Skip to content

Replace hasSimd macro with explicit when dispatch #599

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
treeform wants to merge 1 commit into
base: master
Choose a base branch
from
Draft

Replace hasSimd macro with explicit when dispatch #599

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
348 changes: 211 additions & 137 deletions src/pixie/images.nim
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,20 +56,36 @@ proc fill*(image: Image, color: SomeColor) {.inline, raises: [].} =
## Fills the image with the color.
fillUnsafe(image.data, color, 0, image.data.len)

proc isOneColor*(image: Image): bool {.hasSimd, raises: [].} =
proc isOneColor*(image: Image): bool {.raises: [].} =
## Checks if the entire image is the same color.
result = true
let color = cast[uint32](image.data[0])
for i in 0 ..< image.data.len:
if cast[uint32](image.data[i]) != color:
return false
when allowAvx:
if cpuHasAvx2:
return isOneColorAvx2(image)
when allowSse2:
return isOneColorSse2(image)
elif allowNeon:
return isOneColorNeon(image)
else:
result = true
let color = cast[uint32](image.data[0])
for i in 0 ..< image.data.len:
if cast[uint32](image.data[i]) != color:
return false

proc isTransparent*(image: Image): bool {.hasSimd, raises: [].} =
proc isTransparent*(image: Image): bool {.raises: [].} =
## Checks if this image is fully transparent or not.
result = true
for i in 0 ..< image.data.len:
if image.data[i].a != 0:
return false
when allowAvx:
if cpuHasAvx2:
return isTransparentAvx2(image)
when allowSse2:
return isTransparentSse2(image)
elif allowNeon:
return isTransparentNeon(image)
else:
result = true
for i in 0 ..< image.data.len:
if image.data[i].a != 0:
return false

proc isOpaque*(image: Image): bool {.raises: [].} =
## Checks if the entire image is opaque (alpha values are all 255).
Expand Down Expand Up @@ -167,139 +183,179 @@ proc diff*(master, image: Image): (float32, Image) {.raises: [PixieError].} =

proc minifyBy2*(
image: Image, power = 1
): Image {.hasSimd, raises: [PixieError].} =
): Image {.raises: [PixieError].} =
## Scales the image down by an integer scale.
if power < 0:
raise newException(PixieError, "Cannot minifyBy2 with negative power")
if power == 0:
return image.copy()

var src = image
for _ in 1 .. power:
# When minifying an image of odd size, round the result image size up
# so a 99 x 99 src image returns a 50 x 50 image.
let
srcWidthIsOdd = (src.width mod 2) != 0
srcHeightIsOdd = (src.height mod 2) != 0
resultEvenWidth = src.width div 2
resultEvenHeight = src.height div 2
result = newImage(
if srcWidthIsOdd: resultEvenWidth + 1 else: resultEvenWidth,
if srcHeightIsOdd: resultEvenHeight + 1 else: resultEvenHeight
)
for y in 0 ..< resultEvenHeight:
when allowAvx:
if cpuHasAvx2:
return minifyBy2Avx2(image, power)
when allowSse2:
return minifyBy2Sse2(image, power)
elif allowNeon:
return minifyBy2Neon(image, power)
else:
if power < 0:
raise newException(PixieError, "Cannot minifyBy2 with negative power")
if power == 0:
return image.copy()

var src = image
for _ in 1 .. power:
# When minifying an image of odd size, round the result image size up
# so a 99 x 99 src image returns a 50 x 50 image.
let
topRowStart = src.dataIndex(0, y * 2)
bottomRowStart = src.dataIndex(0, y * 2 + 1)
for x in 0 ..< resultEvenWidth:
srcWidthIsOdd = (src.width mod 2) != 0
srcHeightIsOdd = (src.height mod 2) != 0
resultEvenWidth = src.width div 2
resultEvenHeight = src.height div 2
result = newImage(
if srcWidthIsOdd: resultEvenWidth + 1 else: resultEvenWidth,
if srcHeightIsOdd: resultEvenHeight + 1 else: resultEvenHeight
)
for y in 0 ..< resultEvenHeight:
let
a = src.data[topRowStart + x * 2]
b = src.data[topRowStart + x * 2 + 1]
c = src.data[bottomRowStart + x * 2 + 1]
d = src.data[bottomRowStart + x * 2]
mixed = rgbx(
((a.r.uint32 + b.r + c.r + d.r + 2) div 4).uint8,
((a.g.uint32 + b.g + c.g + d.g + 2) div 4).uint8,
((a.b.uint32 + b.b + c.b + d.b + 2) div 4).uint8,
((a.a.uint32 + b.a + c.a + d.a + 2) div 4).uint8
)
result.data[result.dataIndex(x, y)] = mixed

if srcWidthIsOdd:
let rgbx = mix(
src.data[src.dataIndex(src.width - 1, y * 2 + 0)],
src.data[src.dataIndex(src.width - 1, y * 2 + 1)],
0.5
) * 0.5
result.data[result.dataIndex(result.width - 1, y)] = rgbx

if srcHeightIsOdd:
for x in 0 ..< resultEvenWidth:
let rgbx = mix(
src.data[src.dataIndex(x * 2 + 0, src.height - 1)],
src.data[src.dataIndex(x * 2 + 1, src.height - 1)],
0.5
) * 0.5
result.data[result.dataIndex(x, result.height - 1)] = rgbx

if srcWidthIsOdd:
result.data[result.dataIndex(result.width - 1, result.height - 1)] =
src.data[src.dataIndex(src.width - 1, src.height - 1)] * 0.25

# Set src as this result for if we do another power
src = result
topRowStart = src.dataIndex(0, y * 2)
bottomRowStart = src.dataIndex(0, y * 2 + 1)
for x in 0 ..< resultEvenWidth:
let
a = src.data[topRowStart + x * 2]
b = src.data[topRowStart + x * 2 + 1]
c = src.data[bottomRowStart + x * 2 + 1]
d = src.data[bottomRowStart + x * 2]
mixed = rgbx(
((a.r.uint32 + b.r + c.r + d.r + 2) div 4).uint8,
((a.g.uint32 + b.g + c.g + d.g + 2) div 4).uint8,
((a.b.uint32 + b.b + c.b + d.b + 2) div 4).uint8,
((a.a.uint32 + b.a + c.a + d.a + 2) div 4).uint8
)
result.data[result.dataIndex(x, y)] = mixed

if srcWidthIsOdd:
let rgbx = mix(
src.data[src.dataIndex(src.width - 1, y * 2 + 0)],
src.data[src.dataIndex(src.width - 1, y * 2 + 1)],
0.5
) * 0.5
result.data[result.dataIndex(result.width - 1, y)] = rgbx

if srcHeightIsOdd:
for x in 0 ..< resultEvenWidth:
let rgbx = mix(
src.data[src.dataIndex(x * 2 + 0, src.height - 1)],
src.data[src.dataIndex(x * 2 + 1, src.height - 1)],
0.5
) * 0.5
result.data[result.dataIndex(x, result.height - 1)] = rgbx

if srcWidthIsOdd:
result.data[result.dataIndex(result.width - 1, result.height - 1)] =
src.data[src.dataIndex(src.width - 1, src.height - 1)] * 0.25

# Set src as this result for if we do another power
src = result

proc magnifyBy2*(
image: Image, power = 1
): Image {.hasSimd, raises: [PixieError].} =
): Image {.raises: [PixieError].} =
## Scales image up by 2 ^ power.
if power < 0:
raise newException(PixieError, "Cannot magnifyBy2 with negative power")
when allowSse2:
return magnifyBy2Sse2(image, power)
elif allowNeon:
return magnifyBy2Neon(image, power)
else:
if power < 0:
raise newException(PixieError, "Cannot magnifyBy2 with negative power")

let scale = 2 ^ power
result = newImage(image.width * scale, image.height * scale)
let scale = 2 ^ power
result = newImage(image.width * scale, image.height * scale)

for y in 0 ..< image.height:
# Write one row of pixels duplicated by scale
let
sourceRowStart = image.dataIndex(0, y)
resultRowStart = result.dataIndex(0, y * scale)
for x in 0 ..< image.width:
for y in 0 ..< image.height:
# Write one row of pixels duplicated by scale
let
rgbx = image.data[sourceRowStart + x]
resultIdx = resultRowStart + x * scale
for i in 0 ..< scale:
result.data[resultIdx + i] = rgbx
# Copy that row of pixels into (scale - 1) more rows
for i in 1 ..< scale:
copyMem(
result.data[resultRowStart + result.width * i].addr,
result.data[resultRowStart].addr,
result.width * 4
)
sourceRowStart = image.dataIndex(0, y)
resultRowStart = result.dataIndex(0, y * scale)
for x in 0 ..< image.width:
let
rgbx = image.data[sourceRowStart + x]
resultIdx = resultRowStart + x * scale
for i in 0 ..< scale:
result.data[resultIdx + i] = rgbx
# Copy that row of pixels into (scale - 1) more rows
for i in 1 ..< scale:
copyMem(
result.data[resultRowStart + result.width * i].addr,
result.data[resultRowStart].addr,
result.width * 4
)

proc applyOpacity*(image: Image, opacity: float32) {.hasSimd, raises: [].} =
proc applyOpacity*(image: Image, opacity: float32) {.raises: [].} =
## Multiplies alpha of the image by opacity.
let opacity = round(255 * opacity).uint16
if opacity == 255:
return

if opacity == 0:
image.fill(rgbx(0, 0, 0, 0))
return

for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = ((rgbx.r * opacity) div 255).uint8
rgbx.g = ((rgbx.g * opacity) div 255).uint8
rgbx.b = ((rgbx.b * opacity) div 255).uint8
rgbx.a = ((rgbx.a * opacity) div 255).uint8
image.data[i] = rgbx

proc invert*(image: Image) {.hasSimd, raises: [].} =
when allowAvx:
if cpuHasAvx2:
applyOpacityAvx2(image, opacity)
return
when allowSse2:
applyOpacitySse2(image, opacity)
elif allowNeon:
applyOpacityNeon(image, opacity)
else:
let opacity = round(255 * opacity).uint16
if opacity == 255:
return

if opacity == 0:
image.fill(rgbx(0, 0, 0, 0))
return

for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = ((rgbx.r * opacity) div 255).uint8
rgbx.g = ((rgbx.g * opacity) div 255).uint8
rgbx.b = ((rgbx.b * opacity) div 255).uint8
rgbx.a = ((rgbx.a * opacity) div 255).uint8
image.data[i] = rgbx

proc invert*(image: Image) {.raises: [].} =
## Inverts all of the colors and alpha.
for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = 255 - rgbx.r
rgbx.g = 255 - rgbx.g
rgbx.b = 255 - rgbx.b
rgbx.a = 255 - rgbx.a
image.data[i] = rgbx

# Inverting rgbx(50, 100, 150, 200) becomes rgbx(205, 155, 105, 55). This
# is not a valid premultiplied alpha color.
# We need to convert back to premultiplied alpha after inverting.
image.data.toPremultipliedAlpha()

proc ceil*(image: Image) {.hasSimd, raises: [].} =
when allowAvx:
if cpuHasAvx2:
invertAvx2(image)
return
when allowSse2:
invertSse2(image)
elif allowNeon:
invertNeon(image)
else:
for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = 255 - rgbx.r
rgbx.g = 255 - rgbx.g
rgbx.b = 255 - rgbx.b
rgbx.a = 255 - rgbx.a
image.data[i] = rgbx

# Inverting rgbx(50, 100, 150, 200) becomes rgbx(205, 155, 105, 55). This
# is not a valid premultiplied alpha color.
# We need to convert back to premultiplied alpha after inverting.
image.data.toPremultipliedAlpha()

proc ceil*(image: Image) {.raises: [].} =
## A value of 0 stays 0. Anything else turns into 255.
for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = if rgbx.r == 0: 0 else: 255
rgbx.g = if rgbx.g == 0: 0 else: 255
rgbx.b = if rgbx.b == 0: 0 else: 255
rgbx.a = if rgbx.a == 0: 0 else: 255
image.data[i] = rgbx
when allowAvx:
if cpuHasAvx2:
ceilAvx2(image)
return
when allowSse2:
ceilSse2(image)
elif allowNeon:
ceilNeon(image)
else:
for i in 0 ..< image.data.len:
var rgbx = image.data[i]
rgbx.r = if rgbx.r == 0: 0 else: 255
rgbx.g = if rgbx.g == 0: 0 else: 255
rgbx.b = if rgbx.b == 0: 0 else: 255
rgbx.a = if rgbx.a == 0: 0 else: 255
image.data[i] = rgbx

proc blur*(
image: Image, radius: float32, outOfBounds: SomeColor = color(0, 0, 0, 0)
Expand Down Expand Up @@ -473,13 +529,31 @@ proc blendLineOverwrite(

proc blendLineNormal(
a, b: ptr UncheckedArray[ColorRGBX], len: int
) {.hasSimd.} =
for i in 0 ..< len:
a[i] = blendNormal(a[i], b[i])

proc blendLineMask(a, b: ptr UncheckedArray[ColorRGBX], len: int) {.hasSimd.} =
for i in 0 ..< len:
a[i] = blendMask(a[i], b[i])
) =
when allowAvx:
if cpuHasAvx2:
blendLineNormalAvx2(a, b, len)
return
when allowSse2:
blendLineNormalSse2(a, b, len)
elif allowNeon:
blendLineNormalNeon(a, b, len)
else:
for i in 0 ..< len:
a[i] = blendNormal(a[i], b[i])

proc blendLineMask(a, b: ptr UncheckedArray[ColorRGBX], len: int) =
when allowAvx:
if cpuHasAvx2:
blendLineMaskAvx2(a, b, len)
return
when allowSse2:
blendLineMaskSse2(a, b, len)
elif allowNeon:
blendLineMaskNeon(a, b, len)
else:
for i in 0 ..< len:
a[i] = blendMask(a[i], b[i])

proc blendRect(a, b: Image, pos: Ivec2, blendMode: BlendMode) =
let
Expand Down
Loading
Loading