The code was original target for LLVM 10. I have updated it
to use LLVM 14. This does not fully complete the LLVM code because
there are still missing ops: array32, from_fp16, to_fp16,
load64, store64, load128, store128.
args.gn:
extra_cflags = [
"-DSKVM_LLVM",
"-isystem",
"../llvm-project/include",
]
extra_ldflags = [
"-L../llvm-project/lib",
"-lLLVM",
]
Running:
env DYLD_LIBRARY_PATH=/Users/herb/bin/llvm/lib ./out/SkVM-LLVM/dm
Change-Id: I79e3d3d62b29ed8bdad22890d4db5bc160ba4bf6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/460316
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Rather that recursively running Wang's formula and chopping, we run
Wang's once, divide by the maximum number of tessellation segments,
then chop in a flat loop. The next step will be to share this code
with other tessellators.
Bug: skia:12524
Change-Id: I6c5156f56885aee9f90b3be611fbef229740612d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465362
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
With paint conversion changed, this CL cleans up several SkShader
classes, based on the fact that input alpha will always be opaque.
Bug: skia:11942
Change-Id: I91c919a69b6dfe4fea40b799888dd6179defdd70
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/463799
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Long term plan is to expose a plugin (standalone or with bodymovin) that allows motion artists to write sksl into a composition.
This is the first step where we test how we'd read in the json data under the hood.
Change-Id: I300d3af5d01e12f5b495970f89fd12b5f464a9a1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464368
Commit-Queue: Jorge Betancourt <jmbetancourt@google.com>
Reviewed-by: Florin Malita <fmalita@google.com>
This adds check to make sure that the results in the last
add of the lerp are in range. Also, Smooth out types.
Change-Id: I853835e530f6b6790e16464db12964d68ab9ef8d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/453718
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Herb Derby <herb@google.com>
This is a more cautious attempt of https://skia-review.googlesource.com/c/skia/+/464876
GrAlignTo renamed to SkAlignTo and moved to SkTypes.h
(with the rest of our align helpers).
GR_MAKE_BITFIELD_OPS and GR_DECL_BITFIELD_OPS_FRIENDS
renamed to SK_* and moved to private/SkMacros.h
This avoids our public includes using a src header file
and messing with SkUtils.h, which apparently was used
by Android. That should be cleaned up in a separate effort.
Change-Id: I86d5e1fb6a7834034534266a6c340bc7757f9abb
Bug: skia:12584
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/466176
Reviewed-by: Brian Osman <brianosman@google.com>
All the fields of GrSubRunNoCachePainter are const, so
this can be hoisted out of the loop instead of creating
a new one each loop.
Change-Id: Ib430ad27f5ee1804316b1bf7737e9de80dc2f7b7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465382
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Change-Id: I7512491f55c10118f0ab058500f6ce9b5b8545cd
Bug: oss-fuzz:40557
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/466296
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Change-Id: Id4d9e105f7a1b7ff2cd3bfc9b20cd2309c6aca84
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465394
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This flag is no longer needed, so remove it and the code it protected.
Change-Id: I679eb3cb4cee81c4205309012931e134e108f232
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465696
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Change-Id: I87cc0e6d9dc60a0ceaf32c453d2278091ea3c5e5
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465393
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: Ic657319cf4f3ba2c3c23e2923cf121138d16dbf2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/457397
Commit-Queue: Joe Gregorio <jcgregorio@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
In Metal on some GPUs, `test_for_preserving_PM_conversions` can fail
when half4s are used for premul conversion.
Change-Id: I95a737a88aa8f5a998e3201516900977d5223666
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465798
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Additionally, the default clip state was quick rejecting all the draws.
Bug: skia:12466
Change-Id: I8eaa95418bf24af2a5718c8cf21e97cdd7f80a34
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465476
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
This is a reland of 75e1f4c028
Original change's description:
> Lift recursive curve culling out of tessellators
>
> We need recursive chopping/culling logic in order to draw
> astronomically large paths. But rather than do that at the same time
> the tessellators chop curves, this CL moves that logic into an
> SkPath -> SkPath transformation that runs ahead of time (and only if
> the path is extremely large to begin with). This will enable us to
> remove recursion from the tessellators and quickly determine ahead of
> time the size of buffers they need.
>
> Bug: skia:12524
> Change-Id: Ib2800fb23054f1548501811203173e58273fbc83
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/463936
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
Bug: skia:12524
Change-Id: Idf54f0c2bddaaddc9fc17bee99c910f3961682a5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465361
Reviewed-by: Greg Daniel <egdaniel@google.com>
Auto-Submit: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This will hopefully improve performance on lower-end GPUs.
Change-Id: I9c2ee6dc31acd08bec0bfb5f59edc3cf90163f9e
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465078
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Change-Id: Ic30c48dce0cb0072f07defcdb0b9e60b94f50818
Bug: oss-fuzz:40479
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465392
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
When the flag is disabled, everything works the same as before and all
uniforms are copied to buffers as 32-bit values. Enabling the flag will
cause shorts and halfs to be written to the uniform buffer as 16-bit
values (converting as needed).
Change-Id: I7e7c191f98b14b55a2587dc1499e8a6ff72f573c
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464920
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Bug: skia:12466
Change-Id: Ia0da94fc300ab678712e9bb79a011694fea8595e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464677
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
The iPhone 6 only has 1GB of RAM. (The iPhone 7 and 8 have 2GB.)
Change-Id: I4efd00f21feb9c7b6e42d99936989573d26acef3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465390
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Erik Rose <erikrose@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Erik Rose <erikrose@google.com>
This is a reland of c77620c308
Original change's description:
> Move alpha modulation into paint conversion (Step 1)
>
> Skia has a rule that says (roughly), any SkShader is modulated by the
> paint alpha. The CPU backend implements this precisely - the entire
> shader sees the opaque paint color, and the alpha is applied once at
> the root. The GPU backend has previously made each SkShader responsible
> for doing this modulation. That led to two kinds of problems:
>
> 1) Some shaders forgot to do the modulation at all.
> 2) In shader trees, it's possible for the alpha to get applied
> multiple times, or to child effects where it isn't desired.
>
> A consequence of #2 is that the GPU implementation of things like
> blend-shaders have to jump through hoops to invoke their children with
> opaque versions of their input colors.
>
> This CL allows us to remove the explicit alpha modulation from a variety
> of shaders, and should also cleanup of the blend shader logic. However,
> to get the CL past chrome's layout tests, we need to guard the change.
> So this version just makes the major change so paint conversion, in a
> way that will work universally. Once chrome switches to the new method,
> it will be safe to clean up the various shader implementations.
>
> Bug: skia:11942
> Change-Id: I518534becdd5d10cbd78cf3ff7d4a46dd1faabf9
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464116
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
Cq-Do-Not-Cancel-Tryjobs: true
Bug: skia:11942
Change-Id: I3bf184932adbb425959b02d28af1e6b7f5d127d5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464928
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:12466
Change-Id: Ib23a501664ebba581376fa0cb1ce5c6537e5d349
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465136
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
We used to just assert, but it turns out that Android has tests that
explicitly set out-of-range alpha values (and disable our asserts).
Bug: skia:11942
Bug: b/186543004
Change-Id: I167de7aba261ca3df579f9a0bddb0d90605e0094
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465381
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The fuzzer discovered that SkSL could create an out-of-range int literal
by casting from a floating point literal. We were only doing range
checks when the starting literal was an integer. Since we now assert
when an out-of-range int literal is created (as of
http://review.skia.org/464124), the fuzzer can detect this error.
Change-Id: Ie66f60ddbe7b4fbe5b648c17292c59a4ba079716
Bug: oss-fuzz:40456
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465385
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This guards against unexpected results when dfdy is used in complex
expressions. In practice, I'm not aware of this causing any trouble.
Change-Id: Ia476e57936969d248273856a94d5c403b47c29b4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465379
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This guards against unexpected results when dfdy is used in complex
expressions. In practice, I'm not aware of this causing any trouble.
Change-Id: I639bef465d7907049d79681a49f9be67b4c435a6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465378
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit 9fc189f1cb.
Reason for revert: shader compile failure on AndroidOne-GPU-Mali400MP2 devices
Original change's description:
> Wrap 'u_rtFlip.y * dfdy()' in parentheses.
>
> This guards against unexpected results when dfdy is used in complex
> expressions. In practice, I'm not aware of this causing any trouble.
>
> Change-Id: I58d4762871481fdb4c173b570e4d5d6edf657af7
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465077
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: Idfaa9316d657717d5ee7117837c9cc9c3d4ee189
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465377
Auto-Submit: Greg Daniel <egdaniel@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
This reverts commit 65a726bb49.
Reason for revert: You cannot include a src file in an include file. This ends up using c++17 features in our includes. Breaks rolls.
Original change's description:
> Move GR_MAKE_BITFIELD_OPS and GrAlignTo to non-GPU files
>
> These have been renamed SK_MAKE_BITFIELD_OPS and SkAlignTo
> because nothing seemed particularly GPU/Ganesh specific to them.
>
> I moved the latter to SkTypes.h because we have other align
> code there and former to src/SkUtils.h because I didn't know
> where else it should go.
>
> The primary motivation was removing the GrTypesPriv.h
> include from src/core/SkBlockAllocator.h. I had attempted
> some amount of #if SK_SUPPORT_GPU, but that's not as clean
> here because both our CPU and GPU backends use the
> SkBlockAllocator (as far as I could tell).
>
> This also moves sk_memset* from SkUtils.h to SkOpts.h, because
> SkOpts.h requires bringing in RasterPipeline, which seemed
> like overkill.
>
> Change-Id: I5163ef5064ad3840a15b7e873930d60e2620bf9d
> Bug: skia:12584
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464876
> Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:12584
Change-Id: I1b772bbbc6f150d737bb53fa4e5f45d1581929fa
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465376
Auto-Submit: Greg Daniel <egdaniel@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
This guards against unexpected results when dfdy is used in complex
expressions. In practice, I'm not aware of this causing any trouble.
Change-Id: I58d4762871481fdb4c173b570e4d5d6edf657af7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465077
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This uniform uses a float2 type, not half2. Previously, this was only a
latent problem, since we treated halfs and floats the same when setting
up uniform buffers.
Change-Id: I54f7033561542cf984c08070b2c3eca4173ca69c
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465216
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Bug: skia:12466
Change-Id: I12c64396791d8e0b8891b82927b1c8811a6e164f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464385
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
We need to know a uniform's type to convert it to the uniform-buffer's
expected format when uniforms are set. We have this data in other
arrays, but it's not accessible to the program data manager.
We don't necessarily need to use a bitfield here, if we think there's
ever a possibility of having a uniform buffer > 16MB in size.
Change-Id: I88b3936de5319d65c76891ec0f1d87badc71d2dd
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465076
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The SortKey will only hold the uniqueID of the program and uniform data. When actually executing the draw we will need to reconnect with the actual data.
Bug: skia:12466
Change-Id: Iea7f0a99d471ea7fe2a3864bdd60255b09289088
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464926
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Change-Id: Ifedf6172c8152b809ef8faeae4dd0e7b87b142ee
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464927
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:12466
Change-Id: I3c3eda26d0c09a58108a5b7bdd1bca0e63973f17
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/462887
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Bug: skia:12466
Change-Id: If31d25a1f4a4805069f6556c6802f33fa0bdfcc1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464380
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
These have been renamed SK_MAKE_BITFIELD_OPS and SkAlignTo
because nothing seemed particularly GPU/Ganesh specific to them.
I moved the latter to SkTypes.h because we have other align
code there and former to src/SkUtils.h because I didn't know
where else it should go.
The primary motivation was removing the GrTypesPriv.h
include from src/core/SkBlockAllocator.h. I had attempted
some amount of #if SK_SUPPORT_GPU, but that's not as clean
here because both our CPU and GPU backends use the
SkBlockAllocator (as far as I could tell).
This also moves sk_memset* from SkUtils.h to SkOpts.h, because
SkOpts.h requires bringing in RasterPipeline, which seemed
like overkill.
Change-Id: I5163ef5064ad3840a15b7e873930d60e2620bf9d
Bug: skia:12584
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464876
Reviewed-by: Brian Osman <brianosman@google.com>
