This patch is to avoid the high overhead that exists when trying to
kick ever single draw during multidraw.
glMultiDrawArrays performance profiling:
342.5 thousand draws/second -> 40 million draws/second
Special thanks to Arthur Huillet for helping getting this profiled
in irc.
Signed-off-by: Yusuf Khan <yusisamerican@gmail.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
---
v2: fix typos pointed out by Arthur
v3: nvc0_draw_vbo -> nvc0_draw_single_vbo, intialize count
v4: remove num_draws from wrapped function and add mutex assert
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28641>
Having this in one place is better.
When I wrote the old I wasn't aware that checking the kill flag on definitions
is the same as checking zero uses.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29512>
On newer hardware, the hi operation reads the lo half of the inline constant.
On older hardware, it reads the hi half (zero).
I tested this on Navi31 for gfx11 and Raphael for gfx10.
Foz-DB Navi31:
Totals from 4 (0.01% of 79395) affected shaders:
CodeSize: 36832 -> 36448 (-1.04%)
Latency: 20362 -> 20334 (-0.14%)
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29512>
Indeed, the access to the array is done with a 4x multiplier.
The size of the array should be calculated accordingly.
For instance, this issue is triggered on radeonsi with
"piglit/bin/arb_direct_state_access-gettextureimage-formats -auto -fbo":
==3419==ERROR: AddressSanitizer: stack-buffer-overflow on address 0x7ffc31f804b0 at pc 0x7fac7ef81b2d bp 0x7ffc31f803d0 sp 0x7ffc31f803c8
WRITE of size 1 at 0x7ffc31f804b0 thread T0
#0 0x7fac7ef81b2c in do_span_rgba_unorm8 ../src/mesa/main/mipmap.c:160
#1 0x7fac7ef83549 in do_row ../src/mesa/main/mipmap.c:258
#2 0x7fac7ef83986 in make_2d_mipmap ../src/mesa/main/mipmap.c:371
#3 0x7fac7ef8670b in generate_mipmap_compressed ../src/mesa/main/mipmap.c:1062
#4 0x7fac7ef8670b in _mesa_generate_mipmap ../src/mesa/main/mipmap.c:1119
#5 0x7fac7e5472aa in check_gen_mipmap ../src/mesa/main/teximage.c:2910
#6 0x7fac7e5472aa in check_gen_mipmap ../src/mesa/main/teximage.c:2904
#7 0x7fac7e5472aa in teximage ../src/mesa/main/teximage.c:3315
#8 0x7fac7e5472aa in teximage_err ../src/mesa/main/teximage.c:3342
#9 0x7fac7e550cfa in _mesa_TexImage2D ../src/mesa/main/teximage.c:3413
Address 0x7ffc31f804b0 is located in stack of thread T0 at offset 96 in frame
#0 0x7fac7ef814ff in do_span_rgba_unorm8 ../src/mesa/main/mipmap.c:132
This frame has 3 object(s):
[32, 96) 'result' (line 145) <== Memory access at offset 96 overflows this variable
[128, 384) 'rowA' (line 144)
[448, 704) 'rowB' (line 144)
Fixes: dd8fb7139d ("mesa/main: rewrite mipmap generation code")
Signed-off-by: Patrick Lerda <patrick9876@free.fr>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29572>
Usually this is lowered in NIR, but GFX12 needs to use an intrinsic.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Acked-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29466>
It was set to "always run" for amd common files changes when I obviously
meant for it to be manual and messed up my copy/paste when I wrote that.
Fixes: ebaede788e ("amd/ci: limit radv jobs to radv + aco files changes")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29550>
The intent behind these asserts is to ensure the layer is within
bounds, so we rather check it is within the image layer count than
within the layerCount of the image subresource passed by the API.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29544>
If the cache is disabled then we need to destroy the pipelines
manually when they are no longer needed. Do that by adding them
as private objects to the command buffer.
Fixes: 4f26303dbb ('v3dv: add debug option to disable custom pipeline caches for meta operations')
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29544>
Make some helper functions panvk_cmd_buffer agnostic to prepare for moving
them to the bifrost subdirectory.
Signed-off-by: Rebecca Mckeever <rebecca.mckeever@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29369>
Add helpers to fill/prepare the push_uniforms array. While at it
move the push_uniforms field out of the descriptor_state.
Signed-off-by: Rebecca Mckeever <rebecca.mckeever@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29369>
