- 28 Jul, 2022 9 commits
-
-
Italo Nicola authored
-
Italo Nicola authored
This commit can be dropped. Signed-off-by:Italo Nicola <italonicola@collabora.com>
-
Italo Nicola authored
Signed-off-by: -
Italo Nicola authored
Signed-off-by:
Gert Wollny <gert.wollny@collabora.com>
-
Italo Nicola authored
This is needed for separable shaders when two consecutive stages have a different amount of inputs. Signed-off-by:
-
Italo Nicola authored
Some of the fields are not needed for separable shaders. Signed-off-by:
-
Italo Nicola authored
Implement support for compiling shaders in single-stage programs on the host, in an attempt to reduce the number of unnecessary recompilation in games that also make use of this feature. Signed-off-by:
-
Italo Nicola authored
The `vrend_use_program` function is now a trivial wrapper, but we will keep it that way since in the next couple commits we will have to modify it to add support for program pipelines. Signed-off-by:
-
Italo Nicola authored
This function isn't used everytime we want to link shaders, it's only used to force a link command as soon as possible. This name is more descriptive. Signed-off-by:
-
- 26 Jul, 2022 1 commit
-
-
Gert Wollny authored
Signed-off-by:
-
- 21 Jul, 2022 1 commit
-
-
Signed-off-by:
Ryan Neph <ryanneph@google.com> Reviewed-by:
Yiwei Zhang <zzyiwei@chromium.org> Part-of: <!859>
-
- 19 Jul, 2022 2 commits
-
-
Ryan Neph authored
vrend_state.current_hw_context must be invalidated when query checking requires a sub context switch, otherwise gl context switching doesn't actually occur. Fixes: 48525f0e ("vrend: Track sub-ctx in queries") Signed-off-by:
-
Add the sub context id to the query structure and set it current before checking the query. This is necessary as switching context is not enough to restore the GL context associated to a query object. Signed-off-by:
Antonio Caggiano <antonio.caggiano@collabora.com>
Reviewed-by:
-
- 13 Jul, 2022 2 commits
-
-
Corentin Noël authored
Signed-off-by:
Corentin Noël <corentin.noel@collabora.com> Reviewed-by:
David Heidelberg <david.heidelberg@collabora.com>
Part-of: <virgl/virglrenderer!768> -
Corentin Noël authored
Do not hardcode the project id to be able to use forks of mesa which do not share the same project id. Signed-off-by:
David Heidelberg <david.heidelberg@collabora.com>
Part-of: <!768>
-
- 11 Jul, 2022 2 commits
-
-
Italo Nicola authored
This includes a patch disabling SSOs for builtin shaders and a bugfix for OES_EGL_image_external_essl3. I also added a few flaky tests to the expectations. Signed-off-by:
-
We didn't handle the start instance, and we also have to take into accoun that the start instance is not zero, but the instance count is one. This fixes rendering errors with Transport Fever. Signed-off-by:
-
- 08 Jul, 2022 1 commit
-
-
We need to signal to the guest that we are reporting the right number of uniforms now. Signed-off-by:
-
- 07 Jul, 2022 1 commit
-
-
Igor Torrente authoredSigned-off-by:
Igor Torrente <igor.torrente@collabora.com>
Part-of: <!850>
-
- 06 Jul, 2022 3 commits
-
-
Corentin Noël authored
This contains a cts update and a change to propagate the max uniforms. Part-of: <!849>
-
There is no need to handle it as input or output. Signed-off-by:
-
The patch is only allowed as TCS output and TES input, so there is no need to query the prev and next shaders type to deduct a name. Instead always use "patch". Signed-off-by:
-
- 05 Jul, 2022 1 commit
-
-
Gert Wollny authored
We did a mistake in the guest that the TGSI_SEPARABLE_SHADER property would be send without the host actually understanding this proeprty, so add a flag that can signal this support. Hold off in setting it until we have some actual code to deal with the property and possibly with additional information. Signed-off-by:
-
- 28 Jun, 2022 3 commits
-
-
Corentin Noël authored
Allows to only reduce the max number of uniform blocks by 1 instead of slightly reducing the number of uniforms. Signed-off-by:
-
Corentin Noël authored
We are now using an UBO internally in any case. Signed-off-by:
-
Many of these are actual tests we want to try. Part-of: <virgl/virglrenderer!842>
-
- 27 Jun, 2022 2 commits
-
-
If the client explicitly asks for dma_buf export, without the real dma_buf export support, we must fallback to use gbm bo import path, instead of using opaque fd export. Signed-off-by:
Chad Versace <chadversary@chromium.org> Part-of: <!825>
-
If there's external memory import, then the virgl resource has to exist already. We don't need to force export since the resource_map will use the existing virgl resource directly. Signed-off-by:
-
- 24 Jun, 2022 3 commits
-
-
Signed-off-by:
-
Signed-off-by:
-
Copy in auto-generated protocol bindings. Signed-off-by:
-
- 21 Jun, 2022 2 commits
-
-
Gert Wollny authored
v2: Update expectations Signed-off-by:
-
Gert Wollny authored
With that we can pin the precise specifier to the values that are created by the operations that are tagged as precise without touching all temporaries. Signed-off-by:
-
- 17 Jun, 2022 5 commits
-
-
Treating all GL errors as fatal has lead to several unexpected production breakages, where ignored errors were otherwise non-disruptive. For the short-term, we need an option to disable fatal error checking, while figuring out a long-term plan for mitigating production breakages. Signed-off-by:
-
Ryan Neph authored
Need this to get app-specific driver configurations with Venus. Without process isolation, Mesa's dri-conf matching still won't be correct, but there are other options that can be explored later when multiple venus contexts share a process. Signed-off-by:
-
This is not used anymore. Signed-off-by:
-
This reinstates 99f8cd3d and removes the suffix from the variable that encodes the array ID. It can't be assumed that the array IDs are always in the same order for all shaders, but since we removed the use of enhanced layouts, the array id encoding is no longer relevant, because for each SID we can have only one array. v2: fix leftover array suffix from interface matching Closes: #266 Signed-off-by:
-
Corentin Noël authored
This makes mesa transform atomic counters into ssbo which can be handled better in OpenGL ES as atomicCounter*ARB operations are not available there. Signed-off-by:
-
- 16 Jun, 2022 2 commits
-
-
Use the new summary feature and use the pkgconfig module to generate the .pc file. Increase minimum version to 0.53 (aligned with the Mesa requirements) Part-of: <!802>
-
Allow the compiler to check that we are using the right enum members and to always use the right allocation in structures. Signed-off-by:
-
