected, but these options are quick helps
for the common case of x86/x86_64 using GCC.

16 years agoi915: apply commit a0a5e8cfc04c14873441b50f7d594ef11806b9a8 from 965.
Xiang, Haihao [Tue, 25 Dec 2007 09:22:19 +0000 (17:22 +0800)]
i915: apply commit a0a5e8cfc04c14873441b50f7d594ef11806b9a8 from 965.
fix #11925

16 years agomesa: fix a bad cast in put_values_z24.
Xiang, Haihao [Tue, 25 Dec 2007 06:16:05 +0000 (14:16 +0800)]
mesa: fix a bad cast in put_values_z24.
The values passed to put_values_z24 are GLuint,
not GLubyte. fix #13543

16 years ago__driConfigOptions must be PUBLIC.
Adam Jackson [Tue, 25 Dec 2007 00:16:24 +0000 (19:16 -0500)]
__driConfigOptions must be PUBLIC.

16 years agoR300: RV410 SE chips have half the pipes of regular RV410
Alex Deucher [Mon, 24 Dec 2007 16:59:27 +0000 (11:59 -0500)]
R300: RV410 SE chips have half the pipes of regular RV410

This fixes 3D rendering on x700 SE chips.  Reported
by Kano.

16 years agoautoconf: Documentation for using the autoconf'd build
Dan Nicholson [Mon, 24 Dec 2007 00:38:18 +0000 (16:38 -0800)]
autoconf: Documentation for using the autoconf'd build

Most of the options available from configure are documented on the
autoconf.html. This page is reached as an alternative provided on the
install.html page. An FAQ about why there is no configure script has
been removed.

16 years agofix GL_LINE_LOOP with drivers using own render pipeline stage (#12410, #13527)
Roland Scheidegger [Sat, 22 Dec 2007 17:54:01 +0000 (18:54 +0100)]
fix GL_LINE_LOOP with drivers using own render pipeline stage (#12410, #13527)

primitive needs to include the begin/end flags (broken since vbo-0.2). Should
fix missing first/last line segment on gamma, i810, i915, mga, r200, radeon,
s3v, savage, unichrome (r300 already correct). Tested on r200, fixes #13527.

16 years agoSilence compiler warnings from XML error macros.
Kristian Høgsberg [Fri, 21 Dec 2007 20:31:00 +0000 (15:31 -0500)]
Silence compiler warnings from XML error macros.

16 years ago[965] Fix and enable separate stencil.
Eric Anholt [Fri, 21 Dec 2007 19:39:33 +0000 (11:39 -0800)]
[965] Fix and enable separate stencil.

Note that this does not enable GL_EXT_stencil_two_side, because Mesa's computed
_TestTwoSide ends up respecting only STENCIL_TEST_TWO_SIDE_EXT (defaults to
GL_FALSE), even if the application uses only GL 2.0 / ATI entrypoints.

16 years ago[intel] Move some pixel path support from drivers to shared.
Eric Anholt [Thu, 20 Dec 2007 19:44:40 +0000 (11:44 -0800)]
[intel] Move some pixel path support from drivers to shared.

16 years agointel: cast a pointer to unsigned long, avoid potential error.
Xiang, Haihao [Fri, 21 Dec 2007 09:03:55 +0000 (17:03 +0800)]
intel: cast a pointer to unsigned long, avoid potential error.

16 years ago[965] Enable EXT_framebuffer_object.
Eric Anholt [Thu, 20 Dec 2007 19:29:39 +0000 (11:29 -0800)]
[965] Enable EXT_framebuffer_object.

To do so, merge the remainnig necessary code from the buffers, blit, span, and
screen code to shared, and replace it with those.

16 years ago[965] Actually enable SGIS_generate_mipmap.
Eric Anholt [Thu, 20 Dec 2007 19:28:10 +0000 (11:28 -0800)]
[965] Actually enable SGIS_generate_mipmap.

16 years ago[intel] Fix and reenable (software) SGIS_generate_mipmap
Eric Anholt [Wed, 19 Dec 2007 22:26:14 +0000 (14:26 -0800)]
[intel] Fix and reenable (software) SGIS_generate_mipmap

The core problem was that _mesa_generate_mipmap was not respecting RowStride
of the source image.  Additionally, the intel private data associated with the
images (level and face) was not being initialized for the
_mesa_generate_mipmap-generated images.

16 years ago[intel] Allow driver hooks to be NULL in intel_buffers.c and just update flags.
Eric Anholt [Thu, 20 Dec 2007 16:24:45 +0000 (08:24 -0800)]
[intel] Allow driver hooks to be NULL in intel_buffers.c and just update flags.

The 965 driver relies on flag checking instead of these hooks, and will be
using this code soon.

16 years ago[i915] Move meta_draw_quad into the vtbl with other meta operations.
Eric Anholt [Thu, 20 Dec 2007 16:19:42 +0000 (08:19 -0800)]
[i915] Move meta_draw_quad into the vtbl with other meta operations.

16 years agoreturn correct size from glGetActiveUniform (bug 13751)
Brian [Thu, 20 Dec 2007 16:05:37 +0000 (09:05 -0700)]
return correct size from glGetActiveUniform (bug 13751)

16 years agoi915: avoid dead lock in intel_meta_draw_poly. fix #13696
Xiang, Haihao [Thu, 20 Dec 2007 08:49:25 +0000 (16:49 +0800)]
i915: avoid dead lock in intel_meta_draw_poly. fix #13696

16 years agoDon't try to build nonexistent i915tex driver on linux-x86-64
Dan Nicholson [Fri, 7 Dec 2007 19:26:41 +0000 (11:26 -0800)]
Don't try to build nonexistent i915tex driver on linux-x86-64

16 years agodocs: Fix links in contribute section
Dan Nicholson [Fri, 14 Dec 2007 17:59:16 +0000 (09:59 -0800)]
docs: Fix links in contribute section

The contribute section had a broken hyperlink, masking the mailing list
text.

16 years ago[915] Set cliprects in the drawbuffer software fallback case as well.
Eric Anholt [Wed, 19 Dec 2007 02:54:59 +0000 (18:54 -0800)]
[915] Set cliprects in the drawbuffer software fallback case as well.

Otherwise, we may violate cliprect asssertions on clearing the buffers, which
isn't affected by the fallback.

16 years agoi965: allocate GRF registers before building subroutines,
Xiang, Haihao [Wed, 19 Dec 2007 02:22:28 +0000 (10:22 +0800)]
i965: allocate GRF registers before building subroutines,
it ensures there are sufficient registers for all subroutines.

16 years agoi965: restore the flag after building the subroutine of the
Xiang, Haihao [Wed, 19 Dec 2007 01:59:37 +0000 (09:59 +0800)]
i965: restore the flag after building the subroutine of the
GS thread. fix #13240

16 years agoadded glSecondaryColor3fv_func
Brian [Tue, 18 Dec 2007 23:30:42 +0000 (16:30 -0700)]
added glSecondaryColor3fv_func

16 years agoadded two-side test
Brian [Tue, 18 Dec 2007 23:25:48 +0000 (16:25 -0700)]
added two-side test

16 years agoTest GL_VERTEX_PROGRAM_TWO_SIDE and frag shader gl_FrontFacing features
Brian [Tue, 18 Dec 2007 23:21:49 +0000 (16:21 -0700)]
Test GL_VERTEX_PROGRAM_TWO_SIDE and frag shader gl_FrontFacing features

16 years agoadded twoside demo
Brian [Tue, 18 Dec 2007 23:20:10 +0000 (16:20 -0700)]
added twoside demo

16 years agosimplify update two-side lighting test (follow-on to previous front/back-face changes)
Brian [Tue, 18 Dec 2007 23:12:23 +0000 (16:12 -0700)]
simplify update two-side lighting test (follow-on to previous front/back-face changes)

16 years agofix NEED_SECONDARY_COLOR for vert/frag progs
Brian [Tue, 18 Dec 2007 23:10:11 +0000 (16:10 -0700)]
fix NEED_SECONDARY_COLOR for vert/frag progs

16 years agoFix a quadstrip front/back-face inconsistancy.
Brian [Sat, 15 Dec 2007 00:34:01 +0000 (17:34 -0700)]
Fix a quadstrip front/back-face inconsistancy.

Only set front material colors to make it noticable if front/back-face
determination is incorrect anywhere.

16 years agoFix mismatched map/unmap of buffers in swrast read/drawpixels error paths.
Eric Anholt [Tue, 18 Dec 2007 22:51:42 +0000 (14:51 -0800)]
Fix mismatched map/unmap of buffers in swrast read/drawpixels error paths.

16 years ago[915] Free dri_bufmgr after mesa context data.
Eric Anholt [Tue, 18 Dec 2007 22:17:27 +0000 (14:17 -0800)]
[915] Free dri_bufmgr after mesa context data.

Fixes a crash when buffer objects are left around until context destroy.

16 years ago[915] Make polygon stipple use pre-unpacked pixel data.
Eric Anholt [Tue, 18 Dec 2007 22:14:44 +0000 (14:14 -0800)]
[915] Make polygon stipple use pre-unpacked pixel data.

This fixes a crash when stippling using data from a PBO.

16 years ago[915] Fix clear color when clearing with triangles.
Eric Anholt [Tue, 18 Dec 2007 18:41:44 +0000 (10:41 -0800)]
[915] Fix clear color when clearing with triangles.

The diffuse color format is always ARGB32, regardless of the destination
surface format.

16 years ago[INTEL] Fix 965 to use new centralized mipmap pitch function
Keith Packard [Tue, 18 Dec 2007 18:18:33 +0000 (10:18 -0800)]
[INTEL] Fix 965 to use new centralized mipmap pitch function

16 years ago[Intel] Centralize mipmap pitch computations.
Keith Packard [Tue, 18 Dec 2007 06:43:48 +0000 (22:43 -0800)]
[Intel] Centralize mipmap pitch computations.

mipmap pitches must account for the device alignment requirements, which
used to be fairly simple; just align to a 4-byte boundary. However, to allow
textures to be drawn to under TTM, they now need to be aligned to a 64-byte
boundary. Placing all of the alignment constraints in a single function
allows this new constraint to be applied uniformly.

There was some pitch constraining code in intel_miptree_create, but that was
modifying the pitch long after the miptree had been layed out, so it only
served to wreck the mipmap and cause rendering errors.

16 years ago[i915] Remove redundant set_draw_region code (like the comment says).
Eric Anholt [Tue, 18 Dec 2007 01:01:07 +0000 (17:01 -0800)]
[i915] Remove redundant set_draw_region code (like the comment says).

16 years ago[intel] Improve INTEL_DEBUG=blit description of clearing.
Eric Anholt [Tue, 18 Dec 2007 00:56:59 +0000 (16:56 -0800)]
[intel] Improve INTEL_DEBUG=blit description of clearing.

16 years ago[intel] Fix copy'n'pasteo in decoding of the blit clear packet.
Eric Anholt [Tue, 27 Nov 2007 18:57:28 +0000 (10:57 -0800)]
[intel] Fix copy'n'pasteo in decoding of the blit clear packet.

16 years ago[965] Add decode of 3DSTATE_DRAWING_RECTANGLE.
Eric Anholt [Tue, 27 Nov 2007 18:16:58 +0000 (10:16 -0800)]
[965] Add decode of 3DSTATE_DRAWING_RECTANGLE.

16 years ago[965] Allow draw or depth regions to be NULL.
Eric Anholt [Tue, 18 Dec 2007 00:50:09 +0000 (16:50 -0800)]
[965] Allow draw or depth regions to be NULL.

With FBOs, we end up wanting to do 3D metaops against one or the other without
having to find the other one to fill in if we're not going to draw to it.

16 years ago[965] Simplify scissor handling by using DrawBuffer values.
Eric Anholt [Mon, 17 Dec 2007 22:51:24 +0000 (14:51 -0800)]
[965] Simplify scissor handling by using DrawBuffer values.

16 years ago[965] fix bad conflict resolution in debug code.
Eric Anholt [Mon, 17 Dec 2007 22:42:21 +0000 (14:42 -0800)]
[965] fix bad conflict resolution in debug code.

16 years ago[965] Replace our own depth constants in intel context with GL context ones.
Eric Anholt [Mon, 17 Dec 2007 22:28:54 +0000 (14:28 -0800)]
[965] Replace our own depth constants in intel context with GL context ones.

16 years ago[965] Fix software fallbacks with region-backed textures.
Eric Anholt [Mon, 17 Dec 2007 21:47:52 +0000 (13:47 -0800)]
[965] Fix software fallbacks with region-backed textures.

16 years ago[intel] Cleanup of */intel_blit.c to bring the two closer.
Eric Anholt [Mon, 17 Dec 2007 21:19:33 +0000 (13:19 -0800)]
[intel] Cleanup of */intel_blit.c to bring the two closer.

16 years ago[965] Output the buffer type in INTEL_DEBUG=bat surface state decode.
Eric Anholt [Mon, 17 Dec 2007 19:11:18 +0000 (11:11 -0800)]
[965] Output the buffer type in INTEL_DEBUG=bat surface state decode.

16 years agoi915: Fix issues with glDrawBuffer(GL_NONE).
Michel Dänzer [Mon, 17 Dec 2007 11:20:56 +0000 (12:20 +0100)]
i915: Fix issues with glDrawBuffer(GL_NONE).

Don't dereference NULL renderbuffer pointer, and make sure the software
fallback sticks.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=13694 .

16 years agoi965: check NULL pointer
Xiang, Haihao [Mon, 17 Dec 2007 06:42:42 +0000 (14:42 +0800)]
i965: check NULL pointer

16 years ago[i915] Fix missing symbol from 965 changes.
Eric Anholt [Sun, 16 Dec 2007 21:00:23 +0000 (13:00 -0800)]
[i915] Fix missing symbol from 965 changes.

16 years ago[965] Fully initialize the texture surface key data (padding around GLboolean)
Eric Anholt [Sun, 16 Dec 2007 20:09:22 +0000 (12:09 -0800)]
[965] Fully initialize the texture surface key data (padding around GLboolean)

16 years ago[965] Enable ARB_pixel_buffer_object, and disable broken imaging extension.
Eric Anholt [Sun, 16 Dec 2007 19:46:10 +0000 (11:46 -0800)]
[965] Enable ARB_pixel_buffer_object, and disable broken imaging extension.

While I haven't tested the imaging extension, this matches what 915 does.

16 years ago[965] Move to using shared texture management code.
Eric Anholt [Sun, 16 Dec 2007 00:12:17 +0000 (16:12 -0800)]
[965] Move to using shared texture management code.

This removes the delayed texture upload optimization from 965, in exchange for
bringing us closer to PBO support.  It also disables SGIS_generate_mipmap,
which didn't seem to be working before anyway, according to the lodbias demo.

16 years ago[intel] Whitespace and comment changes to bring intel_mipmap_tree.c closer.
Eric Anholt [Sat, 15 Dec 2007 21:09:58 +0000 (13:09 -0800)]
[intel] Whitespace and comment changes to bring intel_mipmap_tree.c closer.

16 years ago[intel] Merge intel_buffer_objects to shared.
Eric Anholt [Sat, 15 Dec 2007 20:47:23 +0000 (12:47 -0800)]
[intel] Merge intel_buffer_objects to shared.

965 gains fixed TTM typing of the buffer object buffers and unused PBO
functions, and 915 gains buffer size == 0 fixes from 965.

16 years ago[965] Use shared intel_regions.c.
Eric Anholt [Sat, 15 Dec 2007 00:00:48 +0000 (16:00 -0800)]
[965] Use shared intel_regions.c.

This adds (so far) unused PBO functions, and holding the lock while writing
to regions (which may be shared static screen regions).

16 years ago[intel] Fix uninitialized data in screen-region buffer objects.
Eric Anholt [Fri, 14 Dec 2007 22:56:01 +0000 (14:56 -0800)]
[intel] Fix uninitialized data in screen-region buffer objects.

16 years ago[intel] Remove excessive validation debugging.
Eric Anholt [Fri, 14 Dec 2007 22:52:15 +0000 (14:52 -0800)]
[intel] Remove excessive validation debugging.

16 years ago[intel] Initialize debug flag for dri_bufmgrs
Eric Anholt [Fri, 14 Dec 2007 22:51:55 +0000 (14:51 -0800)]
[intel] Initialize debug flag for dri_bufmgrs

16 years ago[intel] Remove useless intel_region_idle.
Eric Anholt [Fri, 14 Dec 2007 21:28:39 +0000 (13:28 -0800)]
[intel] Remove useless intel_region_idle.

The idling it was trying to ensure was covered by the
intel_miptree_image_map()->intel_region_map() that immediately followed it.

16 years ago[intel] warnings cleanup
Eric Anholt [Fri, 14 Dec 2007 20:52:17 +0000 (12:52 -0800)]
[intel] warnings cleanup

16 years agofix polygon cull regression
Brian [Fri, 14 Dec 2007 21:34:15 +0000 (14:34 -0700)]
fix polygon cull regression

16 years agoBuild rain demo (Gonzo <andreas.wendleder@gmail.com>)
Brian [Wed, 12 Dec 2007 15:05:12 +0000 (08:05 -0700)]
Build rain demo (Gonzo <andreas.wendleder@gmail.com>)

16 years ago[intel] Remove the relocation buffer lists and just cache one per buffer.
Eric Anholt [Wed, 12 Dec 2007 22:38:38 +0000 (14:38 -0800)]
[intel] Remove the relocation buffer lists and just cache one per buffer.

Each buffer object now has a relocation buffer pointer, which contains the
relocations for the buffer if there are any.  At the point where we have to
create a new type of relocation entry, we can change the code over to allowing
multiple relocation lists, but trying to anticipate what that'll look like
now just increases complexity.

This is a 30% performance improvement on 965.

16 years ago[965] Replace the state cache suballocator with direct dri_bufmgr use.
Eric Anholt [Fri, 14 Dec 2007 19:02:48 +0000 (11:02 -0800)]
[965] Replace the state cache suballocator with direct dri_bufmgr use.

The user-space suballocator that was used avoided relocation computations by
using the general and surface state base registers and allocating those types
of buffers out of pools built on top of single buffer objects.  It also
avoided calls into the buffer manager for these small state allocations, since
only one buffer object was being used.

However, the buffer allocation cost appears to be low, and with relocation
caching, computing relocations for buffers is essentially free.  Additionally,
implementing the suballocator required a don't-fence-subdata flag to disable
waiting on buffer maps so that writing new data didn't block on rendering using
old data, and careful handling when mapping to update old data (which we need
to do for unavoidable relocations with FBOs).  More importantly, when the
suballocator filled, it had no replacement algorithm and just threw out all
of the contents and forced them to be recomputed, which is a significant cost.

This is the first step, which just changes the buffer type, but doesn't yet
improve the hash table to not result in full recompute on overflow.  Because
the buffers are all allocated out of the general buffer allocator, we can
no longer use the general/surface state bases to avoid relocations, and they
are set to 0 instead.