Convert 32 ABGR pixels (128 bytes) to 32 Y values.

Convert AR30 to AB30.

Convert AR30 to ABGR.

24 instructions

16x2 pixels -> 8x1. width is number of argb pixels. e.g. 16.

16x2 pixels -> 8x1. width is number of argb pixels. e.g. 16.

18 instructions.

16x2 pixels -> 8x1. width is number of argb pixels. e.g. 16.

16x2 pixels -> 8x1. width is number of argb pixels. e.g. 16.

Add 2 ARGB images and store to destination.

Add 2 rows of ARGB pixels together, 8 pixels at a time.

Add 2 rows of ARGB pixels together, 4 pixels at a time.

Add 2 rows of ARGB pixels together, 8 pixels at a time.

Add 2 rows of ARGB pixels together, 8 pixels at a time.

Add 2 rows of ARGB pixels together, 4 pixels at a time. TODO(fbarchard): Port this to posix, neon and other math functions.

Add 2 rows of ARGB pixels together, 4 pixels at a time.

Copy pixels from rotated source to destination row with a slope.

Attenuate 8 pixels at a time.

Multiply source RGB by alpha and store to destination. This code mimics the SSSE3 version for better testability.

Multiply source RGB by alpha and store to destination. This code mimics the SSSE3 version for better testability.

Attenuate 8 pixels at a time.

Attenuate 8 pixels at a time.

Attenuate 4 pixels at a time.

Alpha Blend 2 ARGB images and store to destination.

Get a blender that optimized for the CPU and pixel count. As there are 6 blenders to choose from, the caller should try to use the same blend function

Blend src_argb0 over src_argb1 and store to dst_argb. dst_argb may be src_argb0 or src_argb1. This code mimics the SSSE3 version for better testabilit

Blend src_argb0 over src_argb1 and store to dst_argb. dst_argb may be src_argb0 or src_argb1. This code mimics the SSSE3 version for better testabilit

dr * (256 - sa) / 256 + sr = dr - dr * sa / 256 + sr

dr * (256 - sa) / 256 + sr = dr - dr * sa / 256 + sr

Blend 8 pixels at a time.

Blend 8 pixels at a time

Blur ARGB image. Caller should allocate CumulativeSum table of width * height * 16 bytes aligned to 16 byte boundary. height can be radius * 2 + 2 to

Apply color matrix to a row of image. Matrix is signed. TODO(fbarchard): Consider adding rounding (+32).

Apply color matrix to a row of image. Matrix is signed. TODO(fbarchard): Consider adding rounding (+32).

Tranform 8 ARGB pixels (32 bytes) with color matrix. TODO(fbarchard): Was same as Sepia except matrix is provided. This function needs to saturate.

Tranform 8 ARGB pixels (32 bytes) with color matrix. TODO(fbarchard): Was same as Sepia except matrix is provided. This function needs to saturate.

Tranform 8 ARGB pixels (32 bytes) with color matrix. Same as Sepia except matrix is provided. TODO(fbarchard): packuswbs only use half of the reg. To

Tranform 8 ARGB pixels (32 bytes) with color matrix. Same as Sepia except matrix is provided.

Apply a color table each ARGB pixel. Table contains 256 ARGB values.

Apply color table to a row of image.

Tranform ARGB pixels with color table.

Copy Alpha from one ARGB image to another.

width in pixels

width in pixels

Copy a planar Y channel to the alpha channel of a destination ARGB image.

width in pixels

width in pixels

Scan an opaque argb image and return fourcc based on alpha offset. Returns FOURCC_ARGB, FOURCC_BGRA, or 0 if unknown.

Extract just the alpha channel from ARGB.