Lots of x86 instructions can take their right hand side argument from
memory directly rather than a register. We can use this to avoid the
need to allocate a register for many constants.
The strategy in this CL is one of several I've been stewing over, the
simplest of those strategies I think. There are some trade offs
particularly on ARM; this naive ARM implementation means we'll load&op
every time, even though the load part of the operation can logically be
hoisted. From here on I'm going to just briefly enumerate a few other
approaches that allow the optimization on x86 and still allow the
immediate splats to hoist on ARM.
1) don't do it on ARM
A very simple approach is to simply not perform this optimization on
ARM. ARM has more vector registers than x86, and so register pressure
is lower there. We're going to end up with splatted constants in
registers anyway, so maybe just let that happen the normal way instead
of some roundabout complicated hack like I'll talk about in 2). The
only downside in my mind is that this approach would make high-level
program descriptions platform dependent, which isn't so bad, but it's
been nice to be able to compare and diff debug dumps.
2) split Op::splat up
The next less-simple approach to this problem could fix this by
splitting splats into two Ops internally, one inner Op::immediate that
guantees at least the constant is in memory and is compatible with
immediate-aware Ops like mul_f32_imm, and an outer Op::constant that
depends on that Op::immediate and further guarantees that constant has
been broadcast into a register to be compatible with non-immediate-aware
ops like div_f32. When building a program, immediate-aware ops would
peek for Op::constants as they do today for Op::splats, but instead of
embedding the immediate themselves, they'd replace their dependency with
the inner Op::immediate.
On x86 these new Ops would work just as advertised, with Op::immediate a
runtime no-op, Op::constant the usual vbroadcastss. On ARM
Op::immediate needs to go all the way and splat out a register to make
the constant compatible with immediate-aware ops, and the Op::constant
becomes a noop now instead. All this comes together to let the
Op::immediate splat hoist up out of the loop while still feeding
Op::mul_f32_imm and co. It's a rather complicated approach to solving
this issue, but I might want to explore it just to see how bad it is.
3) do it inside the x86 JIT
The conceptually best approach is to find a way to do this peepholing
only inside the JIT only on x86, avoiding the need for new
Op::mul_f32_imm and co. ARM and the interpreter don't benefit from this
peephole, so the x86 JIT is the logical owner of this optimization.
Finding a clean way to do this without too much disruption is the least
baked idea I've got here, though I think the most desirable long-term.
Cq-Include-Trybots: skia.primary:Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Debug-All-SK_USE_SKVM_BLITTER,Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Release-All-SK_USE_SKVM_BLITTER
Change-Id: Ie9c6336ed08b6fbeb89acf920a48a319f74f3643
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/254217
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
This centralizes the logic for whether a color type is normalized.
The change in SkRasterPipelineBlitter is minor, only now properly
treating kR16G16_float_SkColorType as unnormalized.
The change in SkColorSpaceXformSteps is more extreme given the way
it had been written, with all the newer color types now correct.
I think I'm making the right call on kA16_float_SkColorType?
Are there equivalent sites to update in Ganesh?
Change-Id: I32a40b31b86c5fde0dea2528122a4deda91c5545
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/253668
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Rm unused function SkColorTypeIsGray.
Expect kGray_8 readback to work in tests.
Bug: skia:8962
Change-Id: Iad9d2be3e0a2e594e62dc681b79c59e0a833116a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/245162
Auto-Submit: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This initializes more formats correctly with solid colors and relies
less on GrPixelConfig.
It also includes the changes to get gray8 read pixels working
and almost everything to initialize GL_RGB8 textures with data. Minor
stuff to follow to get RGB8 fully working and update test expectations.
Bug: skia:8962
Bug: skia:6718
Bug: skia:9358
Change-Id: Ic044b4c4badc37f14fb46c898cd3b3c21a6fc7fd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/251199
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
typename 'B0' is defined in sys/termios.h
Bug: none
Change-Id: I6a2e6eb1a91f123459087280b70493dab795bbd3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/249917
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Herb Derby <herb@google.com>
When run on cooperative threads, we need to notify the threading system
when we're going to block (in any way, sleep or spin) so that it doesn't
wake that thread up expecting it to make forward progress.
This should cover SkOnce, SkSpinlock, SkSemaphore, and by extension
SkMutex. Not sure if there are others to hit, but really the only way
to find out is by deadlocked stack traces. This CL (obviously?) does
nothing to mark synchronization primitives used by Skia's dependencies,
and in general I don't think we can do anything about them.
See cr/275261423 for more background and discussion. It's not clear to
me that marking these symbols as weak is necessary, so I figured I'd
just try not doing that and seeing what breaks. I can always follow up
with weak symbols if proven necessary.
Bug: skia:9577
Change-Id: I2c03fe92c58ad506dd8a68bdc90a09b28f965149
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/251221
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
This CL begins pulling some of the work forward into onPrePrepare.
Change-Id: If049e0662db51b465b8b82aafebeef2323bddfd4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/249802
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Uses SkISize instead of separate width/height variables in many places.
No functional change.
Change-Id: If87b2c57e43d810f0820c4e3c9ef8e6b8ebd10ba
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/249886
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:6718
Change-Id: Ia6363d74f016b076fa5e19a57d223f7c47b723c0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/248999
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This change also allows for the remove of GrPixelConfigIsOpaque function.
Bug: skia:6718
Change-Id: I7e7b70f02d911eda67640d648fb6348091e0f55d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/248698
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:6718
Change-Id: I4d9c31bc0ca42a80e652de29462531587efef031
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/248257
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:6718
Change-Id: I0f9c7d89e732c59bdbe76ab45d42934029106efb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/247684
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Add first, a ctor that will make an SkZip with more constness,
a safe default ctor, and allow SkSpan to be used as r-values.
Change-Id: Iaa89eee3a8994c3a0bd1def1a41e0f066dc872b9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/246822
Commit-Queue: Herb Derby <herb@google.com>
Auto-Submit: Herb Derby <herb@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
kLinesAdjacency is no longer used while kPath has historically been left out.
Change-Id: I84be64a2e8ec79ff299a2bced73ded0be04df38e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/246307
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Change-Id: I9b27c073122e615cf2db289de5607d6636658803
Bug: skia:6718
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/246077
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
SkIota adds an index to some collection you already use:
std::vector<int> v;
for (auto t : SkMakeIota(v)) {
size_t i; int v;
std::tie(i, v) = t;
...
}
Change-Id: I18390fc084eb8509cfaeb50d9720541759a2986d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/225546
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Everything except for SkImageInfo.h is mechanical
Change-Id: I2d775c79467fb15f6022e80d21b4a9151272fe2a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/242896
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
These two formats are now exercised in the wacky_yuv_formats GM
Change-Id: I07a5649401b7928e8c4666e39793915ac1d45646
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/242560
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This CL adds:
kAlpha_F16_SkColorType
kRG_F16_SkColorType
kRGBA_16161616_SkColorType,
which should be it for a while.
Bug: skia:9121
Change-Id: I81b9d46a202a76e9b7d7ca86495d72dbdae32576
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/241357
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Also trim includes in SkSLCompiler while I'm there
Change-Id: I87b27199b84e5e856e0a914f44368313a6f226e4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/242085
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This also switches GrColorType::kR_16 to kAlpha_16 to more closely match raster.
Bug: skia:9121
Change-Id: I03c6e6c52c90aa4223478c5ea6c8b2ed8558f677
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/239930
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
As part of this (clean up), also removed support for serialized-paths
older than version-4, which was introduced Feb 2018.
Change-Id: I2dc74a52bb8bdd7ea0cb2d8a78b644ca783eb31f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/239102
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Reed <reed@google.com>
This is a reland of 804f6a0fe7
Original change's description:
> Initiate MSAA resolves during DAG generation
>
> Adds an "fIsMSAADirty" flag to GrRenderTargetProxy and switches to
> resolving MSAA in GrTextureResolveRenderTask. This completes our push
> to resolve textures outside of render passes.
>
> For the time being, we only store a dirty flag on the proxy and still
> rely on the GrRenderTarget itself to track the actual dirty rect. This
> will be followed by a CL that moves the dirty rect out of
> GrRenderTarget and into the proxy.
>
> Bug: skia:
> Change-Id: I21219a58028bdb4590940210e565133093cd34b3
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235672
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
Bug: skia:
Change-Id: I805b3af1404eb7919ae937cff3dfa97921e32c69
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/237482
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Change-Id: Icd331e8946d80652750b8b6ea0db65f5f676ac3e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/238058
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
This reduces the number of map lookups that need to happen.
Change-Id: I068819f2576bf644a5c3550d48e69413e19179d3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/237217
Reviewed-by: Khushal Sagar <khushalsagar@chromium.org>
Commit-Queue: Herb Derby <herb@google.com>
This reverts commit 804f6a0fe7.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> Initiate MSAA resolves during DAG generation
>
> Adds an "fIsMSAADirty" flag to GrRenderTargetProxy and switches to
> resolving MSAA in GrTextureResolveRenderTask. This completes our push
> to resolve textures outside of render passes.
>
> For the time being, we only store a dirty flag on the proxy and still
> rely on the GrRenderTarget itself to track the actual dirty rect. This
> will be followed by a CL that moves the dirty rect out of
> GrRenderTarget and into the proxy.
>
> Bug: skia:
> Change-Id: I21219a58028bdb4590940210e565133093cd34b3
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235672
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@google.com,csmartdalton@google.com
Change-Id: Ife557caa840edfb64cbcafc272dc3012cfb43702
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/237242
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Adds an "fIsMSAADirty" flag to GrRenderTargetProxy and switches to
resolving MSAA in GrTextureResolveRenderTask. This completes our push
to resolve textures outside of render passes.
For the time being, we only store a dirty flag on the proxy and still
rely on the GrRenderTarget itself to track the actual dirty rect. This
will be followed by a CL that moves the dirty rect out of
GrRenderTarget and into the proxy.
Bug: skia:
Change-Id: I21219a58028bdb4590940210e565133093cd34b3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235672
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Also don't specify redundant width/height to SkGpuDevice.
Change-Id: I389df5c4b073c2c05632ba6b7c95b02a22dfaf98
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235824
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Unfortunately, boringssl also declares a function named sk_free making
decltype give an error since the function is overloaded if both Skia and
BoringSSL headers are included in the same TU (as happens in ChromeOS
builds). In theory one can use decltype and specify which overload to
use, but this is far more verbose than just specifying the type.
Change-Id: I7dfefbce0a75c7ff6791b58287f26f2cc47c8eca
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235828
Reviewed-by: Ben Wagner <bungeman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
SkFunctionWrapper was made to be a simple abstraction over existing
resource release functions which generally follow a specific pattern of
returning void and taking a pointer to the underlying type. However,
this has been observed to be an unnecessary limit. This makes it more
generic while also making the call sites a little less brittle.
This change also uncovered an issue in msvc v19.20 to v19.22, see
https://developercommunity.visualstudio.com/content/problem/698192/in-templateusing-decltype-is-void.html
To work around this, several otherwise redundant '&' are used.
There was an attempt to take references to functions instead of pointers
to functions which greatly simplifies the intermediate wrappers.
However, that uncovered another issue in msvc, see
https://developercommunity.visualstudio.com/content/problem/699878/function-to-pointer.html
Change-Id: I54ab945ed9d9cfd0204d4d6650c2fde47cc9e175
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235105
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Herb Derby <herb@google.com>
We will use this flag to check whether a proxy needs manual msaa
resolve, and if so, install a render task to do just that when it is
dirty and about to be read from.
For the time being, we are leaving GrRenderTarget::getResovleType in
place. This will allow us to assert that requiresManualMSAAResolve
works as expected. Long term, we plan to delete the former in favor of
the latter.
Bug: skia:
Change-Id: I155fa2a6d876edd339232c7c8208015d5da2c8ee
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/234436
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This also makes the isTexturable that takes a color explicitly called
isFormatTexturableAndUploadable. A lot of the uses of isConfigTexturable
were changed to the Uploadable version of the check, even though this is
not where we want to land in the long run. In the long run the uploadability
will be checked via calls to supportedWritePixelsColorType and all the
isTexturable calls should purely be about texturing. However, until those
changes land we keep it safe and keep the same functionality that
isConfigTexturable had by checking uploadability as well.
Bug: skia:6718
Change-Id: I3563f33f49811923da80e676fa3036ae46c4da70
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/234323
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This is a reland of fe19203eb7
Original change's description:
> Initiate regeneration of mipmaps from proxy DAG land
>
> This allows us to resolve all the textures before executing a command
> buffer, rather than interrupting execution as is currently done in the
> GL backend.
>
> Change-Id: I998546b312d24cf47fc2c837e7c94fbf95571af0
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/230636
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
Change-Id: Ic904d0b1bcb451bcb74cfae2204fb7297eaff108
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/234016
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This is a stop gap. We should probably base the precision decision on the texture format.
Also removes all code used to add sampler precision to program keys. The precision that
was added to the key did not affect the generated GLSL.
Bug: skia:8863
Bug: skia:6718
Change-Id: Ibdb702e1aca5d48d83e2f24cb24010a0b7270871
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/234322
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This is probably overkill but yields an officially thread-safe means of calling defaultBackendFormat.
Change-Id: Ibf344acf11102d938e335e9ea38bfaee575e00c5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/234324
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
