- Feb 21, 2025
-
-
Emma Anholt authored
This generates one extra instruction to set the rounding mode to RTE due to f2f16_rtne in the lowering. This changes the result for fquantize2f16(65505.0) from 65536 to 65504, which is perhaps better but still incorrect (should be +inf).
-
- Feb 19, 2025
-
-
Just set it explicitly to false to keep the old behavior before the recent core changes in dc1b719e. Signed-off-by:
Pavel Ondračka <pavel.ondracka@gmail.com> Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <!33545>
-
This was already done in the fp paths, but was missed here. With this in place drivers can report compile failures also from other shader stages besides fragment shaders. Signed-off-by:
-
See comment. Fixes: 666a6eb8 ("radeonsi/gfx12: disable display dcc for front buffer rendering") Closes: #12552 Reviewed-by:
-
Label names are duplicated, so free them before resetting or deleting the cmdbuffer/queue. Fixes leaks when testing dEQP-VK.api.debug_utils.long_labels. Reviewed-by:
Faith Ekstrand <faith.ekstrand@collabora.com> Reviewed-by:
Samuel Pitoiset <samuel.pitoiset@gmail.com> Signed-off-by:
Juan A. Suarez Romero <jasuarez@igalia.com> Part-of: <!33611>
-
They operate on VAs. Signed-off-by:
-
Signed-off-by:
-
VK_KHR_acceleration_structure requires the BDA feature to be enabled and this enables the global BO list. Signed-off-by:
-
The BOs are already added in radv_CmdCopyImage2KHR(). Signed-off-by:
-
Signed-off-by:
-
Signed-off-by:
-
This broke with the descriptor rework and it will never work because uniform buffers are bindless now. Reviewed-by:
Mike Blumenkrantz <michael.blumenkrantz@gmail.com> Part-of: <!32963>
-
Reviewed-by:
-
Reviewed-by:
-
Reviewed-by:
-
Reviewed-by:
-
Emitting scalar instructions reduces compile time. Reviewed-by:
-
Reviewed-by:
-
There are no dynamic swizzles so it is enough to apply them during translation. This also gets rid of the extract/insertvalue spam. Reviewed-by:
-
Annotating access that does not need bounds checking allows the backend to omit bounds checking which results in code that compiles and runs faster. Reviewed-by:
-
The driver woiuld have to run nir_opt_algebraic before instruction selection. Reviewed-by:
-
This will be used by lavapipe for skipping bounds checking when robustness is disabled. Reviewed-by:
-
Speeds up compilation and it should also run faster. Reviewed-by:
-
Crashes divergence analysis. Reviewed-by:
-
This moves lowering vulkan resource indices out of gallivm into lavapipe where it should be. Reviewed-by:
-
The number of lavapipe nir passes keeps growing so it's a good idea to add some separation before it becomes a mess. Reviewed-by:
-
Reviewed-by:
-
Loading from a "zero" address instead speeds up compilation. Reviewed-by:
-
Lowering them to 32 bit integers increases compile time by a lot because llvm will optimize them to i1. Reviewed-by:
-
Reviewed-by:
-
Reviewed-by:
-
The soa backend will work on very different looking NIR in the future. Reviewed-by:
-
It shouldn't be only added at creation time. Cc: mesa-stable Signed-off-by:
-
Symlinking $CARGO_HOME to /usr/local/bin made rustup uninstaller delete the entire folder, causing mysterious build errors, so let's do the traditional .cargo/env sourcing to make rustup available to the rest of the build scripts. Also make sure that required scripts run the shell's rcfile to be able to setup the PATH correctly. Signed-off-by:
Guilherme Gallo <guilherme.gallo@collabora.com> Part-of: <!33519>
-
Use `rustup self uninstall -y` instead of manually removing folders to ensure a proper cleanup of the rustup installation, including cargo and init command injections in shell rc files. Failing to do so can cause issues, such as bash failing to run in a `set -e` environment due to a missing `$HOME/.cargo/env`, for example. Signed-off-by:
-
as well as the prior versions promoted from. Signed-off-by:
Yiwei Zhang <zzyiwei@chromium.org> Part-of: <!33594>
-
This is purely on the driver side. Signed-off-by:
-
Signed-off-by:
-
Signed-off-by:
-
Signed-off-by:
-
