583a4210e5
virgl/ci: Update virglrenderer
...
Update virglrenderer to the latest version on time.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16627 >
2022-05-23 06:16:07 +00:00
d12b97122e
docs/asahi: Add initial driver docs
...
Emma asked for some docs on how to use wrap for when wrap gets landed.
This should do :-)
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 18:22:41 -04:00
1b7304f44b
asahi: Add wrap library
...
Add a library that wraps the key IOKit entrypoints used in the macOS
UABI for AGX. Our wrapped routines print information about the kernel
calls made and dump work submitted to the GPU using agxdecode. This code
has two major use cases:
1. Debugging Mesa, particularly around the undocumented macOS
user-kernel interface. Logs from Mesa may compared to Metal to check
that the UABI is being used correcrly.
2. Reverse-engineering the hardware, using this as glue to get at the
"interesting" GPU memory.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:58:07 -04:00
d814711155
asahi: Guard against encoder overflows
...
We don't seem to hit this yet, but it's possible so add an assert to
rule it out for failing tests.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
1d0d27aac7
asahi: Allocate a larger segment list
...
This ought to be dynamically allocated but because this is macOS kernel
UABI only gunk, let's just hack around it...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
ac09dcd358
asahi: Handle RECT textures as 2D
...
Rectangle textures are just 2D textures with unnormalized coordinates,
but we already handle unnormalized coordinates in the sampler state. So
we just need to alias RECT and 2D.
Fixes GALLIUM_HUD.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
689be69681
asahi: Fix pipeline decoding
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
8a1ee708a4
asahi: Improve render target dumping
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
6eb62c5f50
asahi: Sync some more unknown fields
...
I don't understand these structs well. Good news, though-- Asahi Lina
confirmed this is all software-defined crap in the macOS UABI.
Unfortunately, they seem to correspond to physical registers.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
d8227f09d0
asahi: Drop some unknowns
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
03a3a7843e
asahi: Identify suballocated size
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
14d966734f
asahi: Identify suballocation mechanism
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
ca9b1d48fe
asahi: Decode resource allocation requests
...
We already know much of this structure, let's pretty print to simplify
our traces.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
0c120cf141
asahi: Decode resource allocation responses
...
We already know much of this structure, let's pretty print to simplify
our traces.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
a5e774ef9e
asahi: Align segment list unknowns with Metal
...
No observable change.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
8da2fdc889
asahi: Compute segment list length
...
Quiesces agxdecode error introduced in previous commit.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
1e42e4a390
asahi: Validate segment list length
...
This is easy now!
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
343f4f886f
asahi: Split unknown field in segment list header
...
Seen as 0x8000.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
6b12292754
asahi: Add segment list pretty printer
...
Validate all the new expectations and print all the fields. This should
make differences between the drivers obvious, I hope.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:11:08 -04:00
db32b4a064
asahi: Sync some names with Project Zero
...
This should clarify a few things I didn't get independently
investigating the interface. Of coruse, I got other pieces... the sum of
the parts is better :-)
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:10:57 -04:00
6e59474ced
asahi: Identify partial render pipeline
...
Needed if a partial render is incurred from overflowing the parameter
buffer (too much geometry).
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
65500b19df
asahi: Implement polygon offset
...
This is pretty simple now that the hardware is understood. The hardware
interfaces parallels that of scissors, so the scissor path is reused
with minor modifications to accommodate the new functionality.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
666f307d41
asahi: Add a depth bias structure
...
This is simple and corresponds directly to the Metal inputs. However,
the alignment is a bit tricky, so let's add formal XML for it.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
cb760a00c2
asahi: Identify depth bias enable bit
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
7a0c220401
asahi: Identify "set depth bias index" field
...
Grouped together with the "set scissor index" field, which is natural.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
ff84c07b21
asahi: Identify depth bias array
...
"Inspired" by PowerVR code, by comparison.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
a39fb009d2
asahi: Note unknown field
...
...used internally with visibility tests, together with a weird
vertex/compute-like shader used to zero the visibility buffer.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
baa54da7d6
asahi: Relax Draw "command" check
...
Other values in the lower byte seen with multiple draws and
visibility testing.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
d8541d52d2
asahi: Add XML used for visibility tests
...
Aka occlusion queries. There is an annoying limitation in the hardware
(reflected in Metal) that only a single buffer may be bound per render
pass, with the per-draw settings merely specifying an offset.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
46a7c10349
asahi: Dump "unknown 4a" structure
...
A few fields are known at least, let's decode them correctly instead of
falling back to a hex dump.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:04 -04:00
a5e6a5e585
asahi: Identify more depth pipeline fields
...
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-22 17:09:00 -04:00
6aadd72bd4
asahi: Account for every byte of the cmdbuf
...
This adds the remaining XML. I don't know how much of this is correct,
but it nominally accounts for every byte. So there shouldn't be more
surprises in the command buffer after this.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:33 -04:00
c61e660323
asahi: Handle 3rd deflake address
...
Context switching is so broken. Just trying to get closer. Adding some
XML here to see if we're missing something else obvious.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:33 -04:00
3be0314e5a
asahi: Flush on partial clears
...
Not the fastest thing, but should be conformant. (Panfrost to this day
has this effective behaviour...)
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:33 -04:00
95a18d1c1c
asahi: Pass depth/stencil clear values to firmware
...
These need to be format-packed, but as we only support Z32F_S8 right
now, that's trivial.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
43f6c08a3b
asahi: Handle uncompressed Z32F depth buffers
...
This uses a subset of the depth/stencil infrastructure we built out to
support writing back tiled, uncompressed Z32F depth buffers to memory.
Texturing from this format is already supported.
This gets glmark2 -bshadow working.
v2: Fix partial renders
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
c955dd726e
asahi: Don't store to unbound render targets
...
Otherwise we have a state leak with depth-only attachments, as in
glmark2 -bshadow.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
a0fc622be5
asahi: Validate depth/stencil formats
...
Don't go advertising formats we don't actually support.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
e6c809d767
asahi: Use z24_in_z32f lowering
...
Thanks Dave!
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
0ea174f38c
asahi: Run lower_blend for all fragment shaders
...
Unify the paths so that colour masks are respected even when regular
blending is off (or when logic ops are used!) Fixes
dEQP-GLES2.functional.color_clear.masked_rgb
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512 >
2022-05-21 22:01:32 -04:00
aa32b96c51
turnip: fix assert for max xfb outputs
...
this is a counter, not an index, so use <=
cc: mesa-stable
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16620 >
2022-05-21 16:48:54 +00:00
2a8e6a4d1a
turnip: disable UBWC for SNORM formats
...
In copy_format, we treat snorm as unorm to avoid clamping. But snorm
and unorm are UBWC incompatible for special values such as all 0's or
all 1's. Disable UBWC for snorm.
For reference, I dumped the first byte of an UBWC blocks and it was
color UNORM SNORM
all black 0x01 0x31
all white 0x0d 0x11
@flto clarified that bit 4 is unset for fast clear encoded blocks. It
looks like fast clear is not used for SNORM.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6480
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16534 >
2022-05-21 15:27:42 +00:00
d7dc27645e
egl: dereference XCB drawable pointers
...
eglCreatePlatformWindowSurface[EXT] and
eglCreatePlatformPixmapSurface[EXT] should be passed (xcb_window_t *)
and (xcb_pixmap_t *), so we must dereference these types before using
them as drawables. We already do something similar with X11 drawable
pointers.
Signed-off-by: Jeffrey Knockel <jeff@jeffreyknockel.com >
Reviewed-by: Eric Engestrom <eric@engestrom.ch >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16269 >
2022-05-21 14:42:06 +00:00
cd8c083ab5
clc: disable opaque pointers until they are supported
...
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479 >
2022-05-21 12:26:37 +00:00
b6ed3c6ea2
clc: fix compiler features_macro CTS Test
...
Even with that alone we can't pass the test, as LLVM enables some
extensions based on the SPIR target we choose.
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479 >
2022-05-21 12:26:37 +00:00
bcc2df4890
clc: speed up compilation by not relying on opencl-c.h
...
This depends on LLVM change: https://reviews.llvm.org/D125401
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479 >
2022-05-21 12:26:37 +00:00
e5a052f75f
clc: drop parsingComplete check
...
This relies too much on the properties of the SPIRV-LLVM-Translator and is
required to load SPIR-Vs found in the OpenCL CTS.
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479 >
2022-05-21 12:26:37 +00:00
c0cf7f578a
clc: parse localSize and localSizeHint
...
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479 >
2022-05-21 12:26:37 +00:00
e8eb6d13a5
turnip: fix tu6_pack_border_color for z24
...
The value should be at the bottom 24 bits, not at the top.
dEQP-VK.pipeline.sampler.* still passes. This fixes most of
dEQP-GLES31.functional.texture_border_clamp.formats.*depth* on angle.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16570 >
2022-05-21 00:54:28 +00:00
2f2ddfa0ac
panfrost: Move patched_s out of the pan_blitter_views struct
...
The struct is returned from a function, so in debug builds the address
may change after returning, and pointers to patched_s will be broken.
Pass the pointer to the patched stencil view as a parameter to
pan_preload_get_views to avoid this.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16343 >
2022-05-20 23:17:07 +00:00
Corentin Noël