This reverts commit 5572737d95.
Reason for revert: Adreno 4xx bots hitting compatibility assert in InitialTextureClear test
Original change's description:
> Pass GrColorType to the GrGpu::wrapRenderableBackendTexture chain of calls
>
> This is a step towards reducing our reliance-on/use-of the GrPixelConfig stored in the GrBackendTexture.
>
> Bug: skia:6718
> Change-Id: I2170032bfbbb57423c4bb0d901ad014c61d38131
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/223701
> Commit-Queue: Robert Phillips <robertphillips@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
TBR=egdaniel@google.com,robertphillips@google.com
Change-Id: I24cf6b3aa0dfca2e935a36592860ad91171b21a7
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:6718
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227637
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This is a step towards reducing our reliance-on/use-of the GrPixelConfig stored in the GrBackendTexture.
Bug: skia:6718
Change-Id: I2170032bfbbb57423c4bb0d901ad014c61d38131
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/223701
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Change-Id: If581c8ceeaa76985535cb7b6772742f0011cfe8e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227436
Commit-Queue: Hal Canary <halcanary@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Auto-Submit: Hal Canary <halcanary@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Change-Id: I171a680aac554a0015d1854c46b35e9c9785fdf3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227061
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Previously we would just run coverage count transformations on
everything, including cached literal coverage atlases. This was
wasteful since it isn't necessary if the atlas already has literal
coverage. MSAA mode will introduce even more atlases that don't need
coverage count transformations, so it's definitely time to clean this
up.
Bug: skia:
Change-Id: Ifc72eaa7cbd4ab5e4ef4acb5610117ae9f54e4c1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227144
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Basically the same deal as aarch64:
- a bunch of instructions to rewrite control
flow to be two loops, body and tail
- a bunch of instructions to support scalar
loads and stores in the tail
We can now remove the JIT::mask field.
I've removed the SkUNREACHABLE I'd put in for the ARM code... as
written the interpreter is still reachable by the loser if two threads
race to JIT the program. Medium term I plan to move JIT compilation to
a more proactive time, eliminating the need for the lock and letting the
interpreter become truly unreachable.
I had a little bit of a false start with what instructions to use for
scalar load8 and store8, first starting with instructions that loaded
via GP registers, then remembering vpinsrb and vpextrb can take a memory
argument, loading into xmm directly. I've left the first instructions I
used in the file, still implemented but only used from the unit tests.
They're pretty common and will probably be useful some day.
Change-Id: I471b13026af4b1c6e861a53159f9df5f0285447c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227178
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
I had been setting the REX R bit to select high registers,
but you actually set the B bit. Don't know how I got that
wrong before... the leading byte should be 49 not 4c.
$ cat test.s
foo:
addq $7, %r8
$ clang -c test.s && objdump -d test.o
0000000000000000 <foo>:
0: 49 83 c0 07 add $0x7,%r8
Change-Id: I039e1c4f4ea20523a1e2cc9bcf5f6d9321a6223b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227177
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This reverts commit 8c5c54ceaf.
Reason for revert: bad ANGLE diffs, Chrome roll failing
Original change's description:
> Use saveLayer-determined CT in SkGpuDevice::onCreateDevice
>
> Currently Ganesh always uses the prev render target CT, but saveLayer
> may force a different CT, depending on flags.
>
> This fixes handling of saveLayer/kF16ColorType (added in
> https://skia-review.googlesource.com/c/skia/+/227066).
>
> Change-Id: I8e73fdfdf18eb61f97fadd0504a4bd0f4a71977d
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227258
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Florin Malita <fmalita@chromium.org>
TBR=bsalomon@google.com,fmalita@chromium.org
Change-Id: I0c3e74be64d5286b8d95b2aabaa464ad97fcaccf
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227262
Reviewed-by: Florin Malita <fmalita@google.com>
Commit-Queue: Florin Malita <fmalita@google.com>
Currently Ganesh always uses the prev render target CT, but saveLayer
may force a different CT, depending on flags.
This fixes handling of saveLayer/kF16ColorType (added in
https://skia-review.googlesource.com/c/skia/+/227066).
Change-Id: I8e73fdfdf18eb61f97fadd0504a4bd0f4a71977d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227258
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Florin Malita <fmalita@chromium.org>
Add a tail loop to handle elements one at a time.
Just like in the interpreter, the only instructions
that need to be changed are the loads and stores,
16 byte -> 4 byte and 4 byte -> 1 byte.
With this we can mark the interpreter as SkUNREACHABLE,
and it even completely compiles away, saving a few KB.
Example profile for the SkVMTool float-squaring program
running N=15 over and over:
Samples│
│ skvm-jit-3663518994():
42 │40: cmp x0, #0x4
│44: ↓ b.lt 60
51 │48: ldr q0, [x1]
197 │4c: mul v0.4s, v0.4s, v0.4s
135 │50: str q0, [x1]
│54: add x1, x1, #0x10
43 │58: sub x0, x0, #0x4
│5c: b.al 40
150 │60: ↓ cbz x0, 7c
67 │64: ldr s0, [x1]
130 │68: mul v0.4s, v0.4s, v0.4s
135 │6c: str s0, [x1]
18 │70: add x1, x1, #0x4
17 │74: sub x0, x0, #0x1
20 │78: b.al 60
124 │7c: ← ret
Change-Id: I153d7bc247942366a686e30a9cad60c935f754ed
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227138
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
These didn't work correctly, and they're extremely tricky to get right
in the vectorized execution model (vs. structured control flow). As a
side effect, determine the maximum stack depth used for the execution
masking - the same idea will be used for the primary stack in a later
CL. Add a unit test to verify the new restriction, and fix two places
that were relying on this feature before.
In addition, boolean external values need to be masks. I may implement
this in the code-gen at some point, but this is already a fringe
feature, so just fix the one unit test for now.
Change-Id: I9607ffaf67c7795dbf42e4009180aea8f3e65c44
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226849
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This turned out to be quite an easy transformation
with yesterday's work already done. No codegen changes.
Change-Id: Ife19ab7731514c54cfed963a6d2e9b1ec2246997
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227137
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Also rename from GrGLSizedInternalFormat since the compressed formats
aren't really sized internal formats.
Change-Id: I8744f1f4b8156300ab69d89066b44913bce120a3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226956
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Found by a GCC warning that
fPhase = that.fPhase; // line 193
was sometimes copying over uninitialized data.
Seems pretty harmless to initialize.
Change-Id: I9011ccf965bea4a702e0cca61f89e0aa644c55a6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227122
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
We were checking 'externalType' twice, and never checking
'internalFormat'. We now check each of them once.
Change-Id: Iae0578a4828fd2a9224e4924245bea0708d8c931
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227028
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit aed8009a6d.
Reason for revert: Flutter's version of Wuffs has been updated. See https://github.com/flutter/engine/pull/9466#issuecomment-510639898
Original change's description:
> Revert "Update Wuffs version"
>
> This reverts commit 42ece2b7c9.
>
> Reason for revert: Requiring the latest version of wuffs broke the flutter roll.
>
> Original change's description:
> > Update Wuffs version
> >
> > The primary purpose of this commit is to track upstream Wuffs more
> > closely.
> >
> > A side effect is to pull in the Wuffs commit
> > 5bea867f72
> > "Allow an LZW literal width of 1", which eliminates a difference between
> > the old third_party/gif decoder and the new third_party/wuffs decoder.
> >
> > As the CodecTest.cpp comment says, the GIF spec explicitly says that the
> > LZW literal width should be at least 2, but in practice, GIF encoders
> > violate the spec. After that upstream commit, Wuffs has followed other
> > GIF decoders in being more liberal in what it accepts.
> >
> > Codec_InvalidAnimated therefore no longer has a separate "#ifdef
> > SK_HAS_WUFFS_LIBRARY" section. The first frame of the test's GIF image
> > data, being the required frame of the third frame, no longer has an
> > invalid LZW literal width according to Wuffs.
> >
> > Bug: skia:8235
> > Change-Id: Ie94537f5232128ffc1d1547f4c0b84992e54ab02
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226476
> > Commit-Queue: Leon Scroggins <scroggo@google.com>
> > Reviewed-by: Leon Scroggins <scroggo@google.com>
>
> TBR=scroggo@google.com,nigeltao@google.com
>
> Change-Id: I9e636e81f57eefd836a53738872ddb9f5c9b13c3
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:8235
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226697
> Reviewed-by: Leon Scroggins <scroggo@google.com>
> Commit-Queue: Leon Scroggins <scroggo@google.com>
TBR=scroggo@google.com,nigeltao@google.com
Change-Id: Ibeeea1cf9c2e210b5e49dec65037ec8a494209de
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:8235
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226851
Reviewed-by: Leon Scroggins <scroggo@google.com>
Reviewed-by: Nigel Tao <nigeltao@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Support forward references in Label.
In addition to tracking the current Label offset (used for
backward references essentially just the same as before this CL)
we also store a list of instructions that refer to each Label.
When a Label moves, each instruction gets a new displacement.
To make this a little easier, remove the 8-bit jump form on x86...
this way all x86 displacements are 32-bit and and all ARM 19-bit.
For now only cbz() supports this, just to start somewhere.
More to do but it's worth an early design review.
Change-Id: I23d2bcd7742965ab694ae4828f53409cb9fc807f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226937
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Adds a separate resolve texture to GrMtlRenderTarget, which can be used
to do a resolve for the main multisample color texture. The resolve is
handled by setting a special Store action for the RenderCommandEncoder.
Bug: skia:8243
Change-Id: I1ffd756c01a9b363116ffefee2c4c50ba9a3e637
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/225536
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Also adds the following to the format table:
*base internal format
*sized internal format
*compressed internal format
*the internal format to use with glTexImage
*the external type to use with glTexImage when there is no data to upload
(i.e., no GrColorType for data)
Bug: skia:6718
Change-Id: Ica51a8d4588bd24078c7d61805f6eef9c02ae14d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226558
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
MakeFromBackendTextureAsRenderTarget is planned to be deprecated, so we
should use MakeFromBackendTexture with a sampleCount parameter instead.
On Vulkan, this ran into issues because we assumed an allocation for the
VkImage and the swapchain doesn't provide us with one. Fixed so we don't
need an allocation for Borrowed textures.
Bug: skia:
Change-Id: Ib26888020e093f4a734a4159eae898539c2273b7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226839
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Change-Id: I231949c90342a44e9c0b6030818139d3164819da
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226837
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This adds a bunch of instructions we'll need to handle the N < 4 tail
within the JIT code on ARM.
- ldrb/strb are 1-byte load and stores
- sub subtracts without setting flags
- cmp just sets flags (actually just subs with an xzr destination)
- add b and b.lt, just like b.ne
- cbz and cbnz... we only need cbz but I accidentally did cbnz first
Once I add support for forward jumps, we'll be able to use these
instructions to restructure the loop to
entry:
hoisted setup
loop:
if N < 4, jump tail (cmp N,#4; b.lt tail)
... handle 4 values ...
jump loop (b loop)
tail:
if N == 0, jump end (cbz N, end)
... handle 1 value ...
jump tail (b tail)
end:
ret
Change-Id: I62d2d190f670f758197a25d99dfde13362189993
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226828
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
We conflate luminance (dot of linear r,g,b) and luma (dot of
gamma-encoded r',g',b') all over the place in Skia. This one is
possibly the most confusing of any of them, in that the dot-product
coefficients for BT.709 luma were somewhat arbitrarily chosen to be the
same as the coefficients for luminance.
So, big old ridiculously new long name to make it clear it works
for both luma and luminance, but that it's hardcoded to BT.709.
My quick reference when I forget things:
https://en.wikipedia.org/wiki/Luma_(video)
Change-Id: I5a6567de296795f558acc5dd3c39974b8035234e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226762
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
As the powers-of-2 get larger the coarse binning can burn a lot of VRAM.
Granted it isn't the best metric but, with this CL, the number of textures created and scratch textures reused remains unchanged when running the GMs.
Change-Id: I84abbbae0ed01aabb387671b5ee0e4fcdb82b671
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226226
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This reverts commit 42ece2b7c9.
Reason for revert: Requiring the latest version of wuffs broke the flutter roll.
Original change's description:
> Update Wuffs version
>
> The primary purpose of this commit is to track upstream Wuffs more
> closely.
>
> A side effect is to pull in the Wuffs commit
> 5bea867f72
> "Allow an LZW literal width of 1", which eliminates a difference between
> the old third_party/gif decoder and the new third_party/wuffs decoder.
>
> As the CodecTest.cpp comment says, the GIF spec explicitly says that the
> LZW literal width should be at least 2, but in practice, GIF encoders
> violate the spec. After that upstream commit, Wuffs has followed other
> GIF decoders in being more liberal in what it accepts.
>
> Codec_InvalidAnimated therefore no longer has a separate "#ifdef
> SK_HAS_WUFFS_LIBRARY" section. The first frame of the test's GIF image
> data, being the required frame of the third frame, no longer has an
> invalid LZW literal width according to Wuffs.
>
> Bug: skia:8235
> Change-Id: Ie94537f5232128ffc1d1547f4c0b84992e54ab02
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226476
> Commit-Queue: Leon Scroggins <scroggo@google.com>
> Reviewed-by: Leon Scroggins <scroggo@google.com>
TBR=scroggo@google.com,nigeltao@google.com
Change-Id: I9e636e81f57eefd836a53738872ddb9f5c9b13c3
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:8235
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226697
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Fix: Use non-zero rowbytes when uploading zeros for initial texture clear
This reverts commit f16020ba1b.
Bug: chromium:981254
Change-Id: Iafd5893dd1b397ec1d91c64d48d46059e62488a7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226557
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: b/137017906
Change-Id: I98bfc9607fa78536e1cf44707636558ed1980b4a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226616
Commit-Queue: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This is basically a no-op CL, with two changes designed to make
SkVM lifetime analysis easier to reason about:
1) rename Analysis.life to Analysis.death,
since it indicates the instruction when the value dies
2) use death == 0 to mark dead code instead of an NA sentinel
The life->death change really is just a naming thing, something I
realized makes a little more sense this way around after talking about
how this code works out loud. 'death' holds the time of a Val's death.
The second change also isn't very important, but I realized that there
is a perfectly good in-band value to mark dead code, 0, and there's no
need to use NA as a sentinel. If we mark an instruction as death == 0,
that indicates the value is needed only until instruction 0 executes,
i.e. never. They're sort of pre-dead. This cuts one of the overloads
for what NA means in SkVM.cpp; now it only means "no value".
All the code that tested against the NA sentinel now tests against 0
instead. We could go a step further and rewrite the tests to be death >
id for live code or death < id for dead code, but that amounts to
roughly the same thing in the end: instructions either live for some
time with a death that is later than their own ID, or are dead with
death == 0. There is some small ambiguity around whether we should mark
store instruction's deaths as id or id+1 and which of the tests should
be <= or >=, but checking against 0 makes that all moot, and I think the
checks also stand out more clearly with the literal '0' in the code.
This is a little warmup to refamiliarize myself with the code, with an
ultimate goal of moving hoisting and register assignment to the backends
so they can be influenced by instruction selection: platform specific
ways to handle immediates or splats, choosing destination registers and
instructions that play well with the available argument registers, etc.
Change-Id: I6978abf0bd01dcd0e7a142d632826e7692060ade
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226549
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: I23a88524d741cecefa16ae7e364d2294db1c6030
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226508
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The primary purpose of this commit is to track upstream Wuffs more
closely.
A side effect is to pull in the Wuffs commit
5bea867f72
"Allow an LZW literal width of 1", which eliminates a difference between
the old third_party/gif decoder and the new third_party/wuffs decoder.
As the CodecTest.cpp comment says, the GIF spec explicitly says that the
LZW literal width should be at least 2, but in practice, GIF encoders
violate the spec. After that upstream commit, Wuffs has followed other
GIF decoders in being more liberal in what it accepts.
Codec_InvalidAnimated therefore no longer has a separate "#ifdef
SK_HAS_WUFFS_LIBRARY" section. The first frame of the test's GIF image
data, being the required frame of the third frame, no longer has an
invalid LZW literal width according to Wuffs.
Bug: skia:8235
Change-Id: Ie94537f5232128ffc1d1547f4c0b84992e54ab02
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226476
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
This reverts commit abb5a315af.
Two fixes:
1) GMs pass valid rowBytes when calling directly to GrGpu.
2) Check for non-null data before trying to set UNPACK_ROW_LENGTH
Bug: chromium:981254
Change-Id: I24e46b0d2b14562d6b84a29fefe3410ce5c06c94
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226498
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
I think this is the minimum rule that's easy to understand when writing
SkSL for the interpreter that ensures we'll be able to statically
determine total stack usage of a particular function.
While writing the new test, I also noticed that we still return
(invalid) byte code, even when there are errors. Fixed that.
Change-Id: I625a8592c9ba1656074e5f0d4227d41968af7b37
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226218
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
