MyouProject/pixie
3
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

better

Browse source
This commit is contained in:
Ryan Oldenburg 2022-07-30 23:49:05 -05:00
parent c4cd380676
commit 0245eeebea
1 changed files with 17 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

26
src/pixie/simd/neon.nim
View file

@ -58,9 +58,9 @@ proc isOneColorNeon*(image: Image): bool {.simd.} =
rgEq = vandq_u8(rEq, gEq) rgEq = vandq_u8(rEq, gEq)
baEq = vandq_u8(bEq, aEq) baEq = vandq_u8(bEq, aEq)
rgbaEq = vandq_u8(rgEq, baEq) rgbaEq = vandq_u8(rgEq, baEq)
mask = mask = vget_lane_u64(cast[uint64x1](
cast[uint64](vget_low_u64(cast[uint64x2](rgbaEq))) and vand_u8(vget_low_u8(rgbaEq), vget_high_u8(rgbaEq)
cast[uint64](vget_high_u64(cast[uint64x2](rgbaEq))) )), 0)
if mask != uint64.high: if mask != uint64.high:
return false return false
i += 16 i += 16
@ -82,12 +82,16 @@ proc isTransparentNeon*(image: Image): bool {.simd.} =
result = true result = true
let iterations = (image.data.len - i) div 16 let
vecZero = vmovq_n_u8(0)
iterations = (image.data.len - i) div 16
for _ in 0 ..< iterations: for _ in 0 ..< iterations:
let let
alphas = vld4q_u8(image.data[i].addr).val[3] alphas = vld4q_u8(image.data[i].addr).val[3]
eq = vceqq_u64(cast[uint64x2](alphas), vmovq_n_u64(0)) eq = vceqq_u8(alphas, vecZero)
mask = cast[uint64](vget_low_u64(eq)) and cast[uint64](vget_high_u64(eq)) mask = vget_lane_u64(cast[uint64x1](
vand_u8(vget_low_u8(eq), vget_high_u8(eq)
)), 0)
if mask != uint64.high: if mask != uint64.high:
return false return false
i += 16 i += 16
@ -109,12 +113,16 @@ proc isOpaqueNeon*(data: var seq[ColorRGBX], start, len: int): bool {.simd.} =
inc i inc i
p += 4 p += 4
let iterations = (start + len - i) div 16 let
vec255 = vmovq_n_u8(255)
iterations = (start + len - i) div 16
for _ in 0 ..< iterations: for _ in 0 ..< iterations:
let let
alphas = vld4q_u8(data[i].addr).val[3] alphas = vld4q_u8(data[i].addr).val[3]
eq = vceqq_u64(cast[uint64x2](alphas), vmovq_n_u64(uint64.high)) eq = vceqq_u8(alphas, vec255)
mask = cast[uint64](vget_low_u64(eq)) and cast[uint64](vget_high_u64(eq)) mask = vget_lane_u64(cast[uint64x1](
vand_u8(vget_low_u8(eq), vget_high_u8(eq)
)), 0)
if mask != uint64.high: if mask != uint64.high:
return false return false
i += 16 i += 16