src/mesa: add GL_NV_half_float extension support (v2)

This patch adds support for GL_NV_half_float
extension.

v2: fix main_test failure

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6219>

dri_util: Update internal_format to GL_RGB8 for MESA_FORMAT_B8G8R8X8_UNORM

Port the change done for RGBX8888 in
02a1f95386b43bf46cd1c8297d0955242f554fa2.

If XR24 images are considered to be VIEW_CLASS_24_BITS-compatible, it's
reasonable to assume that XB24 images would be as well.

Fixes: bf576772ab4d ("dri_util: add driImageFormatToSizedInternalGLFormat function")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6095>

aco: execute branch instructions in WQM if necessary

It could happen that only the branch condition was computed in WQM
and not the branch instruction.
There is now some rendundancy which should be cleaned up.

Fixes: 3817fa7a4d1f51c385b28a2e45a1edf227526028 ('aco: fix WQM handling in nested loops')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6260>

nir: nir_range_analysis needs to be updated for vec16

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6275>

aco: don't move memory accesses to before control barriers

Fixes random failures of dEQP-VK.image.qualifiers.volatile.cube_array.r32i
and similar tests on Vega.

fossil-db (Navi):
Totals from 6 (0.00% of 135946) affected shaders:
VMEM: 1218 -> 1110 (-8.87%); split: +2.46%, -11.33%
SMEM: 174 -> 189 (+8.62%)
Copies: 84 -> 87 (+3.57%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: cd392a10d05 ('radv/aco,aco: use scoped barriers')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6174>

radv: limit LATE_ALLOC_GS to prevent a GPU hang on GFX10

Found by inspection, doesn't fix anything known.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6279>

radv/gfx10: add missing initialization of registers

Found by inspection.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6277>

etnaviv: completely turn off MSAA

MSAA worked before etnaviv landed in upstream mesa but got
broken over time. Disable MSAA completely until it is fixed again.

Fixes problems/crashes with applications that want to make use of MSAA.

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5608>

nir/lower_ssbo: Don't set align_* for atomics

Fixes crashes when lowering atomic SSBOs:

   run: ../src/compiler/nir/nir.h:1875: nir_intrinsic_align_mul: Assertion `info->index_map[NIR_INTRINSIC_ALIGN_MUL] > 0' failed.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6204>

anv: add a check for depthStencilState before using it

v2: move the code under existing correct check!

Fixes: e4590c075009 ("anv: depth/stencil dynamic state support")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3375
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6240>

anv: fix up dynamic clip emission

There were 2 issues :

   * We were not emitting the clip state when the pipeline changed

   * On Gen7 we did not program the front facing & cull mode dynamic
     values in the clip state

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c34d8ac26e0a21 ("anv: handle dynamic viewport count")
Closes https://gitlab.freedesktop.org/mesa/mesa/-/issues/3379

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6265>

anv: centralize vk to gen arrays

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6265>

freedreno/decode: try harder to not crash in disasm

Move the handling for catching asserts when we start decoding garbage
into disasm-a3xx.  This way it can also cover other cases where cffdec
tries to disassemble memory, such as SP_xS_OBJ_START.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6242>

freedreno/crashdec: handle section name typos

The fixes tag isn't so much because it was incorrect before, but because
I'm going to send a kernel patch to fix the typo, and that will break
old crashdec.

Fixes: 1ea4ef0d3be ("freedreno: slurp in decode tools")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6242>

freedreno/ir3: add more disasm stats

Add tracking for # of instructions per category, similar to the last
patch.  Also add a few other shader-db stats that were missing on the
disasm side, to make it easier to compare to shaders from cmdstream
traces.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6243>

freedreno/ir3: add tracking for # of instructions per category

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6243>

gallium/u_transfer_helper: add util functions for doing deinterleaving during map

in some cases (e.g., zink and d3d12) we only want to split the depth and stencil
buffers when we're mapping them, and we can handle packed buffers in other cases,
so being able to reuse the u_transfer_helper functionality is still desired but
only if we can preserve the underlying buffer the rest of the time

Kenneth Graunke notes during post-review:
Vulkan reads/copies on packed Z24S8 only return depth, so we need to use separate
Z24 and S8 reads and do packing tricks.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5338>

gallium: add pipe_transfer_usage for z/s only mappings

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5338>

intel/fs: work around gen12 lower-precision source modifier limitation

GEN:BUG:1604601757 prevents source modifiers for multiplication of
lower precision integers.

lower_mul_dword_inst() splits 32x32 multiplication into 32x16, and
needs to eliminate source modifiers in this case.

Closes: #3329
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

intel/fs: Assert if lower_source_modifiers converts 32x16 to 32x32 multiplication

Lowering source modifiers will convert a 16bit source to a 32bit
value.  In the case of integer multiplication, this will reverse
previous lowering performed by lower_mul_dword_inst.

Assert to prevent an illegal DxD operation (and GPU hang).

Reviewed-by: Francisco Jerez <currojerez@riseup.net>

android: pan/bi: Separate disasm/compiler targets

Fixes the following building errors:

ld.lld: error: undefined symbol: bi_interp_mode_name
>>> referenced by bi_print.c:207 (external/mesa/src/panfrost/bifrost/bi_print.c:207)
...
ld.lld: error: undefined symbol: bi_round_mode_name
>>> referenced by bi_print.c:285 (external/mesa/src/panfrost/bifrost/bi_print.c:285)

ld.lld: error: undefined symbol: bi_clause_type_name
>>> referenced by disassemble.c:142 (external/mesa/src/panfrost/bifrost/disassemble.c:142)
...
ld.lld: error: undefined symbol: bi_ldst_type_name
>>> referenced by disassemble.c:0 (external/mesa/src/panfrost/bifrost/disassemble.c:0)

Fixes: 14bb72c68 ("pan/bi: Separate disasm/compiler targets")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6261>

android: pan/mdg: Separate disassembler and compiler targets

Fixes the following building errors:

ld.lld: error: undefined symbol: mir_print_constant_component
>>> referenced by disassemble.c:356 (external/mesa/src/panfrost/midgard/disassemble.c:356)
...
ld.lld: error: undefined symbol: mir_print_constant_component
>>> referenced by disassemble.c:416 (external/mesa/src/panfrost/midgard/disassemble.c:416)

Fixes: b792d613e ("pan/mdg: Separate disassembler and compiler targets")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6261>

android: panfrost: Move pandecode into lib/

Fixes the following building errors:

target  C: libpanfrost_decode <= external/mesa/src/panfrost/pandecode/common.c
...
clang: error: no such file or directory: 'external/mesa/src/panfrost/pandecode/common.c'
clang: error: no input files

Fixes: d62a6e7c5 ("panfrost: Move pandecode into lib/")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6261>

android: panfrost: Rename encoder/ to lib/

Fixes the following building errors:

target  C: libpanfrost_encoder <= external/mesa/src/panfrost/encoder/pan_attributes.c
...
clang: error: no such file or directory: 'external/mesa/src/panfrost/encoder/pan_attributes.c'
clang: error: no input files

target  C: libpanfrost_encoder <= external/mesa/src/panfrost/encoder/pan_afbc.c
...
clang: error: no such file or directory: 'external/mesa/src/panfrost/encoder/pan_afbc.c'
clang: error: no input files

Fixes: 1c62b5528 ("panfrost: Rename encoder/ to lib/")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6261>

aco: set constant_data_offset correctly in the case of merged shaders

setup_nir() is done for all shaders before any of them are selected, so
constant_data_offset could be incorrect for the first shader.

Fixes incorrect geometry in Mafia III and Max Payne 3.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2768
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6205>

radeon/vcn: fix jpeg decode for navi10

Fixes an issue where the JPEG decode would timeout when decoding certain JPEG files.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <boyuan.zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6249>

radv: Do not consider layouts fast-clearable on compute queue.

We cannot decompress from the compute queue. While I'm pretty sure
VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL is only useful on the
graphics queue, I cannot find a VU that prevents the transition
from happening on another queue, so we need to be careful here.

This patch ensures we do the decompression on the barrier that changes
the queue ownership.

Another problem was that DCC images were considered fast-clearable
when not DCC compressed, which resulted in a mess with concurrent
queue ownership.

Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3387
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6252>

radv: Add forcecompress debug flag.

Enables DCC/HTILE/CMASK/FMASK when supported, not just when we think
it is beneficial.

This is helpful to detect compression bugs with CTS.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6252>

intel/perf: export performance counters sorted by [group|set] and name

It's a lot easier to deal with them in RenderDoc when they are
in some meaningful order.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

intel/perf: split load_oa_metrics

Move oa_metrics_available out of load_oa_metrics and call
build_unique_counter_list outside.

This change is a preparation for the next patch. It should
not have any functional impact.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

intel/perf: fix performance counters availability after glFinish

Currently Linux kernel gathers performance counters at fixed
intervals (~5-10ms), so if application uses AMD_performance_monitor
extension and immediately after glFinish() asks GL driver for HW
performance counter values it might not get any data (values == 0).

Fix this by moving the "read counters from kernel" code from
"is query ready" to "get counter values" callback with a loop around
it. Unfortunately it means that the "read counters from kernel"
code can spin for up to 10ms.

Ideally kernel should gather performance counters whenever we ask
it for counter values, but for now we have deal with what we have.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

intel/perf: streamline error handling in read_oa_samples_until

No functional changes.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

intel/perf: fix how pipeline stats are stored

It matters only when counters are not ordered by offset.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

intel/perf: fix calculation of used counter space

It matters only when counters are not ordered by offset.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5788>

radv: report a better error message when QueueWaitIdle() failed

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6230>

radv: report errors back to the application via VK_EXT_debug_report

Help for debugging.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6230>

radv: rework the error function helpers a bit

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6230>

radv: report the spirv-nir logs back to the application

Via VK_EXT_debug_report to help debugging various SPIRV->NIR issues.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6223>

radv: Fix assert that is too strict.

The added assert fails on MSAA images if we disable FMASK ....

Reordered things.

Fixes: c6aadbae715 "radv: Don't use both DCC and CMASK for single sample images."
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3385
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6232>

radeon/radeon_vce: fix out of target bitrate in CBR mode (H.264)

StoneyRidge may not comply to required target bitrate
when generating H.264 stream in CBR mode.

Signed-off-by: Louis Li <Ching-shih.Li@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4384>

r600: Enable compute shaders for NIR code path

There are still regessions compared to TGSI, but there are also many
fixes.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: Force a minimum of 4 GPRs, it seems to fix atomics

This fixes spec@arb_compute_shader@execution@atomic-counter on
HD 5450

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: handle querying SSBO size

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: Correct ssbo instruction handling

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: correct allocating and emitting of atomics

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: Add a mapping table for atomics

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: add r600 specific lowering pass for atomics and use it

v2: rebase to use global variable lists

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600/sfn: Sort uniforms by binding and offset

This is required to get atomics right.

v2: rebase to use global variable lists

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

r600: Set PIPE_CAP_NIR_ATOMICS_AS_DEREF to true

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

gallium + mesa/st: Add PIPE_CAP_NIR_ATOMICS_AS_DEREF and use it

This cap is useful for drivers that support hardware atomics and need
special handling to resolve their addresses.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6025>

meson: bump required glvnd version

https://github.com/KhronosGroup/EGL-Registry/pull/95 has moved
a couple of extensions defines and functions to the upstream `eglext.h`,
but when 9a74746bd1f3bd28d4c4 sync'ed these files we broke compilation
of apps that require these symbols on systems that don't have the
updated Khronos headers.

On non-GLVND builds, we still provide these headers, so everything's
fine, but on GLVND builds the Khronos headers are external so we need to
make sure we have a libglvnd version that's recent enough.

Fixes: 9a74746bd1f3bd28d4c4 ("EGL: sync headers with Khronos")
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6069>

driconf: fix force_gl_vendor description

The option is not a toggle to "allow GPU vendor to be overridden", it
*is* the override.

Fixes: dca119f12c291d7665d7 ("mesa/gallium: add dric option to allow overriding GL vendor string")
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6207>

egl/entrypoint-check: add check that GLVND and plain EGL have the same entrypoints

Cc: mesa-stable
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4448>

egl/entrypoint-check: split sort-check into a function

Cc: mesa-stable
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4448>

Correct a typo in threads_win32.h

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6238>

pick-ui: specify git commands in "resolve cherry pick" message

Cc: mesa-stable
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6134>

zink: Fix 32-bit compilation

`sizeof(void*) != sizeof(VkShaderModule)` on x86 hence this fails with `-Werror=int-conversion`

Fixes: 0f059d550bd "zink: split up creating zink_shader objects and VkShaderModule objects"

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6239>

i965: add support for force_gl_vendor

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3363
Reviewed-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6198>

freedreno/ir3: don't install ir3_compiler cmdline tool

It is mostly just useful to us.. and it is big since it links in nir and
most of the rest of gallium.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6234>

freedreno/registers: add some missing regs to build

Needed for installed version of crashdec/cffdump.

Fixes: 9c33c538981 ("freedreno/registers: install gzip'd register database")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6234>

meson/freedreno: Fix lua requirement

Freedreno needs at least Lua 5.2, but the current code will report found
for 5.1, which doesn't actually work.

Fixes: caa107cb8db930bc59cd557a325a6a0bc4a86565
       ("freedreno/decode: move dependencies up a level")

Reviewed-by: Rob Clark <robclark@freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6229>

radeonsi: various fixes for gfx10.3

The magic numbers fix sample shading.
The bypass flag is optional.

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: remove the NGG hack decreasing LDS usage to deal with overflows

The LDS size can't overflow anymore, so we can use the correct max LDS size.

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: add a common function for getting the size of gs_ngg_scratch

The next commit will use it.

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: don't count unusable vertices to the NGG LDS size

Now we get optimal LDS usage.

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: fix applying the NGG minimum vertex count requirement

The code applied the restriction too late, which could overflow LDS size,
which started happening more often after the minimum vertex count was
increased for Sienna.

Incorporate the clamping into the previous code for rounding up the counts.
Now the LDS size can never overflow, but it may use vector lanes less
efficiently (max_gsprims can be decreased more), which will be addressed
in the next commit.

Fixes: 4ecc39e1aa1 ("radeonsi/gfx10: NGG geometry shader PM4 and upload")

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: increase minimum NGG vertex count requirement per workgroup on gfx 10.3

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: use the same units for esgs_ring_size and ngg_emit_size

for consistency

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: use correct wave size in gfx10_ngg_calculate_subgroup_info

Fixes: 88efb63cafc ("radeonsi/gfx10: implement Wave32")

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

Revert "radeonsi: honor a user-specified pitch on gfx10.3"

This reverts commit c4b5fd9ab096a0bb5106b93191b13c81cc32243b.

It breaks mipmapping. This is only meant to be used by OpenCL, which allows
setting a user pitch for linear images. In all other cases, don't support
a custom pitch.

Fixes: c4b5fd9ab096a0bb51 "radeonsi: honor a user-specified pitch on gfx10.3"

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

ac/gpu_info: set num_tiles_pipes on gfx10+ too

Based on PAL.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>

radeonsi: enable ETC2 hw acceleration on Raven2

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6213>

zink: pre-hash gfx-pipeline-state

Store a hash in `zink_gfx_pipeline_state` to keep track of state
changes and avoid to recompute it when the state has not changed.

Signed-off-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6061>

zink: rename zink_gfx_program::stages to 'modules'

we've been confusing 'stages' and 'shaders' over and over for a long time,
so maybe having a totally different name will help here

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: always compile shaders in pipeline order

in order to accurately perform slot/location mapping that's consistent across
stages, we need to go through the stages in order so that we can pass each successive
slot map allocation along to the next compiled stage

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: start using per-stage flags for new shaders, refcount shader modules

we don't want to recompile shaders if we don't have to, so we can set bitflags
upon receiving new shader states and then compile only the stages that have
changed while refcounting the unchanged stages

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: use ZINK_SHADER_COUNT instead of PIPE_SHADER_TYPES - 1 everywhere

this is just for convenience and consistency

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: split up creating zink_shader objects and VkShaderModule objects

the actual VkShaderModule is only needed when we're creating a program to
draw with, so this can be split off for "uncompiled" and "compiled" shader
objects which will facilitate implementing shader keys

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: refcount zink_gfx_program objects

now that we're tracking these by shader, we want to ensure that they live through
each render pass successfully if there's no flush regardless of the timing when the
shader objects are destroyed. this becomes useful when we split up shader create and compile
functionality in future patches, at which point program refcounts can be changed
during successive draw calls, potentially resulting in a program being destroyed at that
point when it shouldn't be

with this patch, each shader used by the program gets a reference, with the renderpass
batch itself becoming the owner of the program such that it will be deleted
when the draw state gets invalidated and a new program is created

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: adjust zink_shader struct to contain full streamout info

move remapped register_index -> slot into a separate value on the struct
in order to preserve the register_index value

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

zink: move shader state methods for pipe_context into zink_program.c

just moving these so all the shader code can be in one place

Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5970>

intel/dump_gpu: add an option to capture a single frame

We use the driver identifier buffer to get the information about the
current frame.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: fix --platform option

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

anv: track the current frame and write it into the driver identifier BO

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: only map in GTT buffers not previously mapped

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: add an only-capture option

This option allows for smaller aub files to be generated by only
storing the BOs flagged with EXEC_OBJECT_CAPTURE.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: set default device_override

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: further track mapping of BOs

We can go further in tracking what BOs are written to by the driver by
tracking when a buffer in unmapped. A BO could be mmap, written, unmap
and never be written to again. In such case we can just write the BO's
content on the first exec buf after unmap and never write it again.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

intel/dump_gpu: only write BOs mapped by the driver

Track what BOs are mapped by the driver and only write those. We can
safely assume that when not mapped there is no data to save.

v2: Don't forget to return the ret (Rafael)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2201>

zink: set primitive restart cap

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5912>

zink: use util_draw_vbo_without_prim_restart for unsupported prim modes

this fixes up primitive restart functionality for the primitive types that
vulkan doesn't support using primitive restart with

fixes #2873

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5912>

zink: move 8bit index handling out of u_primconvert path

putting this through util_translate_prim_restart_ib reduces our reliance on
u_primconvert, which ideally we don't want to be using

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5912>

zink: basic primitive restart support for strip/fan topologies

this conditionally handles rewriting the index buffer to use vk-compatible
restart indexes and then enables it in the pipeline for supported draw
modes

fixes #3174

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5912>

radv: Add ETC2 support on RAVEN2.

I did a whole bunch of validation and it essentially works.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6215>

radv: allow to force-enable LLVM internally for a specific shader stage

For ACO debugging purposes, developers only.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6208>

etnaviv: call nir_lower_bool_to_bitsize

Starting with commit 6f394343b1f ("nir/algebraic: i2f(f2i()) -> trunc()")
dEQP-GLES2.functional.shaders.operator.binary_operator.div.lowp_int_vertex
ends with an unsuppored flt instruction. Use nir_lower_bool_to_bitsize to
convert this flt to a flt32 which is supported. This fixes the introduced
regression.

Cc: 20.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6182>

panfrost: Drop compiler cmdstream deps

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6196>

panfrost: Treat texture dimension as first-class

Instead of hiding behind a cmdstream enum. The raw values have a very
simple interpretation as dimension.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6196>

panfrost: Inline max rt into compilers

Another dep.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6196>

pan/bi: Drop use of MALI_POSITIVE

Contributing to a dep on the main cmdstream. It's a fancy word for
subtract one...

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6196>