group float timings (to easily disable, they are going away)

src/pixie/blends.nim

@@ -153,7 +153,7 @@ when defined(amd64) and not defined(pixieNoSimd):
 
   type BlenderSimd* = proc(blackdrop, source: M128i): M128i
 
-  proc blendNormalPremultipliedSimd*(backdrop, source: M128i): M128i =
+  proc blendNormalSimd*(backdrop, source: M128i): M128i =
     let
       alphaMask = mm_set1_epi32(cast[int32](0xff000000))
       oddMask = mm_set1_epi16(cast[int16](0xff00))
@@ -187,7 +187,7 @@ when defined(amd64) and not defined(pixieNoSimd):
 
   proc blenderSimd*(blendMode: BlendMode): BlenderSimd =
     case blendMode:
-    of bmNormal: blendNormalPremultipliedSimd
+    of bmNormal: blendNormalSimd
     of bmOverwrite: blendOverwriteSimd
     else:
       raise newException(PixieError, "No SIMD blender for " & $blendMode)

tests/benchmark_blends.nim

@@ -18,20 +18,144 @@ timeIt "blendNormal":
 
 reset()
 
-timeIt "blendNormalFloats":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendNormalFloats(
-      backdrop.data[i].color, source.data[i].color
-    ).rgba
-
-reset()
-
 timeIt "blendDarken":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendDarken(backdrop.data[i], source.data[i])
 
 reset()
 
+timeIt "blendMultiply":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendMultiply(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendLinearBurn":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendLinearBurn(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendColorBurn":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendColorBurn(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendLighten":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendLighten(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendScreen":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendScreen(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendLinearDodge":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendLinearDodge(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendColorDodge":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendColorDodge(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendOverlay":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendOverlay(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendSoftLight":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendSoftLight(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendHardLight":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendHardLight(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendDifference":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendDifference(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendExclusion":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendExclusion(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendHue":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendHue(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendSaturation":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendSaturation(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendColor":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendColor(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendLuminosity":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendLuminosity(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendMask":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendMask(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendSubtractMask":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendSubtractMask(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendIntersectMask":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendIntersectMask(backdrop.data[i], source.data[i])
+
+reset()
+
+timeIt "blendExcludeMask":
+  for i in 0 ..< backdrop.data.len:
+    backdrop.data[i] = blendExcludeMask(backdrop.data[i], source.data[i])
+
+when defined(amd64) and not defined(pixieNoSimd):
+  import nimsimd/sse2
+
+  reset()
+
+  timeIt "blendNormal [simd]":
+    for i in countup(0, backdrop.data.len - 4, 4):
+      let
+        b = mm_loadu_si128(backdrop.data[i].addr)
+        s = mm_loadu_si128(source.data[i].addr)
+      mm_storeu_si128(backdrop.data[i].addr, blendNormalSimd(b, s))
+
+reset()
+
 timeIt "blendDarkenFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendDarkenFloats(
@@ -40,9 +164,11 @@ timeIt "blendDarkenFloats":
 
 reset()
 
-timeIt "blendMultiply":
+timeIt "blendNormalFloats":
   for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendMultiply(backdrop.data[i], source.data[i])
+    backdrop.data[i] = blendNormalFloats(
+      backdrop.data[i].color, source.data[i].color
+    ).rgba
 
 reset()
 
@@ -54,12 +180,6 @@ timeIt "blendMultiplyFloats":
 
 reset()
 
-timeIt "blendLinearBurn":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendLinearBurn(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendLinearBurnFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendLinearBurnFloats(
@@ -68,12 +188,6 @@ timeIt "blendLinearBurnFloats":
 
 reset()
 
-timeIt "blendColorBurn":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendColorBurn(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendColorBurnFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendColorBurnFloats(
@@ -82,12 +196,6 @@ timeIt "blendColorBurnFloats":
 
 reset()
 
-timeIt "blendLighten":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendLighten(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendLightenFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendLightenFloats(
@@ -96,12 +204,6 @@ timeIt "blendLightenFloats":
 
 reset()
 
-timeIt "blendScreen":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendScreen(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendScreenFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendScreenFloats(
@@ -110,12 +212,6 @@ timeIt "blendScreenFloats":
 
 reset()
 
-timeIt "blendLinearDodge":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendLinearDodge(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendLinearDodgeFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendLinearDodgeFloats(
@@ -124,12 +220,6 @@ timeIt "blendLinearDodgeFloats":
 
 reset()
 
-timeIt "blendColorDodge":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendColorDodge(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendColorDodgeFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendColorDodgeFloats(
@@ -138,12 +228,6 @@ timeIt "blendColorDodgeFloats":
 
 reset()
 
-timeIt "blendOverlay":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendOverlay(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendOverlayFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendOverlayFloats(
@@ -152,12 +236,6 @@ timeIt "blendOverlayFloats":
 
 reset()
 
-timeIt "blendSoftLight":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendSoftLight(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendSoftLightFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendSoftLightFloats(
@@ -166,12 +244,6 @@ timeIt "blendSoftLightFloats":
 
 reset()
 
-timeIt "blendHardLight":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendHardLight(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendHardLightFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendHardLightFloats(
@@ -180,12 +252,6 @@ timeIt "blendHardLightFloats":
 
 reset()
 
-timeIt "blendDifference":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendDifference(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendDifferenceFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendDifferenceFloats(
@@ -194,12 +260,6 @@ timeIt "blendDifferenceFloats":
 
 reset()
 
-timeIt "blendExclusion":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendExclusion(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendExclusionFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendExclusionFloats(
@@ -208,12 +268,6 @@ timeIt "blendExclusionFloats":
 
 reset()
 
-timeIt "blendHue":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendHue(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendHueFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendHueFloats(
@@ -222,12 +276,6 @@ timeIt "blendHueFloats":
 
 reset()
 
-timeIt "blendSaturation":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendSaturation(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendSaturationFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendSaturationFloats(
@@ -236,12 +284,6 @@ timeIt "blendSaturationFloats":
 
 reset()
 
-timeIt "blendColor":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendColor(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendColorFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendColorFloats(
@@ -250,12 +292,6 @@ timeIt "blendColorFloats":
 
 reset()
 
-timeIt "blendLuminosity":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendLuminosity(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendLuminosityFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendLuminosityFloats(
@@ -264,12 +300,6 @@ timeIt "blendLuminosityFloats":
 
 reset()
 
-timeIt "blendMask":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendMask(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendMaskFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendMaskFloats(
@@ -278,12 +308,6 @@ timeIt "blendMaskFloats":
 
 reset()
 
-timeIt "blendSubtractMask":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendSubtractMask(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendSubtractMaskFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendSubtractMaskFloats(
@@ -292,12 +316,6 @@ timeIt "blendSubtractMaskFloats":
 
 reset()
 
-timeIt "blendIntersectMask":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendIntersectMask(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendIntersectMaskFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendIntersectMaskFloats(
@@ -306,30 +324,8 @@ timeIt "blendIntersectMaskFloats":
 
 reset()
 
-timeIt "blendExcludeMask":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendExcludeMask(backdrop.data[i], source.data[i])
-
-reset()
-
 timeIt "blendExcludeMaskFloats":
   for i in 0 ..< backdrop.data.len:
     backdrop.data[i] = blendExcludeMaskFloats(
       backdrop.data[i].color, source.data[i].color
     ).rgba
-
-reset()
-
-timeIt "blendNormalPremultiplied":
-  for i in 0 ..< backdrop.data.len:
-    backdrop.data[i] = blendNormalPremultiplied(backdrop.data[i], source.data[i])
-
-when defined(amd64) and not defined(pixieNoSimd):
-  import nimsimd/sse2
-
-  timeIt "blendNormalPremultiplied [simd]":
-    for i in countup(0, backdrop.data.len - 4, 4):
-      let
-        b = mm_loadu_si128(backdrop.data[i].addr)
-        s = mm_loadu_si128(source.data[i].addr)
-      mm_storeu_si128(backdrop.data[i].addr, blendNormalPremultiplied(b, s))