Alpha premul pipeline cleanup

This assumptions are now made:
- Internally float buffers are always linear alpha-premul colors
- Readers should worry about delivering float buffers with that
  assumptions.
- There's an input image setting to say whether it's stored with
  straight/premul alpha on the disk.
- Byte buffers are now assumed have straight alpha, readers should
  deliver straight alpha.

Some implementation details:

- Removed scene's color unpremultiply setting, which was very
  much confusing and was wrong for default settings.
  Now all renderers assumes to deliver premultiplied alpha.

- IMB_buffer_byte_from_float will now linearize alpha when
  converting from buffer.

- Sequencer's effects were changed to assume bytes have got
  straight alpha. Most of effects will work with bytes still,
  however for glow it was more tricky to avoid data loss, so
  there's a commented out glow implementation which converts
  byte buffer to floats first, operates on floats and returns
  bytes back. It's slower and not sure if it should actually
  be used -- who're using glow on alpha anyway?

- Sequencer modifiers should also be working nice with straight
  bytes now.

- GLSL preview will predivide float textures to make nice shading,
  shading with byte textures worked nice (GLSL was assuming straight
  alpha).

- Blender Internal will set alpha=1 to the whole sky. The same
  happens in Cycles and there's no way to avoid this -- sky is
  neither straight nor premul and doesn't fit color pipeline well.

- Straight alpha mode for render result was also eliminated.

- Conversion to correct alpha need to be done before linearizing
  float buffer.

- TIFF will now load and save files with proper alpha mode setting
  in file meta data header.

- Remove Use Alpha from texture mapping and replaced with image
  datablock setting.

  Behaves much more predictable and clear from code point of view
  and solves possible regressions when non-premultiplied images were
  used as textures with ignoring alpha channel.

1 files changed, 210 insertions, 160 deletions

diff --git a/source/blender/blenkernel/intern/seqeffects.c b/source/blender/blenkernel/intern/seqeffects.c
index 3bff209f53c..7dbbca6ddaa 100644
--- a/source/blender/blenkernel/intern/seqeffects.c
+++ b/source/blender/blenkernel/intern/seqeffects.c
@@ -156,42 +156,43 @@ static void init_alpha_over_or_under(Sequence *seq)
 	seq->seq1 = seq2;
 }
 
-static void do_alphaover_effect_byte(float facf0, float facf1, int x, int y,  char *rect1, char *rect2, char *out)
+static void do_alphaover_effect_byte(float facf0, float facf1, int x, int y,  unsigned char *rect1, unsigned char *rect2, unsigned char *out)
 {
-	int fac2, mfac, fac, fac4;
-	int xo, tempc;
-	char *rt1, *rt2, *rt;
+	float fac2, mfac, fac, fac4;
+	int xo;
+	unsigned char *cp1, *cp2, *rt;
+	float tempc[4], rt1[4], rt2[4];
 
 	xo = x;
-	rt1 = (char *) rect1;
-	rt2 = (char *) rect2;
-	rt = (char *) out;
+	cp1 = rect1;
+	cp2 = rect2;
+	rt = out;
 
-	fac2 = (int) (256.0f * facf0);
-	fac4 = (int) (256.0f * facf1);
+	fac2 = facf0;
+	fac4 = facf1;
 
 	while (y--) {
 		x = xo;
 		while (x--) {
-
 			/* rt = rt1 over rt2  (alpha from rt1) */
 
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
+
 			fac = fac2;
-			mfac = 256 - ( (fac2 * rt1[3]) >> 8);
+			mfac = 1.0f - fac2 * rt1[3];
 
-			if (fac == 0) *( (unsigned int *) rt) = *( (unsigned int *) rt2);
-			else if (mfac == 0) *( (unsigned int *) rt) = *( (unsigned int *) rt1);
+			if (fac <= 0.0f) *( (unsigned int *) rt) = *( (unsigned int *) cp2);
+			else if (mfac <= 0.0f) *( (unsigned int *) rt) = *( (unsigned int *) cp1);
 			else {
-				tempc = (fac * rt1[0] + mfac * rt2[0]) >> 8;
-				if (tempc > 255) rt[0] = 255; else rt[0] = tempc;
-				tempc = (fac * rt1[1] + mfac * rt2[1]) >> 8;
-				if (tempc > 255) rt[1] = 255; else rt[1] = tempc;
-				tempc = (fac * rt1[2] + mfac * rt2[2]) >> 8;
-				if (tempc > 255) rt[2] = 255; else rt[2] = tempc;
-				tempc = (fac * rt1[3] + mfac * rt2[3]) >> 8;
-				if (tempc > 255) rt[3] = 255; else rt[3] = tempc;
+				tempc[0] = fac * rt1[0] + mfac * rt2[0];
+				tempc[1] = fac * rt1[1] + mfac * rt2[1];
+				tempc[2] = fac * rt1[2] + mfac * rt2[2];
+				tempc[3] = fac * rt1[3] + mfac * rt2[3];
+
+				premul_float_to_straight_uchar(rt, tempc);
 			}
-			rt1 += 4; rt2 += 4; rt += 4;
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 
 		if (y == 0) break;
@@ -199,22 +200,23 @@ static void do_alphaover_effect_byte(float facf0, float facf1, int x, int y,  ch
 
 		x = xo;
 		while (x--) {
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
+
 			fac = fac4;
-			mfac = 256 - ( (fac4 * rt1[3]) >> 8);
+			mfac = 1.0f - (fac4 * rt1[3]);
 
-			if (fac == 0) *( (unsigned int *) rt) = *( (unsigned int *) rt2);
-			else if (mfac == 0) *( (unsigned int *) rt) = *( (unsigned int *) rt1);
+			if (fac <= 0.0f) *( (unsigned int *) rt) = *( (unsigned int *) cp2);
+			else if (mfac <= 0.0f) *( (unsigned int *) rt) = *( (unsigned int *) cp1);
 			else {
-				tempc = (fac * rt1[0] + mfac * rt2[0]) >> 8;
-				if (tempc > 255) rt[0] = 255; else rt[0] = tempc;
-				tempc = (fac * rt1[1] + mfac * rt2[1]) >> 8;
-				if (tempc > 255) rt[1] = 255; else rt[1] = tempc;
-				tempc = (fac * rt1[2] + mfac * rt2[2]) >> 8;
-				if (tempc > 255) rt[2] = 255; else rt[2] = tempc;
-				tempc = (fac * rt1[3] + mfac * rt2[3]) >> 8;
-				if (tempc > 255) rt[3] = 255; else rt[3] = tempc;
+				tempc[0] = fac * rt1[0] + mfac * rt2[0];
+				tempc[1] = fac * rt1[1] + mfac * rt2[1];
+				tempc[2] = fac * rt1[2] + mfac * rt2[2];
+				tempc[3] = fac * rt1[3] + mfac * rt2[3];
+
+				premul_float_to_straight_uchar(rt, tempc);
 			}
-			rt1 += 4; rt2 += 4; rt += 4;
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 	}
 }
@@ -298,17 +300,17 @@ static void do_alphaover_effect(SeqRenderData context, Sequence *UNUSED(seq), fl
 
 		slice_get_byte_buffers(&context, ibuf1, ibuf2, NULL, out, start_line, &rect1, &rect2, NULL, &rect_out);
 
-		do_alphaover_effect_byte(facf0, facf1, context.rectx, total_lines, (char *) rect1, (char *) rect2, (char *) rect_out);
+		do_alphaover_effect_byte(facf0, facf1, context.rectx, total_lines, rect1, rect2, rect_out);
 	}
 }
 
 /*********************** Alpha Under *************************/
 
-static void do_alphaunder_effect_byte(float facf0, float facf1, int x, int y, char *rect1, char *rect2, char *out)
+static void do_alphaunder_effect_byte(float facf0, float facf1, int x, int y, unsigned char *rect1, unsigned char *rect2, unsigned char *out)
 {
 	int fac2, mfac, fac, fac4;
 	int xo;
-	char *rt1, *rt2, *rt;
+	unsigned char *rt1, *rt2, *rt;
 
 	xo = x;
 	rt1 = rect1;
@@ -460,17 +462,17 @@ static void do_alphaunder_effect(SeqRenderData context, Sequence *UNUSED(seq), f
 
 		slice_get_byte_buffers(&context, ibuf1, ibuf2, NULL, out, start_line, &rect1, &rect2, NULL, &rect_out);
 
-		do_alphaunder_effect_byte(facf0, facf1, context.rectx, total_lines, (char *) rect1, (char *) rect2, (char *) rect_out);
+		do_alphaunder_effect_byte(facf0, facf1, context.rectx, total_lines, rect1, rect2, rect_out);
 	}
 }
 
 /*********************** Cross *************************/
 
-static void do_cross_effect_byte(float facf0, float facf1, int x, int y, char *rect1, char *rect2, char *out)
+static void do_cross_effect_byte(float facf0, float facf1, int x, int y, unsigned char *rect1, unsigned char *rect2, unsigned char *out)
 {
 	int fac1, fac2, fac3, fac4;
 	int xo;
-	char *rt1, *rt2, *rt;
+	unsigned char *rt1, *rt2, *rt;
 
 	xo = x;
 	rt1 = rect1;
@@ -570,7 +572,7 @@ static void do_cross_effect(SeqRenderData context, Sequence *UNUSED(seq), float
 
 		slice_get_byte_buffers(&context, ibuf1, ibuf2, NULL, out, start_line, &rect1, &rect2, NULL, &rect_out);
 
-		do_cross_effect_byte(facf0, facf1, context.rectx, total_lines, (char *) rect1, (char *) rect2, (char *) rect_out);
+		do_cross_effect_byte(facf0, facf1, context.rectx, total_lines, rect1, rect2, rect_out);
 	}
 }
 
@@ -713,31 +715,32 @@ static void free_gammacross(Sequence *UNUSED(seq))
 static void do_gammacross_effect_byte(float facf0, float UNUSED(facf1),  int x, int y, unsigned char *rect1,
                                       unsigned char *rect2, unsigned char *out)
 {
-	int fac1, fac2, col;
+	float fac1, fac2;
 	int xo;
-	unsigned char *rt1, *rt2, *rt;
-	
+	unsigned char *cp1, *cp2, *rt;
+	float rt1[4], rt2[4], tempc[4];
+
 	xo = x;
-	rt1 = (unsigned char *) rect1;
-	rt2 = (unsigned char *) rect2;
-	rt = (unsigned char *) out;
+	cp1 = rect1;
+	cp2 = rect2;
+	rt = out;
 
-	fac2 = (int)(256.0f * facf0);
-	fac1 = 256 - fac2;
+	fac2 = facf0;
+	fac1 = 1.0f - fac2;
 
 	while (y--) {
 		x = xo;
 		while (x--) {
-			col = (fac1 * igamtab1[rt1[0]] + fac2 * igamtab1[rt2[0]]) >> 8;
-			if (col > 65535) rt[0] = 255; else rt[0] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[1]] + fac2 * igamtab1[rt2[1]]) >> 8;
-			if (col > 65535) rt[1] = 255; else rt[1] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[2]] + fac2 * igamtab1[rt2[2]]) >> 8;
-			if (col > 65535) rt[2] = 255; else rt[2] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[3]] + fac2 * igamtab1[rt2[3]]) >> 8;
-			if (col > 65535) rt[3] = 255; else rt[3] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = gammaCorrect(fac1 * invGammaCorrect(rt1[0]) + fac2 * invGammaCorrect(rt2[0]));
+			tempc[1] = gammaCorrect(fac1 * invGammaCorrect(rt1[1]) + fac2 * invGammaCorrect(rt2[1]));
+			tempc[2] = gammaCorrect(fac1 * invGammaCorrect(rt1[2]) + fac2 * invGammaCorrect(rt2[2]));
+			tempc[3] = gammaCorrect(fac1 * invGammaCorrect(rt1[3]) + fac2 * invGammaCorrect(rt2[3]));
+
+			premul_float_to_straight_uchar(rt, tempc);
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 
 		if (y == 0)
@@ -746,16 +749,16 @@ static void do_gammacross_effect_byte(float facf0, float UNUSED(facf1),  int x,
 
 		x = xo;
 		while (x--) {
-			col = (fac1 * igamtab1[rt1[0]] + fac2 * igamtab1[rt2[0]]) >> 8;
-			if (col > 65535) rt[0] = 255; else rt[0] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[1]] + fac2 * igamtab1[rt2[1]]) >> 8;
-			if (col > 65535) rt[1] = 255; else rt[1] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[2]] + fac2 * igamtab1[rt2[2]]) >> 8;
-			if (col > 65535) rt[2] = 255; else rt[2] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
-			col = (fac1 * igamtab1[rt1[3]] + fac2 * igamtab1[rt2[3]]) >> 8;
-			if (col > 65535) rt[3] = 255; else rt[3] = ( (char *)(gamtab + col))[MOST_SIG_BYTE];
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = gammaCorrect(fac1 * invGammaCorrect(rt1[0]) + fac2 * invGammaCorrect(rt2[0]));
+			tempc[1] = gammaCorrect(fac1 * invGammaCorrect(rt1[1]) + fac2 * invGammaCorrect(rt2[1]));
+			tempc[2] = gammaCorrect(fac1 * invGammaCorrect(rt1[2]) + fac2 * invGammaCorrect(rt2[2]));
+			tempc[3] = gammaCorrect(fac1 * invGammaCorrect(rt1[3]) + fac2 * invGammaCorrect(rt2[3]));
+
+			premul_float_to_straight_uchar(rt, tempc);
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 	}
 }
@@ -828,31 +831,34 @@ static void do_gammacross_effect(SeqRenderData context, Sequence *UNUSED(seq), f
 static void do_add_effect_byte(float facf0, float facf1, int x, int y, unsigned char *rect1, unsigned char *rect2,
                                unsigned char *out)
 {
-	int col, xo, fac1, fac3;
-	char *rt1, *rt2, *rt;
+	int xo;
+	unsigned char *cp1, *cp2, *rt;
+	float fac1, fac3;
+	float tempc[4], rt1[4], rt2[4];
 
 	xo = x;
-	rt1 = (char *)rect1;
-	rt2 = (char *)rect2;
-	rt = (char *)out;
+	cp1 = rect1;
+	cp2 = rect2;
+	rt = out;
 
-	fac1 = (int)(256.0f * facf0);
-	fac3 = (int)(256.0f * facf1);
+	fac1 = facf0;
+	fac3 = facf1;
 
 	while (y--) {
 		x = xo;
 
 		while (x--) {
-			col = rt1[0] + ((fac1 * rt2[0]) >> 8);
-			if (col > 255) rt[0] = 255; else rt[0] = col;
-			col = rt1[1] + ((fac1 * rt2[1]) >> 8);
-			if (col > 255) rt[1] = 255; else rt[1] = col;
-			col = rt1[2] + ((fac1 * rt2[2]) >> 8);
-			if (col > 255) rt[2] = 255; else rt[2] = col;
-			col = rt1[3] + ((fac1 * rt2[3]) >> 8);
-			if (col > 255) rt[3] = 255; else rt[3] = col;
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = rt1[0] + fac1 * rt2[0];
+			tempc[1] = rt1[1] + fac1 * rt2[1];
+			tempc[2] = rt1[2] + fac1 * rt2[2];
+			tempc[3] = min_ff(1.0f, rt1[3] + fac1 * rt2[3]);
+
+			premul_float_to_straight_uchar(rt, tempc);
+
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 
 		if (y == 0)
@@ -861,16 +867,17 @@ static void do_add_effect_byte(float facf0, float facf1, int x, int y, unsigned
 
 		x = xo;
 		while (x--) {
-			col = rt1[0] + ((fac3 * rt2[0]) >> 8);
-			if (col > 255) rt[0] = 255; else rt[0] = col;
-			col = rt1[1] + ((fac3 * rt2[1]) >> 8);
-			if (col > 255) rt[1] = 255; else rt[1] = col;
-			col = rt1[2] + ((fac3 * rt2[2]) >> 8);
-			if (col > 255) rt[2] = 255; else rt[2] = col;
-			col = rt1[3] + ((fac3 * rt2[3]) >> 8);
-			if (col > 255) rt[3] = 255; else rt[3] = col;
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = rt1[0] + fac3 * rt2[0];
+			tempc[1] = rt1[1] + fac3 * rt2[1];
+			tempc[2] = rt1[2] + fac3 * rt2[2];
+			tempc[3] = min_ff(1.0f, rt1[3] + fac3 * rt2[3]);
+
+			premul_float_to_straight_uchar(rt, tempc);
+
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 	}
 }
@@ -890,22 +897,28 @@ static void do_add_effect_float(float facf0, float facf1, int x, int y, float *r
 	fac3 = facf1;
 
 	while (y--) {
-		x = xo * 4;
+		x = xo;
 		while (x--) {
-			*rt = *rt1 + fac1 * (*rt2);
+			rt[0] = rt1[0] + fac1 * rt2[0];
+			rt[1] = rt1[1] + fac1 * rt2[1];
+			rt[2] = rt1[2] + fac1 * rt2[2];
+			rt[3] = min_ff(1.0f, rt1[3] + fac1 * rt2[3]);
 
-			rt1++; rt2++; rt++;
+			rt1 += 4; rt2 += 4; rt += 4;
 		}
 
 		if (y == 0)
 			break;
 		y--;
 
-		x = xo * 4;
+		x = xo;
 		while (x--) {
-			*rt = *rt1 + fac3 * (*rt2);
+			rt[0] = rt1[0] + fac1 * rt2[0];
+			rt[1] = rt1[1] + fac1 * rt2[1];
+			rt[2] = rt1[2] + fac1 * rt2[2];
+			rt[3] = min_ff(1.0f, rt1[3] + fac3 * rt2[3]);
 
-			rt1++; rt2++; rt++;
+			rt1 += 4; rt2 += 4; rt += 4;
 		}
 	}
 }
@@ -931,32 +944,35 @@ static void do_add_effect(SeqRenderData context, Sequence *UNUSED(seq), float UN
 
 /*********************** Sub *************************/
 
-static void do_sub_effect_byte(float facf0, float facf1, int x, int y, char *rect1, char *rect2, char *out)
+static void do_sub_effect_byte(float facf0, float facf1, int x, int y, unsigned char *rect1, unsigned char *rect2, unsigned char *out)
 {
-	int col, xo, fac1, fac3;
-	char *rt1, *rt2, *rt;
+	int xo;
+	unsigned char *cp1, *cp2, *rt;
+	float fac1, fac3;
+	float tempc[4], rt1[4], rt2[4];
 
 	xo = x;
-	rt1 = (char *) rect1;
-	rt2 = (char *) rect2;
-	rt = (char *) out;
+	cp1 = rect1;
+	cp2 = rect2;
+	rt = out;
 
-	fac1 = (int) (256.0f * facf0);
-	fac3 = (int) (256.0f * facf1);
+	fac1 = facf0;
+	fac3 = facf1;
 
 	while (y--) {
 		x = xo;
 		while (x--) {
-			col = rt1[0] - ((fac1 * rt2[0]) >> 8);
-			if (col < 0) rt[0] = 0; else rt[0] = col;
-			col = rt1[1] - ((fac1 * rt2[1]) >> 8);
-			if (col < 0) rt[1] = 0; else rt[1] = col;
-			col = rt1[2] - ((fac1 * rt2[2]) >> 8);
-			if (col < 0) rt[2] = 0; else rt[2] = col;
-			col = rt1[3] - ((fac1 * rt2[3]) >> 8);
-			if (col < 0) rt[3] = 0; else rt[3] = col;
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = rt1[0] - fac1 * rt2[0];
+			tempc[1] = rt1[1] - fac1 * rt2[1];
+			tempc[2] = rt1[2] - fac1 * rt2[2];
+			tempc[3] = rt1[3] - fac1 * rt2[3];
+
+			premul_float_to_straight_uchar(rt, tempc);
+
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 
 		if (y == 0)
@@ -965,16 +981,17 @@ static void do_sub_effect_byte(float facf0, float facf1, int x, int y, char *rec
 
 		x = xo;
 		while (x--) {
-			col = rt1[0] - ((fac3 * rt2[0]) >> 8);
-			if (col < 0) rt[0] = 0; else rt[0] = col;
-			col = rt1[1] - ((fac3 * rt2[1]) >> 8);
-			if (col < 0) rt[1] = 0; else rt[1] = col;
-			col = rt1[2] - ((fac3 * rt2[2]) >> 8);
-			if (col < 0) rt[2] = 0; else rt[2] = col;
-			col = rt1[3] - ((fac3 * rt2[3]) >> 8);
-			if (col < 0) rt[3] = 0; else rt[3] = col;
+			straight_uchar_to_premul_float(rt1, cp1);
+			straight_uchar_to_premul_float(rt2, cp2);
 
-			rt1 += 4; rt2 += 4; rt += 4;
+			tempc[0] = rt1[0] - fac3 * rt2[0];
+			tempc[1] = rt1[1] - fac3 * rt2[1];
+			tempc[2] = rt1[2] - fac3 * rt2[2];
+			tempc[3] = rt1[3] - fac3 * rt2[3];
+
+			premul_float_to_straight_uchar(rt, tempc);
+
+			cp1 += 4; cp2 += 4; rt += 4;
 		}
 	}
 }
@@ -1029,7 +1046,7 @@ static void do_sub_effect(SeqRenderData context, Sequence *UNUSED(seq), float UN
 
 		slice_get_byte_buffers(&context, ibuf1, ibuf2, NULL, out, start_line, &rect1, &rect2, NULL, &rect_out);
 
-		do_sub_effect_byte(facf0, facf1, context.rectx, total_lines, (char *) rect1, (char *) rect2, (char *) rect_out);
+		do_sub_effect_byte(facf0, facf1, context.rectx, total_lines, rect1, rect2, rect_out);
 	}
 }
 
@@ -1039,10 +1056,10 @@ static void do_sub_effect(SeqRenderData context, Sequence *UNUSED(seq), float UN
 #define XOFF    8
 #define YOFF    8
 
-static void do_drop_effect_byte(float facf0, float facf1, int x, int y, char *rect2i, char *rect1i, char *outi)
+static void do_drop_effect_byte(float facf0, float facf1, int x, int y, unsigned char *rect2i, unsigned char *rect1i, unsigned char *outi)
 {
 	int height, width, temp, fac, fac1, fac2;
-	char *rt1, *rt2, *out;
+	unsigned char *rt1, *rt2, *out;
 	int field = 1;
 
 	width = x;
@@ -1051,9 +1068,9 @@ static void do_drop_effect_byte(float facf0, float facf1, int x, int y, char *re
 	fac1 = (int) (70.0f * facf0);
 	fac2 = (int) (70.0f * facf1);
 
-	rt2 = (char *) (rect2i + YOFF * width);
-	rt1 = (char *) rect1i;
-	out = (char *) outi;
+	rt2 = (unsigned char *) (rect2i + YOFF * width);
+	rt1 = (unsigned char *) rect1i;
+	out = (unsigned char *) outi;
 	for (y = 0; y < height - YOFF; y++) {
 		if (field) fac = fac1;
 		else fac = fac2;
@@ -1122,12 +1139,12 @@ static void do_mul_effect_byte(float facf0, float facf1, int x, int y, unsigned
                                unsigned char *out)
 {
 	int xo, fac1, fac3;
-	char *rt1, *rt2, *rt;
+	unsigned char *rt1, *rt2, *rt;
 
 	xo = x;
-	rt1 = (char *)rect1;
-	rt2 = (char *)rect2;
-	rt = (char *)out;
+	rt1 = rect1;
+	rt2 = rect2;
+	rt = out;
 
 	fac1 = (int)(256.0f * facf0);
 	fac3 = (int)(256.0f * facf1);
@@ -1539,13 +1556,13 @@ static void do_wipe_effect_byte(Sequence *seq, float facf0, float UNUSED(facf1),
 	WipeZone wipezone;
 	WipeVars *wipe = (WipeVars *)seq->effectdata;
 	int xo, yo;
-	char *rt1, *rt2, *rt;
+	unsigned char *cp1, *cp2, *rt;
 
 	precalc_wipe_zone(&wipezone, wipe, x, y);
 
-	rt1 = (char *)rect1;
-	rt2 = (char *)rect2;
-	rt = (char *)out;
+	cp1 = rect1;
+	cp2 = rect2;
+	rt = out;
 
 	xo = x;
 	yo = y;
@@ -1553,11 +1570,18 @@ static void do_wipe_effect_byte(Sequence *seq, float facf0, float UNUSED(facf1),
 		for (x = 0; x < xo; x++) {
 			float check = check_zone(&wipezone, x, y, seq, facf0);
 			if (check) {
-				if (rt1) {
-					rt[0] = (int)(rt1[0] * check) + (int)(rt2[0] * (1 - check));
-					rt[1] = (int)(rt1[1] * check) + (int)(rt2[1] * (1 - check));
-					rt[2] = (int)(rt1[2] * check) + (int)(rt2[2] * (1 - check));
-					rt[3] = (int)(rt1[3] * check) + (int)(rt2[3] * (1 - check));
+				if (cp1) {
+					float rt1[4], rt2[4], tempc[4];
+
+					straight_uchar_to_premul_float(rt1, cp1);
+					straight_uchar_to_premul_float(rt2, cp2);
+
+					tempc[0] = rt1[0] * check + rt2[0] * (1 - check);
+					tempc[1] = rt1[1] * check + rt2[1] * (1 - check);
+					tempc[2] = rt1[2] * check + rt2[2] * (1 - check);
+					tempc[3] = rt1[3] * check + rt2[3] * (1 - check);
+
+					premul_float_to_straight_uchar(rt, tempc);
 				}
 				else {
 					rt[0] = 0;
@@ -1567,11 +1591,11 @@ static void do_wipe_effect_byte(Sequence *seq, float facf0, float UNUSED(facf1),
 				}
 			}
 			else {
-				if (rt2) {
-					rt[0] = rt2[0];
-					rt[1] = rt2[1];
-					rt[2] = rt2[2];
-					rt[3] = rt2[3];
+				if (cp2) {
+					rt[0] = cp2[0];
+					rt[1] = cp2[1];
+					rt[2] = cp2[2];
+					rt[3] = cp2[3];
 				}
 				else {
 					rt[0] = 0;
@@ -1582,11 +1606,11 @@ static void do_wipe_effect_byte(Sequence *seq, float facf0, float UNUSED(facf1),
 			}
 
 			rt += 4;
-			if (rt1 != NULL) {
-				rt1 += 4;
+			if (cp1 != NULL) {
+				cp1 += 4;
 			}
-			if (rt2 != NULL) {
-				rt2 += 4;
+			if (cp2 != NULL) {
+				cp2 += 4;
 			}
 		}
 	}
@@ -2254,16 +2278,42 @@ static void copy_glow_effect(Sequence *dst, Sequence *src)
 }
 
 static void do_glow_effect_byte(Sequence *seq, int render_size, float facf0, float UNUSED(facf1),  int x, int y,
-                                char *rect1, char *UNUSED(rect2), char *out)
+                                unsigned char *rect1, unsigned char *UNUSED(rect2), unsigned char *out)
 {
-	unsigned char *outbuf = (unsigned char *)out;
-	unsigned char *inbuf = (unsigned char *)rect1;
+#if 0
+	/* XXX: not sure what's better here, on the one hand conversion to floats
+	 *      here is not so much trouble, but on the other hand who're using
+	 *      glow on buffers with alpha?
+	 */
+	unsigned char *outbuf = out;
+	unsigned char *inbuf = rect1;
 	GlowVars *glow = (GlowVars *)seq->effectdata;
-	
+
 	RVIsolateHighlights_byte(inbuf, outbuf, x, y, glow->fMini * 765, glow->fBoost * facf0, glow->fClamp);
 	RVBlurBitmap2_byte(outbuf, x, y, glow->dDist * (render_size / 100.0f), glow->dQuality);
 	if (!glow->bNoComp)
 		RVAddBitmaps_byte(inbuf, outbuf, outbuf, x, y);
+#else
+	float *outbuf, *inbuf;
+	GlowVars *glow = (GlowVars *)seq->effectdata;
+
+	inbuf = MEM_mallocN(4 * sizeof(float) * x * y, "glow effect input");
+	outbuf = MEM_mallocN(4 * sizeof(float) * x * y, "glow effect output");
+
+	IMB_buffer_float_from_byte(inbuf, rect1, IB_PROFILE_SRGB, IB_PROFILE_SRGB, FALSE, x, y, x, x);
+	IMB_buffer_float_premultiply(inbuf, x, y);
+
+	RVIsolateHighlights_float(inbuf, outbuf, x, y, glow->fMini * 3.0f, glow->fBoost * facf0, glow->fClamp);
+	RVBlurBitmap2_float(outbuf, x, y, glow->dDist * (render_size / 100.0f), glow->dQuality);
+	if (!glow->bNoComp)
+		RVAddBitmaps_float(inbuf, outbuf, outbuf, x, y);
+
+	IMB_buffer_float_unpremultiply(outbuf, x, y);
+	IMB_buffer_byte_from_float(out, outbuf, 4, 0.0f, IB_PROFILE_SRGB, IB_PROFILE_SRGB, FALSE, x, y, x, x);
+
+	MEM_freeN(inbuf);
+	MEM_freeN(outbuf);
+#endif
 }
 
 static void do_glow_effect_float(Sequence *seq, int render_size, float facf0, float UNUSED(facf1),  int x, int y,
@@ -2292,7 +2342,7 @@ static ImBuf *do_glow_effect(SeqRenderData context, Sequence *seq, float UNUSED(
 	}
 	else {
 		do_glow_effect_byte(seq, render_size, facf0, facf1, context.rectx, context.recty,
-		                    (char *) ibuf1->rect, (char *) ibuf2->rect, (char *) out->rect);
+		                    (unsigned char *) ibuf1->rect, (unsigned char *) ibuf2->rect, (unsigned char *) out->rect);
 	}
 
 	return out;
@@ -2735,7 +2785,7 @@ static ImBuf *do_speed_effect(SeqRenderData context, Sequence *UNUSED(seq), floa
 	}
 	else {
 		do_cross_effect_byte(facf0, facf1, context.rectx, context.recty,
-		                     (char *) ibuf1->rect, (char *) ibuf2->rect, (char *) out->rect);
+		                     (unsigned char *) ibuf1->rect, (unsigned char *) ibuf2->rect, (unsigned char *) out->rect);
 	}
 	return out;
 }
@@ -2761,8 +2811,8 @@ static void do_overdrop_effect(SeqRenderData context, Sequence *UNUSED(seq), flo
 
 		slice_get_byte_buffers(&context, ibuf1, ibuf2, NULL, out, start_line, &rect1, &rect2, NULL, &rect_out);
 
-		do_drop_effect_byte(facf0, facf1, x, y, (char *) rect1, (char *) rect2, (char *) rect_out);
-		do_alphaover_effect_byte(facf0, facf1, x, y, (char *) rect1, (char *) rect2, (char *) rect_out);
+		do_drop_effect_byte(facf0, facf1, x, y, rect1, rect2, rect_out);
+		do_alphaover_effect_byte(facf0, facf1, x, y, rect1, rect2, rect_out);
 	}
 }