mesa.git
10 years agonv50: TXF already has integer arguments, don't try to convert from f32
Ilia Mirkin [Fri, 29 Nov 2013 06:18:57 +0000 (01:18 -0500)]
nv50: TXF already has integer arguments, don't try to convert from f32

Fixes the texelFetch piglit tests

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
10 years agollvmpipe: clamp fragment shader depth write to the current viewport depth range.
Matthew McClure [Tue, 26 Nov 2013 18:50:27 +0000 (10:50 -0800)]
llvmpipe: clamp fragment shader depth write to the current viewport depth range.

With this patch, generate_fs_loop will clamp any fragment shader depth writes
to the viewport's min and max depth values. Viewport selection is determined
by the geometry shader output for the viewport array index. If no index is
specified, then the default viewport index is zero. Semantics for this path
can be found in draw_clamp_viewport_idx and lp_clamp_viewport_idx.

lp_jit_viewport was created to store viewport information visible to JIT code,
and is validated when the LP_NEW_VIEWPORT dirty flag is set.

lp_rast_shader_inputs is responsible for passing the viewport_index through
the rasterizer stage to fragment stage (via lp_jit_thread_data).

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
10 years agowayland: Add support for eglSwapInterval
Neil Roberts [Fri, 15 Nov 2013 13:50:50 +0000 (13:50 +0000)]
wayland: Add support for eglSwapInterval

The Wayland EGL platform now respects the eglSwapInterval value. The value is
clamped to either 0 or 1 because it is difficult (and probably not useful) to
sync to more than 1 redraw.

The main change is that if the swap interval is 0 then Mesa won't install a
frame callback so that eglSwapBuffers can be executed as often as necessary.
Instead it will do a sync request after the swap buffers. It will block for
sync complete event in get_back_bo instead of the frame callback. The
compositor is likely to send a release event while processing the new buffer
attach and this makes sure we will receive that before deciding whether to
allocate a new buffer.

If there are no buffers available then instead of returning with an error,
get_back_bo will now poll the compositor by repeatedly sending sync requests
every 10ms. This is a last resort and in theory this shouldn't happen because
there should be no reason for the compositor to hold on to more than three
buffers. That means whenever we attach the fourth buffer we should always get
an immediate release event which should come in with the notification for the
first sync request that we are throttled to.

When the compositor is directly scanning out from the application's buffer it
may end up holding on to three buffers. These are the one that is is currently
scanning out from, one that has been given to DRM as the next buffer to flip
to, and one that has been attached and will be given to DRM as soon as the
previous flip completes. When we attach a fourth buffer to the compositor it
should replace that third buffer so we should get a release event immediately
after that. This patch therefore also changes the number of buffer slots to 4
so that we can accomodate that situation.

If DRM eventually gets a way to cancel a pending page flip then the compositors
can be changed to only need to hold on to two buffers and this value can be
put back to 3.

This also moves the vblank configuration defines from platform_x11.c to the
common egl_dri2.h header so they can be shared by both platforms.

10 years agowayland: Block for the frame callback in get_back_bo not dri2_swap_buffers
Neil Roberts [Fri, 15 Nov 2013 13:50:49 +0000 (13:50 +0000)]
wayland: Block for the frame callback in get_back_bo not dri2_swap_buffers

Consider a typical game-style main loop which might be like this:

while (1) {
draw_something();
eglSwapBuffers();
}

In this case the game is relying on eglSwapBuffers to throttle to a sensible
frame rate. Previously this game would end up using three buffers even though
it should only need two. This is because Mesa decides whether to allocate a
new buffer in get_back_bo which would be before it has tried to read any
events from the compositor so it wouldn't have seen any buffer release events
yet.

This patch just moves the block for the frame callback to get_back_bo.
Typically the compositor will send a release event immediately after one of
the attaches so if we block for the frame callback here then we can be sure to
have completed at least one roundtrip and received that release event after
attaching the previous buffer before deciding whether to allocate a new one.

dri2_swap_buffers always calls get_back_bo so even if the client doesn't
render anything we will still be sure to block to the frame callback. The code
to create the new frame callback has been moved to after this call so that we
can be sure to have cleared the previous frame callback before requesting a
new one.

10 years agoglapi: Do not include dlfcn.h on Windows.
Vinson Lee [Fri, 11 Oct 2013 01:13:25 +0000 (18:13 -0700)]
glapi: Do not include dlfcn.h on Windows.

This patch fixes this MinGW build error.

  CC     glapi_gentable.lo
glapi_gentable.c:47:19: fatal error: dlfcn.h: No such file or directory

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agor600/llvm: Allow arbitrary amount of temps in tgsi to llvm
Vincent Lejeune [Sun, 1 Dec 2013 23:54:44 +0000 (00:54 +0100)]
r600/llvm: Allow arbitrary amount of temps in tgsi to llvm

10 years agofreedreno/a3xx: add adreno 330 support
Rob Clark [Sat, 7 Dec 2013 13:47:10 +0000 (08:47 -0500)]
freedreno/a3xx: add adreno 330 support

Signed-off-by: Rob Clark <robclark@freedesktop.org>
10 years agofreedreno/a3xx/compiler: add ROUND
Rob Clark [Sat, 7 Dec 2013 13:01:29 +0000 (08:01 -0500)]
freedreno/a3xx/compiler: add ROUND

Signed-off-by: Rob Clark <robclark@freedesktop.org>
10 years agomesa: Require per-sample shading if the `sample` qualifier is used.
Chris Forbes [Fri, 29 Nov 2013 08:44:13 +0000 (21:44 +1300)]
mesa: Require per-sample shading if the `sample` qualifier is used.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agoglsl: Populate gl_fragment_program::IsSample bitfield
Chris Forbes [Fri, 29 Nov 2013 08:30:58 +0000 (21:30 +1300)]
glsl: Populate gl_fragment_program::IsSample bitfield

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agomesa: add IsSample bitfield to gl_fragment_program
Chris Forbes [Fri, 29 Nov 2013 08:30:00 +0000 (21:30 +1300)]
mesa: add IsSample bitfield to gl_fragment_program

Drivers will need to look at this to decide if they need to do
per-sample fragment shader dispatch.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agoglsl: Put `sample`-qualified varyings in their own packing classes
Chris Forbes [Fri, 29 Nov 2013 08:28:32 +0000 (21:28 +1300)]
glsl: Put `sample`-qualified varyings in their own packing classes

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agoglsl: Add ir support for `sample` qualifier; adjust compiler and linker
Chris Forbes [Fri, 29 Nov 2013 08:26:10 +0000 (21:26 +1300)]
glsl: Add ir support for `sample` qualifier; adjust compiler and linker

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agoglsl: Add frontend support for `sample` auxiliary storage qualifier
Chris Forbes [Fri, 29 Nov 2013 08:21:56 +0000 (21:21 +1300)]
glsl: Add frontend support for `sample` auxiliary storage qualifier

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
10 years agoi965: Don't flag gather quirks for Gen8+
Chris Forbes [Fri, 29 Nov 2013 09:04:58 +0000 (22:04 +1300)]
i965: Don't flag gather quirks for Gen8+

My understanding is that Broadwell retains the same SCS mechanism
that Haswell has, so even if the underlying issue with this format
is not fixed, the w/a will be applied in SCS rather than needing
shader code.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965/Gen7: Allow CMS layout for multisample textures
Chris Forbes [Fri, 29 Nov 2013 21:03:10 +0000 (10:03 +1300)]
i965/Gen7: Allow CMS layout for multisample textures

Now that all the pieces are in place, this should provide
a nice performance boost for apps using multisample textures.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965/vs: Sample from MCS surface when required
Chris Forbes [Sat, 30 Nov 2013 00:49:50 +0000 (13:49 +1300)]
i965/vs: Sample from MCS surface when required

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965/fs: Sample from MCS surface when required
Chris Forbes [Fri, 29 Nov 2013 22:16:38 +0000 (11:16 +1300)]
i965/fs: Sample from MCS surface when required

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965: Add shader opcode for sampling MCS surface
Chris Forbes [Fri, 29 Nov 2013 21:32:16 +0000 (10:32 +1300)]
i965: Add shader opcode for sampling MCS surface

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965/Gen7: Include bitfield in the sampler key for CMS layout
Chris Forbes [Fri, 29 Nov 2013 21:01:12 +0000 (10:01 +1300)]
i965/Gen7: Include bitfield in the sampler key for CMS layout

We need to emit extra shader code in this case to sample the
MCS surface first; we can't just blindly do this all the time
since IVB will sometimes try to access the MCS surface even if
disabled.

V3: Use actual MSAA layout from the texture's mt, rather
then computing what would have been used based on the format.
This is simpler and less fragile - there's at least one case where
we might want to have a texture's MSAA layout change based on what
the app does (CMS SINT falling back to UMS if the app ever attempts
to render to it with a channel disabled.)

This also obsoletes V2's 1/10 -- compute_msaa_layout can now remain
an implementation detail of the miptree code.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965/Gen7: Move decision to allocate MCS surface into intel_mipmap_create
Chris Forbes [Sat, 30 Nov 2013 22:44:39 +0000 (11:44 +1300)]
i965/Gen7: Move decision to allocate MCS surface into intel_mipmap_create

This gives us correct behavior for both renderbuffers (which previously
worked) and multisample textures (which would never get an MCS surface
allocated, even if CMS layout was selected)

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965/Gen7: emit mcs info for multisample textures
Chris Forbes [Sat, 30 Nov 2013 22:30:45 +0000 (11:30 +1300)]
i965/Gen7: emit mcs info for multisample textures

Previously this was only done for render targets.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965/wm: Set copy of sample mask in 3DSTATE_PS correctly for Haswell
Chris Forbes [Sat, 30 Nov 2013 22:07:46 +0000 (11:07 +1300)]
i965/wm: Set copy of sample mask in 3DSTATE_PS correctly for Haswell

The bspec says:

"SW must program the sample mask value in this field so that it matches
with 3DSTATE_SAMPLE_MASK"

I haven't observed this to actually fix anything, but stumbled across it
while adding the rest of the support for CMS layout for multisample
   textures.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965: refactor sample mask calculation
Chris Forbes [Sat, 30 Nov 2013 22:03:41 +0000 (11:03 +1300)]
i965: refactor sample mask calculation

Haswell needs a copy of the sample mask in 3DSTATE_PS; this makes that
convenient.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoglsl: Don't emit empty declaration warning for a struct specifier
Ian Romanick [Wed, 27 Nov 2013 19:22:27 +0000 (11:22 -0800)]
glsl: Don't emit empty declaration warning for a struct specifier

The intention is that things like

   int;

will generate a warning.  However, we were also accidentally emitting
the same warning for things like

  struct Foo { int x; };

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68838
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: Aras Pranckevicius <aras@unity3d.com>
Cc: "9.2 10.0" <mesa-stable@lists.freedesktop.org>
10 years agost/xa: Bump major version number to 2
Thomas Hellstrom [Thu, 5 Dec 2013 11:55:43 +0000 (03:55 -0800)]
st/xa: Bump major version number to 2

For some reason this was left out when the version was changed...

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
10 years agonvc0: fixup gk110 and up not being listed in various switch statements
Ben Skeggs [Thu, 5 Dec 2013 23:09:42 +0000 (09:09 +1000)]
nvc0: fixup gk110 and up not being listed in various switch statements

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agoi965: Replace non-standard INLINE macro with "inline".
Kenneth Graunke [Mon, 2 Dec 2013 21:39:40 +0000 (13:39 -0800)]
i965: Replace non-standard INLINE macro with "inline".

These are identical: main/compiler.h defines INLINE to "inline".

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965: Don't use GL types in files shared with intel-gpu-tools.
Kenneth Graunke [Mon, 25 Nov 2013 23:51:24 +0000 (15:51 -0800)]
i965: Don't use GL types in files shared with intel-gpu-tools.

sed -i -e 's/GLuint/unsigned/g' -e 's/GLint/int/g' \
       -e 's/GLfloat/float/g' -e 's/GLubyte/uint8_t/g' \
       -e 's/GLshort/int16_t/g' \
       brw_eu* brw_disasm.c brw_structs.h

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965: Drop trailing whitespace from the rest of the driver.
Kenneth Graunke [Mon, 25 Nov 2013 23:46:34 +0000 (15:46 -0800)]
i965: Drop trailing whitespace from the rest of the driver.

Performed via:
$ for file in *; do sed -i 's/  *//g'; done

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965: Drop trailing whitespace from files shared with intel-gpu-tools.
Kenneth Graunke [Mon, 25 Nov 2013 23:39:03 +0000 (15:39 -0800)]
i965: Drop trailing whitespace from files shared with intel-gpu-tools.

Performed via s/  *$//g.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agotools/trace: More tweaks to state dumping.
José Fonseca [Thu, 5 Dec 2013 13:29:29 +0000 (13:29 +0000)]
tools/trace: More tweaks to state dumping.

- Ignore buffer format (it is totally arbitrary)
- Initialize state.
- Handle begin/end_query statements.

10 years agotrace: Reorder dumping of pipe_rasterizer_state.
José Fonseca [Thu, 5 Dec 2013 13:25:38 +0000 (13:25 +0000)]
trace: Reorder dumping of pipe_rasterizer_state.

Such that it matches the pipe_rasterizer_state declaration, making it
easier to double-check that all state is being actually dumped.

Trivial.

10 years agotrace: Dump pipe_sampler_state::seamless_cube_map.
José Fonseca [Thu, 5 Dec 2013 13:24:59 +0000 (13:24 +0000)]
trace: Dump pipe_sampler_state::seamless_cube_map.

Trivial.

10 years agoradeonsi: Remove some stale XXX / FIXME comments
Michel Dänzer [Wed, 4 Dec 2013 04:37:07 +0000 (13:37 +0900)]
radeonsi: Remove some stale XXX / FIXME comments

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
10 years agoi965: Emit better code for ir_unop_sign.
Matt Turner [Mon, 25 Nov 2013 06:44:32 +0000 (22:44 -0800)]
i965: Emit better code for ir_unop_sign.

total instructions in shared programs: 1550449 -> 1550048 (-0.03%)
instructions in affected programs:     15207 -> 14806 (-2.64%)

Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
10 years agoi965/fs: New peephole optimization to flatten IF/BREAK/ENDIF.
Matt Turner [Wed, 23 Oct 2013 00:51:28 +0000 (17:51 -0700)]
i965/fs: New peephole optimization to flatten IF/BREAK/ENDIF.

total instructions in shared programs: 1550713 -> 1550449 (-0.02%)
instructions in affected programs:     7931 -> 7667 (-3.33%)

Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965/fs: Emit a MOV instead of a SEL if the sources are the same.
Matt Turner [Mon, 28 Oct 2013 00:09:41 +0000 (17:09 -0700)]
i965/fs: Emit a MOV instead of a SEL if the sources are the same.

One program affected.

instructions in affected programs:     436 -> 428 (-1.83%)

Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965/fs: Extend SEL peephole to handle only matching MOVs.
Matt Turner [Wed, 30 Oct 2013 04:39:52 +0000 (21:39 -0700)]
i965/fs: Extend SEL peephole to handle only matching MOVs.

Before this patch, the following code would not be optimized even though
the first two instructions were common to the then and else blocks:

   (+f0) IF
   MOV dst0 ...
   MOV dst1 ...
   MOV dst2 ...
   ELSE
   MOV dst0 ...
   MOV dst1 ...
   MOV dst3 ...
   ENDIF

This commit extends the peephole to handle this case.

No shader-db changes.

Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/fs: New peephole optimization to generate SEL.
Matt Turner [Wed, 23 Oct 2013 00:51:28 +0000 (17:51 -0700)]
i965/fs: New peephole optimization to generate SEL.

fs_visitor::try_replace_with_sel optimizes only if statements whose
"then" and "else" bodies contain a single MOV instruction. It also
could not handle constant arguments, since they cause an extra MOV
immediate to be generated (since we haven't run constant propagation,
there are more than the single MOV).

This peephole fixes both of these and operates as a normal optimization
pass.

fs_visitor::try_replace_with_sel is still arguably necessary, since it
runs before pull constant loads are lowered.

total instructions in shared programs: 1559129 -> 1545833 (-0.85%)
instructions in affected programs:     167120 -> 153824 (-7.96%)
GAINED:                                13
LOST:                                  6

Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoi965/fs: Add SEL() convenience function.
Matt Turner [Wed, 23 Oct 2013 02:04:14 +0000 (19:04 -0700)]
i965/fs: Add SEL() convenience function.

Reviewed-by: Paul Berry <stereotype441@gmail.com>
10 years agoglsl: Use fabs() on floating point values.
Matt Turner [Tue, 26 Nov 2013 06:18:28 +0000 (22:18 -0800)]
glsl: Use fabs() on floating point values.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
10 years agoi965: Print conditional mod in dump_instruction().
Matt Turner [Sun, 20 Oct 2013 22:48:14 +0000 (15:48 -0700)]
i965: Print conditional mod in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Externalize conditional_modifier for use in dump_instruction().
Matt Turner [Mon, 2 Dec 2013 21:15:45 +0000 (13:15 -0800)]
i965: Externalize conditional_modifier for use in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Print argument types in dump_instruction().
Matt Turner [Mon, 2 Dec 2013 21:10:29 +0000 (13:10 -0800)]
i965: Print argument types in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Externalize reg_encoding for use in dump_instruction().
Matt Turner [Mon, 2 Dec 2013 20:58:45 +0000 (12:58 -0800)]
i965: Externalize reg_encoding for use in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/vec4: Don't print swizzles for immediate values.
Matt Turner [Mon, 2 Dec 2013 20:43:50 +0000 (12:43 -0800)]
i965/vec4: Don't print swizzles for immediate values.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/vec4: Print negate and absolute value for src args.
Matt Turner [Mon, 2 Dec 2013 20:41:16 +0000 (12:41 -0800)]
i965/vec4: Print negate and absolute value for src args.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/vec4: Add support for printing HW_REGs in dump_instruction().
Matt Turner [Tue, 26 Nov 2013 06:17:29 +0000 (22:17 -0800)]
i965/vec4: Add support for printing HW_REGs in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/fs: Print ARF registers properly in dump_instruction().
Matt Turner [Mon, 25 Nov 2013 23:37:18 +0000 (15:37 -0800)]
i965/fs: Print ARF registers properly in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Don't print extra (null) arguments in dump_instruction().
Matt Turner [Mon, 25 Nov 2013 23:15:25 +0000 (15:15 -0800)]
i965: Don't print extra (null) arguments in dump_instruction().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoglsl: Remove silly OR(..., 0x0) from ldexp() lowering.
Matt Turner [Wed, 27 Nov 2013 22:23:50 +0000 (14:23 -0800)]
glsl: Remove silly OR(..., 0x0) from ldexp() lowering.

I translated copysign(0.0f, x) a little too literally.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Allow commuting the operands of ADDC for const propagation.
Matt Turner [Thu, 28 Nov 2013 00:14:14 +0000 (16:14 -0800)]
i965: Allow commuting the operands of ADDC for const propagation.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/fs: Rename register_coalesce_2() -> register_coalesce().
Matt Turner [Sat, 30 Nov 2013 06:16:14 +0000 (22:16 -0800)]
i965/fs: Rename register_coalesce_2() -> register_coalesce().

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/fs: Remove now useless register_coalesce() pass.
Matt Turner [Sat, 30 Nov 2013 06:14:14 +0000 (22:14 -0800)]
i965/fs: Remove now useless register_coalesce() pass.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/fs: Let register_coalesce_2() eliminate self-moves.
Matt Turner [Fri, 29 Nov 2013 19:28:54 +0000 (11:28 -0800)]
i965/fs: Let register_coalesce_2() eliminate self-moves.

This is the last thing that register_coalesce() still handled.

total instructions in shared programs: 1561060 -> 1560908 (-0.01%)
instructions in affected programs:     15758 -> 15606 (-0.96%)

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Allow constant propagation into ASR and BFI1.
Matt Turner [Tue, 26 Nov 2013 21:49:31 +0000 (13:49 -0800)]
i965: Allow constant propagation into ASR and BFI1.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Document cur_* variables.
Matt Turner [Mon, 2 Dec 2013 18:29:49 +0000 (10:29 -0800)]
i965/cfg: Document cur_* variables.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Remove ip & cur from brw_cfg.
Matt Turner [Sun, 1 Dec 2013 04:38:48 +0000 (20:38 -0800)]
i965/cfg: Remove ip & cur from brw_cfg.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Clean up cfg_t constructors.
Matt Turner [Fri, 29 Nov 2013 07:24:44 +0000 (23:24 -0800)]
i965/cfg: Clean up cfg_t constructors.

parent_mem_ctx was unused since db47074a, so remove the two wrappers
around create() and make create() the constructor.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Throw out confusing make_list method.
Matt Turner [Tue, 26 Nov 2013 23:25:44 +0000 (15:25 -0800)]
i965/cfg: Throw out confusing make_list method.

make_list is just a one-line wrapper and was confusingly called by
NULL objects. E.g., cur_if == NULL; cur_if->make_list(mem_ctx).

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Include only needed headers.
Matt Turner [Sun, 1 Dec 2013 00:39:43 +0000 (16:39 -0800)]
i965/cfg: Include only needed headers.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Remove unnecessary endif_stack.
Matt Turner [Fri, 29 Nov 2013 07:39:02 +0000 (23:39 -0800)]
i965/cfg: Remove unnecessary endif_stack.

Unnecessary since last commit.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Rework to make IF & ELSE blocks flow into ENDIF.
Matt Turner [Fri, 29 Nov 2013 05:33:05 +0000 (21:33 -0800)]
i965/cfg: Rework to make IF & ELSE blocks flow into ENDIF.

Previously we made the basic block following an ENDIF instruction a
successor of the basic blocks ending with IF and ELSE. The PRM says that
IF and ELSE instructions jump *to* the ENDIF, rather than over it.

This should be immaterial to dataflow analysis, except for if, break,
endif sequences:

   START B1 <-B0 <-B9
0x00000100: cmp.g.f0(8)     null            g15<8,8,1>F     g4<0,1,0>F
0x00000110: (+f0) if(8) 0 0                 null            0x00000000UD
   END B1 ->B2 ->B4
   START B2 <-B1
   break
0x00000120: break(8) 0 0                    null            0D
   END B2 ->B10
   START B3
0x00000130: endif(8) 2                      null            0x00000002UD
   END B3 ->B4

The ENDIF block would have no parents, so dataflow analysis would
generate incorrect results, preventing copy propagation from eliminating
some instructions.

This patch changes the CFG to make ENDIF start rather than end basic
blocks, so that it can be the jump target of the IF and ELSE
instructions.

It helps three programs (including two fs8/fs16 pairs).

total instructions in shared programs: 1561126 -> 1561060 (-0.00%)
instructions in affected programs:     837 -> 771 (-7.89%)

More importantly, it allows copy propagation to handle more cases.
Disabling the register_coalesce() pass before this patch hurts 58
programs, while afterward it only hurts 11 programs.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Keep pointers to IF/ELSE/ENDIF instructions in the cfg.
Matt Turner [Wed, 30 Oct 2013 23:51:32 +0000 (16:51 -0700)]
i965/cfg: Keep pointers to IF/ELSE/ENDIF instructions in the cfg.

Useful for finding the associated control flow instructions, given a
block ending in one.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965/cfg: Add code to dump blocks and cfg.
Matt Turner [Thu, 28 Nov 2013 19:03:14 +0000 (11:03 -0800)]
i965/cfg: Add code to dump blocks and cfg.

Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agomesa: Remove GL_MESA_texture_array cruft from gl.h
Ian Romanick [Wed, 20 Nov 2013 21:52:18 +0000 (13:52 -0800)]
mesa: Remove GL_MESA_texture_array cruft from gl.h

glext.h has had all the necessary bits for years.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Remove support for GL_MESA_texture_array
Ian Romanick [Wed, 20 Nov 2013 21:48:36 +0000 (13:48 -0800)]
mesa: Remove support for GL_MESA_texture_array

This extension enabled the use of texture array with fixed-function and
assembly fragment shaders.  No applications are known to use this
extension.

NOTE: This patch regresses GL_TEXTURE_1D_ARRAY and GL_TEXTURE_2D_ARRAY
cases of the copyteximage piglit test.  The test is incorrectly using
texture arrays with fixed function while only requiring the
GL_EXT_texture_array extension.  A fix for the test has been posted to
the piglit mailing list.

http://lists.freedesktop.org/archives/piglit/2013-November/008639.html

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Use a single enable for GL_EXT_texture_array and GL_MESA_texture_array
Ian Romanick [Wed, 20 Nov 2013 21:41:23 +0000 (13:41 -0800)]
mesa: Use a single enable for GL_EXT_texture_array and GL_MESA_texture_array

Every driver that enables one also enables the other.  The difference
between the two is MESA adds support for fixed-function and assembly
fragment shaders, but EXT only adds support for GLSL.  The MESA
extension was created back when Mesa did not support GLSL.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Minor clean-up of target_enum_to_index
Ian Romanick [Wed, 20 Nov 2013 21:05:35 +0000 (13:05 -0800)]
mesa: Minor clean-up of target_enum_to_index

Constify the gl_context parameter, and remove suffixes from enums that
have non-suffix versions.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Silence GCC warning in count_tex_size
Ian Romanick [Wed, 20 Nov 2013 20:59:22 +0000 (12:59 -0800)]
mesa: Silence GCC warning in count_tex_size

main/texobj.c: In function 'count_tex_size':
main/texobj.c:886:23: warning: unused parameter 'key' [-Wunused-parameter]

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Silence GCC warning in _mesa_test_texobj_completeness
Ian Romanick [Wed, 20 Nov 2013 20:58:37 +0000 (12:58 -0800)]
mesa: Silence GCC warning in _mesa_test_texobj_completeness

main/texobj.c: In function '_mesa_test_texobj_completeness':
main/texobj.c:553:34: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
main/texobj.c:553:193: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
main/texobj.c:553:254: warning: signed and unsigned type in conditional expression [-Wsign-compare]
main/texobj.c:553:148: warning: signed and unsigned type in conditional expression [-Wsign-compare]

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Add missing API check for GL_TEXTURE_3D
Ian Romanick [Sat, 23 Nov 2013 20:13:50 +0000 (12:13 -0800)]
mesa: Add missing API check for GL_TEXTURE_3D

There are no 3D textures in OpenGL ES 1.x.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: Add missing checks for GL_TEXTURE_CUBE_MAP_ARRAY
Ian Romanick [Wed, 20 Nov 2013 21:22:05 +0000 (13:22 -0800)]
mesa: Add missing checks for GL_TEXTURE_CUBE_MAP_ARRAY

That enum requires GL_ARB_texture_cube_map_array, and it is only
available on desktop GL.  It looks like this has been an un-noticed
issue since GL_ARB_texture_cube_map_array support was added in commit
e0e7e295.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agowayland: Add an extension to create wl_buffers from EGLImages
Neil Roberts [Mon, 28 Oct 2013 15:07:03 +0000 (15:07 +0000)]
wayland: Add an extension to create wl_buffers from EGLImages

This adds an extension called EGL_WL_create_wayland_buffer_from_image
which adds the following single function:

struct wl_buffer *
eglCreateWaylandBufferFromImageWL(EGLDisplay dpy, EGLImageKHR image);

The function creates a wl_buffer which shares its contents with the given
EGLImage. The expected use case for this is in a nested Wayland compositor
which is using subsurfaces to present buffers from its clients. Using this
extension it can attach the client buffers directly to the subsurface without
having to blit the contents into an intermediate buffer. The compositing can
then be done in the parent compositor.

The extension is only implemented in the Wayland EGL platform because of
course it wouldn't make sense anywhere else.

10 years agoegl/wayland: Damage INT32_MAX x INT32_MAX region for eglSwapBuffers
Kristian Høgsberg [Thu, 5 Dec 2013 00:13:35 +0000 (16:13 -0800)]
egl/wayland: Damage INT32_MAX x INT32_MAX region for eglSwapBuffers

If we're not using EGL_EXT_swap_buffers_with_damage, we have to
damage the full extent.  EGL operates on buffer coordinates, but
wl_surface.damage takes surface coordinates.  EGL doesn't know the
buffer transformation (rotated or scaled) and can't post accurate
damage in surface coordinates.  The damage event however is clipped to
the surface extents so we can just damage the maximum rectangle.

In case of EGL_EXT_swap_buffers_with_damage, the application knows
the buffer transform and is expected to pass in rectangles in
surface space.

https://bugs.freedesktop.org/show_bug.cgi?id=70250
Cc: "10.0" mesa-stable@lists.freedesktop.org
10 years agoEnable throttling in SwapBuffers
Axel Davy [Tue, 3 Dec 2013 16:04:10 +0000 (17:04 +0100)]
Enable throttling in SwapBuffers

flush_with_flags, when available, allows the driver to throttle.
Using this suppress input lag issues that can be observed in heavy
rendering situations on non-intel cards.

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Cc: "10.0" mesa-stable@lists.freedesktop.org
10 years agoegl/wayland: Send commit after flushing the driver context
Kristian Høgsberg [Wed, 4 Dec 2013 20:08:35 +0000 (12:08 -0800)]
egl/wayland: Send commit after flushing the driver context

This typically won't make a difference, since we only send the requests at
wl_display_flush() time.  There might be a small race
with another thread calling wl_display_flush() after our commit request,
but before we flush the DRI driver.  Moving the commit below the DRI
driver flush call looks more natural and eliminates the small race.

Cc: "10.0" mesa-stable@lists.freedesktop.org
10 years agoegl/wayland: Flush the wl_display at the end of SwapBuffers
Axel Davy [Tue, 3 Dec 2013 16:38:09 +0000 (17:38 +0100)]
egl/wayland: Flush the wl_display at the end of SwapBuffers

We would like the compositor to receive the commited buffer
as soon as possible, so it has the time to treat it, and
release old ones. We shouldn't rely on the client
to flush the queue for us.

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Cc: "10.0" mesa-stable@lists.freedesktop.org
10 years agomesa: reduce memory used for short display lists
Brian Paul [Sun, 24 Nov 2013 14:04:33 +0000 (07:04 -0700)]
mesa: reduce memory used for short display lists

Display lists allocate memory in chunks of 256 tokens (1KB) at a time.
If an app creates many short display lists or uses glXUseXFont() this
can waste quite a bit of memory.

This patch uses realloc() to trim short lists and reduce the memory
used.

Also, null/zero-out some list construction fields in _mesa_EndList().

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
10 years agomesa: update/remove display list comments
Brian Paul [Wed, 4 Dec 2013 16:45:38 +0000 (09:45 -0700)]
mesa: update/remove display list comments

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
10 years agomesa: remove gl_dlist_node::next pointer to reduce dlist memory use
Brian Paul [Wed, 4 Dec 2013 16:45:38 +0000 (09:45 -0700)]
mesa: remove gl_dlist_node::next pointer to reduce dlist memory use

Now, sizeof(gl_dlist_node)==4 even on 64-bit systems.  This can
halve the memory used by some display lists on 64-bit systems.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
10 years agomesa: begin reducing memory used by display lists
Brian Paul [Wed, 4 Dec 2013 16:45:38 +0000 (09:45 -0700)]
mesa: begin reducing memory used by display lists

This is a first step in reducing memory used by display lists on
64-bit systems.  On 64-bit systems, the gl_dlist_node union type
is 8 bytes because of the 'data' and 'next' fields.  This causes
every display list node/token to occupy 8 bytes instead of 4 as
originally designed.  This basically doubles the memory used by
some display lists on 64-bit systems.

The fix is to remove the 64-bit 'data' and 'next' pointer fields
from the union and instead store them as a pair of 32-bit values.
Easily done with a few helper functions.

The next patch will take care of the 'next' field.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
10 years agonouveau: Add lots of comments to the buffer transfer logic
Ilia Mirkin [Mon, 2 Dec 2013 02:13:17 +0000 (21:13 -0500)]
nouveau: Add lots of comments to the buffer transfer logic

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
10 years agonv50: wait on the buf's fence before sticking it into pushbuf
Ilia Mirkin [Thu, 28 Nov 2013 21:23:15 +0000 (16:23 -0500)]
nv50: wait on the buf's fence before sticking it into pushbuf

This resolves some rendering issues in source games.
See https://bugs.freedesktop.org/show_bug.cgi?id=64323

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "9.2 10.0" <mesa-stable@lists.freedesktop.org>
10 years agonouveau: avoid leaking fences while waiting
Ilia Mirkin [Fri, 29 Nov 2013 23:49:44 +0000 (18:49 -0500)]
nouveau: avoid leaking fences while waiting

This fixes a memory leak in some situations. Also avoids emitting an
extra fence if the kick handler does the call to nouveau_fence_next
itself.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "9.2 10.0" <mesa-stable@lists.freedesktop.org>
10 years agonv50: fix a small leak on context destroy
Ilia Mirkin [Fri, 29 Nov 2013 09:36:41 +0000 (04:36 -0500)]
nv50: fix a small leak on context destroy

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
10 years agodocs: put MD5 sums in 9.2.4 relnotes file
Brian Paul [Wed, 4 Dec 2013 14:47:05 +0000 (07:47 -0700)]
docs: put MD5 sums in 9.2.4 relnotes file

Signed-off-by: Brian Paul <brianp@vmware.com>
10 years agodocs: use --disable-dri3 for VMware guest driver build
Brian Paul [Wed, 4 Dec 2013 14:39:59 +0000 (07:39 -0700)]
docs: use --disable-dri3 for VMware guest driver build

For the time being at least.  Suggested by Adrian Rangel.

Signed-off-by: Brian Paul <brianp@vmware.com>
10 years agomesa: modified _mesa_align_free() to accept NULL pointer
Siavash Eliasi [Wed, 4 Dec 2013 04:50:00 +0000 (21:50 -0700)]
mesa: modified _mesa_align_free() to accept NULL pointer

So that it acts like ordinary free().  This lets us remove a bunch of
if statements where the function is called.

v2:
- Avoiding compile error on MSVC and possible warnings on other compilers.
- Added comment regards passing NULL pointer being safe.

Reviewed-by: Brian Paul <brianp@vmware.com>
10 years agomesa: don't leak performance monitors on context destroy
Ilia Mirkin [Fri, 29 Nov 2013 10:52:27 +0000 (05:52 -0500)]
mesa: don't leak performance monitors on context destroy

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "10.0" <mesa-stable@lists.freedesktop.org>
10 years agonv50: Fix GPU_READING/WRITING bit removal
Ilia Mirkin [Mon, 2 Dec 2013 04:13:56 +0000 (23:13 -0500)]
nv50: Fix GPU_READING/WRITING bit removal

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
CC: "9.1, 9.2, 10.0" <mesa-stable@lists.freedesktop.org>
10 years agopipe-loader: Fix llvmpipe.la path
Michel Dänzer [Wed, 4 Dec 2013 02:56:10 +0000 (11:56 +0900)]
pipe-loader: Fix llvmpipe.la path

Fixes

 make[3]: *** No rule to make target `.../src/gallium/drivers/softpipe/libllvmpipe.la', needed by `pipe_swrast.la'.  Stop.

10 years agoi965: Fix BRW_BATCH_STRUCT to specify RENDER_RING, not UNKNOWN_RING.
Kenneth Graunke [Wed, 27 Nov 2013 00:32:15 +0000 (16:32 -0800)]
i965: Fix BRW_BATCH_STRUCT to specify RENDER_RING, not UNKNOWN_RING.

I missed this in the boolean -> enum conversion.  C cheerfully casts
false -> 0 -> UNKNOWN_RING.  On Gen4-5, this causes the render ring
prelude hook to get called in the middle of the batch, which is crazy.

BRW_BATCH_STRUCT is not used on Gen6+.

Fixes regressions since 395a32717df494353703f3581edcd3ba380f16d6
("i965: Introduce an UNKNOWN_RING state.").

Fixes "fips -v glxgears" on Ironlake.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoRevert "i965: Move brw_emit_query_begin() to the render ring prelude."
Kenneth Graunke [Wed, 27 Nov 2013 00:32:14 +0000 (16:32 -0800)]
Revert "i965: Move brw_emit_query_begin() to the render ring prelude."

This reverts commit a4bf7f6b6e612626c4e4fc21507ac213a7ba4b00.
It breaks occlusion queries on Gen4-5.  Doing this right will likely
require larger changes, which should be done at a future date.

Some Piglit tests still passed due to other bugs; fixing those revealed
this problem.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agoi965: Fix OACONTROL assertion failures on Ironlake.
Kenneth Graunke [Wed, 27 Nov 2013 00:32:13 +0000 (16:32 -0800)]
i965: Fix OACONTROL assertion failures on Ironlake.

I guarded half of the callers to start/stop_oa_counters with generation
checks, but missed the other half (which were added later).  OACONTROL
doesn't exist on Ironlake, so we better not write it.  Also, there's no
need---Ironlake's performance counters are always running.

This patch moves the generation checks into start/stop_oa_counters,
rather than requiring the caller to do them.

Fixes assertion failures in Piglit's AMD_performance_monitor/measure.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
10 years agogallium/radeon: use PRIu64 macro for printing uint64_t
Emil Velikov [Sat, 30 Nov 2013 19:53:53 +0000 (19:53 +0000)]
gallium/radeon: use PRIu64 macro for printing uint64_t

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
10 years agopipe-loader: build llvmpipe on top of softpipe
Emil Velikov [Sun, 17 Nov 2013 20:07:33 +0000 (20:07 +0000)]
pipe-loader: build llvmpipe on top of softpipe

One can select if they want to fallback to softpipe.
Current approach makes this not possible, whereas other
targets (dri-swrast) handle this approapriately.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
10 years agomesa: resolve typo DTXn/DXTn
Emil Velikov [Sun, 17 Nov 2013 17:06:23 +0000 (17:06 +0000)]
mesa: resolve typo DTXn/DXTn

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
10 years agoautomake: include only one copy VERSION in tarball
Emil Velikov [Mon, 2 Dec 2013 19:42:51 +0000 (19:42 +0000)]
automake: include only one copy VERSION in tarball

The VERSION file is tracked by git (git ls-files), thus
adding it to EXTRA_FILES will result in a duplicate copy
within the final tarball.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72230
Cc: "10.0" <mesa-stable@lists.freedesktop.org>
Reported-by: Patrick Steinhardt <ps@pks.im>
Tested-by: Patrick Steinhardt <ps@pks.im>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>