glthread: add GL_DRAW_INDIRECT_BUFFER tracking and generator support

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: don't increment variable_data if it's the last variable-size param

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: don't insert _mesa_post_marshal_hook into every function

Let the developer decide that in the python script.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: simplify repeated function sequences in marshal_generated.c

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: use int instead of size_t where it's OK

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: reduce pointer dereferences in glthread_unmarshal_batch

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: inline _mesa_unmarshal_dispatch_cmd and convert the switch to a table

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: don't prefix variable_data with const

Not all variable data that is constant is declared with const, such as
glDeletePerfMonitorsAMD.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

glthread: don't generate the sync fallback if the call size is not variable

marshal_generated.c is 12% smaller.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3948>

docs: update news, calendar, and link release notes for 20.0.1

Also fix a couple of dates that are wrong.

Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4075>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4075>

docs: Add sha256sums for 20.0.1

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4075>

docs: add relnotes for 20.0.1

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4075>

docs: update releasing to cover updated post_version.py

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2505>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2505>

bin/post_version.py: Make the git commit as well.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2505>

bin/post_version.py: Pretty print the html

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2505>

bin/post_version.py: Update the release calendar as well

Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2505>

docs: Update release notes with current process

There's a lot of stuff here that's out of date, update it to something
more modern.

Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4066>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4066>

docs/submittingpatches: Fix confusing typo + missing pronoun

Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4066>

gitlab-ci: add a job that runs Fossilize on RADV/Polaris10

RADV_FORCE_FAMILY forces creating a null device that allows RADV
to be instanced without AMDGPU.

The Fossilize database only contains pipelines from the Sascha
Vulkan triangle demos at the moment. I will add more once this
is merged.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3960>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3960>

gitlab-ci: enable building the test image for VK unconditionally

It was diabled because RADV is the only driver that tests Vulkan
and running CTS on my personal machine and without recovery is
not safe enough for CI (too long and too unstable).

Now that we are going to test Fossilize with RADV, it's needed to
build the test image for VK unconditionally. As RADV now supports
creating NULL devices, the fossilize jobs can run everywhere.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3960>

gitlab-ci: add Fossilize support to detect compiler regressions

Fossilize is equivalent to vkpipeline-db but it's definitely more
robust. This is based on the CI traces system.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3960>

gitlab-ci: build Fossilize in the test image for VK

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3960>

aco: only reserve sgprs for vcc if it's used

pipeline-db (Vega):

Totals:
SGPRS: 5186302 -> 5075616 (-2.13 %)
VGPRS: 3704580 -> 3704580 (0.00 %)
Spilled SGPRs: 144859 -> 144859 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 4124 -> 4124 (0.00 %) dwords per thread
Code Size: 247315944 -> 247315944 (0.00 %) bytes
LDS: 1311 -> 1311 (0.00 %) blocks
Max Waves: 674560 -> 674562 (0.00 %)

Totals from affected shaders:
SGPRS: 536992 -> 426306 (-20.61 %)
VGPRS: 356404 -> 356404 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 8498748 -> 8498748 (0.00 %) bytes
LDS: 8 -> 8 (0.00 %) blocks
Max Waves: 113832 -> 113834 (0.00 %)

There are some small code size changes in a few RotTR shaders and a small
increase in max_waves in two Detroit: Become Human shaders.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3906>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3906>

aco: improve control flow handling in GFX6-9 NOP pass

Fixes Detroit: Become Human hang. Also affects World of Warships.

pipeline-db (Tahiti):
Totals from affected shaders:
SGPRS: 0 -> 0 (0.00 %)
VGPRS: 0 -> 0 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 0 -> 0 (0.00 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 0 -> 0 (0.00 %)

pipeline-db (Polaris):
Totals from affected shaders:
SGPRS: 17168 -> 17168 (0.00 %)
VGPRS: 11296 -> 11296 (0.00 %)
Spilled SGPRs: 1870 -> 1870 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 1472628 -> 1473292 (0.05 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 628 -> 628 (0.00 %)

pipeline-db (Vega):
Totals from affected shaders:
SGPRS: 17168 -> 17168 (0.00 %)
VGPRS: 11296 -> 11296 (0.00 %)
Spilled SGPRs: 1870 -> 1870 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 1409716 -> 1410380 (0.05 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 0 -> 0 (0.00 %)

Max Waves is lower than it should be because of a null winsys bug.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4004>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4004>

aco: consider non-hazard writes in handle_raw_hazard_internal

I think this helps GFX6 in particular because code like this is common:
s_add_i32       s4, 0x60, s3
s_mov_b32       s5, 0
s_load_dwordx4  s[4:7], s[4:5], 0x0
s_buffer_load_dword s4, s[4:7], 0xcc

pipeline-db (Tahiti):
Totals from affected shaders:
SGPRS: 1923878 -> 1923878 (0.00 %)
VGPRS: 1528964 -> 1528964 (0.00 %)
Spilled SGPRs: 476 -> 476 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 88723604 -> 88528880 (-0.22 %) bytes
LDS: 241 -> 241 (0.00 %) blocks
Max Waves: 145402 -> 145402 (0.00 %)

pipeline-db (Polaris):
Totals from affected shaders:
SGPRS: 428128 -> 428128 (0.00 %)
VGPRS: 353092 -> 353092 (0.00 %)
Spilled SGPRs: 119251 -> 119251 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 57580468 -> 57563964 (-0.03 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 11631 -> 11631 (0.00 %)

piepline-db (Vega):
Totals from affected shaders:
SGPRS: 425016 -> 425016 (0.00 %)
VGPRS: 349588 -> 349588 (0.00 %)
Spilled SGPRs: 117835 -> 117835 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 54890792 -> 54874432 (-0.03 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 54 -> 54 (0.00 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4004>

aco: improve get_wait_states()

pipeline-db (Tahiti):
Totals from affected shaders:
SGPRS: 21208 -> 21208 (0.00 %)
VGPRS: 22388 -> 22388 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 3278596 -> 3277004 (-0.05 %) bytes
LDS: 19 -> 19 (0.00 %) blocks
Max Waves: 238 -> 238 (0.00 %)

pipeline-db (Polaris):
Totals from affected shaders:
SGPRS: 64 -> 64 (0.00 %)
VGPRS: 96 -> 96 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 5200 -> 5192 (-0.15 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 10 -> 10 (0.00 %)

pipeline-db (Vega):
Totals from affected shaders:
SGPRS: 0 -> 0 (0.00 %)
VGPRS: 0 -> 0 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 0 -> 0 (0.00 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 0 -> 0 (0.00 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4004>

aco: add new NOP insertion pass for GFX6-9

This new pass is more similar to the GFX10 pass and should be able to
handle control flow better.

No pipeline-db changes.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4004>

iris: Enable HiZ and stencil CCS for blorp blit destinations

Now that blorp blits write to depth and stencil as depth and stencil, we
can leave HiZ and stencil CCS enabled for blorp blit destinations.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

iris: Enable CCS for copies from HiZ+CCS depth buffers

Ever since b274469daae, BLORP is able to sample from whatever the
sampler supports.  In c0c899cf7892, we added HiZ support for copies from
HiZ compressed depth buffers but forgot HiZ+CCS.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

anv: Enable HiZ for VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

blorp: Write to depth/stencil images as depth/stencil when possible

On Gen4 and G45 and earlier, we have to handle weird offsetting to write
to depth and stencil due to a lack of proper depth mipmapping support in
hardware.  On Gen6, we have to deal with strange HiZ and stencil
layouts.  Prior to Gen9, we also had to do crazy things for stencil
writes because we didn't support GL_ARB_shader_stencil_export and
friends in hardware.  However, starting with Gen7 for depth and Gen9 for
stencil, we can easily write out with the "right" hardware.  This allows
us to leave HiZ and other compression enabled for blorp_blit() and
blorp_copy() operations.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

iris: Allow HiZ on blit sources

Ever since 95cc5438ebf, BLORP has been able to read from HiZ-compressed
depth buffers as long as the sampler supports HiZ.  This just makes iris
stop doing the unneeded resolve.

Closes: #2583
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

isl: Set 3DSTATE_DEPTH_BUFFER::Depth correctly for 3D surfaces

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3717>

docs: Update stable process around using fixes: and gitlab

Currently the docs still recommend using
mesa-stable@lists.freedesktop.org, which is pretty awful. We really
don't want a second mailing list and it's mostly full of junk because of
CC: tags anyway.
This changes the preferred actions to be:
1) use a fixes: tag ahead of time
2) use a Cc tag ahead of time if fixes isn't appropriate
3) Use a gitlab MR against the staging/ branch for post-merge/backport
   nominations

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3056>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3056>

turnip: fix tile->slot calculation

Fixes HW binning cases when the horizontal number of tiles isn't divisible
by the horizontal number of pipes (only happens with more than 32 tiles).

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3142>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3142>

turnip: improve binning pipe layout config

The old code looks the same as GL driver, but we get things like
pipe_count = {32, 1}, which seems bad.

This uses similar logic as for tiles which produces a balanced pipe_count
width/height.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3142>

Revert "spirv: Use a simpler and more correct implementaiton of tanh()"

This reverts commit da1c49171d0df185545cfbbd600e287f7c6160fa.

The reduced formula has precision problems on fp16 around 0.  Bring
back the old formula, but make sure to keep the clamping.

Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4054>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4054>

Revert "glsl: Use a simpler formula for tanh"

This reverts commit 9807f502eb7a023be619a14119388b2a43271b0e.

The simplified formula doesn't pass the tanh dEQP tests when we lower
to fp16 math.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4054>

pan/bi: Add bi_print_shader

Woot! That's the last of it! IR printing is now complete*

*until the IR gets updated when new shiny things are added.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_print_block

Almost there...

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_print_clause

Again for post-sched purposes.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_print_bundle for printing bi_bundle

Post-schedule, nops are significnat here.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_instruction printing

So we can debug the IR in memory before code emit has happened. We'd
like to have a complete dump of the IR -- neglecting this with Midgard
was one of those mistakes I've regretted so let's get this right for the
first time around.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Move bi_interp_mode_name to bi_print

Instead of open-coding it in the middle of the disassembler.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add BIR manipulation routines to bir.c

New file.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Move some print routines out of the disasm

These are generally useful for debug of the compiler IR even prior to
code emit; let's share these.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add IR iteration macros

Copypaste from Midgard, for the most part.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add quirks system

Modeled after the Midgard system. Already we know of two
compiler-visible differences between G52 and G71, so let's keep track so
we can eventually port the compiler to other Bifrost systems.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add high-latency property for classes

This is required to know how to schedule legally, and also influences
some issues relating to RA.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add CSEL condition

Along with src_types, this is enough to represent CSEL.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_branch data

For BI_BRANCH, of course. Meshes well with the cfg.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Extract bifrost_branch structure

It's in the disassembler as bitfields, let's extract to a proper
structure so we can see what's there.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add pred/successors to build CFG

We'll want this for analysis passes or something, probably.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add constants to bi_clause

Scheduling will have to pay attention to this.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add EXTRACT, MAKE_VEC synthetic ops

These allow translating between the vector I/O and scalar ALUs,
facilitated by an RA dance to ensured contiguous registers are used.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add source type for conversions

We should now be able to unambiguously represent conversions.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add swizzles

Requires a new field on bifrost_instruction, as well as a new class
property and a new class for the dedicated swizzle ops.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Clarify special op scheduling

They're encoded on ADD but eat the full cycle.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add clause header fields to bi_clause

These will be filled out during scheduling (and possibly RA), to be used
when emitting code.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add class-specific ops

For disambiguating things like min and max within the MINMAX class.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add constant field to bi_instruction

Now that we can index it.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add special indices

For fixed registers, uniforms, and constants, which bypass the usual SSA
mechanism to map well to the ISA.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add dest_type field to bifrost_instruction

A number of opcodes within a class are disambiguated by type/size, and
whether modifiers make sense or not depends on whether the instruction
acts like a float.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_clause, bi_bundle abstractions

These will be used during and after scheduling.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add PAN_SCHED_* flags

Class (mostly) determines scheduling options.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_load_vary structure

For ld_vary in the IR.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Pull out bifrost_load_var

We're not using this structure yet but we want everything in the ISA
ready for us.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bi_load structure

Fills out the class for LD_ATTR, LD_VAR_ADDR

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add bifrost_minmax_mode field

We'll open up a union for class specific data, since this is interesting
only to BI_MINMAX. (And even then...)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add a bifrost_roundmode field

And a class property signaling it's okay to use.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Factor out enum bifrost_minmax_mode

We'll want it from the compiler-side.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add BI_GENERIC property

I don't want to have 20 class-specific structures floating around. So
let's derive them all from a common generic ALU type.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add modifiers to bi_instruction

Now that we can check if we support them via the class.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add class properties

We need to keep track of what specific classes support. For now just
track floating point modifiers.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add src/dest fields to bifrost_instruction

...along with some helpers to generate indices. The indexing scheme is
mostly a copypaste from Midgard, except we specifically reserve 0 as the
sentinel (midgard uses ~0 for this which has always been a pain point).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Add the control flow graph

We're starting to build up the IR data structures in preparation to get
everything piped through.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Stub out new compiler

Just enough to pipe in the NIR shader.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

pan/bi: Gut old compiler

We're making some pretty dramatic design pivots so this early on it'll
be easier to start from scratch, I think.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

panfrost: Add note about preloaded varyings

There's a magic bit in preload_regs which controls this. It doesn't
appear to be supported on G71 but it is on G52. I'd guess G72 supports
it too but I don't have a way to check this.

Needless to say, we'll need a quirks database for this.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4061>

aco: fix image load/store with lod and 1D images

Make sure to add the lod value if non-null as the 2nd operand.

Fixes dEQP-VK.image.load_store_lod.with_format.1d.* on all gens
except GFX9.

Fixes: 4d49a7ac737 ("aco: handle nir_intrinsic_image_deref_{load,store} with lod")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4060>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4060>

gitlab-ci: Distribute jobs across more stages

The stages and mapping of jobs to them are somewhat arbitrary; the goal
is to avoid having to scroll through large numbers of jobs.

v2: (Pierre-Eric Pelloux-Prayer)
* Use even more stages for test jobs
* Give somewhat meaningful names to stages

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3995>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3995>

gitlab-ci: Drop "test-" prefix from llvmpipe/softpipe job names

Redundant.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3995>

vbo: merge draws even when begin==0 or end==0

Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: merge more primitive types for glBegin/End (v2)

v2: clean it up more

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

mesa: deduplicate draw indirect functions

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

mesa: optimize get_index_size

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

mesa: remove _mesa_index_buffer::index_size in favor of index_size_shift

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Suggested-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

Revert "mesa: check for z=0 in _mesa_Vertex3dv()"

This reverts commit f04d7439a0ad6e13ff2912ff824553b6bcf511a4.

It no longer helps performance and the current vbo implementation is
faster anyway.

The app that hit this was a CAD program called Spazio3D. It made pretty
terrible use of the OpenGL API and we sent them some tips for improvements.
I'm assuming they've fixed this by now.

Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: fold code from vbo_exec_fixup_vertex to vbo_exec_wrap_upgrade_vertex

Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: clean up conditional blocks in ATTR_UNION

Move the A != 0 code to the first block.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: handle GS and tess primitive types when splitting Begin/End

Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: clean up vbo_copy_vertices

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: deduplicate copy_vertices functions

There are some differences in exec, but those look like bug fixes not ported
to vbo_save.

Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

vbo: don't look at the second draw's count when merging 2 glBegin/End draws

Only the first count needs to be aligned.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

mesa: replace some index_size multiplications and divisions with shifts

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

mesa: add index_size_shift = log2(index_size) into _mesa_index_buffer

for faster division

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Ian Romanick <ian.d.romanic@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4052>

android: r600/sfn: Add GDS instructions

Fixes the following building errors:

external/mesa/src/gallium/drivers/r600/sfn/sfn_emitssboinstruction.cpp:59: error: undefined reference to 'r600::GDSInstr::GDSInstr(r600::ESDOp, r600::GPRVector const&, std::__1::shared_ptr<r600::Value> const&, std::__1::shared_ptr<r600::Value> const&, std::__1::shared_ptr<r600::Value> const&, int)'
...
external/mesa/src/gallium/drivers/r600/sfn/sfn_emitssboinstruction.cpp:256: error: undefined reference to 'r600::RatInstruction::RatInstruction(r600::ECFOpCode, r600::RatInstruction::ERatOp, r600::GPRVector const&, r600::GPRVector const&, int, std::__1::shared_ptr<r600::Value> const&, int, int, int, bool)'

Fixes: 32d3435a ("r600/sfn: Add GDS instructions")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>

android: r600/sfn: fix includes and libmesa_nir dependency

Fixes the following building errors:

In file included from external/mesa/src/gallium/drivers/r600/sfn/sfn_debug.cpp:28:
In file included from external/mesa/src/gallium/drivers/r600/sfn/sfn_debug.h:34:
In file included from external/mesa/src/compiler/nir/nir.h:41:
In file included from external/mesa/src/compiler/nir_types.h:36:
external/mesa/src/compiler/glsl_types.h:38:10: fatal error: 'main/config.h' file not found
#include "main/config.h"
         ^~~~~~~~~~~~~~~
1 error generated.

In file included from external/mesa/src/gallium/drivers/r600/sfn/sfn_debug.cpp:28:
In file included from external/mesa/src/gallium/drivers/r600/sfn/sfn_debug.h:34:
external/mesa/src/compiler/nir/nir.h:50:10: fatal error: 'nir_opcodes.h' file not found
#include "nir_opcodes.h"
         ^~~~~~~~~~~~~~~
1 error generated.

Fixes: f718ac62 ("r600/sfn: Add a basic nir shader backend")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>

android: aco: fix PIPE_FORMAT related building errors

Fixes the following building errors:

In file included from external/mesa/src/amd/compiler/aco_dead_code_analysis.cpp:25:
In file included from external/mesa/src/amd/compiler/aco_ir.h:33:
In file included from external/mesa/src/compiler/nir/nir.h:40:
external/mesa/src/util/format/u_format.h:33:10: fatal error: 'pipe/p_format.h' file not found
#include "pipe/p_format.h"
         ^~~~~~~~~~~~~~~~~
...
In file included from external/mesa/src/amd/compiler/aco_dominance.cpp:31:
In file included from external/mesa/src/amd/compiler/aco_ir.h:33:
In file included from external/mesa/src/compiler/nir/nir.h:40:
external/mesa/src/util/format/u_format.h:33:10: fatal error: 'pipe/p_format.h' file not found
#include "pipe/p_format.h"
         ^~~~~~~~~~~~~~~~~
...
In file included from external/mesa/src/amd/compiler/aco_instruction_selection.cpp:31:
In file included from external/mesa/src/amd/common/ac_shader_util.h:32:
In file included from external/mesa/src/compiler/nir/nir.h:40:
external/mesa/src/util/format/u_format.h:33:10: fatal error: 'pipe/p_format.h' file not found
#include "pipe/p_format.h"
         ^~~~~~~~~~~~~~~~~
3 errors generated.

Fixes: 8d07d661 ("glsl,nir: Switch the enum representing shader image formats to PIPE_FORMAT.")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>

nir: Flush to zero with OOB low exponents in ldexp

Reviewed-by: Arcady Goldmints-Orlov <agoldmints@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>