This reverts commit 198393b2de.
Reason for revert: android crash
Original change's description:
> Notify RTC when OpsTask is closed so it can drop ownership
>
> Change-Id: I95d32ed89447995541f33bf80730876ce9c0747a
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284519
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@google.com
Change-Id: I5133fa1b8f90182864ffbee3b60bfd5781dc16bd
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284728
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Change-Id: I1012a4b6c6eb67e01923f767baeb78ebc18a0fd5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284477
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This is a reland of 1283d55f35
... this time, also checking for HSW feature set.
Original change's description:
> Reland "gather8/16 JIT support"
>
> This is a reland of 54659e51bc
>
> ... now expecting not to JIT when under ASAN/MSAN.
>
> Original change's description:
> > gather8/16 JIT support
> >
> > The basic strategy is one at a time, inserting 8- or 16-bit values
> > into an Xmm register, then expanding to 32-bit in a Ymm at the end
> > using vpmovzx{b,w}d instructions.
> >
> > Somewhat annoyingly we can only pull indices from an Xmm register,
> > so we grab the first four then shift down the top before the rest.
> >
> > Added a unit test to get coverage where the indices are reused and
> > not consumed directly by the gather instruction. It's an important
> > case, needing to find another register for accum that can't just be
> > dst(), but there's no natural coverage of that anywhere.
> >
> > Change-Id: I8189ead2364060f10537a2f9364d63338a7e596f
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284311
> > Reviewed-by: Herb Derby <herb@google.com>
> > Commit-Queue: Mike Klein <mtklein@google.com>
>
> Change-Id: I67f441615b312b47e7a3182e85e0f787286d7717
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284472
> Reviewed-by: Herb Derby <herb@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: Id0e53ab67f7a70fe42dccca1d9912b07ec11b54d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284504
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This reverts commit 1283d55f35.
Reason for revert: one more try...
Original change's description:
> Reland "gather8/16 JIT support"
>
> This is a reland of 54659e51bc
>
> ... now expecting not to JIT when under ASAN/MSAN.
>
> Original change's description:
> > gather8/16 JIT support
> >
> > The basic strategy is one at a time, inserting 8- or 16-bit values
> > into an Xmm register, then expanding to 32-bit in a Ymm at the end
> > using vpmovzx{b,w}d instructions.
> >
> > Somewhat annoyingly we can only pull indices from an Xmm register,
> > so we grab the first four then shift down the top before the rest.
> >
> > Added a unit test to get coverage where the indices are reused and
> > not consumed directly by the gather instruction. It's an important
> > case, needing to find another register for accum that can't just be
> > dst(), but there's no natural coverage of that anywhere.
> >
> > Change-Id: I8189ead2364060f10537a2f9364d63338a7e596f
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284311
> > Reviewed-by: Herb Derby <herb@google.com>
> > Commit-Queue: Mike Klein <mtklein@google.com>
>
> Change-Id: I67f441615b312b47e7a3182e85e0f787286d7717
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284472
> Reviewed-by: Herb Derby <herb@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,herb@google.com
Change-Id: I953fcd2aef308fd901880618fa540ac9f6d88e84
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284503
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This is a reland of 54659e51bc
... now expecting not to JIT when under ASAN/MSAN.
Original change's description:
> gather8/16 JIT support
>
> The basic strategy is one at a time, inserting 8- or 16-bit values
> into an Xmm register, then expanding to 32-bit in a Ymm at the end
> using vpmovzx{b,w}d instructions.
>
> Somewhat annoyingly we can only pull indices from an Xmm register,
> so we grab the first four then shift down the top before the rest.
>
> Added a unit test to get coverage where the indices are reused and
> not consumed directly by the gather instruction. It's an important
> case, needing to find another register for accum that can't just be
> dst(), but there's no natural coverage of that anywhere.
>
> Change-Id: I8189ead2364060f10537a2f9364d63338a7e596f
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284311
> Reviewed-by: Herb Derby <herb@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: I67f441615b312b47e7a3182e85e0f787286d7717
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284472
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This reverts commit 54659e51bc.
Reason for revert: ASAN
Original change's description:
> gather8/16 JIT support
>
> The basic strategy is one at a time, inserting 8- or 16-bit values
> into an Xmm register, then expanding to 32-bit in a Ymm at the end
> using vpmovzx{b,w}d instructions.
>
> Somewhat annoyingly we can only pull indices from an Xmm register,
> so we grab the first four then shift down the top before the rest.
>
> Added a unit test to get coverage where the indices are reused and
> not consumed directly by the gather instruction. It's an important
> case, needing to find another register for accum that can't just be
> dst(), but there's no natural coverage of that anywhere.
>
> Change-Id: I8189ead2364060f10537a2f9364d63338a7e596f
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284311
> Reviewed-by: Herb Derby <herb@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,herb@google.com
Change-Id: I912273e6ffc9258537ba806951a5964be0218d58
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284471
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
The basic strategy is one at a time, inserting 8- or 16-bit values
into an Xmm register, then expanding to 32-bit in a Ymm at the end
using vpmovzx{b,w}d instructions.
Somewhat annoyingly we can only pull indices from an Xmm register,
so we grab the first four then shift down the top before the rest.
Added a unit test to get coverage where the indices are reused and
not consumed directly by the gather instruction. It's an important
case, needing to find another register for accum that can't just be
dst(), but there's no natural coverage of that anywhere.
Change-Id: I8189ead2364060f10537a2f9364d63338a7e596f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284311
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: I6d48d4aebfffcf106ab22b511e43f76a4f8972c6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284276
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Renames the provider to SkMatrixProvider, which is now also able to
provide the local-to-device matrix. Everywhere that does paint
conversion and FP generation now has access to the entire matrix
provider, instead of just the CTM.
This will allow the SkSL FP (and others) to fetch other matrix state.
Change-Id: Iffb00bae0d438da0e8de3eebe75183ed6d440fd6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284040
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Array-stride calculation did not check for zero-sized objects before
performing a modulo. This would lead to undefined behavior (crash)
while attempting to build the AST.
Change-Id: I84b4662978955d49a3ca28f6bb577d15c87cccb6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284354
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit 73b86c1ade.
Reason for revert: wasn't the problem, relanding
Original change's description:
> Revert "Rename GrShape to GrStyledShape"
>
> This reverts commit f3f08af010.
>
> Reason for revert: maybe this is breaking the ios perf bot; it is the first CL that caused the bot to fail, but I can't really say why this would break them.
>
> Original change's description:
> > Rename GrShape to GrStyledShape
> >
> > Change-Id: Ic457e634b4b95356f5615cff3fce1ca7d7677c26
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284036
> > Reviewed-by: Robert Phillips <robertphillips@google.com>
> > Commit-Queue: Michael Ludwig <michaelludwig@google.com>
>
> TBR=bsalomon@google.com,robertphillips@google.com,michaelludwig@google.com
>
> # Not skipping CQ checks because original CL landed > 1 day ago.
>
> Change-Id: I2e5adbfc820811fbbde9cb57af28f86a7ba40bd9
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284231
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
TBR=bsalomon@google.com,robertphillips@google.com,michaelludwig@google.com
Change-Id: Ifdd52f8bacb5d66a7bf58efd328675c4c443ac8a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284376
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
This is a reland of 16abfa5e42
Original change's description:
> Revert "Revert "GrDisplacementEffect uses GrTextureEffect.""
>
> Also further limits the number of FP stages on ANGLE D3D9 to avoid
> varying limit.
>
> This reverts commit 0b406736c1.
>
> Bug: skia:10139
>
> Change-Id: Ifda0b932ab8ce176f1a1cd1ca9c2deaf1fda20bd
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284048
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:10139
Change-Id: I542506a87fadeeb54e2f749b199806d2155c6cab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284230
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit f3f08af010.
Reason for revert: maybe this is breaking the ios perf bot; it is the first CL that caused the bot to fail, but I can't really say why this would break them.
Original change's description:
> Rename GrShape to GrStyledShape
>
> Change-Id: Ic457e634b4b95356f5615cff3fce1ca7d7677c26
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284036
> Reviewed-by: Robert Phillips <robertphillips@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
TBR=bsalomon@google.com,robertphillips@google.com,michaelludwig@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I2e5adbfc820811fbbde9cb57af28f86a7ba40bd9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284231
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Funnel the ARM instructions through two op() helpers, one for 3-arg
vector instructions, another for all others (0,1,2 arg, optional imm).
More consistent use of (immN & N_mask) to make things clearer.
Add missing imm12 offset to load and store instructions, with tests.
Notice they're in element counts, so we can go up to 4096 16-byte stack
entries, not 256 entries like you might think.
Change-Id: I99a3ad30b7b0926f93da671f00d89759934e65b4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284255
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Going to be easier to work on stack/register things if we
don't have to keep thinking of aarch64 as a special case.
This just sets up the frames, will follow up with JITMode::Stack.
Change-Id: Ic0df4c5deb9c7d55eb73a62e4b6b1c9919996974
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284243
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: I04e6c829fb122dc8b2192911bda94e5106927bff
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/275437
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Change-Id: I123caca8649097316f52f6fb0aed2e6ed0d5090a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284256
Commit-Queue: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit 16abfa5e42.
Reason for revert: Appears to have made ANGLE and ios unhappy
Original change's description:
> Revert "Revert "GrDisplacementEffect uses GrTextureEffect.""
>
> Also further limits the number of FP stages on ANGLE D3D9 to avoid
> varying limit.
>
> This reverts commit 0b406736c1.
>
> Bug: skia:10139
>
> Change-Id: Ifda0b932ab8ce176f1a1cd1ca9c2deaf1fda20bd
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284048
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com,michaelludwig@google.com
Change-Id: I0e4b135a59f713342bf5c6c258779975762ccd1f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:10139
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284217
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Also further limits the number of FP stages on ANGLE D3D9 to avoid
varying limit.
This reverts commit 0b406736c1.
Bug: skia:10139
Change-Id: Ifda0b932ab8ce176f1a1cd1ca9c2deaf1fda20bd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284048
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Move all the non-vector instructions together,
and convert them to use Operand where possible.
In general that can be any of
- (Operand, imm)
- (Operand, GP64)
- (GP64, Operand)
and that means there are two ways to encode (GP64,GP64)
instructions, so there's a disambiguator added.
Our measure of sucess is eliminating calls to rex()
except from our one helper, and so far, so good.
I haven't seen a need for Label Operands yet, and they're
only useful as (GP64, Operand) style arguments (can't
really be destinations in read-only memory) but we could
add support pretty easily if we find the need.
Tweak one test to avoid int/pointer ambiguity about 0.
Changed some of the instructions to always use a REX
prefix just to make it easier to funnel everything
through one place. movzbl -> movzbq, etc.
Change-Id: I606f94e76e0ef8f491409f23748f5c8dcb607491
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284023
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: Ic457e634b4b95356f5615cff3fce1ca7d7677c26
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/284036
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Rename YmmOperand to Operand, focusing on that side of things for
now. And delete unused GP64Operand... might not need to return.
Big refactor around W and L bits and the helper op() functions.
Lots more is now funneled through a single core op() function.
Support Xmm and GP64 (direct moves) as Operands too.
As a rule of thumb I measured my progress by counting vex() calls.
Ideally we call it only in that centralized op().
I think I got as close as we can get, with only vgatherdps calling vex()
itself. Given its weird encoding, there's no good way to work
vgatherdps into the abstraction. It's close to Mem{base,0,index,scale},
but the index is a Ymm register, and there isn't any corresponding
special cases for it like there is normally for rsp in SIB.
Change-Id: I48e4583293e1df386a18d37ad54197016ce13251
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283806
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
This replaces most vmovups variants with two: load to register from
flexible operand, or store from register to flexible operand.
And upgrade the zero-extending loads too to finish off load_store().
More to come in small steps.
Change-Id: I80645f264ee91662260046c8e0a45ba6d1bf98c6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283753
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This operation came when updating aggregate inner and outer bounds
for Ganesh' new clip stack (particularly when accounting for the effect
of a difference operation). This geometric operation is theoretically
more general purpose so I moved it out to SkRectPriv.
Change-Id: Ibd76f9b95efc1790ecda1038779c124155031d8f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283756
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
This introduces Mem, a way of expressing x86 addressing:
addr = base reg + offset imm + (scale imm * index reg)
using the usual x86 convention of index = rsp to indicate no index.
And then, this introduces GP64Operand and YmmOperand, which are
generalizations like YmmOrLabel that fold over all the types of
arguments available at that position. (YmmOperand replaces YmmOrLabel).
There's still much to do, but I've started by generalizing most
of the Ymm instructions to take YmmOperand, and added some new
unit tests for vmovdqa to make sure all the various modes work.
Change-Id: Ie6cc1186310ff39c52a2a061431a91d10816c98a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283344
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Also make GrDynamicAtlas directly use the GrSurface callback type rather
than go through a springboard.
Change-Id: I3e3c155bbc1e2e794e0d6828b0ea3c64c6a7f3e9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283226
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This allows fragment processors to sample their children with their
local coordinate system transformed by a matrix.
Change-Id: Ifa848bbd85b939bbc5751fec5cf8f89ee904bf39
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282590
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
I propose landing this, but then pause on extending math functions until
we develop a clearer migration story for sksl.
Change-Id: Id42ec37071da058e6e7809abe1ed0570d48df8e7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283229
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Reed <reed@google.com>
These are neat but mostly just a distraction for now.
I've left all the assembly in place and unit tested
to make putting these back easy when we want to.
Change-Id: Id2bd05eca363baf9c4e31125ee79e722ded54cb7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283307
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
One new instruction movzwl needed.
Change-Id: Ic70ba34d667eb6d570aeca88c4243e0c3309525f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283305
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
- let skvm tell us if FMAs are supported
- unguard previously LLVM-only tests
- simplify testing JIT and interpreter
We're getting close enough to always being able to JIT that carefully
marking what JITs and what doesn't is more annoying than helpful.
Now just test the JIT if present, and always test the interpreter.
Change-Id: I83762b38e0773ccaee795ae0fc9907e86628d73e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283275
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Reed <reed@google.com>
These are the dregs of the omnibus GrDrawOpAtlas CL
Change-Id: I0723346db293dd6650b905f661053955d2f46a1d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279908
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This looks to be the culprit behind the failing/flaky metal bots. Not
sure what the direct cause is because it was different tests that were
failing and I would have thought we'd eventually submit at the end of
ProgramsTest when tearing things down. Maybe we just created such a
backlog of work submitted to the actual GPU that the GPU got put into
a bad state?
Bug: skia:10118
Change-Id: I318428e228e08fbcaaa12b23ad2869ecc35cc786
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283136
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Starting on atan2, but there is a lot of quadrant clean-up, so will
do that in a separate CL.
Change-Id: Ie1e70051a6ecb19a2e521b56ed09796e8e745276
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/283016
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Reed <reed@google.com>
Change-Id: I22c120db2535929bd20df3068cca1aecc57ae746
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282744
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Reed <reed@google.com>
Bug: skia:10118
Change-Id: I53e3b9f1bd28a00276a3d35b5160aa0cfec30cfd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282417
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
There apparently were other tests that had similar issues and this fix
is much better and is more future proof.
Change-Id: I4835c7e5772b9e70249a69255aae8808be172eef
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282681
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>
Change-Id: I6f47bb8aa9d743925ea98d2614058137914e0531
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282637
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>
Will need followups on Vertices and SkSL to use these handles.
Change-Id: If775cb01168f601541e889bfa2421129e505b4a4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282416
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Tests min() / max() float behavior fairly exhaustively.
We sometimes specialize into min_f32_imm and max_f32_imm, so it's
important to test with constant values as each argument to cover that
specialization, and to test with both as non-constant values to cover
when that specialization does not apply.
Change-Id: Ib021fd5a6d322058af2f504048b9ed02d0510732
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282315
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Change-Id: Ib1c0570d747bf9f46be3486f37eba3af53ed1e3d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281642
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Change-Id: I90f12cb305ff8daf64b07e5f47bb3a158df95bee
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282120
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Extending text to grapheme edges should correct glyph range
Bug: skia:10087
Change-Id: I254901aaaa40c2782d1afbd5d5390599bdd7c922
Change-Id: I1d51076656d09e4d2e35e3ddad28bfd60fc87081
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281756
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Julia Lavrova <jlavrova@google.com>
This is a reland of 17f05c737e
Original change's description:
> Rename instanceAttribSupport -> drawInstancedSupport
>
> Change-Id: I7d8ff8597849f2b910928867842857e25a12b4b1
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281582
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com
Change-Id: I2e15ac72d86747c30e71bb50d30da48a5a342772
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/282118
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Having GrAtlasManager.h in GrContextPriv.h was needlessly propagating dependence on that header.
Change-Id: Idf5836f1e217ecd2da91f751b488a63a884c02ce
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281739
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
This is a prerequisite change for the upcoming sample(child, matrix)
function. By itself, this CL doesn't really change anything; it just
adds an ownership tracking feature which sample(child, matrix) depends
on.
Change-Id: I98b12e5fb062a2535af367931e7a932ea9c63a59
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281337
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
saveCamera() is no longer experimental
In a separate CL, will stage changes to concat virtual to take M44.
Change-Id: Iaf37ce2f24ab1223c54aeb1e79eaebf18f87fece
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281589
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Reed <reed@google.com>
bit_clear is at least useful as a special case for select(),
which helps with code readability.
Add is_NaN() and use these all together in sweep gradient.
Change-Id: I57a54f8956f85e0db0662b33f8446b8dc7342d8d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281685
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
- new this-> convention: never use it when calling common public
Builder methods like splat(), bit_and(), etc like you'd see in
normal user code, but always use it when calling private methods
like this->push(), this->isImm(), this->allImm().
- use c++17 if-statements to scope this->allImm() variables tighter.
- check for x.id == y.id cases where applicable, including a tweak
to min() and max() to make them able to hit the special case.
- add special cases for I32 +,-,*, and remove an old unimportant
unit test that assumed we didn't fold these.
- add special cases for select(), and use select() in a few more
places where it's clearer and now just as efficient.
Change-Id: Idaac9250ac5a95a48d33eeba1cc4380c8c91629d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281678
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: Ie877e995642fe6be48b7e242aa955e87ecc14fb0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281385
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
bit_clear() is just another bit_and(),
and bytes() is a way of expression pshufb
that we never really use (yet).
Can always add them back later, but there's
some extra complexity to think about for each
that I'd like to not think about now:
- common sub-expression elimination between bit_and and bit_clear
- large constant management JIT'ing bytes
Change-Id: I3a54afa963231fec1d5de949acc647e3430ed0d8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281557
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Bug: skia:10078
Change-Id: I3ce0d97f8ada55403cc3f88bb16659085449ea29
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281207
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
It's clearer to see the flow of data this way and to read each pass'
implementation without all the pointer indirection, and move semantics
should let this be just as efficient.
Change-Id: I1ac211fbe54bec37de6d126eec0c211573c2a568
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281218
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This finishes up the main refactoring.
Still some follow-ups I want to try.
I got tired of typing usage.users(id) so I converted that to operator[],
which I think is clear for now. If we add more methods that don't refer
to the users, we can undo?
Change-Id: I0ac563cfb1899f7a3f8b2cb6d50ca1646dd05071
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281216
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
The color type used to influence the GrPixelConfig and the swizzles
used with the proxy and its GrTextures. So we had to create a new proxy
if the same API texture was used to fulfill a promise image of a
different color type. That has all been removed or lifted out of proxies
and into views. So now it is fine to reuse the same proxy/GrTexture with
a different color type.
Also remove unit test that checked required a new GrTexture be made if
a different color type was used.
Bug: skia:10078
Change-Id: Ib74c419b5d54d382ea95720af7026640e35013f0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281056
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Seems nicer to keep encapsulated in a program->program pass
so nothing upstream of it has to think about liveness.
I will be circling back around to profiling the cost of these
tempoaries, copies, etc. I just want to start writing them as
if cost were no object first.
Change-Id: I1d1187b521fbe963e720e0d8de90316a549f7797
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281182
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
We already checked the SkSafeRange earlier (after its last use). Here we
need to be checking the SkSafeMath object. Oops.
The counts are directly copied from the constructor to the vertices'
fields - the user already knows what they are (and the tests are
verifying a tautology).
Bug: skia:9984
Change-Id: I33c62e02825718e497834b0dfa40d0d56e46a197
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280963
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Adds structure to the per-vertex (now custom) data.
Attributes currently just have a type, but the next
step is to augment that with semantic flags to handle
transformation logic in the vertex shader.
Added unit tests and GMs that exercise attributes of
varying float counts, as well as normalized bytes.
Bug: skia:9984
Change-Id: I02402d40b66a6e15b39f71125004efb98bc06295
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280338
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is purely to make debugging easier, as the shaders now make more sense
to read.
Change-Id: Ie8e9b9d36112fda3ff997d088955de5a0e2c787a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281017
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Instead of copying fIndex and marching it forward,
we can tick down our existing uses counts backward,
saving one temporary std::vector.
Our implementation does guarantee the Instructions
returned by users() are sorted, so let's lean into
that... that means we can find the death time of any
instruction simply by looking at users().back()
(if there are any, of course).
Everything else is names and formatting, the biggest
being renaming Uses -> Usage. There's enough mention
of "users" and "uses" contrasting with each other that
I think it makes sense for the type to have the nice
middle-ground neutral name Usage, reflecting the arrow
and not which way we're thinking about it pointing.
Change-Id: I32ea9af6eb6430a162bee6da4810a599e8ed0dfd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/281003
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
It is also used for writes to surfaces other than fragment shader
output. e.g. clears.
Change-Id: Id1eb79be6d1a8aed936456bffa59dee32661cec8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280344
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Change-Id: Ic6c0de262e13c3c3ea9e7777ccdd04f22ff6ae0a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280345
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Liveness tracks all the live instructions in the instruction stream.
Uses maps this value to instructions that use it.
Uses is overkill for the current schedule, but will be needed for
spilling.
Change-Id: Id20b7b7a90901e156d323bb612c5908f91405967
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277744
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Rather than two separate (partially overlapping) ways of accessing the
private portions of SkVertices, use a single privileged helper class
(similar to GrContextPriv).
Change-Id: I76b14b63088658ed8726719cce126577e5a52078
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280601
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Reed <reed@google.com>
This reverts commit 8176508cc6.
Replaces incorrect check for Read pixels support with Write check.
Change-Id: Idc80aaaa41ba35014339df450bb8b583fb6dab51
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280399
Auto-Submit: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Allow push_back to take an rvalue to move into the allocator, which has
the nice side effect of making the test for it a bit better.
Change-Id: I7e313cf75c1ad65fbf765b10c4717672063b5570
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280412
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Change-Id: I60a3569b47b599b710c0f3a9522241748f15360d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280409
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Reed <reed@google.com>
It was too easy to get into circular include chains. Added static
asserts to ensure we keep our quad AA flags in sync. Also, IWYU.
Change-Id: I01aefa264aa56420ab5a46a8ecd9e63c021c79ab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/280405
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Instead separately check texturability and uploadability using existing
queries.
Also remove unused renderable param from onGetDefaultBackendFormat()
Change-Id: I8751e6d62263ddabd713c850beb788620a78bf10
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279996
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
GCC 9 warns about this pessimizing-move. It isn't clear why clang isn't
as well.
GCC 9 has a working redundant-move diagnostic. Clang has an old (C++11
style) redundant-move diagnostic which only warns when moving a
parameter. The GCC warning conflicts with Clang's
return-std-move-in-c++11, which we want to keep until we can drop
support for older compilers. So just disable redundant-move warnings
until we can remove return-std-move-in-c++11.
This change allows us to compile without warnings on gcc 9.3.0.
Change-Id: If21fcfb2944ce49e27fc84d40805752895ae68cf
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279958
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Precusor step for making this public and adding a method to
GrBackendFormat to query its channels.
Bug: skia:10078
Change-Id: I2d8fa6586721c35961bc328a15eef8e2ebd4406e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279422
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This effectively reverts https://codereview.chromium.org/972153010/
Note that we have no such hard limit on SkShaders, and it's never
been a problem. Complex SkPaint objects can generate too many FPs
for us to handle, but those are rare, and better managed by just
reporting the shader compile error (which we do).
Change-Id: Iee5dc3d65ec130f2ce0a29e55fbe3c25b00dc828
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279836
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Also renames GrD3DSurfaceResource to GrD3DTextureResource. This makes
things consistent with our naming convention in Vulkan, and with the
terminology used in Direct3D.
Bug: skia:9935
Change-Id: I6e6d2066ac70eb8a0d63c1b5731f31851a3017d5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279338
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Ambiguity has come up re: what "tessellation" means in Ganesh. I
believe that, in the context of a graphics library, "tessellation"
should refer to the hardware pipeline feature of submitting patches
and tessellating them on the GPU. This CL therefore renames classes
that triangulate things on the CPU to call it "triangulation".
Change-Id: Ic8515ea6a33000f1b638a852d5122bc9bd6b38f2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/279236
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Previously, the GrAllocator header defined three types: GrAllocator,
GrTAllocator, and GrSTAllocator. GrAllocator was generic, and would
manage the blocks of memory (fixed size to hold N * sizeof(item)). It
stored an SkTArray of pointers to each block. GrTAllocator wrapped
the GrAllocator to add template support for a specific T. GrSTAllocator
extended GrTAllocator to include a template arg for inline storage.
In this CL, GrAllocator is no longer defined, and its memory functionality
is replaced with the use of GrBlockAllocator. For the most part, the
implementation of GrTAllocator on top of GrBlockAllocator remains the
same, although there is explicit array to the block pointers so indexing
is slightly different. GrSTAllocator is also removed entirely, so that
GrTAllocator's template includes initial storage size.
The iteration over the allocated items
is updated to wrap GrBlockAllocator's block iterator, and then iterate
by index within each block. Documentation on GrAllocator already recommended
using the iterator instead of random access, and these changes further
reinforce it. It turns out that many of the use cases of GrAllocator
were written to use random access, so many of the files outside of
GrAllocator.h have just been updated to use the new for-range iteration
instead of a random access for loop.
Change-Id: I28b0bc277c323fd7035d4a8442ae67f058b2b64c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/272058
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Properly tagging links in PDFs just required a small change: at the time
an annotation is drawn, we need to keep track of the bounding rect, the
url and/or named destination, and now additionally the node ID. Instead
of keeping track of separate maps of rects/urls and rects/dests, this
patch combines them into a vector of SkPDFLinks.
Then, at the time the annotation is written to the file, an object
reference to that annotation can be saved and passed to the document
structure tree, where it can be included as an additional child of
the node with that node ID.
That allows for properly tagging the link and optionally giving it alt
text (like for an image link).
Bug: chromium:1039816
Change-Id: I97e47d3c70949020c3e0a69b8c9fc743748f3a7b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277426
Commit-Queue: Dominic Mazzoni <dmazzoni@chromium.org>
Reviewed-by: Ben Wagner <bungeman@google.com>
... in support of programs for colorspacexforms
Change-Id: I72ace09f10511ef8994038a4af3feab8bc1a299e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/278466
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit bc0fe058f8.
Reason for revert: red
Original change's description:
> Make TessellationTestOp and GrPipelineDynamicStateTestOp surface their programInfos at record time
>
> These aren't high priority but it would be nice to have all the ops go through the same system.
>
> Bug: skia:9455
> Change-Id: Idbd8b0829faf12703c82f5a016f5b0e7c3efb762
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277757
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Robert Phillips <robertphillips@google.com>
TBR=egdaniel@google.com,robertphillips@google.com
Change-Id: I579ad16f144d17afaec9ee1b9000eaaa5c04228c
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:9455
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/278508
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
These aren't high priority but it would be nice to have all the ops go through the same system.
Bug: skia:9455
Change-Id: Idbd8b0829faf12703c82f5a016f5b0e7c3efb762
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277757
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Updated to use sentinel GL context even when GL backend is not built.
This reverts commit 1171d314ef.
Change-Id: Ia94bbe4865ddd4e898446c13886877c539f0eb0b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277976
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This makes onPrePrepare match onPrepare & onExecute. Most of the new method bodies will-have-to/should be filled in anyway.
Change-Id: Ifc897feaeb2d8fe6eec3ac3a8e91f99393ad6758
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/277542
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
