[Mesa-dev] [RFC PATCH] gallium: add interface for EQAA
Roland Scheidegger
sroland at vmware.com
Tue May 1 14:48:22 UTC 2018
Am 01.05.2018 um 01:43 schrieb Marek Olšák:
> From: Marek Olšák <marek.olsak at amd.com>
>
> This is a hypothetical interface for EQAA (a superset of CSAA). CSAA could be
> exposed via GL_NV_framebuffer_multisample_coverage. EQAA additionally removes
> the restriction that the number of samples in all FBO attachments must match,
> which means it allows arbitrary sample counts in each FBO attachment.
The idea sounds ok to me, as long as you don't expose the additional
modes with more samples as ordinary msaa modes.
(I'm quite sure at some point nvidia exposed 16x msaa modes through
ordinary means, which then gave you really 4 color samples and 12
coverage-only samples, in some not-very-compliant way. Though IIRC the
extension is actually older than the ability to do per-sample msaa fb
reads, so it wasn't as problematic back then.)
> ---
> src/gallium/docs/source/screen.rst | 17 +++++++++++++++--
> src/gallium/include/pipe/p_defines.h | 1 +
> src/gallium/include/pipe/p_state.h | 3 ++-
> 3 files changed, 18 insertions(+), 3 deletions(-)
>
> diff --git a/src/gallium/docs/source/screen.rst b/src/gallium/docs/source/screen.rst
> index 3837360fb40..28934c2f7b9 100644
> --- a/src/gallium/docs/source/screen.rst
> +++ b/src/gallium/docs/source/screen.rst
> @@ -398,20 +398,22 @@ The integer capabilities:
> * ``PIPE_CAP_LOAD_CONSTBUF``: True if the driver supports TGSI_OPCODE_LOAD use
> with constant buffers.
> * ``PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS``: Any TGSI register can be used as
> an address for indirect register indexing.
> * ``PIPE_CAP_TILE_RASTER_ORDER``: Whether the driver supports
> GL_MESA_tile_raster_order, using the tile_raster_order_* fields in
> pipe_rasterizer_state.
> * ``PIPE_CAP_MAX_COMBINED_SHADER_OUTPUT_RESOURCES``: Limit on combined shader
> output resources (images + buffers + fragment outputs). If 0 the state
> tracker works it out.
> +* ``PIPE_CAP_EQAA_COLOR_SAMPLE_SUPPORT_MASK``: If the i-th bit is set, EQAA
> + supports (i+1) color samples.
I agree with Nicolai here, this looks strange, this should instead query
for total samples or something.
> * ``PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET``:
> Whether pipe_vertex_buffer::buffer_offset is treated as signed. The u_vbuf
> module needs this for optimal performance in workstation applications.
> * ``PIPE_CAP_CONTEXT_PRIORITY_MASK``: For drivers that support per-context
> priorities, this returns a bitmask of PIPE_CONTEXT_PRIORITY_x for the
> supported priority levels. A driver that does not support prioritized
> contexts can return 0.
> * ``PIPE_CAP_FENCE_SIGNAL``: True if the driver supports signaling semaphores
> using fence_server_signal().
> * ``PIPE_CAP_CONSTBUF0_FLAGS``: The bits of pipe_resource::flags that must be
> @@ -743,22 +745,33 @@ Modern APIs allow using buffers as shader resources.
> (1 for 1D or 1D array textures).
>
> **depth0** the depth of the base mip level of the texture
> (1 for everything else).
>
> **array_size** the array size for 1D and 2D array textures.
> For cube maps this must be 6, for other textures 1.
>
> **last_level** the last mip map level present.
>
> -**nr_samples** the nr of msaa samples. 0 (or 1) specifies a resource
> -which isn't multisampled.
> +**nr_samples**: For Z/S, this is the number of samples. For color, if EQAA
> +is unsupported, this is the number of both coverage samples and color samples.
> +If EQAA is supported, this is the number of coverage samples. 0 and 1
> +specify a resource which isn't multisampled.
I think you should keep nr_samples alone, and re-change the meaning to
actually be "real" samples with all associated data (which is what
everybody expects with msaa).
> +
> +**nr_color_samples**: This is the number of color samples for EQAA, while
> +``nr_samples`` is the number of coverage samples. If the format is Z/S,
> +``nr_color_samples`` is ignored. Constraints:
> +* ``nr_color_samples`` must not be greater than ``nr_samples``.
> +* If ``nr_color_samples`` is equal to ``nr_samples``, it is called MSAA.
> +* If ``nr_color_samples`` is less than ``nr_samples``, it is called EQAA.
> +* If ``nr_color_samples`` is equal to 1, the behavior of the resolve blit is
> +driver-dependent.
>
Hence instead use something like "total_samples" or "coverage samples"
here (albeit the latter could be a bit confusing, unclear if it means
all samples or just the extra ones).
If you'd use something like "number of additional coverage samples"
instead, then you could even leave all state trackers alone, since 0
would automatically mean ordinary msaa without crazy stuff.
And fwiw I'd stick to the CSAA name - coverage at least has some
meaning, whereas "enhanced" (which is what the "E" in EQAA is for
according to some old articles) really is just completely meaningless
marketing term. There really isn't any difference, ok the ability to use
different amount of samples per rt might be something, but you can't
expose this anyway (AMD didn't come up with an extension for this since
the 7+ years it's supported (since NI), so I have some doubts of its
usefulness - besides you could have another cap which then would say if
that's valid or not if you really come up with an extension specifically
for it.)
Unless there's some non-technical reasons you can't use the term CSAA.
Roland
> **usage** one of the :ref:`PIPE_USAGE` flags.
>
> **bind** bitmask of the :ref:`PIPE_BIND` flags.
>
> **flags** bitmask of PIPE_RESOURCE_FLAG flags.
>
>
>
> resource_changed
> diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
> index c4ae0532060..97e1a3a3d42 100644
> --- a/src/gallium/include/pipe/p_defines.h
> +++ b/src/gallium/include/pipe/p_defines.h
> @@ -783,20 +783,21 @@ enum pipe_cap
> PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION,
> PIPE_CAP_POST_DEPTH_COVERAGE,
> PIPE_CAP_BINDLESS_TEXTURE,
> PIPE_CAP_NIR_SAMPLERS_AS_DEREF,
> PIPE_CAP_QUERY_SO_OVERFLOW,
> PIPE_CAP_MEMOBJ,
> PIPE_CAP_LOAD_CONSTBUF,
> PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS,
> PIPE_CAP_TILE_RASTER_ORDER,
> PIPE_CAP_MAX_COMBINED_SHADER_OUTPUT_RESOURCES,
> + PIPE_CAP_EQAA_COLOR_SAMPLE_SUPPORT_MASK,
> PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET,
> PIPE_CAP_CONTEXT_PRIORITY_MASK,
> PIPE_CAP_FENCE_SIGNAL,
> PIPE_CAP_CONSTBUF0_FLAGS,
> PIPE_CAP_PACKED_UNIFORMS,
> };
>
> /**
> * Possible bits for PIPE_CAP_CONTEXT_PRIORITY_MASK param, which should
> * return a bitmask of the supported priorities. If the driver does not
> diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
> index 4dce399f848..4010b92e67b 100644
> --- a/src/gallium/include/pipe/p_state.h
> +++ b/src/gallium/include/pipe/p_state.h
> @@ -503,41 +503,42 @@ struct pipe_box
> int16_t depth;
> };
>
>
> /**
> * A memory object/resource such as a vertex buffer or texture.
> */
> struct pipe_resource
> {
> struct pipe_reference reference;
> - struct pipe_screen *screen; /**< screen that this texture belongs to */
>
> unsigned width0; /**< Used by both buffers and textures. */
> uint16_t height0; /* Textures: The maximum height/depth/array_size is 16k. */
> uint16_t depth0;
> uint16_t array_size;
>
> enum pipe_format format:16; /**< PIPE_FORMAT_x */
> enum pipe_texture_target target:8; /**< PIPE_TEXTURE_x */
> unsigned last_level:8; /**< Index of last mipmap level present/defined */
> unsigned nr_samples:8; /**< for multisampled surfaces, nr of samples */
> + unsigned nr_color_samples:8; /**< Number of color samples for EQAA. */
> unsigned usage:8; /**< PIPE_USAGE_x (not a bitmask) */
>
> unsigned bind; /**< bitmask of PIPE_BIND_x */
> unsigned flags; /**< bitmask of PIPE_RESOURCE_FLAG_x */
>
> /**
> * For planar images, ie. YUV EGLImage external, etc, pointer to the
> * next plane.
> */
> struct pipe_resource *next;
> + struct pipe_screen *screen; /**< screen that this texture belongs to */
> };
>
>
> /**
> * Transfer object. For data transfer to/from a resource.
> */
> struct pipe_transfer
> {
> struct pipe_resource *resource; /**< resource to transfer to/from */
> unsigned level; /**< texture mipmap level */
>
More information about the mesa-dev
mailing list