Computer shaders that access tileY resources (textures) benefit
from Y-locality accesses. Easiest way to implement this is walk
local ids in Y-major fashion, instead of X-major fashion. Y-major
local ids will reduce partial writes and increase cache locality
for tileY accesses since tileY resources cachelines progress in
Y direction.
Improves performance on TGL:
Borderlands3.dxvk-g2 +1.5%
Y-major can introduce a performance drop on CS that use mixture
of buffers and images. This should be fixed in next commit.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10733>
In file included from ../src/intel/ds/intel_pps_perf.cc:8:
../src/intel/ds/intel_pps_perf.h:34:78: error: ‘string’ in namespace ‘std’ does not name a type
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10934>
I forgot that after rebasing on large_consts support that this is now
called after the first time nir_lower_wrmask is called and can generate
partial writemasks that need to be lowered. While we're here, also call
the main optimization loop if things are lowered to scratch because it
generates address arithmetic that may need to be cleaned up.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10922>
There's a bug in ATEST (and LD_TILE?) handling affecting v6, manifesting
as gpu sched timeouts in dmesg and ultimately as random dEQP flakes.
Until this can be sorted out and the G72 jobs are proven to be stable,
we can't have them in the critical merge path.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10923>
I've been meaning to make this change for a while, but I've been holding
out in case dEQP-GLES2 on T860/Bifrost catches a driver issue missed by
dEQP-GLES3 and also missed by T720's dEQP-GLES2 job. Given that hasn't
happened (indeed, dEQP-GLES3 is an effective super set of dEQP-GLES2),
let's save the cycles and skip the redundant jobs.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10923>
The u_resource_vtbl indirection is going to be removed.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10659>
It's only used by buffers and only zink uses it privately for textures too.
This is part of removing u_resource_vtbl.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10659>
This is the initial step towards removing u_resource_vtbl.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10659>
State uploads are bound to the batch, so if the batch changes, we
invalidate all state. Theoretically this isn't totally optimal but in
practice it should be good enough.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10888>
This patch will avoid reallocation,if surface is already protected.
Fixing the comparision logic of boolean value(true \ flase) with
PIPE_BIND_PROTECTED.
Fixes: 81be8b3c2f ("va/picture: make sure destination buffer is protected if needed")
Signed-off-by: SureshGuttula <suresh.guttula@amd.corp-partner.google.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10916>
this was broken for the indirect case if the indirect draw count or
firstIndex was nonzero and also would rewrite the index buffer onto the
wrong offset of the dst buffer
Fixes: 0c85d6c523 ("gallium/util: factor out primitive-restart rewriting logic")
Fixes: 330d0607ed ("gallium: remove pipe_index_buffer and set_index_buffer")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10909>
The wayland EGL platform sets the modifier count to 0 in some cases
to signal that modifiers should not be used, even if a list of modifiers
is present. The loader_dri_create_image helper didn't handle this case
properly and called the modifierful driver interface with a 0 modifier
count, leading to the obvious outcome of the driver being unable to
allocate an image.
Fixes: cb9ae4273d ("dri: add loader_dri_create_image helper")
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10903>
SCS_ONE and ISL_CHANNEL_SELECT_ONE (and SCS_ZERO and
ISL_CHANNEL_SELECT_ZERO) have the same values, so this doesn't matter in
practice. However, some compiler warning options will cause warnings to
be generated.
src/gallium/drivers/iris/iris_state.c: In function ‘fmt_swizzle’:
src/gallium/drivers/iris/iris_state.c:2177:32: warning: implicit conversion from ‘enum GFX125_ShaderChannelSelect’ to ‘enum isl_channel_select’ [-Wenum-conversion]
2177 | case PIPE_SWIZZLE_1: return SCS_ONE;
| ^~~~~~~
src/gallium/drivers/iris/iris_state.c:2178:32: warning: implicit conversion from ‘enum GFX125_ShaderChannelSelect’ to ‘enum isl_channel_select’ [-Wenum-conversion]
2178 | case PIPE_SWIZZLE_0: return SCS_ZERO;
| ^~~~~~~~
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9610>
These aren't real formats. They cannot be used with
RenderbufferObjects, Textures, or anything else visible through the GL
API. The only purpose is to support YUYV textures imported through
EGL_image_external. There is hardware that can sample from this
subsampled format, but the hardware does not do the colorspace
conversion automatically.
v2: Treat these formats as compressed. This is a lie, but it seems to
be a less egregious lie than what I was doing before. It also passes
'ninja test'.
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9610>
