mesa.git
12 years agosvga: Flush the hwtnl primitives before writing to buffers.
José Fonseca [Thu, 27 Oct 2011 18:09:25 +0000 (19:09 +0100)]
svga: Flush the hwtnl primitives before writing to buffers.

svga keeps a small queue of similar primitive draws in order to coalesce
them into a single draw primitive command.

But the buffers referred in primitives not yet emitted were being ignored
in the considerations to flush or not the context.

This fixes piglit vbo-map-remap, vbo-subdata-sync, vbo-subdata-zero, and
Seeker.

Based on investigation and patch from Brian Paul.

Reviewed-By: Brian Paul <brianp@vmware.com>
12 years agost/xa: Plug a pipe context leak.
Thomas Hellstrom [Mon, 31 Oct 2011 13:01:39 +0000 (14:01 +0100)]
st/xa: Plug a pipe context leak.

Forgot to destroy the pipe context on xa context destroy.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
12 years agopipebuffer/debug: Fix a recursive mutex lock
Thomas Hellstrom [Mon, 31 Oct 2011 11:14:05 +0000 (12:14 +0100)]
pipebuffer/debug: Fix a recursive mutex lock

pb_debug_manager_dump was trying to take a lock already
held by all callers.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: José Fonseca <jfonseca@vmware.com>
12 years agoradeon/r200: drop remains of r300/r600 support along with old drm 1.x kernel
Dave Airlie [Sat, 29 Oct 2011 07:39:54 +0000 (08:39 +0100)]
radeon/r200: drop remains of r300/r600 support along with old drm 1.x kernel

This drops all the old drmSupports* checks since KMS does them all, and it
also drop R300_CLASS and R600_CLASS.

Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agollvmpipe: fix a crash in non-SSE path
Chia-I Wu [Fri, 23 Sep 2011 03:40:49 +0000 (11:40 +0800)]
llvmpipe: fix a crash in non-SSE path

It is a typo went unnoticed.

12 years agointel: enable GL_OES_draw_texture
Chia-I Wu [Sun, 21 Aug 2011 13:47:39 +0000 (21:47 +0800)]
intel: enable GL_OES_draw_texture

12 years agometa: add _mesa_meta_DrawTex()
Chia-I Wu [Sun, 21 Aug 2011 13:08:25 +0000 (21:08 +0800)]
meta: add _mesa_meta_DrawTex()

It is set to dd->DrawTex.

Reviewed-by: Brian Paul <brianp@vmware.com>
[olv: set dd->DrawTex in _mesa_init_driver_functions]

12 years agor600g: reset the render condition to NULL when rendering unconditionally
Marek Olšák [Sat, 29 Oct 2011 14:19:45 +0000 (16:19 +0200)]
r600g: reset the render condition to NULL when rendering unconditionally

12 years agor600g: shortcut ctx->screen->ws to ctx->ws
Marek Olšák [Sat, 29 Oct 2011 15:56:12 +0000 (17:56 +0200)]
r600g: shortcut ctx->screen->ws to ctx->ws

12 years agoi965: Remove the prepare() hook from state atoms.
Eric Anholt [Sat, 22 Oct 2011 19:16:44 +0000 (12:16 -0700)]
i965: Remove the prepare() hook from state atoms.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Remove the memcpy()ed atoms array now that everything is emit()-based.
Eric Anholt [Sat, 22 Oct 2011 19:16:05 +0000 (12:16 -0700)]
i965: Remove the memcpy()ed atoms array now that everything is emit()-based.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Merge brw_validate_state() and brw_upload_state() together.
Eric Anholt [Sat, 22 Oct 2011 19:11:27 +0000 (12:11 -0700)]
i965: Merge brw_validate_state() and brw_upload_state() together.

They were called back-to-back at this point.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Remove state upload code for calling prepare() now that there are none.
Eric Anholt [Sat, 22 Oct 2011 19:02:37 +0000 (12:02 -0700)]
i965: Remove state upload code for calling prepare() now that there are none.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Remove some old texturing debug code.
Eric Anholt [Sat, 22 Oct 2011 19:03:36 +0000 (12:03 -0700)]
i965: Remove some old texturing debug code.

It caught one possible bug I recall in my time working on the driver,
and we haven't been setting it for non-fixed-function since the new FS
backend came along.  The bug it caught was likely a confusion about
sampler mappings, which we have tests for these days.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Finally, move the global fallbacks check to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:57:44 +0000 (11:57 -0700)]
i965: Finally, move the global fallbacks check to emit() time.

This was the last prepare() function, and it's the first state atom,
so it must be ready to move.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move index buffer upload to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:54:32 +0000 (11:54 -0700)]
i965: Move index buffer upload to emit() time.

It's consumed by the brw_emit_index_buffer() code at emit() time.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Do a hack job of merging VB prepare()/emit() together.
Eric Anholt [Sat, 22 Oct 2011 18:35:56 +0000 (11:35 -0700)]
i965: Do a hack job of merging VB prepare()/emit() together.

I don't really want to touch this impenetrable code in this series, so
just call the one function from the other, since no other atom cares
about them.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move the WM input sizes calculation to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:34:49 +0000 (11:34 -0700)]
i965: Move the WM input sizes calculation to emit() time.

It's used for program compile.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move program compile to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:33:54 +0000 (11:33 -0700)]
i965: Move program compile to emit() time.

Only 4 other prepare() functions are left, which don't rely on this.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Move CURBE offset calculation to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:32:05 +0000 (11:32 -0700)]
i965/gen4: Move CURBE offset calculation to emit() time.

This is consumed by the unit state.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Fold push constant prepare()/emit() together.
Eric Anholt [Sat, 22 Oct 2011 18:21:51 +0000 (11:21 -0700)]
i965/gen4: Fold push constant prepare()/emit() together.

While other units need to know about our constant buffer offsets,
nothing else cared about which particular BO other than the emit() half.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen6: Move viewport state setup to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:18:56 +0000 (11:18 -0700)]
i965/gen6: Move viewport state setup to emit() time.

Only the emit() for the pointers into the batch later in this file
cares.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Move URB fence recalculate to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:25:15 +0000 (11:25 -0700)]
i965/gen4: Move URB fence recalculate to emit() time.

This is used by the unit state, which is at emit() time.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Fold the gen6/7 URB state prepare()/emit() together.
Eric Anholt [Sat, 22 Oct 2011 18:17:27 +0000 (11:17 -0700)]
i965: Fold the gen6/7 URB state prepare()/emit() together.

No other unit cares about the prepare state, unlike gen4-5.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move VS pull constant upload to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:14:02 +0000 (11:14 -0700)]
i965: Move VS pull constant upload to emit() time.

Only needed by the emit() for VS surfaces.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Fold prepare() and emit() of VS surface state setup together.
Eric Anholt [Sat, 22 Oct 2011 18:12:43 +0000 (11:12 -0700)]
i965: Fold prepare() and emit() of VS surface state setup together.

This rearranges the code a bit, and makes the upload of the binding
table take only as many surfaces as there are in use.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move WM pull constant setup to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:07:40 +0000 (11:07 -0700)]
i965: Move WM pull constant setup to emit() time.

It's needed by the WM surface state setup, which is now emit().

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen7: Fold WM surface state prepare()/emit() together.
Eric Anholt [Sat, 22 Oct 2011 18:07:00 +0000 (11:07 -0700)]
i965/gen7: Fold WM surface state prepare()/emit() together.

These produce BRW_NEW_SURFACES (used by binding table emit()) and
BRW_NEW_NR_WM_SURFACES (used by WM unit emit()).  Fixes a bug where
with no texturing and no color buffer, we wouldn't consider the null
renderbuffer in nr_surfaces.  This was harmless because nr_surfaces is
only used for the prefetch info in the unit state.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Fold WM surface state prepare()/emit() together.
Eric Anholt [Sat, 22 Oct 2011 18:01:40 +0000 (11:01 -0700)]
i965/gen4: Fold WM surface state prepare()/emit() together.

These produce BRW_NEW_SURFACES (used by binding table emit()) and
BRW_NEW_NR_WM_SURFACES (used by WM unit emit()).  Fixes a bug where
with no texturing and no color buffer, we wouldn't consider the null
renderbuffer in nr_surfaces.  This was harmless because nr_surfaces is
only used for the prefetch info in the unit state.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move sampler state to emit() time.
Eric Anholt [Sat, 22 Oct 2011 18:23:59 +0000 (11:23 -0700)]
i965: Move sampler state to emit() time.

This is consumed by the WM unit, which is already at emit().

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Move unit state setup to emit() time.
Eric Anholt [Sat, 22 Oct 2011 17:51:34 +0000 (10:51 -0700)]
i965/gen4: Move unit state setup to emit() time.

It is only needed in time for brw_psp_urb_cbs(), which is also an emit().

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen7: Fold prepare() and emit() of SF CLIP/VP state together.
Eric Anholt [Sat, 22 Oct 2011 17:53:18 +0000 (10:53 -0700)]
i965/gen7: Fold prepare() and emit() of SF CLIP/VP state together.

The prepare() only made state for its emit(), not anybody else.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen4: Move CC VP to emit() time, since it's only needed by CC's emit().
Eric Anholt [Sat, 22 Oct 2011 17:48:26 +0000 (10:48 -0700)]
i965/gen4: Move CC VP to emit() time, since it's only needed by CC's emit().

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Move push constants setup to emit() time.
Eric Anholt [Sat, 22 Oct 2011 17:43:05 +0000 (10:43 -0700)]
i965: Move push constants setup to emit() time.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965/gen6: Move setup of CC state batches to emit time.
Eric Anholt [Sat, 22 Oct 2011 17:37:14 +0000 (10:37 -0700)]
i965/gen6: Move setup of CC state batches to emit time.

This is part of a series trying to eliminate the separate prepare()
hook in state upload.  The prepare() hook existed to support the
check_aperture in between calculating state updates and setting up the
batch, but there should be no reason for that any more.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Check Fallback again after upload.
Eric Anholt [Sat, 22 Oct 2011 18:56:01 +0000 (11:56 -0700)]
i965: Check Fallback again after upload.

As we move state to emit() time from prepare() time, a couple of the
places that flag fallbacks will move here.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Remove the validated BO list, now that it's unused.
Eric Anholt [Sat, 22 Oct 2011 16:40:38 +0000 (09:40 -0700)]
i965: Remove the validated BO list, now that it's unused.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Use the batch save/reset code to avoid needing the BO validate step.
Eric Anholt [Sat, 22 Oct 2011 16:37:57 +0000 (09:37 -0700)]
i965: Use the batch save/reset code to avoid needing the BO validate step.

We were doing the BO validate step in prepare() (brw_validate_state())
hooks of atoms so that we could check_aperture before emitting the
relocation trees during brw_upload_state() that would actually make
the batchbuffer reference too much memory to be executed.  Now that
all relocations occur in the batchbuffer, we can instead
check_aperture after emitting our state into the batchbuffer, and
easily roll back, flush, and retry if we happened to go over the
limits.

This will let us remove the whole prepare() vs emit() split in our
state atoms, which is a source of tricky dependencies and duplicated
code.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agointel: Return error value from intel_batchbuffer_flush().
Eric Anholt [Sat, 22 Oct 2011 17:33:24 +0000 (10:33 -0700)]
intel: Return error value from intel_batchbuffer_flush().

This will let the caller do something sensible on error, if it cares.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi965: Add a note about an unsafe-looking state check.
Eric Anholt [Sat, 22 Oct 2011 02:08:55 +0000 (19:08 -0700)]
i965: Add a note about an unsafe-looking state check.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agointel: Add an interface for saving/restoring the batchbuffer state.
Eric Anholt [Sat, 22 Oct 2011 02:01:17 +0000 (19:01 -0700)]
intel: Add an interface for saving/restoring the batchbuffer state.

This will be used to avoid the prepare() step in the i965 driver's
state setup.  Instead, we can just speculatively emit the primitive
into the batchbuffer, then check if the batch is too big, rollback and
flush, and replay the primitive.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoi915: Move the always_flush_cache code to triangle emit.
Eric Anholt [Fri, 21 Oct 2011 23:32:03 +0000 (16:32 -0700)]
i915: Move the always_flush_cache code to triangle emit.

This could have broken always_flush_cache on i965, since
reserved_space doesn't reflect the size of the workaround flushes, and
we might run out of space.  This should make always_flush_cache more
useful on pre-i965, anyway (since the point is to flush around each
draw call, even within a batchbuffer).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Paul Berry <stereotype441@gmail.com>
12 years agoglsl: Fix compilation of glsl_lexer.ll with MSVC.
Morgan Armand [Sat, 29 Oct 2011 17:37:58 +0000 (10:37 -0700)]
glsl: Fix compilation of glsl_lexer.ll with MSVC.

strtoull is not supported on msvc (as there is no C99 support).

12 years agor600g: Replace needless flush in texture upload.
Mathias Fröhlich [Tue, 26 Jul 2011 05:05:10 +0000 (07:05 +0200)]
r600g: Replace needless flush in texture upload.

Replace pipe->flush() with pipe->texture_barrier() in
the texture upload path for the staging texture.
This should be enough to get data out of the gpu
caches ready to be read for texture fetch.

12 years agor600g: remove one pointless flush
Marek Olšák [Fri, 28 Oct 2011 17:24:16 +0000 (19:24 +0200)]
r600g: remove one pointless flush

It's not useful for anything.
The rest of the patch is just a cleanup resulting
from some of the variables being no longer used.

There are no piglit regressions.

12 years agoradeon/r200: forgot one somehow
Dave Airlie [Sat, 29 Oct 2011 07:15:00 +0000 (08:15 +0100)]
radeon/r200: forgot one somehow

drops last usage.

12 years agoradeon/r200: drop remains of non-libdrm_radeon build
Dave Airlie [Sat, 29 Oct 2011 07:11:37 +0000 (08:11 +0100)]
radeon/r200: drop remains of non-libdrm_radeon build

These wrappers and associated symlinks were from the non-libdrm_radeon build.

Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agomesa/st: get interpolation mode from the fragment shader.
Dave Airlie [Fri, 28 Oct 2011 16:18:15 +0000 (17:18 +0100)]
mesa/st: get interpolation mode from the fragment shader.

With the recent changes to interpolation stuff, we can now get the value
direct from the program instead of just being fail.

fixes some of the glsl-1.30 interpolation tests with softpipe

Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agoglx: Don't enable INTEL_swap_event unconditionally
Adam Jackson [Sat, 29 Oct 2011 00:38:32 +0000 (20:38 -0400)]
glx: Don't enable INTEL_swap_event unconditionally

DRI2 supports this now - and already enables it explicitly - but drisw
does not and should not.  Otherwise toolkits like clutter will only ever
SwapBuffers once and wait forever for an event that's not coming.

Signed-off-by: Adam Jackson <ajax@redhat.com>
12 years agoimplement WGL_ARB_create_context
Morgan Armand [Fri, 28 Oct 2011 13:42:29 +0000 (06:42 -0700)]
implement WGL_ARB_create_context

Signed-off-by: José Fonseca <jfonseca@vmware.com>
12 years agor600g: get backend mask after the context is fully set up
Marek Olšák [Fri, 28 Oct 2011 20:31:34 +0000 (22:31 +0200)]
r600g: get backend mask after the context is fully set up

12 years agoir_to_mesa: Let check_resources halt compilation
Ian Romanick [Tue, 11 Oct 2011 23:27:41 +0000 (16:27 -0700)]
ir_to_mesa: Let check_resources halt compilation

Previously check_resources could fail, but we'd still try to optimize
the shader, do device-specific code generation, etc.  In some cases,
this could explode (especially in the device-specific code
generation).  I haven't found that I could trigger this with the
current code.  When too many samplers were used with the new uniform
handling code, I observed several crashes deep down in the driver.

NOTE: This is candidate for the 7.11 branch.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=41609
Cc: Eric Anholt <eric@anholt.net>
Reviewed-and-tested-by: Kenneth Graunke <kenneth@whitecape.org>
12 years agoi965: Use glsl_type::column_type instead of open-coding it
Ian Romanick [Sat, 22 Oct 2011 01:22:42 +0000 (18:22 -0700)]
i965: Use glsl_type::column_type instead of open-coding it

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
12 years agoglsl: Clean-up spurious error message on bad structure definitions
Ian Romanick [Wed, 26 Oct 2011 00:49:07 +0000 (17:49 -0700)]
glsl: Clean-up spurious error message on bad structure definitions

Previously a shader like

int X;
struct X { int i; };

void main() { gl_Position = vec4(0.0); }

would generate two error message:

0:2(19): error: struct `X' previously defined
0:2(20): error: incomplete declaration

The first one is the real error, and the second is spurious.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
12 years agoglsl: Generate an error for array-of-array declarations
Ian Romanick [Mon, 24 Oct 2011 18:45:50 +0000 (11:45 -0700)]
glsl: Generate an error for array-of-array declarations

Other parts of the code already caught things like 'float x[4][2]'.
However, nothing caught 'float [4] x[2]'.

Fixes piglit test array-multidimensional-new-syntax.vert.

NOTE: This is candidate for the 7.11 branch.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
12 years agor300c/compiler: remove the compiler too
Marek Olšák [Fri, 28 Oct 2011 19:27:22 +0000 (21:27 +0200)]
r300c/compiler: remove the compiler too

Gallium has a fork of this.

12 years agoi965/fs: Use the actual hardware g0 register for texel offset setup.
Kenneth Graunke [Thu, 27 Oct 2011 05:41:07 +0000 (22:41 -0700)]
i965/fs: Use the actual hardware g0 register for texel offset setup.

The idea here is to set up the message header with the Sampler State
pointer which the hardware provides as part of the PS Thread Payload in
register g0.

Unfortunately, the existing code

   fs_reg(GRF, 0, BRW_REGISTER_TYPE_UD))

actually references "virtual GRF 0" rather than the hardware g0.  This
is just some arbitrary GRF temporary which will get register allocated.

So, we ended up setting up the header with garbage.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agou_format: Fix -NaN handling for packing of 10F_11F_11F_REV to match GL specs.
Eric Anholt [Wed, 26 Oct 2011 23:23:50 +0000 (16:23 -0700)]
u_format: Fix -NaN handling for packing of 10F_11F_11F_REV to match GL specs.

Fixes the remainder of piglit GL_EXT_packed_float/pack.c

Reviewed-by: Marek Ol ák <maraeo@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agou_format: Fix clamping of overflow in 10F_11F_11F_REV to match GL specs.
Eric Anholt [Wed, 26 Oct 2011 23:11:27 +0000 (16:11 -0700)]
u_format: Fix clamping of overflow in 10F_11F_11F_REV to match GL specs.

Fixes the 1000000.0 overflow cases of piglit
GL_EXT_packed_float/pack.c

Reviewed-by: Marek Ol ák <maraeo@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agomesa: Don't do [0, 1] clamping on glGetTexImage() of packed float formats.
Eric Anholt [Wed, 26 Oct 2011 23:09:16 +0000 (16:09 -0700)]
mesa: Don't do [0, 1] clamping on glGetTexImage() of packed float formats.

From the GL_EXT_packed_float spec:

    For an RGBA color, if <type> is not one of FLOAT,
    UNSIGNED_INT_5_9_9_9_REV_EXT, or UNSIGNED_INT_10F_11F_11F_REV_EXT,
    or if the CLAMP_READ_COLOR_ARB is TRUE, or CLAMP_READ_COLOR_ARB
    is FIXED_ONLY_ARB and the selected color (or texture) buffer is
    a fixed-point buffer, each component is first clamped to [0,1].
    Then the appropriate conversion formula from table 4.7 is applied
    the component."

(but we previously resolved that the CLAMP_READ_COLOR bit is not
relevant to glGetTexImage())

This fixes most of the cases in piglit GL_EXT_packed_float/pack.

Reviewed-by: Marek Ol ák <maraeo@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agou_format: Fix bitshifting for unpacking from 10F.
Eric Anholt [Tue, 25 Oct 2011 23:53:00 +0000 (16:53 -0700)]
u_format: Fix bitshifting for unpacking from 10F.

This code was copy and pasted from the 11F unpacking, but not updated
for actually being 10 bits instead of 11.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=41206
Reviewed-by: Marek Ol ák <maraeo@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agou_format: Fix bit definition of UF10_MANTISSA_BITS.
Eric Anholt [Tue, 25 Oct 2011 23:50:44 +0000 (16:50 -0700)]
u_format: Fix bit definition of UF10_MANTISSA_BITS.

This is only used in the code for packing to INF, and resulted in an
extra bit set that was set anyway, so it was harmless except for the
confusion caused.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agoglsl: Implement GLSL 1.30's literal integer range restrictions.
Eric Anholt [Mon, 3 Oct 2011 23:59:01 +0000 (16:59 -0700)]
glsl: Implement GLSL 1.30's literal integer range restrictions.

From page 22 (28 of PDF) of GLSL 1.30 spec:
    It is an error to provide a literal integer whose magnitude is too
    large to store in a variable of matching signed or unsigned type.

    Unsigned integers have exactly 32 bits of precision.  Signed integers
    use 32 bits, including a sign bit, in two's complement form.

Fixes piglit int-literal-too-large-0[123].frag.

v2: Take care with INT_MIN, use stroull, and make it a function.

Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agoradeon: Remove the non-libdrm kernel memory manager support.
Eric Anholt [Fri, 14 Oct 2011 21:50:50 +0000 (14:50 -0700)]
radeon: Remove the non-libdrm kernel memory manager support.

We should have never been building this at this point.

12 years agoradeon: Insist on libdrm being present to build.
Eric Anholt [Fri, 14 Oct 2011 21:34:19 +0000 (14:34 -0700)]
radeon: Insist on libdrm being present to build.

There's no sense in building a broken driver.  Previously, there was
the potential of building a DRI1-only driver that would work for DRI1
and fail on DRI2 because the newer libdrm code wasn't present.  Now
the radeon build system should be matching intel and nouveau.

12 years agoconfigure.ac: Fix equality checks in gallium st setup.
Eric Anholt [Fri, 14 Oct 2011 21:33:00 +0000 (14:33 -0700)]
configure.ac: Fix equality checks in gallium st setup.

12 years agodri: Remove driver GenerateMipmap hooks.
Eric Anholt [Fri, 14 Oct 2011 21:11:17 +0000 (14:11 -0700)]
dri: Remove driver GenerateMipmap hooks.

Mesa sets up _mesa_meta_GenerateMipmap as the default hook, which does
this check for fallback and call the fallback itself.

12 years agoradeon: Drop some remaining DRI1 vblank support code.
Eric Anholt [Fri, 14 Oct 2011 21:04:48 +0000 (14:04 -0700)]
radeon: Drop some remaining DRI1 vblank support code.

12 years agointel: remove dead prototype for old DRI1 code.
Eric Anholt [Fri, 14 Oct 2011 20:54:44 +0000 (13:54 -0700)]
intel: remove dead prototype for old DRI1 code.

Noticed while grepping for radeon code.

12 years agoradeon: Simplify cliprects computation now that there's just 1.
Eric Anholt [Fri, 14 Oct 2011 20:52:31 +0000 (13:52 -0700)]
radeon: Simplify cliprects computation now that there's just 1.

This can probably be reduced even further by moving this logic to the
scissor state update or just removing the logic entirely, but I don't
trust myself in radeon quite that much.

12 years agoradeon: Drop the clipping in spans, now that we always have (0,0) -> (w,h).
Eric Anholt [Fri, 14 Oct 2011 20:39:38 +0000 (13:39 -0700)]
radeon: Drop the clipping in spans, now that we always have (0,0) -> (w,h).

12 years agoradeon: Drop the legacy BO manager code.
Eric Anholt [Fri, 14 Oct 2011 20:27:02 +0000 (13:27 -0700)]
radeon: Drop the legacy BO manager code.

12 years agoradeon: Drop the DRI1 zero-copy TFP code.
Eric Anholt [Fri, 14 Oct 2011 20:25:06 +0000 (13:25 -0700)]
radeon: Drop the DRI1 zero-copy TFP code.

12 years agoradeon: Drop the radeon_cs_legacy code now that we rely on kernel mm.
Eric Anholt [Fri, 14 Oct 2011 20:14:22 +0000 (13:14 -0700)]
radeon: Drop the radeon_cs_legacy code now that we rely on kernel mm.

12 years agoradeon: Drop dri2 checks now that it's always true.
Eric Anholt [Fri, 14 Oct 2011 19:49:59 +0000 (12:49 -0700)]
radeon: Drop dri2 checks now that it's always true.

This makes LOCK_HARDWARE empty, so it goes away.

12 years agoradeon: Drop the DRI1 swapbuffers implementation.
Eric Anholt [Fri, 14 Oct 2011 20:01:16 +0000 (13:01 -0700)]
radeon: Drop the DRI1 swapbuffers implementation.

12 years agoradeon: Drop the non-kernel-memory-manager support, and thus DRI1.
Eric Anholt [Fri, 14 Oct 2011 19:22:57 +0000 (12:22 -0700)]
radeon: Drop the non-kernel-memory-manager support, and thus DRI1.

It's past time, and it was going to get in the way of the renderbuffer
mapping refactor.  We dropped all the other DRI1 drivers for this
release, and I can't imagine anybody supporting DRI1 radeon classic in
a new release of Mesa.

Diff produced by treating kernel_mm as true, deleting the DRI1 paths
that produce kernel_mm false, and deleting code.

12 years agor200: Drop the non-kernel-memory-manager and DRI1 code.
Eric Anholt [Fri, 14 Oct 2011 19:26:53 +0000 (12:26 -0700)]
r200: Drop the non-kernel-memory-manager and DRI1 code.

12 years agoradeon: Unifdef RADEON_R300 and RADEON_R600.
Eric Anholt [Thu, 20 Oct 2011 22:15:34 +0000 (15:15 -0700)]
radeon: Unifdef RADEON_R300 and RADEON_R600.

12 years agodocs: Update to note that r300 and r600 have been replaced.
Eric Anholt [Thu, 20 Oct 2011 22:10:05 +0000 (15:10 -0700)]
docs: Update to note that r300 and r600 have been replaced.

12 years agor300c, r600c: Remove these DRI drivers.
Eric Anholt [Thu, 20 Oct 2011 22:06:54 +0000 (15:06 -0700)]
r300c, r600c: Remove these DRI drivers.

They have been superseded by the gallium equivalents.

Acked-by: Michel Dänzer <michel@daenzer.net>
Acked-by: Alex Deucher <alexdeucher@gmail.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Acked-by: Corbin Simpson <mostawesomedude@gmail.com>
12 years agomesa: Remove build infrastructure for r300c and r600c.
Eric Anholt [Thu, 20 Oct 2011 21:55:53 +0000 (14:55 -0700)]
mesa: Remove build infrastructure for r300c and r600c.

These drivers have been superseded by the gallium equivalents.

12 years agoradeon: Delete DRI1 screen init code and thus support for !kernel_mm.
Eric Anholt [Fri, 14 Oct 2011 19:27:48 +0000 (12:27 -0700)]
radeon: Delete DRI1 screen init code and thus support for !kernel_mm.

It's past time, and it was going to get in the way of the renderbuffer
mapping refactor.  We dropped all the other DRI1 drivers for this
release, and I can't imagine anybody supporting DRI1 radeon classic in
a new release of Mesa.

Cleanup of the resulting dead code to follow.

Acked-by: Alex Deucher <alexander.deucher@amd.com>
12 years agodri: Drop _dri_texformats that just obfuscate MESA_FORMAT names.
Eric Anholt [Fri, 14 Oct 2011 18:17:39 +0000 (11:17 -0700)]
dri: Drop _dri_texformats that just obfuscate MESA_FORMAT names.

The remaining _dri_texformats are the ones that are variable depending
on the endianness of the system.

12 years agoradeon: Use _mesa_get_format_base_format for winsys renderbuffer setup.
Eric Anholt [Fri, 14 Oct 2011 00:17:04 +0000 (17:17 -0700)]
radeon: Use _mesa_get_format_base_format for winsys renderbuffer setup.

12 years agoradeon: Remove dead swrast renderbuffer setup code.
Eric Anholt [Fri, 14 Oct 2011 00:14:11 +0000 (17:14 -0700)]
radeon: Remove dead swrast renderbuffer setup code.

This was from the stub code in the initial commit of this file.

12 years agoglsl: Remove pointless uses of glsl_type::get_base_type().
Kenneth Graunke [Tue, 25 Oct 2011 19:55:54 +0000 (12:55 -0700)]
glsl: Remove pointless uses of glsl_type::get_base_type().

These are effectively doing type->get_base_type()->base_type, which is
equivalent to type->base_type.  Just use that, as it's simpler.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
12 years agor600g: don't flush before reading query results
Marek Olšák [Fri, 28 Oct 2011 17:05:44 +0000 (19:05 +0200)]
r600g: don't flush before reading query results

Taken care of by the winsys.

12 years agor600g: only maintain the list of active queries
Marek Olšák [Fri, 28 Oct 2011 16:27:00 +0000 (18:27 +0200)]
r600g: only maintain the list of active queries

And not all existing queries. The only reason we have that list is to be able
to suspend and resume the active ones.

This reduces looping over queries when suspending and resuming.
The queries no longer have to track some of their states.

12 years agor600g: cleanup r600_query_result
Marek Olšák [Fri, 28 Oct 2011 10:44:19 +0000 (12:44 +0200)]
r600g: cleanup r600_query_result

12 years agor600g: don't flush in r600_get_backend_mask
Marek Olšák [Fri, 28 Oct 2011 10:37:54 +0000 (12:37 +0200)]
r600g: don't flush in r600_get_backend_mask

The winsys does the flush in buffer_map.

12 years agor600g: remove redundant variable r600_query::buffer_size
Marek Olšák [Fri, 28 Oct 2011 10:34:57 +0000 (12:34 +0200)]
r600g: remove redundant variable r600_query::buffer_size

12 years agor300g: remove useless code
Marek Olšák [Fri, 28 Oct 2011 15:44:02 +0000 (17:44 +0200)]
r300g: remove useless code

12 years agogallium/auxiliary/util: Solaris also has standard Unix sockets
Alan Coopersmith [Fri, 21 Oct 2011 23:17:56 +0000 (16:17 -0700)]
gallium/auxiliary/util: Solaris also has standard Unix sockets

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
12 years agor300/compiler: Fix scheduler bug causing texture corruption
Tom Stellard [Sun, 23 Oct 2011 01:30:46 +0000 (18:30 -0700)]
r300/compiler: Fix scheduler bug causing texture corruption

We weren't setting TEX_SEM_WAIT on instructions that read the value of a
TEX instruction and also wrote the same register as the TEX instruction.

This is the sequence we were miscompiling:

1: TEX temp[0], input[2].xy__, 2D[0]
...
16: src0.xyz = temp[22], src1.xyz = temp[0], src2.xyz = temp[19]
      MAD temp[0].xyz, src0.xxx, src1.xyz, src2.xxx

https://bugs.freedesktop.org/show_bug.cgi?id=42090

12 years agoi965/gen6+: Add support for noperspective interpolation.
Paul Berry [Sat, 22 Oct 2011 16:33:16 +0000 (09:33 -0700)]
i965/gen6+: Add support for noperspective interpolation.

This required the following changes:

- WM setup now makes the appropriate set of barycentric coordinates
  (perspective vs. noperspective) available to the fragment shader,
  based on whether the shader requires perspective interpolation,
  noperspective interpolation, both, or neither.

- The fragment shader backend now uses the appropriate set of
  barycentric coordiantes when interpolating, based on the
  interpolation mode returned by
  ir_variable::determine_interpolation_mode().

- SF setup now uses gl_fragment_program::InterpQualifier to determine
  which attributes are to be flat shaded (as opposed to the old logic,
  which only flat shaded colors).

- CLIP setup now ensures that the clipper outputs non-perspective
  barycentric coordinates when they are needed by the fragment shader.

Fixes the remaining piglit tests of interpolation qualifiers that were
failing:
- interpolation-flat-*-smooth-none
- interpolation-flat-other-flat-none
- interpolation-noperspective-*
- interpolation-smooth-gl_*Color-flat-*

Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agoi965/gen6+: Rename GEN6_CLIP_BARYCENTRIC_ENABLE.
Paul Berry [Sat, 22 Oct 2011 16:33:12 +0000 (09:33 -0700)]
i965/gen6+: Rename GEN6_CLIP_BARYCENTRIC_ENABLE.

The name was misleading.  The actual effect of the bit is to cause
the clipper to emit *non-perspective* barycentric coordinate
information (which is only needed when doing noperspective
interpolation).

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agoi965/fs: use determine_interpolation_mode().
Paul Berry [Fri, 21 Oct 2011 14:56:08 +0000 (07:56 -0700)]
i965/fs: use determine_interpolation_mode().

This patch changes how fs_visitor::emit_general_interpolation()
decides what kind of interpolation to do.  Previously, it used the
shade model to determine how to interpolate colors, and used smooth
interpolation on everything else.  Now it uses
ir_variable::determine_interpolation_mode(), so that it respects GLSL
1.30 interpolation qualifiers.

Fixes piglit tests interpolation-flat-*-smooth-{distance,fixed,vertex}
and interpolation-flat-other-flat-{distance,fixed,vertex}.

Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agoi965/gen6+: Parameterize barycentric interpolation modes.
Paul Berry [Sat, 22 Oct 2011 00:20:32 +0000 (17:20 -0700)]
i965/gen6+: Parameterize barycentric interpolation modes.

This patch modifies the fragment shader back-end so that instead of
using a single delta_x/delta_y register pair to store barycentric
coordinates, it uses an array of such register pairs, one for each
possible intepolation mode.

When setting up the WM, we intstruct it to only provide the
barycentric coordinates that are actually needed by the fragment
shader--that is computed by brw_compute_barycentric_interp_modes().
Currently this function returns just
BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC, because this is the only
interpolation mode we support.  However, that will change in a later
patch.

Reviewed-by: Eric Anholt <eric@anholt.net>
12 years agoi965/fs: Fix split_virtual_grfs() when delta_xy not in a virtual register.
Paul Berry [Fri, 21 Oct 2011 21:16:25 +0000 (14:16 -0700)]
i965/fs: Fix split_virtual_grfs() when delta_xy not in a virtual register.

This patch modifies the special case in
fs_visitor::split_virtual_grfs() that prevents splitting from being
applied to the delta_x/delta_y register pair (this register pair needs
to remain contiguous so that it can be used by the PLN instruction).

When gen>=6, this register pair is in a fixed location, not a virtual
register, so it was in no danger of being split.  And
split_virtual_grfs' attempt not to split it was preventing some other
unrelated register from being split.

Reviewed-by: Eric Anholt <eric@anholt.net>