MyouProject/pixie
2
0
Fork 0
Code Issues Pull requests Actions Packages Projects Releases Wiki Activity

4x faster minifyBy2

Browse source
This commit is contained in:
Ryan Oldenburg 2021-06-18 00:59:31 -05:00
parent eb0d607796
commit 2870fca47e
1 changed files with 46 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

47
src/pixie/images.nim
View file

@ -212,7 +212,52 @@ proc minifyBy2*(image: Image, power = 1): Image =
for _ in 1 .. power:
result = newImage(src.width div 2, src.height div 2)
for y in 0 ..< result.height:
for x in 0 ..< result.width:
var x: int
when defined(amd64) and not defined(pixieNoSimd):
let
oddMask = mm_set1_epi16(cast[int16](0xff00))
first32 = cast[M128i]([uint32.high, 0, 0, 0])
for _ in countup(0, result.width - 4, 2):
let
top = mm_loadu_si128(src.data[src.dataIndex(x * 2, y * 2 + 0)].addr)
btm = mm_loadu_si128(src.data[src.dataIndex(x * 2, y * 2 + 1)].addr)
topShifted = mm_srli_si128(top, 4)
btmShifted = mm_srli_si128(btm, 4)
topEven = mm_andnot_si128(oddMask, top)
topOdd = mm_srli_epi16(mm_and_si128(top, oddMask), 8)
btmEven = mm_andnot_si128(oddMask, btm)
btmOdd = mm_srli_epi16(mm_and_si128(btm, oddMask), 8)
topShiftedEven = mm_andnot_si128(oddMask, topShifted)
topShiftedOdd = mm_srli_epi16(mm_and_si128(topShifted, oddMask), 8)
btmShiftedEven = mm_andnot_si128(oddMask, btmShifted)
btmShiftedOdd = mm_srli_epi16(mm_and_si128(btmShifted, oddMask), 8)
topAddedEven = mm_add_epi16(topEven, topShiftedEven)
btmAddedEven = mm_add_epi16(btmEven, btmShiftedEven)
topAddedOdd = mm_add_epi16(topOdd, topShiftedOdd)
bottomAddedOdd = mm_add_epi16(btmOdd, btmShiftedOdd)
addedEven = mm_add_epi16(topAddedEven, btmAddedEven)
addedOdd = mm_add_epi16(topAddedOdd, bottomAddedOdd)
addedEvenDiv4 = mm_srli_epi16(addedEven, 2)
addedOddDiv4 = mm_srli_epi16(addedOdd, 2)
merged = mm_or_si128(addedEvenDiv4, mm_slli_epi16(addedOddDiv4, 8))
# merged [0, 1, 2, 3] has the correct values for the next two pixels
# at index 0 and 2 so shift those into position and store
zero = mm_and_si128(merged, first32)
two = mm_and_si128(mm_srli_si128(merged, 8), first32)
zeroTwo = mm_or_si128(zero, mm_slli_si128(two, 4))
mm_storeu_si128(result.data[result.dataIndex(x, y)].addr, zeroTwo)
x += 2
for x in x ..< result.width:
let
a = src.getRgbaUnsafe(x * 2 + 0, y * 2 + 0)
b = src.getRgbaUnsafe(x * 2 + 1, y * 2 + 0)