The LLVM backend forgot to set config->lds_size, which is used
for reporting LDS stats.
Fixes: cf89bdb9ba "radv: align the LDS size in calculate_tess_lds_size()"
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9098>
The calculate_tess_lds_size function already returns the size in blocks
of the encoding granule, but we forgot to adjust config->lds_size.
This variable is not used to actually set the LDS size used for TCS,
but by ACO to make scheduling decisions.
Fossil DB stats on Sienna Cichlid:
Please note that the +3729.43% is NOT a regression.
The real LDS size used didn't change, it was just reported incorrectly.
Totals from 1342 (0.96% of 139391) affected shaders:
VGPRs: 60880 -> 80240 (+31.80%); split: -0.05%, +31.85%
CodeSize: 3378456 -> 3381224 (+0.08%); split: -0.23%, +0.31%
LDS: 687104 -> 26312192 (+3729.43%)
MaxWaves: 29794 -> 23962 (-19.57%)
Instrs: 644194 -> 644610 (+0.06%); split: -0.32%, +0.39%
Cycles: 2675068 -> 2676804 (+0.06%); split: -0.31%, +0.38%
VMEM: 428840 -> 517418 (+20.66%); split: +22.53%, -1.88%
SMEM: 91831 -> 88587 (-3.53%); split: +5.70%, -9.23%
VClause: 22740 -> 19384 (-14.76%); split: -16.18%, +1.42%
SClause: 19116 -> 18373 (-3.89%); split: -4.34%, +0.46%
Copies: 66662 -> 63448 (-4.82%); split: -5.55%, +0.73%
Fixes: cf89bdb9ba "radv: align the LDS size in calculate_tess_lds_size()"
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9098>
It's way easier to identify a pipeline compared to a random CPU addr.
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/9083>
This enables VK_EXT_debug_marker only if SQTT is enabled, otherwise
it's just useless.
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/9099>
This currently hangs and I don't know why. It looks better to
print a message instead of hanging the whole system.
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/9097>
RGP actually crashes if pipeline bind markers are emitted without
PSO correlation objects.
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/9097>
If we emit a lot of markers, we might reach the limit easily.
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/9097>
This seems much cleaner and will help for future work. Also, this
might help if we want to dynamically change some SQTT parameters
in the future.
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/9097>
We can't invoke gfx6 addrlib (overridden by SI_FORCE_FAMILY) with a gfx9
family ID.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9028>
Instead of assuming WGP mode on GFX10+ in different places, add a member
to Program that can be used instead.
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/8761>
If each SIMD has to get an different number of waves, report the maximum.
One example of a situation is when a single-wave workgroup uses more than
max_lds_per_simd. This change causes radv_get_max_waves() to report a
single wave per SIMD instead of none.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>
This fixes a issue in radv_get_max_waves() where it aligned the LDS
allocation to 512 bytes instead of 1024 on GFX10.3.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>
Always return the wave32 waves instead of wave64 waves because the wave32
wave count is more precise in the case of wave32.
This also fixes usage of lds_per_wave in wave32.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>
num_simd_per_compute_unit was the number of SIMDs per compute unit, but
lds_size_per_workgroup was the bytes of LDS per WGP.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>
Previously we were calculating the inverse swizzle instead and doing something
horrible to get 0/1 right, and then "fixing" our table.
Let's do it right an align with the mesa-wide table.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8797>
