aarch64 added vector-wise add/mul/min/max instructions.
We can use min and max to implement allTrue() and anyTrue(),
respectively.
(This CL is mostly so I don't forget these intrinsics exist.)
In assembly, these actually compile to two instructions,
the folding operation into a vector register, then a move
from the vector register to a general purpose register.
Change-Id: Ia6a999ac250740de765e871094e911979a8711c7
Reviewed-on: https://skia-review.googlesource.com/116482
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Bug: skia:
Change-Id: Ieedd05ced376a7604936e9d2729fc20a8669496e
Reviewed-on: https://skia-review.googlesource.com/115531
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Bug: skia:
Change-Id: I2adb983d68625d327e7c00e53b6ae4703b46252f
Reviewed-on: https://skia-review.googlesource.com/104761
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Bug: skia:
Change-Id: I0377e6a1dd8259e944f7902a5c68af524fa588c7
Reviewed-on: https://skia-review.googlesource.com/79382
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
and test it.
Change-Id: Ib0c2cf93c63d8d3c36a7d4d60bbec4ecede29bc7
Reviewed-on: https://skia-review.googlesource.com/78480
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Add mulHi to base SkNx, and specialize implementations for Sk4u for
neon and sse.
Add casts for converting from uint8_t by 4 to uint32_t by 4.
Cq-Include-Trybots: skia.primary:Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: I29a32e2ad9812a47fff841ceca334e562362836f
Reviewed-on: https://skia-review.googlesource.com/57960
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Herb Derby <herb@google.com>
Every type now nominally has Load4() and Store4() methods.
The ones that we use are implemented.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3046
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Change-Id: I7984f0c2063ef8acbc322bd2e968f8f7eaa0d8fd
Reviewed-on: https://skia-review.googlesource.com/3046
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reason for revert:
Hitting an assert
Original issue's description:
> Support Float32 output from SkColorSpaceXform
>
> * Adds Float32 support to SkColorSpaceXform
> * Changes API to allows clients to ask for F32, updates clients to
> new API
> * Adds Sk4f_load4 and Sk4f_store4 to SkNx
> * Make use of new xform in SkGr.cpp
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2339233003
> CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/43d6651111374b5d1e4ddd9030dcf079b448ec47TBR=brianosman@google.com,mtklein@google.com,scroggo@google.com,mtklein@chromium.org,bsalomon@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Review-Url: https://codereview.chromium.org/2347473007
* Adds Float32 support to SkColorSpaceXform
* Changes API to allows clients to ask for F32, updates clients to
new API
* Adds Sk4f_load4 and Sk4f_store4 to SkNx
* Make use of new xform in SkGr.cpp
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2339233003
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review-Url: https://codereview.chromium.org/2339233003
It's become clear we need to sometimes deal with values <0 or >1.
I'm not yet convinced we care about NaN or +-inf.
We had some fairly clever tricks and optimizations here for NEON
and SSE. I've thrown them out in favor of a single implementation.
If we find the specializations mattered, we can certainly figure out
how to extend them to this new range/domain.
This happens to add a vectorized float -> half for ARMv7, which was
missing from the _01 version. (The SSE strategy was not portable to
platforms that flush denorm floats to zero.)
I've tested the full float range for FloatToHalf on my desktop and a 5x.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145663003
CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot
Committed: https://skia.googlesource.com/skia/+/3296bee70d074bb8094b3229dbe12fa016657e90
Review-Url: https://codereview.chromium.org/2145663003
Reason for revert:
Unit tests fail on Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast
Original issue's description:
> Expand _01 half<->float limitation to _finite. Simplify.
>
> It's become clear we need to sometimes deal with values <0 or >1.
> I'm not yet convinced we care about NaN or +-inf.
>
> We had some fairly clever tricks and optimizations here for NEON
> and SSE. I've thrown them out in favor of a single implementation.
> If we find the specializations mattered, we can certainly figure out
> how to extend them to this new range/domain.
>
> This happens to add a vectorized float -> half for ARMv7, which was
> missing from the _01 version. (The SSE strategy was not portable to
> platforms that flush denorm floats to zero.)
>
> I've tested the full float range for FloatToHalf on my desktop and a 5x.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145663003
> CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/3296bee70d074bb8094b3229dbe12fa016657e90TBR=msarett@google.com,mtklein@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Review-Url: https://codereview.chromium.org/2151023003
It's become clear we need to sometimes deal with values <0 or >1.
I'm not yet convinced we care about NaN or +-inf.
We had some fairly clever tricks and optimizations here for NEON
and SSE. I've thrown them out in favor of a single implementation.
If we find the specializations mattered, we can certainly figure out
how to extend them to this new range/domain.
This happens to add a vectorized float -> half for ARMv7, which was
missing from the _01 version. (The SSE strategy was not portable to
platforms that flush denorm floats to zero.)
I've tested the full float range for FloatToHalf on my desktop and a 5x.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145663003
CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review-Url: https://codereview.chromium.org/2145663003
- rearrange a bit
- fewer macros
- hooks for all operators
- add left and right scalar operator overrides
- add +=, &=, <<=, etc.
- add SkNx_split() and SkNx_join()
- simplify the many rsqrt() and invert() options to just what we actually use
This refactoring pointed out that our float <-> int NEON conversions are not specialized, so I've implemented them. It seems nice that this is an error rather than silently falling back to serial code.
It's unclear to me if split/join want to be external, static methods, or non-static methods (SkNx_join(), Sk4f::Join(), x.join()). Time will tell?
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1812233003
CQ_EXTRA_TRYBOTS=client.skia.android:Test-Android-GCC-Nexus5-CPU-NEON-Arm7-Release-Trybot;client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review URL: https://codereview.chromium.org/1812233003
- trim unused specializations (Sk4i, Sk2d) and apis (SkNx_dup)
- expand apis a little
* v[0] == v.kth<0>()
* SkNx_shuffle can now convert to different-sized vectors, e.g. Sk2f <-> Sk4f
- remove anonymous namespace
I believe it's safe to remove the anonymous namespace right now.
We're worried about violating the One Definition Rule; the anonymous namespace protected us from that.
In Release builds, this is mostly moot, as everything tends to inline completely.
In Debug builds, violating the ODR is at worst an inconvenience, time spent trying to figure out why the bot is broken.
Now that we're building with SSE2/NEON everywhere, very few bots have even a chance about getting confused by two definitions of the same type or function. Where we do compile variants depending on, e.g., SSSE3, we do so in static inline functions. These are not subject to the ODR.
I plan to follow up with a tedious .kth<...>() -> [...] auto-replace.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1683543002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review URL: https://codereview.chromium.org/1683543002
SkNx_cast() can cast between any of our vector types,
provided they have the same number of elements.
Any types should work with the default implementation,
and we can drop in specializations as needed, like the
SSE and NEON Sk4f -> Sk4i I included here as an example.
To make this work, I made some internal name changes:
SkNi<N,T> -> SkNx<N, T>
SkNf<N> -> SkNx<N, float>
User aliases (Sk4f, Sk16b, etc.) stay the same.
We can land this first (it's PS1) if that makes things easier.
BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review URL: https://codereview.chromium.org/1464623002
- remove float -> int conversion, keeping float -> byte
- remove support for doubles
I was thinking of specializing Sk8f for AVX. This will help keep the complexity down.
This may cause minor diffs in radial gradients: toBytes() rounds where castTrunc() truncated. But I don't see any diffs in Gold.
https://gold.skia.org/search2?issue=1411563008&unt=true&query=source_type%3Dgm&master=false
BUG=skia:4117
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review URL: https://codereview.chromium.org/1411563008
While investigating my bug (skia:4052) I saw this TODO and figured
it'd make me feel better about an otherwise unsuccessful investigation.
This speeds up HardLight and Overlay (same code) by about 15% with SSE, mostly
by rewriting the logic from 1 cheap comparison and 2 expensive div255() calls
to 2 cheap comparisons and 1 expensive div255().
NEON speeds up by a more modest ~3%.
BUG=skia:
Review URL: https://codereview.chromium.org/1230663005
Mostly this is about ergonomics, making it easier to do good operations and hard / impossible to do bad ones.
- SkAlpha / SkPMColor constructors become static factories.
- Remove div255TruncNarrow(), rename div255RoundNarrow() to div255(). In practice we always want to round, and the narrowing to 8-bit is contextually obvious.
- Rename fastMulDiv255Round() approxMulDiv255() to stress it's approximate-ness over its speed. Drop Round for the same reason as above... we should always round.
- Add operator overloads so we don't have to keep throwing in seemingly-random Sk4px() or Sk4px::Wide() casts.
- use operator*() for 8-bit x 8-bit -> 16-bit math. It's always what we want, and there's generally no 8x8->8 alternative.
- MapFoo can take a const Func&. Don't think it makes a big difference, but nice to do.
BUG=skia:
Review URL: https://codereview.chromium.org/1202013002
I haven't figured out a pithy way to have these apply to only classes
originating from SkNx, so let's just remove them. There aren't too
many use cases, and it's not really any less readable without them.
Semantically, this is a no-op.
BUG=skia:
Review URL: https://codereview.chromium.org/1167153002
0x8001 / 0x7fff don't seem to work, but we were close: 0x8000 does.
I plan to use this to implement the Difference xfermode,
and it seems generally handy.
BUG=skia:
Review URL: https://codereview.chromium.org/1133933004
patch from issue 1072303005 at patchset 40001 (http://crrev.com/1072303005#ps40001)
This looks quite launchable. radial_gradient3, min of 100 samples:
N5: 985µs -> 946µs
MBP: 395µs -> 279µs
On my MBP, most of the meat looks like it's now in reading the cache and writing to dst one color at a time. Is that something we could do in float math rather than with a lookup table?
BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.8-Clang-Arm7-Debug-Android-Trybot,Build-Ubuntu-GCC-Arm7-Release-Android_NoNeon-Trybot
Committed: https://skia.googlesource.com/skia/+/abf6c5cf95e921fae59efb487480e5b5081cf0ec
Review URL: https://codereview.chromium.org/1109643002
Reason for revert:
compile failures.
Original issue's description:
> Mike's radial gradient CL with better float -> int.
>
> patch from issue 1072303005 at patchset 40001 (http://crrev.com/1072303005#ps40001)
>
> This looks quite launchable. radial_gradient3, min of 100 samples:
> N5: 985µs -> 946µs
> MBP: 395µs -> 279µs
>
> On my MBP, most of the meat looks like it's now in reading the cache and writing to dst one color at a time. Is that something we could do in float math rather than with a lookup table?
>
> BUG=skia:
>
> CQ_EXTRA_TRYBOTS=client.skia.android:Test-Android-GCC-Nexus5-CPU-NEON-Arm7-Debug-Trybot,Test-Android-GCC-Nexus9-CPU-Denver-Arm64-Debug-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/abf6c5cf95e921fae59efb487480e5b5081cf0ecTBR=reed@google.com,robertphillips@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Review URL: https://codereview.chromium.org/1109883003
patch from issue 1072303005 at patchset 40001 (http://crrev.com/1072303005#ps40001)
This looks quite launchable. radial_gradient3, min of 100 samples:
N5: 985µs -> 946µs
MBP: 395µs -> 279µs
On my MBP, most of the meat looks like it's now in reading the cache and writing to dst one color at a time. Is that something we could do in float math rather than with a lookup table?
BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia.android:Test-Android-GCC-Nexus5-CPU-NEON-Arm7-Debug-Trybot,Test-Android-GCC-Nexus9-CPU-Denver-Arm64-Debug-Trybot
Review URL: https://codereview.chromium.org/1109643002
These will underly the SkPMFloat-like class for uint16_t components.
Sk4h will back a single-pixel version, and Sk8h any larger number than that.
BUG=skia:
Review URL: https://codereview.chromium.org/1088883005
The primary feature this delivers is SkNf and SkNd for arbitrary power-of-two N. Non-specialized types or types larger than 128 bits should now Just Work (and we can drop in a specialization to make them faster). Sk4s is now just a typedef for SkNf<4, SkScalar>; Sk4d is SkNf<4, double>, Sk2f SkNf<2, float>, etc.
This also makes implementing new specializations easier and more encapsulated. We're now using template specialization, which means the specialized versions don't have to leak out so much from SkNx_sse.h and SkNx_neon.h.
This design leaves us room to grow up, e.g to SkNf<8, SkScalar> == Sk8s, and to grown down too, to things like SkNi<8, uint16_t> == Sk8h.
To simplify things, I've stripped away most APIs (swizzles, casts, reinterpret_casts) that no one's using yet. I will happily add them back if they seem useful.
You shouldn't feel bad about using any of the typedef Sk4s, Sk4f, Sk4d, Sk2s, Sk2f, Sk2d, Sk4i, etc. Here's how you should feel:
- Sk4f, Sk4s, Sk2d: feel awesome
- Sk2f, Sk2s, Sk4d: feel pretty good
No public API changes.
TBR=reed@google.com
BUG=skia:3592
Review URL: https://codereview.chromium.org/1048593002
