mesa.git
5 years agonir: add a vectorization pass
Connor Abbott [Sun, 15 Nov 2015 01:26:47 +0000 (20:26 -0500)]
nir: add a vectorization pass

This effectively does the opposite of nir_lower_alus_to_scalar, trying
to combine per-component ALU operations with the same sources but
different swizzles into one larger ALU operation. It uses a similar
model as CSE, where we do a depth-first approach and keep around a hash
set of instructions to be combined, but there are a few major
differences:

1. For now, we only support entirely per-component ALU operations.
2. Since it's not always guaranteed that we'll be able to combine
equivalent instructions, we keep a stack of equivalent instructions
around, trying to combine new instructions with instructions on the
stack.

The pass isn't comprehensive by far; it can't handle operations where
some of the sources are per-component and others aren't, and it can't
handle phi nodes. But it should handle the more common cases, and it
should be reasonably efficient.

[Alyssa: Rebase on latest master, updating with respect to typeless
moves]

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
5 years agopanfrost: Add support for TXS instructions
Boris Brezillon [Mon, 17 Jun 2019 20:13:04 +0000 (22:13 +0200)]
panfrost: Add support for TXS instructions

This patch adds support for nir_texop_txs instructions which are needed
to support the OpenGL textureSize() function. This is also needed to
support RECT texture sampling which is currently lowered to 2D sampling +
a TXS() instruction by the nir_lower_tex() helper.

Changes in v2:
* Split options for the 1st and 2nd tex lowering passes

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Prepare things to support non-native texture ops
Boris Brezillon [Mon, 17 Jun 2019 19:47:46 +0000 (21:47 +0200)]
panfrost: Prepare things to support non-native texture ops

We are about to add support for the TXS (texture size) op which is not
implemented using a midgard texture instruction. Let's rename emit_tex()
into emit_texop_native() and repurpose emit_tex() as a dispatcher.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Move sysval upload logic out of panfrost_emit_for_draw()
Boris Brezillon [Fri, 14 Jun 2019 08:41:17 +0000 (10:41 +0200)]
panfrost: Move sysval upload logic out of panfrost_emit_for_draw()

We're about to add more sysval types, and panfrost_emit_for_draw()
is big enough, so let's move the sysval upload logic in a separate
function.

We also add one sub-function per sysval type to keep the
panfrost_upload_sysvals() small/readable.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Make the sysval logic more generic
Boris Brezillon [Fri, 14 Jun 2019 07:59:20 +0000 (09:59 +0200)]
panfrost: Make the sysval logic more generic

We are about to add support for nir_texop_txs which requires adding a
sysval/uniform containing the texture size. Let's change the
emit_sysval_read() prototype to take a nir_instr object instead of
a nir_intrinsic_instr one so we can re-use this function when emitting
a sysval for a txs instruction.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agonir/lower_tex: Add a way to lower TXS(non-0-LOD) instructions
Boris Brezillon [Mon, 17 Jun 2019 09:43:13 +0000 (11:43 +0200)]
nir/lower_tex: Add a way to lower TXS(non-0-LOD) instructions

The V3D driver has an open-coded solution for this, and we need the
same thing for Panfrost, so let's add a generic way to lower TXS(LOD)
into max(TXS(0) >> LOD, 1).

Changes in v2:
* Use == 0 instead of !
* Rework the minification logic as suggested by Jason
* Assign cursor pos at the beginning of the function
* Patch the LOD just after retrieving the old value

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agonir/lower_tex: Update ->sampler_dim value before calling get_texture_size()
Boris Brezillon [Mon, 17 Jun 2019 09:31:51 +0000 (11:31 +0200)]
nir/lower_tex: Update ->sampler_dim value before calling get_texture_size()

get_texture_size() will create a txs instruction with ->sampler_dim set
to the original tex->sampler_dim. The condition to call lower_rect()
only checks the value of ->sampler_dim and whether lower_rect is
requested or not. This leads to an infinite loop when calling
nir_lower_tex() with the same options until it returns false.

In order to avoid that, let's move the tex->sampler_dim patching before
get_texture_size() is called. This way the txs instruction will have
->sampler_dim set to GLSL_SAMPLER_DIM_2D and nir_lower_tex() won't try
to lower it on the subsequent passes.

Changes in v2:
* Add Jason R-b
* Add a comment explaining why we patch ->sampler_dim at the beginning
  of the lower_rect() func

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agonir/lower_tex: Actually report when projector lowering happened
Boris Brezillon [Mon, 17 Jun 2019 09:23:33 +0000 (11:23 +0200)]
nir/lower_tex: Actually report when projector lowering happened

The code considers that projector lowering was done even if it's not
really the case. Change the project_src() prototype to return a bool
encoding whether projector lowering happened or not and update the
progress var accordingly in nir_lower_tex_block().

---
Changes in v2:
* Add Jason R-b
* Drop the part suggesting that nir_lower_rect() could be called in
  a do-while(progress) loop.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Adapt to constant name change in UABI
Tomeu Vizoso [Fri, 31 May 2019 07:12:59 +0000 (09:12 +0200)]
panfrost: Adapt to constant name change in UABI

We hadn't updated the kernel header after the driver got into mainline.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: ci: Update results
Tomeu Vizoso [Tue, 18 Jun 2019 13:15:19 +0000 (15:15 +0200)]
panfrost: ci: Update results

Alyssa fixed some failing tests last night.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agoradv: adjust the DCC base VA for mipmapped color attachments
Samuel Pitoiset [Tue, 18 Jun 2019 09:51:31 +0000 (11:51 +0200)]
radv: adjust the DCC base VA for mipmapped color attachments

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: fix color decompressions for FMASK/CMASK
Samuel Pitoiset [Tue, 18 Jun 2019 10:02:12 +0000 (12:02 +0200)]
radv: fix color decompressions for FMASK/CMASK

Only skip levels without DCC when it's a DCC decompression.
Whoops.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: do not decompress levels without DCC with the graphics path
Samuel Pitoiset [Tue, 18 Jun 2019 08:30:45 +0000 (10:30 +0200)]
radv: do not decompress levels without DCC with the graphics path

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: do not decompress levels without DCC with the compute path
Samuel Pitoiset [Tue, 18 Jun 2019 08:30:44 +0000 (10:30 +0200)]
radv: do not decompress levels without DCC with the compute path

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: check if DCC is enabled per mip not for the whole image
Samuel Pitoiset [Tue, 18 Jun 2019 08:30:43 +0000 (10:30 +0200)]
radv: check if DCC is enabled per mip not for the whole image

In other words, make use of radv_dcc_enabled() instead of
radv_image_has_dcc() all over the places.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agov3d: implement simultaneous peripheral access exceptions for V3D 4.1+
Iago Toral Quiroga [Mon, 17 Jun 2019 08:15:54 +0000 (10:15 +0200)]
v3d: implement simultaneous peripheral access exceptions for V3D 4.1+

Shader-db results:

total instructions in shared programs: 9117550 -> 9102719 (-0.16%)
instructions in affected programs: 1752873 -> 1738042 (-0.85%)
helped: 7076
HURT: 478
helped stats (abs) min: 1 max: 22 x̄: 2.19 x̃: 2
helped stats (rel) min: 0.07% max: 13.89% x̄: 1.70% x̃: 1.07%
HURT stats (abs)   min: 1 max: 7 x̄: 1.41 x̃: 1
HURT stats (rel)   min: 0.09% max: 10.17% x̄: 0.86% x̃: 0.54%
95% mean confidence interval for instructions value: -2.00 -1.92
95% mean confidence interval for instructions %-change: -1.58% -1.50%
Instructions are helped.

total max-temps in shared programs: 1327774 -> 1327728 (<.01%)
max-temps in affected programs: 1025 -> 979 (-4.49%)
helped: 47
HURT: 2
helped stats (abs) min: 1 max: 2 x̄: 1.02 x̃: 1
helped stats (rel) min: 2.63% max: 20.00% x̄: 7.67% x̃: 5.26%
HURT stats (abs)   min: 1 max: 1 x̄: 1.00 x̃: 1
HURT stats (rel)   min: 4.17% max: 4.17% x̄: 4.17% x̃: 4.17%
95% mean confidence interval for max-temps value: -1.06 -0.82
95% mean confidence interval for max-temps %-change: -8.89% -5.49%
Max-temps are helped.

Reviewed-by: Eric Anholt <eric@anholt.net>
5 years agov3d: only flush jobs accessing the query BO when reading query results
Iago Toral Quiroga [Mon, 17 Jun 2019 06:21:32 +0000 (08:21 +0200)]
v3d: only flush jobs accessing the query BO when reading query results

Reviewed-by: Eric Anholt <eric@anholt.net>
5 years agov3d: add a helper function to flush jobs using a BO
Iago Toral Quiroga [Fri, 14 Jun 2019 10:06:25 +0000 (12:06 +0200)]
v3d: add a helper function to flush jobs using a BO

v2: use _mesa_set_search() (Eric)

Reviewed-by: Eric Anholt <eric@anholt.net>
5 years agoiris: Support more RGBX pipe formats.
Kenneth Graunke [Sun, 9 Jun 2019 00:17:20 +0000 (17:17 -0700)]
iris: Support more RGBX pipe formats.

Without them, the state tracker falls back to an RGBA format, but it
doesn't always manage to override the swizzle for us.  So we lose the
information that the API expects an X channel, where alpha is garbage
and reads back as 1.  We have no equivalent ISL RGBX format for these,
so we just use RGBA directly and override the swizzle in all cases.

5 years agoglsl: Fix out of bounds read in shader_cache_read_program_metadata
Kenneth Graunke [Sat, 8 Jun 2019 06:00:40 +0000 (23:00 -0700)]
glsl: Fix out of bounds read in shader_cache_read_program_metadata

The VaryingNames array has NumVaryings entries.  But BufferStride is
a small array of MAX_FEEDBACK_BUFFERS (4) entries.  Programs with
more than 4 varyings would read out of bounds.

Also, BufferStride is set based on the shader itself, which means that
it's inherently already included in the hash, and doesn't need to be
included again.  At the point when shader_cache_read_program_metadata
is called, the linker hasn't even set those fields yet.  So, just drop
it entirely.

Fixes valgrind errors in KHR-GL45.transform_feedback.linking_errors_test.

Fixes: 6d830940f78 glsl/shader_cache: Allow shader cache usage with transform feedback
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
5 years agoanv: Set STATE_BASE_ADDRESS upper bounds on gen7
Jason Ekstrand [Mon, 17 Jun 2019 22:01:48 +0000 (17:01 -0500)]
anv: Set STATE_BASE_ADDRESS upper bounds on gen7

This should fix floating-point border color on all gen7 HW.  Integer is
still thoroughly busted on gen7 because it doesn't exist on IVB and it's
crazy on HSW.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
5 years agoradv: Disable linear tiled compressed textures.
Bas Nieuwenhuizen [Mon, 17 Jun 2019 19:46:35 +0000 (21:46 +0200)]
radv: Disable linear tiled compressed textures.

Support got removed in the new addrlib update.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
5 years agoanv:Use VK_EXT_separate_stencil_usage to avoid stencil shadows on gen7
Jason Ekstrand [Mon, 17 Jun 2019 14:39:08 +0000 (09:39 -0500)]
anv:Use VK_EXT_separate_stencil_usage to avoid stencil shadows on gen7

Whenever stencil texturing is not required (most of the time), we can
use VK_EXT_separate_stencil_usage to only create the shadow image when
VK_IMAGE_USAGE_SAMPLED_BIT is required for stencil.  Of course, this
depends on applications to use the extension but hopefully DXVK and
similar translators are doing so and that covers most of the apps.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoanv: Add stencil texturing support for gen7
Jason Ekstrand [Mon, 17 Jun 2019 02:21:16 +0000 (21:21 -0500)]
anv: Add stencil texturing support for gen7

Intel hardware didn't get support for sampling from W-tiled (required
for stencil) images until Broadwell so we can't directly sample from
stencil.  Instead, if we want to support stencil texturing on gen7
hardware, we have to keep a texture-capable shadow copy around and use
BLORP to update when stencil changes.  The one thing this commit does
not implement is self-dependencies with stencil input attachments.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99493
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoanv/blorp: Update shadow images when clearing or uploading
Jason Ekstrand [Mon, 17 Jun 2019 06:53:50 +0000 (01:53 -0500)]
anv/blorp: Update shadow images when clearing or uploading

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoanv/cmd_buffer: Add a stencil transition helper
Jason Ekstrand [Mon, 17 Jun 2019 02:55:25 +0000 (21:55 -0500)]
anv/cmd_buffer: Add a stencil transition helper

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoanv/blorp: Take an aspect in anv_image_copy_to_shadow
Jason Ekstrand [Mon, 17 Jun 2019 02:36:21 +0000 (21:36 -0500)]
anv/blorp: Take an aspect in anv_image_copy_to_shadow

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoanv/formats: Re-arrange the way se set some flag bits
Jason Ekstrand [Mon, 17 Jun 2019 02:20:41 +0000 (21:20 -0500)]
anv/formats: Re-arrange the way se set some flag bits

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
5 years agoiris: Make resource_copy_region handle packed depth-stencil resources.
Kenneth Graunke [Mon, 17 Jun 2019 21:35:31 +0000 (16:35 -0500)]
iris: Make resource_copy_region handle packed depth-stencil resources.

Also copy along the separate stencil buffer if needed.

Fixes Piglit's arb_copy_image-formats.

5 years agoiris: Order CS stall and TC invalidate for format reinterpretation hacks
Kenneth Graunke [Mon, 17 Jun 2019 11:55:07 +0000 (06:55 -0500)]
iris: Order CS stall and TC invalidate for format reinterpretation hacks

This should ensure the TC invalidate happens after the stall.

Fixes KHR-GL43.copy_image.functional which does a CopyImage (blorp_copy)
from a buffer (using R8G8B8A8_UINT), then GetTexImage to read back the
original image (using R10G10B10A2_UNORM).

5 years agoiris: Be more aggressive at post-format-reintepret TC invalidate hack
Kenneth Graunke [Mon, 17 Jun 2019 14:39:46 +0000 (09:39 -0500)]
iris: Be more aggressive at post-format-reintepret TC invalidate hack

When copying/blitting with format reinterpretation, we invalidate the
texture cache before/after.  Before is so the source of the copy works,
and after is to get rid of our new data in the "wrong" format to protect
future attempts to sample.

When I ported these hacks to iris, I tried to be cautious by only
bothering with the hacks if the batch referenced the BO.  This makes
some sense for the before case.  If it isn't referenced, the texture
cache can't really have any data for the BO (since it's also invalidated
between batches).  But we still need to do the after case regardless,
as we've just polluted the cache with hazardous entries.

5 years agovirgl: Assume sRGB write control for older guest kernels or virglrenderer hosts
Gert Wollny [Mon, 17 Jun 2019 06:44:14 +0000 (08:44 +0200)]
virgl: Assume sRGB write control for older guest kernels or virglrenderer hosts

When the host virglrenderer is an older version that doesn't check the sRGB write
control feature, or when the guest kernel doesn't support CAPS v2, then the guest
will only report support for GL 2.1 on a GL 3.3 host, even though it was supporting
3.3 with earlier guest mesa versions.

By also checking the host feature check version this regression can be avoided.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110921
Fixes: 2845939d6a72
   virgl: Set sRGB write control CAP based on host capabilities

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
5 years agofreedreno/a6xx: disallow UBWC for x24s8
Rob Clark [Fri, 14 Jun 2019 16:12:46 +0000 (09:12 -0700)]
freedreno/a6xx: disallow UBWC for x24s8

Fixes:
  dEQP-GLES31.functional.stencil_texturing.format.depth24_stencil8_2d
  dEQP-GLES31.functional.stencil_texturing.format.stencil_index8_2d
  dEQP-GLES31.functional.stencil_texturing.misc.compare_mode_effect

Signed-off-by: Rob Clark <robdclark@chromium.org>
5 years agofreedreno/a6xx: un-swap X24S8_UINT
Rob Clark [Thu, 13 Jun 2019 18:58:30 +0000 (11:58 -0700)]
freedreno/a6xx: un-swap X24S8_UINT

The stencil is actually in the .w component, but we used to use SWAP to
remap the channels.  This doesn't work when tiled/ubwc.

Fixes:
  dEQP-GLES31.functional.stencil_texturing.format.depth24_stencil8_2d_array
  dEQP-GLES31.functional.stencil_texturing.format.depth24_stencil8_cube
  dEQP-GLES31.functional.stencil_texturing.format.stencil_index8_2d_array
  dEQP-GLES31.functional.stencil_texturing.format.stencil_index8_cube
  dEQP-GLES31.functional.stencil_texturing.misc.base_level
  dEQP-GLES31.functional.texture.border_clamp.formats.stencil_index8.nearest_size_pot
  dEQP-GLES31.functional.texture.border_clamp.formats.stencil_index8.nearest_size_npot
  dEQP-GLES31.functional.texture.border_clamp.formats.depth24_stencil8_sample_stencil.nearest_size_pot
  dEQP-GLES31.functional.texture.border_clamp.formats.depth24_stencil8_sample_stencil.nearest_size_npot
  dEQP-GLES31.functional.texture.border_clamp.sampler.uint_stencil

Signed-off-by: Rob Clark <robdclark@chromium.org>
5 years agoradv: add mipmaps support for DCC decompression on compute
Samuel Pitoiset [Mon, 17 Jun 2019 08:53:24 +0000 (10:53 +0200)]
radv: add mipmaps support for DCC decompression on compute

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: add mipmaps support for color decompressions (DCC/FMASK/CMASK)
Samuel Pitoiset [Fri, 14 Jun 2019 07:21:58 +0000 (09:21 +0200)]
radv: add mipmaps support for color decompressions (DCC/FMASK/CMASK)

And some cleanups.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: set the DCC/FCE predicates from the base level
Samuel Pitoiset [Fri, 14 Jun 2019 13:17:06 +0000 (15:17 +0200)]
radv: set the DCC/FCE predicates from the base level

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: load the fast color clear values from the base level
Samuel Pitoiset [Fri, 14 Jun 2019 08:07:27 +0000 (10:07 +0200)]
radv: load the fast color clear values from the base level

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: store the DCC predicate for each mip
Samuel Pitoiset [Fri, 14 Jun 2019 13:15:09 +0000 (15:15 +0200)]
radv: store the DCC predicate for each mip

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: store the FCE predicate for each mip
Samuel Pitoiset [Fri, 14 Jun 2019 13:07:24 +0000 (15:07 +0200)]
radv: store the FCE predicate for each mip

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: store the fast color clear values for each mip
Samuel Pitoiset [Fri, 14 Jun 2019 08:21:56 +0000 (10:21 +0200)]
radv: store the fast color clear values for each mip

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: allocate DCC metadata for each mip
Samuel Pitoiset [Fri, 14 Jun 2019 12:52:28 +0000 (14:52 +0200)]
radv: allocate DCC metadata for each mip

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agogallium: Remove unused util_ringbuffer
Caio Marcelo de Oliveira Filho [Wed, 12 Jun 2019 23:14:52 +0000 (16:14 -0700)]
gallium: Remove unused util_ringbuffer

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
5 years agollvmpipe: Don't use u_ringbuffer for lp_scene_queue
Caio Marcelo de Oliveira Filho [Wed, 12 Jun 2019 22:32:30 +0000 (15:32 -0700)]
llvmpipe: Don't use u_ringbuffer for lp_scene_queue

Inline the ring buffer and signal logic into lp_scene_queue instead of
using a u_ringbuffer.  The code ends up simpler since there's no need
to handle serializing data from / to packets.

This fixes a crash when compiling Mesa with LTO, that happened because
of util_ringbuffer_dequeue() was writing data after the "header
packet", as shown below

    struct scene_packet {
       struct util_packet header;
       struct lp_scene *scene;
    };

    /* Snippet of old lp_scene_deque(). */
    packet.scene = NULL;
    ret = util_ringbuffer_dequeue(queue->ring,
                                  &packet.header,
                                  sizeof packet / 4,
    return packet.scene;

but due to the way aliasing analysis work the compiler didn't
considered the "&packet->header" to alias with "packet->scene".  With
the aggressive inlining done by LTO, this would end up always
returning NULL instead of the content read by
util_ringbuffer_dequeue().

Issue found by Marco Simental and iThiago Macieira.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110884
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
5 years agopanfrost/midgard: Simplify 2D array logic
Alyssa Rosenzweig [Mon, 17 Jun 2019 19:41:41 +0000 (12:41 -0700)]
panfrost/midgard: Simplify 2D array logic

It shouldn't matter if we stick a z in for non-arrays, anyway.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Handle non-zero component in store
Alyssa Rosenzweig [Mon, 17 Jun 2019 19:35:57 +0000 (12:35 -0700)]
panfrost/midgard: Handle non-zero component in store

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Apply writemask to LUTs
Alyssa Rosenzweig [Mon, 17 Jun 2019 18:49:44 +0000 (11:49 -0700)]
panfrost/midgard: Apply writemask to LUTs

Fixes LUT instructions with NIR registers.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agoamd: update addrlib
Marek Olšák [Fri, 14 Jun 2019 21:55:38 +0000 (17:55 -0400)]
amd: update addrlib

Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradeonsi: reduce MAX_GEOMETRY_OUTPUT_VERTICES
Nicolai Hähnle [Tue, 19 Jun 2018 11:53:01 +0000 (13:53 +0200)]
radeonsi: reduce MAX_GEOMETRY_OUTPUT_VERTICES

This fixes piglit spec@glsl-1.50@gs-max-output on gfx9.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agopanfrost: Cleanup default blend mode
Alyssa Rosenzweig [Mon, 17 Jun 2019 17:22:37 +0000 (10:22 -0700)]
panfrost: Cleanup default blend mode

Just encode the Mali magic number for `replace` rather than awkwardly
forcing Gallium structures through.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Don't accidentally include blend shader
Alyssa Rosenzweig [Mon, 17 Jun 2019 17:08:47 +0000 (10:08 -0700)]
panfrost: Don't accidentally include blend shader

Some residual dirty state can leak through across frames; zero this out.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Use typeless moves internally
Alyssa Rosenzweig [Mon, 17 Jun 2019 16:40:14 +0000 (09:40 -0700)]
panfrost/midgard: Use typeless moves internally

We switch all fmov to (i)mov, following the NIR switch. This simplifies
some code surrounding blend shaders and should have no functional
changes elsewhere.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agovirgl: better support for PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE
Chia-I Wu [Fri, 10 May 2019 04:44:33 +0000 (21:44 -0700)]
virgl: better support for PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE

When the resource to be mapped is busy and the backing storage can
be discarded, reallocate the backing storage to avoid waiting.

In this new path, we allocate a new buffer, emit a state change,
write, and add the transfer to the queue .  In the
PIPE_TRANSFER_DISCARD_RANGE path, we suballocate a staging buffer,
write, and emit a copy_transfer (which may allocate, memcpy, and
blit internally).  The win might not always be clear.  But another
win comes from that the new path clears res->valid_buffer_range and
does not clear res->clean_mask.  This makes it much more preferable
in scenarios such as

  access = enough_space ? GL_MAP_UNSYNCHRONIZED_BIT :
                          GL_MAP_INVALIDATE_BUFFER_BIT;
  glMapBufferRange(..., GL_MAP_WRITE_BIT | access);
  memcpy(...); // append new data
  glUnmapBuffer(...);

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
5 years agovirgl: add virgl_rebind_resource
Chia-I Wu [Thu, 16 May 2019 22:42:01 +0000 (15:42 -0700)]
virgl: add virgl_rebind_resource

We are going support reallocating the HW resource for a
virgl_resource.  When that happens, the virgl_resource needs to be
rebound to the context.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
5 years agovirgl: save virgl_hw_res in virgl_transfer
Chia-I Wu [Wed, 15 May 2019 22:34:44 +0000 (15:34 -0700)]
virgl: save virgl_hw_res in virgl_transfer

When PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE is properly supported,
virgl_transfer might refer to a different virgl_hw_res than
virgl_resource does.  We need to save the virgl_hw_res and use the
saved one.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
5 years agovirgl: add resource_reference to virgl_winsys
Chia-I Wu [Wed, 15 May 2019 22:28:52 +0000 (15:28 -0700)]
virgl: add resource_reference to virgl_winsys

It works similar to pipe_resource_reference but is for virgl_hw_res.
It can also replace resource_unref.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
5 years agopanfrost/midgard: Add rounding mode specific opcodes
Alyssa Rosenzweig [Wed, 5 Jun 2019 22:03:02 +0000 (15:03 -0700)]
panfrost/midgard: Add rounding mode specific opcodes

This adds a set of opcodes for performing moves and type conversions
with respect to particular rounding modes, required for OpenCL.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Drop draws with complete scissor
Alyssa Rosenzweig [Mon, 17 Jun 2019 16:26:34 +0000 (09:26 -0700)]
panfrost: Drop draws with complete scissor

The hardware support for scissoring requires minimally 1 pixel to be
drawn. If the scissor culls *everything*, we need to drop the draw
entirely early on.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Disable pipelining temporarily
Alyssa Rosenzweig [Mon, 17 Jun 2019 16:25:52 +0000 (09:25 -0700)]
panfrost: Disable pipelining temporarily

Pipelined rendering is important for performance but is not working
right these days. Disable it for correctness until the panfrost_job
refactor is enabled and we can do it right.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/mfbd: Handle rendering to linear mipmap
Alyssa Rosenzweig [Wed, 12 Jun 2019 22:23:19 +0000 (15:23 -0700)]
panfrost/mfbd: Handle rendering to linear mipmap

In anticipation of more general mipmapping support, we implemented
support for rendering to linear mipmaps (a very simple case).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Implement sampling from non-zero initial levels
Alyssa Rosenzweig [Wed, 12 Jun 2019 22:07:09 +0000 (15:07 -0700)]
panfrost: Implement sampling from non-zero initial levels

In preparation for more complex mipmap operations. glGenerateMipmap() in
particular, as implemented by u_blitter, requires reading from non-zero
initial mip levels.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Resource management for linear 2D texture arrays
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:31:14 +0000 (16:31 -0700)]
panfrost: Resource management for linear 2D texture arrays

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Adjust swizzles for 2D arrays
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:30:29 +0000 (16:30 -0700)]
panfrost/midgard: Adjust swizzles for 2D arrays

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Set array_size to permit array textures
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:26:49 +0000 (16:26 -0700)]
panfrost: Set array_size to permit array textures

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Decode array textures
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:23:32 +0000 (16:23 -0700)]
panfrost: Decode array textures

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Implement 3D texture resource management
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:11:56 +0000 (16:11 -0700)]
panfrost: Implement 3D texture resource management

Passes dEQP-GLES3.functional.texture.format.unsized.*3d*

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Specify 3D in texture descriptor
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:07:15 +0000 (16:07 -0700)]
panfrost: Specify 3D in texture descriptor

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Fix 3D texture masks/swizzles
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:03:39 +0000 (16:03 -0700)]
panfrost/midgard: Fix 3D texture masks/swizzles

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost/midgard: Add swizzle_of/mask_of helpers
Alyssa Rosenzweig [Fri, 14 Jun 2019 23:03:01 +0000 (16:03 -0700)]
panfrost/midgard: Add swizzle_of/mask_of helpers

These make manipulating vectors in the Midgard compiler easier.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Enable helper invocations when texturing
Alyssa Rosenzweig [Fri, 7 Jun 2019 23:00:49 +0000 (16:00 -0700)]
panfrost: Enable helper invocations when texturing

it turns out we have explicit control over helper invocations; if a
particular bit in the fragment shader descriptor is set, helper
invocations are launched; if it clear, they are not. Helper invocations
are required whenever computing derivatives, whether explicitly
(dFdx/dFdy) *or* implicitly (any texturing). Accordingly, we set this
bit when texturing to fix edge case behaviour (literally, haha).

Thank you to Jason Ekstrand and Ilia Mirkin for pointing out the
representative dEQP test failed along triangle edges and for suggesting
helper invocations / derivatives as a list of suspect pieces (which led
to discovering the helper invocations enable bit in the first place).

Ideally we would use the new NIR analysis pass for this, but that hasn't
landed quite yet.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Handle missing texture case
Alyssa Rosenzweig [Tue, 11 Jun 2019 21:21:14 +0000 (14:21 -0700)]
panfrost: Handle missing texture case

In some cases, Gallium can give us bad info about the texture count,
counting some NULL textures. We pass Gallium's info to the hardware
blindly, which can confuse the hardware in edge cases. This patch
adjusts accordingly.

5 years agopanfrost: Remove forced flush on clears
Alyssa Rosenzweig [Fri, 14 Jun 2019 19:48:06 +0000 (12:48 -0700)]
panfrost: Remove forced flush on clears

This worked around a bug in oooold versions of Panfrost. Nowadays, its
presence is, at best, *creating* bugs. Let's wack it.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Flush scanout too
Alyssa Rosenzweig [Fri, 14 Jun 2019 19:26:19 +0000 (12:26 -0700)]
panfrost: Flush scanout too

In a poorly coded app, the framebuffer can be partially drawn, an FBO
switched, switch back to the framebuffer and keep drawing, etc.
Reordering would fix this, but for now we need to just be careful about
flushing scanout too.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Improve viewport (clipping) robustness
Alyssa Rosenzweig [Fri, 14 Jun 2019 19:25:26 +0000 (12:25 -0700)]
panfrost: Improve viewport (clipping) robustness

On more complex apps (possibly using desktop GL specific extensions?),
our viewport code was getting wacky results for unclear reasons. Let's
be a little less wacky.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Disable the tiler for clear-only jobs
Alyssa Rosenzweig [Fri, 14 Jun 2019 18:23:24 +0000 (11:23 -0700)]
panfrost: Disable the tiler for clear-only jobs

To do so, we route some basic information through to the FBD creation
routines (currently just a binary toggle of "has draws?"). Eventually,
more refactoring will enable dynamic hierarchy mask selection, but right
now we do the most basic.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Identify and decode mfbd_flags
Alyssa Rosenzweig [Fri, 14 Jun 2019 18:14:01 +0000 (11:14 -0700)]
panfrost: Identify and decode mfbd_flags

Previously known as the unk3 field.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Stub out hierarchy mask selection
Alyssa Rosenzweig [Fri, 14 Jun 2019 15:14:04 +0000 (08:14 -0700)]
panfrost: Stub out hierarchy mask selection

Quite a bit of refactoring in the main driver will be necessary to make
use of this effectively, so the implementation is incomplete.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Rename misc_0 -> tiler_polygon_list
Alyssa Rosenzweig [Fri, 14 Jun 2019 14:24:26 +0000 (07:24 -0700)]
panfrost: Rename misc_0 -> tiler_polygon_list

Just for readability.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Sanity check tiler polygon list size
Alyssa Rosenzweig [Fri, 14 Jun 2019 14:21:05 +0000 (07:21 -0700)]
panfrost: Sanity check tiler polygon list size

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Compute and use polygon list body size
Alyssa Rosenzweig [Fri, 14 Jun 2019 14:08:51 +0000 (07:08 -0700)]
panfrost: Compute and use polygon list body size

This is a bit of a hack, but it gets the point across.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Use polygon list header size computation
Alyssa Rosenzweig [Thu, 13 Jun 2019 22:15:53 +0000 (15:15 -0700)]
panfrost: Use polygon list header size computation

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Calculate polygon list header size
Alyssa Rosenzweig [Thu, 13 Jun 2019 22:11:21 +0000 (15:11 -0700)]
panfrost: Calculate polygon list header size

As per the notes at the beginning of pan_tiler.c, we implement a routine
to calculate the size of the polygon list header given the framebuffer
dimensions and the provided hierarchy mask.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Add pan_tiler.h header
Alyssa Rosenzweig [Fri, 14 Jun 2019 15:14:53 +0000 (08:14 -0700)]
panfrost: Add pan_tiler.h header

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Document tile size heuristic
Alyssa Rosenzweig [Thu, 13 Jun 2019 18:25:18 +0000 (11:25 -0700)]
panfrost: Document tile size heuristic

I'm not sure how the blob does it, but this seems to be a dead simple
test and roughly corresponds to what I've noticed from the blob, so
maybe it's good enough.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Rename tiler fields per tiler research
Alyssa Rosenzweig [Thu, 13 Jun 2019 17:25:32 +0000 (10:25 -0700)]
panfrost: Rename tiler fields per tiler research

Following the research into Midgard's hierarchical tiling
infrastructure, we now understand (in broad stokes) the purpose of each
tiler field in the MFBD. Additionally, we understand more of the tiling
fields in the SFBD and in Bifrost's structures, although this knowledge
is still incomplete.

Update the names, decoder, and comments to reflect this new
understanding.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Add notes about the tiler allocations
Alyssa Rosenzweig [Thu, 13 Jun 2019 16:40:41 +0000 (09:40 -0700)]
panfrost: Add notes about the tiler allocations

This explains how the polygon list is allocated, updating the headers
appropiately to sync the terminology.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agopanfrost: Integrate kernel names for tiler FBD
Alyssa Rosenzweig [Wed, 12 Jun 2019 16:33:06 +0000 (09:33 -0700)]
panfrost: Integrate kernel names for tiler FBD

These names are from the replay workaround in kbase; they begin to shine
some light on the meaning of these fields. In particular, we now
understand why the "tiler_meta" field has the effect it does on
performance in certain scenes (controlling tile granularity).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
5 years agoradv: Add asserts that buffer descriptors are created with valid buffer formats.
Bas Nieuwenhuizen [Sat, 15 Jun 2019 14:10:22 +0000 (16:10 +0200)]
radv: Add asserts that buffer descriptors are created with valid buffer formats.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
5 years agoradv: Decompress DCC when the image format is not allowed for buffers.
Bas Nieuwenhuizen [Sat, 15 Jun 2019 16:05:05 +0000 (18:05 +0200)]
radv: Decompress DCC when the image format is not allowed for buffers.

Otherwise the buffer loads/stores in the bufimage meta operations fail.

If we decompress DCC then we can use the "canonical" format compatible
with the not-supported format.

CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
5 years agoradv: make sure to init the DCC decompress compute path state
Samuel Pitoiset [Thu, 13 Jun 2019 10:44:03 +0000 (12:44 +0200)]
radv: make sure to init the DCC decompress compute path state

This fixes a segfault when forcing DCC decompressions on compute
because internal meta objects are not created since the on-demand
stuff.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoac: make ac_compute_cmask() a static function
Samuel Pitoiset [Thu, 13 Jun 2019 15:17:23 +0000 (17:17 +0200)]
ac: make ac_compute_cmask() a static function

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-By: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: rely on ac_compute_cmask() for CMASK info
Samuel Pitoiset [Thu, 13 Jun 2019 15:17:22 +0000 (17:17 +0200)]
radv: rely on ac_compute_cmask() for CMASK info

Instead of re-computing in the driver. The 3d and cube flags
are correctly set, so the same values should returned by
ac_compute_surface().

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agoradv: silent a compiler warning in radv_CmdPushDescriptorSetKHR()
Samuel Pitoiset [Mon, 17 Jun 2019 07:53:26 +0000 (09:53 +0200)]
radv: silent a compiler warning in radv_CmdPushDescriptorSetKHR()

Trivial.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
5 years agopanfrost: ci: Speed things up a bit by skipping a git clone
Tomeu Vizoso [Mon, 17 Jun 2019 05:56:00 +0000 (07:56 +0200)]
panfrost: ci: Speed things up a bit by skipping a git clone

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
5 years agopanfrost: ci: Exclude all blend tests from results
Tomeu Vizoso [Thu, 13 Jun 2019 18:34:42 +0000 (20:34 +0200)]
panfrost: ci: Exclude all blend tests from results

As they randomly fail on T760.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
5 years agoac: update llvm.amdgcn.icmp intrinsic name for LLVM 9+
Samuel Pitoiset [Fri, 14 Jun 2019 10:00:26 +0000 (12:00 +0200)]
ac: update llvm.amdgcn.icmp intrinsic name for LLVM 9+

LLVM r363339 changed llvm.amdgcn.icmp.i* to llvm.amdgcn.icmp.i64.i*.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-By: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
5 years agolima: lower fmod in ppir and gpir
Erico Nunes [Sat, 15 Jun 2019 13:41:05 +0000 (15:41 +0200)]
lima: lower fmod in ppir and gpir

Since commit 4f3c82c72c5 fmod is no longer being lowered in nir, and
ends up crashing lima programs with "unsupported nir_op: fmod" in both
ppir and gpir.
There seems to be no mod operation in hardware in utgard and there is an
optimization in nir to lower fmod to instructions that lima already
implements, so let's use that.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
5 years agofreedreno/a6xx: re-enable UBWC for depth/stencil
Rob Clark [Wed, 12 Jun 2019 20:24:33 +0000 (13:24 -0700)]
freedreno/a6xx: re-enable UBWC for depth/stencil

Now that we can blit depth/stencil in a way that plays nicely with UBWC,
re-enable it.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@gmail.com>
5 years agofreedreno/a6xx: handle z24s8/z24x8 blits with u_blitter
Rob Clark [Tue, 11 Jun 2019 17:38:19 +0000 (10:38 -0700)]
freedreno/a6xx: handle z24s8/z24x8 blits with u_blitter

Now that it can turn these blits into rendering to RB6_Z24_UNORM_S8_UINT
it can properly handle cases where only one of depth+stencil is being
blit.  And this avoids lying about he format, which completely doesn't
work when UBWC is used.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@gmail.com>
5 years agofreedreno/a6xx: handle fallback for rewritten blits ourself
Rob Clark [Thu, 13 Jun 2019 17:11:57 +0000 (10:11 -0700)]
freedreno/a6xx: handle fallback for rewritten blits ourself

For re-written z/s blits, we want to use the re-written `pipe_blit_info`
even if we have to fallback to 3d pipe (`u_blitter`).  So handle that
fallback ourself.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@gmail.com>