7794 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Alexander	d9cc462106	Merge branch 'simd/rgba-convert' into simd/5.3.x	2018-10-17 14:52:37 +03:00
Alexander	b646ac278f	Merge branch 'simd/resample' into simd/5.3.x	2018-10-17 14:52:32 +03:00
Alexander	dd99b65d78	Merge branch 'simd/filters' into simd/5.3.x	2018-10-17 14:52:26 +03:00
Alexander	32f3dff6f5	Merge branch 'simd/box-blur' into simd/5.3.x	2018-10-17 14:51:58 +03:00
Alexander	0fc8680360	Merge branch 'simd/alpha-composite' into simd/5.3.x	2018-10-17 14:51:45 +03:00
Alexander	8c38010f7d	Speedup other 2L convertions	2018-10-05 13:57:28 +03:00
Alexander	87385595ce	RGB → L 2.2 times faster	2018-10-05 13:57:28 +03:00
Alexander	ca27f8197b	fix rounding and speedup a bit	2018-10-05 13:57:28 +03:00
Alexander	7c7d7018b1	use 16bit arithmetics	2018-10-05 13:57:28 +03:00
Alexander	7f2b368e85	sse4 version (still 1.4x faster than previous avx2 implementation)	2018-10-05 13:57:28 +03:00
Alexander	89ddb0d95a	use float div instead of gather	2018-10-05 13:57:28 +03:00
homm	a92659f65c	fix RGBa → RGBA conversion on AVX2	2018-10-05 13:57:28 +03:00
homm	a880dd08e9	RGBa → RGBA convert using gather	2018-10-05 13:57:28 +03:00
homm	880fede485	avx2 implementation	2018-10-05 13:57:28 +03:00
homm	096aaa1e6c	faster implementation	2018-10-05 13:57:28 +03:00
homm	fdef92c60a	sse4 implementation	2018-10-05 13:57:28 +03:00
Alexander	adc2e0302d	move files	2018-10-05 13:55:10 +03:00
Alexander	ef1692649d	add parentheses around var declarations	2018-10-05 13:55:10 +03:00
Alexander	80a64c013e	optimize coefficients loading for horizontal pass ... wtf is xmax / 2 optimize coefficients loading for vertical pass	2018-10-05 13:55:10 +03:00
homm	b7b3b26483	SIMD resample: unrolled SSE4 & AVX2	2018-10-05 13:55:10 +03:00
Alexander	ff5ed4f6d5	move files	2018-10-05 13:52:49 +03:00
Alexander	1713b71c0a	fix memory access for: ... 3x3f_u8 3x3i_4u8 5x5i_4u8	2018-10-05 13:52:49 +03:00
Alexander	94ea64c416	5x5i_4u8 AVX2	2018-10-05 13:52:49 +03:00
Alexander	3da294ca21	advanced 5x5i_4u8 SSE4	2018-10-05 13:52:49 +03:00
Alexander	96b367c571	5x5i_4u8 SSE4	2018-10-05 13:52:49 +03:00
Alexander	7bd48c8f63	finish 3x3i_4u8	2018-10-05 13:52:49 +03:00
Alexander	cb68d00256	avx2 version	2018-10-05 13:52:49 +03:00
Alexander	3b7b833f45	rearrange operations	2018-10-05 13:52:49 +03:00
Alexander	c4085db81e	reduce number of registers	2018-10-05 13:52:49 +03:00
Alexander	0b3550c24f	Rearrange instruction for speedup	2018-10-05 13:52:49 +03:00
Alexander	e4c9528d55	better loading	2018-10-05 13:52:49 +03:00
Alexander	8695387f05	better macros	2018-10-05 13:52:49 +03:00
Alexander	3e8574ae26	3x3i	2018-10-05 13:52:49 +03:00
Alexander	44c56befbd	move ImagingFilterxxx functions to separate files	2018-10-05 13:52:49 +03:00
Alexander	98bed5abae	fix offset	2018-10-05 13:52:49 +03:00
Alexander	db69139906	5x5 single channel SSE4 (tests failed)	2018-10-05 13:52:49 +03:00
Alexander	cdde46ae17	consider last pixel in AVX	2018-10-05 13:52:49 +03:00
Alexander	5ca47243f8	unroll AVX (with no profit)	2018-10-05 13:52:49 +03:00
Alexander	c30554ca64	Macros for AVX	2018-10-05 13:52:49 +03:00
Alexander	0d36fd05ee	unroll AVX 2 times	2018-10-05 13:52:49 +03:00
Alexander	3c3623265c	First AVX try	2018-10-05 13:52:49 +03:00
Alexander	ee7158d8d5	3x3 SSE4 singleband: 2 lines	2018-10-05 13:52:49 +03:00
Alexander	9966e832e0	reuse loaded values	2018-10-05 13:52:49 +03:00
Alexander	32c372a616	faster 3x3 singleband SSE4	2018-10-05 13:52:49 +03:00
Alexander	86c8aac6f8	3x3 SSE4 singleband	2018-10-05 13:52:49 +03:00
Alexander	bef019f9cf	use macros in 3x3	2018-10-05 13:52:49 +03:00
Alexander	78e99deaef	use macros	2018-10-05 13:52:49 +03:00
Alexander	328bf4593e	rearrange 3x3 filter to match 5x5	2018-10-05 13:52:48 +03:00
Alexander	8a351e1e31	improve locality in 5x5 filter	2018-10-05 13:52:48 +03:00
Alexander	9c8a9014c4	a bit faster 5x5 filter	2018-10-05 13:52:48 +03:00