Commit graph

24 commits

Author SHA1 Message Date
Natalie Vock
f0aa383e09 radv/rt: Use ds_bvh_stack_rtn
Improves Quake 2 RTX performance by 5% on RDNA3.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35269>
2025-07-15 21:34:40 +00:00
Natalie Vock
8815845271 radv/rt/gfx12: Always overwrite origin/dir
They're unchanged if we don't test against instance nodes. This makes
image_bvh8_intersect_ray kill its direction/origin operands, improving
RA.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35269>
2025-07-15 21:34:38 +00:00
Natalie Vock
e236a731e4 radv/rt: Enable pointer flags on GFX11+
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Allows hardware to do some of the culling work, as well as early-cull
box nodes with CullOpaque/CullNonOpaque ray masks when all children are
(not) opaque.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32417>
2025-06-28 10:31:38 +00:00
Konstantin Seurer
97f6287827 radv: Use the BVH8 format on GFX12
Reviewed-by: Natalie Vock <natalie.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34273>
2025-04-17 20:20:40 +00:00
Konstantin Seurer
9d157173b2 radv: Refactor create_bvh_descriptor
Make it a bit more extendable since GFX12 introduced more fields.

Reviewed-by: Natalie Vock <natalie.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34273>
2025-04-17 20:20:40 +00:00
Konstantin Seurer
676e26aed5 radv: Fix rayTracingPositionFetch with multiple geometies
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
The fix adds more indirections to avoid increasing register pressure by
tracking the primitive address.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34460>
2025-04-11 22:26:08 +00:00
Timur Kristóf
666f10c174 radv: Add missing includes and remove unnecessary NIR includes.
RADV won't compile without the added includes after we
stop including the full nir.h from the VK common functions.

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
2025-02-12 22:33:07 +01:00
Marek Olšák
82047fa82f amd: drop support for LLVM 15, 16, 17
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33211>
2025-02-01 04:22:30 +00:00
Eric Engestrom
dd2629b8b8 radv,lvp: fix url to VkAabbPositionsKHR docs
The current URL redirect to a page that does not contain any information
about this struct, so let's fix that.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33159>
2025-01-28 14:28:59 +00:00
Ivan Avdeev
14e3231b56 radv: add a flag to indicate ray tracing support
Determine whether the device has hardware raytracing support early, and
then use this result where needed, instead of checking for `gfx_level`
every time.

This is a prerequisite for CYAN_SKILLFISH chip enablement. This chip is
still GFX10, not GFX10_3, but has hardware support for accelerated
`image_bvh{,64}_intersect_ray` instructions. Just checking for `gfx_level`
is insufficient for it.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33109>
2025-01-20 08:27:11 +00:00
Arseny Kapoulkine
6a3944a195 radv: On GFX11, use box sorting heuristic based on ray flags
Use box with largest ray interval for shadow rays (terminate on first
hit) as it maximizes the probability of finding some object in that box;
for reflection (closest hit) rays, use midpoint instead, which defers
processing of larger boxes the ray origin is in in favor of smaller
boxes closer to origin.

Since the sorting mode must be uniform, when terminate_on_first_hit flag
is divergent, we leave it as closest.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32416>
2024-12-01 05:20:58 +00:00
Friedrich Vock
64c406774f radv/rt: Skip all AABB code when no_skip_aabbs is not set
This avoids having to execute the load_global just to throw the results
away and ignore the node.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31443>
2024-10-03 15:22:08 +00:00
Konstantin Seurer
00c94e0cd4 radv: Workaround apps using ray tracing when it is unsupported
Emitting bvh64_intersect_ray_amd will crash the compiler on pre-GFX10_3
hardware.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11786
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30886>
2024-09-23 14:02:28 +00:00
Georg Lehmann
97aa3464b9 radv/rt: remove one VALU from traversal loop
Not much, but something, I guess?

Foz-DB Navi31:
Totals from 93 (0.12% of 79395) affected shaders:
MaxWaves: 1338 -> 1354 (+1.20%)
Instrs: 3689907 -> 3689721 (-0.01%); split: -0.01%, +0.00%
CodeSize: 18921812 -> 18922920 (+0.01%); split: -0.00%, +0.01%
VGPRs: 9012 -> 8988 (-0.27%)
Latency: 23153748 -> 23167640 (+0.06%)
InvThroughput: 4490882 -> 4493136 (+0.05%)
Copies: 287888 -> 287728 (-0.06%)
VALU: 2022082 -> 2021916 (-0.01%)
SALU: 458904 -> 459064 (+0.03%)

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30342>
2024-07-25 20:24:23 +00:00
Mike Blumenkrantz
2aaa6ebba1 build/amd: add amd-use-llvm build option
this allows amd drivers to disable llvm support while still allowing
llvmpipe/lavapipe to be built

by disabling llvm support in amd drivers, the load times for these drivers
decreases by 5-10ms

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28969>
2024-05-30 19:05:00 +00:00
Samuel Pitoiset
87fde606cf radv: use SPDX-License-Identifier
Only files under src/amd/vulkan/** are concerned.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28599>
2024-04-08 07:17:31 +00:00
Samuel Pitoiset
896c9cf486 radv: remove radv_device::physical_device
Get the logical device object using the base object.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28500>
2024-04-01 16:35:09 +00:00
Samuel Pitoiset
310597cab6 radv: rename radv_physical_device::rad_info to info
The extra rad_ prefix isn't necessary and it's longer to type.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28500>
2024-04-01 16:35:09 +00:00
Konstantin Seurer
0e8f2eca9e radv/rt: Inline constant information about ray flags
Totals from 4 (1.06% of 379) affected shaders:
Instrs: 2868 -> 2588 (-9.76%)
CodeSize: 15184 -> 13544 (-10.80%)
Latency: 69280 -> 86922 (+25.46%)
InvThroughput: 11546 -> 14486 (+25.46%)
Copies: 680 -> 592 (-12.94%)
Branches: 148 -> 128 (-13.51%)
PreSGPRs: 216 -> 172 (-20.37%)

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27137>
2024-03-07 20:07:07 +00:00
Friedrich Vock
7792ee1c15 radv/rt: Fix frontface culling with emulated RT
We need to preserve the divisor's sign for front/backface detection to
work correctly.

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28034>
2024-03-07 10:34:52 +00:00
Konstantin Seurer
767f628079 radv/rra: Dump basic ray history tokens
This only dumps the begin tokens. Tokens are written to a buffer
containing a 12 byte header at the beginning.

We use an intermediate format for the ray history tokens because the RRA
format is very inefficient.

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25548>
2024-02-03 12:59:47 +01:00
Konstantin Seurer
f27a973283 radv/rt: Use doubles inside intersect_ray_amd_software_tri
Increasing precision lets us avoid duplicate hits on shared edges. The
previous biasing WA made a bunch of tests fail.

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27243>
2024-01-30 10:26:43 +01:00
Konstantin Seurer
c511b8968a radv: Implement VK_KHR_ray_tracing_position_fetch
Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26895>
2024-01-05 18:20:20 +00:00
Samuel Pitoiset
f30fa9dec8 radv: move radv_rt_{common,shader} files to nir/
These files contains NIR lowering code for RT.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26851>
2024-01-03 09:40:29 +00:00
Renamed from src/amd/vulkan/radv_rt_common.c (Browse further)