SkHalfToFloat_01 / SkFloatToHalf_01

SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1685133005

Committed: https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed: https://skia.googlesource.com/skia/+/fff055cc5f9ca5015678f4f13a4f842084bd62d5

[mtklein_C, 2016-02-10 20:27:17]

Description was changed from

==========
HalfToFloat_01

BUG=skia:
==========

to

==========
HalfToFloat_01

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:50:31]

Description was changed from

==========
HalfToFloat_01

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:52:26]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01


curr/maxrss	loops	min	median	mean	max	stddev	samples   	config	bench
   8/8  
MB	1	254µs	264µs	260µs	265µs	2%	▁▁▁▇████▅▁	nonrendering	xferu64_bw_1_opaque_u16
   8/8  
MB	1	1.75ms	1.83ms	1.83ms	2.1ms	6%	█▃▃▃▃▃▁▁▁▁	nonrendering	xferu64_bw_1_alpha_u16
   8/8  
MB	1	2.66ms	2.78ms	2.78ms	3.02ms	4%	▃▃▃▁▁▄█▃▃▃	nonrendering	xferu64_aa_1_opaque_u16
   8/8  
MB	1	3.26ms	3.41ms	3.36ms	3.55ms	3%	▁▁▁▁▃▅▅█▅▅	nonrendering	xferu64_aa_1_alpha_u16
   8/8  
MB	2	268µs	268µs	268µs	272µs	1%	▁▁▁▅█▁▁▁▁▁	nonrendering	xferu64_bw_1_opaque_f16
   8/8  
MB	1	3.24ms	3.29ms	3.32ms	3.57ms	3%	▁▁▁▂▁▁█▄▄▄	nonrendering	xferu64_bw_1_alpha_f16
   8/8  
MB	1	4.03ms	4.06ms	4.13ms	4.38ms	3%	▁▁▁▂▁▁█▅▅▅	nonrendering	xferu64_aa_1_opaque_f16
   8/8  
MB	1	4.67ms	4.7ms	4.76ms	4.97ms	2%	▁▁▂▃▁█▆▆▂▁	nonrendering	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:54:50]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01


curr/maxrss	loops	min	median	mean	max	stddev	samples   	config	bench
   8/8  
MB	1	254µs	264µs	260µs	265µs	2%	▁▁▁▇████▅▁	nonrendering	xferu64_bw_1_opaque_u16
   8/8  
MB	1	1.75ms	1.83ms	1.83ms	2.1ms	6%	█▃▃▃▃▃▁▁▁▁	nonrendering	xferu64_bw_1_alpha_u16
   8/8  
MB	1	2.66ms	2.78ms	2.78ms	3.02ms	4%	▃▃▃▁▁▄█▃▃▃	nonrendering	xferu64_aa_1_opaque_u16
   8/8  
MB	1	3.26ms	3.41ms	3.36ms	3.55ms	3%	▁▁▁▁▃▅▅█▅▅	nonrendering	xferu64_aa_1_alpha_u16
   8/8  
MB	2	268µs	268µs	268µs	272µs	1%	▁▁▁▅█▁▁▁▁▁	nonrendering	xferu64_bw_1_opaque_f16
   8/8  
MB	1	3.24ms	3.29ms	3.32ms	3.57ms	3%	▁▁▁▂▁▁█▄▄▄	nonrendering	xferu64_bw_1_alpha_f16
   8/8  
MB	1	4.03ms	4.06ms	4.13ms	4.38ms	3%	▁▁▁▂▁▁█▅▅▅	nonrendering	xferu64_aa_1_opaque_f16
   8/8  
MB	1	4.67ms	4.7ms	4.76ms	4.97ms	2%	▁▁▂▃▁█▆▆▂▁	nonrendering	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01


Getting close to _u16 performance:

curr/maxrss	loops	min	median	mean	max	stddev	samples   	config	bench
   8/8  
MB	1	254µs	264µs	260µs	265µs	2%	▁▁▁▇████▅▁	nonrendering	xferu64_bw_1_opaque_u16
   8/8  
MB	1	1.75ms	1.83ms	1.83ms	2.1ms	6%	█▃▃▃▃▃▁▁▁▁	nonrendering	xferu64_bw_1_alpha_u16
   8/8  
MB	1	2.66ms	2.78ms	2.78ms	3.02ms	4%	▃▃▃▁▁▄█▃▃▃	nonrendering	xferu64_aa_1_opaque_u16
   8/8  
MB	1	3.26ms	3.41ms	3.36ms	3.55ms	3%	▁▁▁▁▃▅▅█▅▅	nonrendering	xferu64_aa_1_alpha_u16
   8/8  
MB	2	268µs	268µs	268µs	272µs	1%	▁▁▁▅█▁▁▁▁▁	nonrendering	xferu64_bw_1_opaque_f16
   8/8  
MB	1	3.24ms	3.29ms	3.32ms	3.57ms	3%	▁▁▁▂▁▁█▄▄▄	nonrendering	xferu64_bw_1_alpha_f16
   8/8  
MB	1	4.03ms	4.06ms	4.13ms	4.38ms	3%	▁▁▁▂▁▁█▅▅▅	nonrendering	xferu64_aa_1_opaque_f16
   8/8  
MB	1	4.67ms	4.7ms	4.76ms	4.97ms	2%	▁▁▂▃▁█▆▆▂▁	nonrendering	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:55:18]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01


Getting close to _u16 performance:

curr/maxrss	loops	min	median	mean	max	stddev	samples   	config	bench
   8/8  
MB	1	254µs	264µs	260µs	265µs	2%	▁▁▁▇████▅▁	nonrendering	xferu64_bw_1_opaque_u16
   8/8  
MB	1	1.75ms	1.83ms	1.83ms	2.1ms	6%	█▃▃▃▃▃▁▁▁▁	nonrendering	xferu64_bw_1_alpha_u16
   8/8  
MB	1	2.66ms	2.78ms	2.78ms	3.02ms	4%	▃▃▃▁▁▄█▃▃▃	nonrendering	xferu64_aa_1_opaque_u16
   8/8  
MB	1	3.26ms	3.41ms	3.36ms	3.55ms	3%	▁▁▁▁▃▅▅█▅▅	nonrendering	xferu64_aa_1_alpha_u16
   8/8  
MB	2	268µs	268µs	268µs	272µs	1%	▁▁▁▅█▁▁▁▁▁	nonrendering	xferu64_bw_1_opaque_f16
   8/8  
MB	1	3.24ms	3.29ms	3.32ms	3.57ms	3%	▁▁▁▂▁▁█▄▄▄	nonrendering	xferu64_bw_1_alpha_f16
   8/8  
MB	1	4.03ms	4.06ms	4.13ms	4.38ms	3%	▁▁▁▂▁▁█▅▅▅	nonrendering	xferu64_aa_1_opaque_f16
   8/8  
MB	1	4.67ms	4.7ms	4.76ms	4.97ms	2%	▁▁▂▃▁█▆▆▂▁	nonrendering	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01


Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:56:17]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01


Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:57:18]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:57:37]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:

    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:57:48]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, vectorized versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:59:09]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:59:25]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for speed, f32->f16 does not round properly.  Instead it truncates.
It's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.  Instead it
truncates, so it's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:59:33]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.  Instead it
truncates, so it's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[mtklein_C, 2016-02-10 21:59:48]

The CQ bit was checked by mtklein@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-10 21:59:59]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/100001

[mtklein_C, 2016-02-10 22:02:17]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[commit-bot: I haz the power, 2016-02-10 22:02:49]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-02-10 22:02:50]

Dry run: Try jobs failed on following builders:
  Build-Mac-Clang-Arm7-Debug-iOS-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Mac-Clang-Arm7...)
  Build-Ubuntu-GCC-Arm64-Debug-Android-Trybot on client.skia.compile
(JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Ubuntu-GCC-Arm...)
  Build-Ubuntu-GCC-Arm7-Debug-Android-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Ubuntu-GCC-Arm...)
  Build-Ubuntu-GCC-Mips-Debug-Android-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Ubuntu-GCC-Mip...)
  Build-Win-MSVC-x86-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86-D...)
  Build-Win-MSVC-x86_64-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86_6...)

[mtklein_C, 2016-02-10 22:05:12]

The CQ bit was checked by mtklein@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-10 22:05:22]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/120001

[commit-bot: I haz the power, 2016-02-10 22:06:15]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-02-10 22:06:16]

Dry run: Try jobs failed on following builders:
  Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Debug-Trybot on client.skia (JOB_FAILED,
http://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX2...)
  Build-Mac-Clang-Arm7-Debug-iOS-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Mac-Clang-Arm7...)

[mtklein_C, 2016-02-10 22:11:14]

The CQ bit was checked by mtklein@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-10 22:11:22]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/160001

[commit-bot: I haz the power, 2016-02-10 22:13:35]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-02-10 22:13:36]

Dry run: Try jobs failed on following builders:
  Build-Win-MSVC-x86-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86-D...)
  Build-Win-MSVC-x86_64-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86_6...)

[mtklein_C, 2016-02-10 22:14:38]

The CQ bit was checked by mtklein@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-10 22:14:46]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/180001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/180001

[commit-bot: I haz the power, 2016-02-10 22:16:58]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-02-10 22:16:58]

Dry run: Try jobs failed on following builders:
  Build-Win-MSVC-x86-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86-D...)

[mtklein_C, 2016-02-10 22:18:22]

The CQ bit was checked by mtklein@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-10 22:18:27]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/200001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/200001

[mtklein_C, 2016-02-10 22:20:44]

mtklein@chromium.org changed reviewers:
+ jvanverth@google.com, reed@google.com

[mtklein_C, 2016-02-10 22:20:45]

The bots may have a little more complaining to do, but it's probably good enough
to take a look at now.

[mtklein_C, 2016-02-10 22:27:55]

(A weirder version of f16 -> f32 requiring more denormalized multiplication,

    __m128i h = _mm_unpacklo_epi16(_mm_loadl_epi64((const __m128i*)&hs),
_mm_setzero_si128());
    const __m128 magic = _mm_castsi128_ps(_mm_set1_epi32(239 << 23));
    return _mm_mul_ps(magic, _mm_castsi128_ps(_mm_slli_epi32(h, 13)));

results in very interesting performance numbers:

    261.13  	xferu64_bw_1_opaque_u16	nonrendering
   1832.58  	xferu64_bw_1_alpha_u16	nonrendering
   2704.76  	xferu64_aa_1_opaque_u16	nonrendering
   3268.43  	xferu64_aa_1_alpha_u16	nonrendering
    262.84  	xferu64_bw_1_opaque_f16	nonrendering
   2061.21  	xferu64_bw_1_alpha_f16	nonrendering
   2906.57  	xferu64_aa_1_opaque_f16	nonrendering
   8183.68  	xferu64_aa_1_alpha_f16	nonrendering

Generally closer to _u16, with one notable regression (probably hitting the
denormalized multiply more often).

[commit-bot: I haz the power, 2016-02-10 22:51:57]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-02-10 22:51:58]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-02-11 13:47:39]

double-woot!

Glad we at least have a chance to add back in < 0 and > 1 if needed in the
future, but certainly not needed today.

lgtm

[mtklein, 2016-02-11 13:55:25]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2016-02-11 13:55:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/200001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/200001

[commit-bot: I haz the power, 2016-02-11 13:56:10]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9
==========

[commit-bot: I haz the power, 2016-02-11 13:56:11]

Committed patchset #11 (id:200001) as
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9

[mtklein, 2016-02-11 14:00:40]

A revert of this CL (patchset #11 id:200001) has been created in
https://codereview.chromium.org/1693443003/ by mtklein@google.com.

The reason for reverting is: Gotta fix
Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD.

[mtklein, 2016-02-11 14:01:48]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[mtklein, 2016-02-11 14:04:06]

The CQ bit was checked by mtklein@google.com

[mtklein, 2016-02-11 14:04:06]

The patchset sent to the CQ was uploaded after l-g-t-m from reed@google.com
Link to the patchset: https://codereview.chromium.org/1685133005/#ps220001
(title: "guard")

[commit-bot: I haz the power, 2016-02-11 14:04:11]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685133005/220001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685133005/220001

[commit-bot: I haz the power, 2016-02-11 14:30:06]

Description was changed from

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
SkHalfToFloat_01 / SkFloatToHalf_01

These are basically inlined, 4-at-a-time versions of our existing functions,
but cut down to avoid any work that's only necessary outside [0,1].

Both f16 and f32 denorms should work fine modulo the usual ARMv7 NEON
denorm==zero caveat.

In exchange for a little speed, f32->f16 does not round properly.
Instead it truncates, so it's never off by more than 1 bit.

Support for finite values >1 or <0 is straightforward to add back.
>1 might already work as-is.

Getting close to _u16 performance:
    micros   	bench
    261.13  	xferu64_bw_1_opaque_u16
   1833.51  	xferu64_bw_1_alpha_u16
   2762.32 ?	xferu64_aa_1_opaque_u16
   3334.29  	xferu64_aa_1_alpha_u16
    249.78  	xferu64_bw_1_opaque_f16
   3383.18  	xferu64_bw_1_alpha_f16
   4214.72  	xferu64_aa_1_opaque_f16
   4701.19  	xferu64_aa_1_alpha_f16


BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/9ea11a4235b3e3521cc8bf914a27c2d0dc062db9

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/fff055cc5f9ca5015678f4f13a4f842084bd62d5
==========

[commit-bot: I haz the power, 2016-02-11 14:30:07]

Committed patchset #12 (id:220001) as
https://skia.googlesource.com/skia/+/fff055cc5f9ca5015678f4f13a4f842084bd62d5