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.

6 files changed, 325 insertions, 255 deletions

diff --git a/source/blender/blenkernel/BKE_blender.h b/source/blender/blenkernel/BKE_blender.h
index ecd38403d72..10528f1b270 100644
--- a/source/blender/blenkernel/BKE_blender.h
+++ b/source/blender/blenkernel/BKE_blender.h
@@ -42,7 +42,7 @@ extern "C" {
  * and keep comment above the defines.
  * Use STRINGIFY() rather than defining with quotes */
 #define BLENDER_VERSION         265
-#define BLENDER_SUBVERSION      4
+#define BLENDER_SUBVERSION      5
 
 /* 262 was the last editmesh release but it has compatibility code for bmesh data */
 #define BLENDER_MINVERSION      262
diff --git a/source/blender/blenkernel/intern/image.c b/source/blender/blenkernel/intern/image.c
index 8630ace4edf..b93f917c9e2 100644
--- a/source/blender/blenkernel/intern/image.c
+++ b/source/blender/blenkernel/intern/image.c
@@ -312,10 +312,6 @@ static void image_assign_ibuf(Image *ima, ImBuf *ibuf, int index, int frame)
 				break;
 
 		ibuf->index = index;
-		if (ima->flag & IMA_CM_PREDIVIDE)
-			ibuf->flags |= IB_cm_predivide;
-		else
-			ibuf->flags &= ~IB_cm_predivide;
 
 		/* this function accepts (link == NULL) */
 		BLI_insertlinkbefore(&ima->ibufs, link, ibuf);
@@ -552,6 +548,26 @@ int BKE_image_scale(Image *image, int width, int height)
 	return (ibuf != NULL);
 }
 
+static void image_init_color_management(Image *ima)
+{
+	ImBuf *ibuf;
+	char name[FILE_MAX];
+
+	BKE_image_user_file_path(NULL, ima, name);
+
+	/* will set input color space to image format default's */
+	ibuf = IMB_loadiffname(name, IB_test | IB_alphamode_detect, ima->colorspace_settings.name);
+
+	if (ibuf) {
+		if (ibuf->flags & IB_alphamode_premul)
+			ima->alpha_mode = IMA_ALPHA_PREMUL;
+		else
+			ima->alpha_mode = IMA_ALPHA_STRAIGHT;
+
+		IMB_freeImBuf(ibuf);
+	}
+}
+
 Image *BKE_image_load(const char *filepath)
 {
 	Image *ima;
@@ -579,6 +595,8 @@ Image *BKE_image_load(const char *filepath)
 	if (BLI_testextensie_array(filepath, imb_ext_movie))
 		ima->source = IMA_SRC_MOVIE;
 
+	image_init_color_management(ima);
+
 	return ima;
 }
 
@@ -666,7 +684,7 @@ static ImBuf *add_ibuf_size(unsigned int width, unsigned int height, const char
 		/* both byte and float buffers are filling in sRGB space, need to linearize float buffer after BKE_image_buf_fill* functions */
 
 		IMB_buffer_float_from_float(rect_float, rect_float, ibuf->channels, IB_PROFILE_LINEAR_RGB, IB_PROFILE_SRGB,
-		                            ibuf->flags & IB_cm_predivide, ibuf->x, ibuf->y, ibuf->x, ibuf->x);
+		                            TRUE, ibuf->x, ibuf->y, ibuf->x, ibuf->x);
 	}
 
 	return ibuf;
@@ -2343,7 +2361,7 @@ void BKE_image_backup_render(Scene *scene, Image *ima)
 static void image_create_multilayer(Image *ima, ImBuf *ibuf, int framenr)
 {
 	const char *colorspace = ima->colorspace_settings.name;
-	int predivide = ima->flag & IMA_CM_PREDIVIDE;
+	int predivide = ima->alpha_mode == IMA_ALPHA_PREMUL;
 
 	ima->rr = RE_MultilayerConvert(ibuf->userdata, colorspace, predivide, ibuf->x, ibuf->y);
 
@@ -2375,6 +2393,18 @@ static void image_initialize_after_load(Image *ima, ImBuf *ibuf)
 
 }
 
+static int imbuf_alpha_flags_for_image(Image *ima)
+{
+	int flag = 0;
+
+	if (ima->flag & IMA_IGNORE_ALPHA)
+		flag |= IB_ignore_alpha;
+	else if (ima->alpha_mode == IMA_ALPHA_PREMUL)
+		flag |= IB_alphamode_premul;
+
+	return flag;
+}
+
 static ImBuf *image_load_sequence_file(Image *ima, ImageUser *iuser, int frame)
 {
 	struct ImBuf *ibuf;
@@ -2389,8 +2419,7 @@ static ImBuf *image_load_sequence_file(Image *ima, ImageUser *iuser, int frame)
 	BKE_image_user_file_path(iuser, ima, name);
 
 	flag = IB_rect | IB_multilayer;
-	if (ima->flag & IMA_DO_PREMUL)
-		flag |= IB_premul;
+	flag |= imbuf_alpha_flags_for_image(ima);
 
 	/* read ibuf */
 	ibuf = IMB_loadiffname(name, flag, ima->colorspace_settings.name);
@@ -2549,15 +2578,14 @@ static ImBuf *image_load_image_file(Image *ima, ImageUser *iuser, int cfra)
 	/* is there a PackedFile with this image ? */
 	if (ima->packedfile) {
 		flag = IB_rect | IB_multilayer;
-		if (ima->flag & IMA_DO_PREMUL) flag |= IB_premul;
+		flag |= imbuf_alpha_flags_for_image(ima);
 
 		ibuf = IMB_ibImageFromMemory((unsigned char *)ima->packedfile->data, ima->packedfile->size, flag,
 		                             ima->colorspace_settings.name, "<packed data>");
 	}
 	else {
 		flag = IB_rect | IB_multilayer | IB_metadata;
-		if (ima->flag & IMA_DO_PREMUL)
-			flag |= IB_premul;
+		flag |= imbuf_alpha_flags_for_image(ima);
 
 		/* get the right string */
 		BKE_image_user_frame_calc(iuser, cfra, 0);
@@ -2777,15 +2805,6 @@ static ImBuf *image_get_render_result(Image *ima, ImageUser *iuser, void **lock_
 
 	ibuf->dither = dither;
 
-	if (iuser->scene->r.color_mgt_flag & R_COLOR_MANAGEMENT_PREDIVIDE) {
-		ibuf->flags |= IB_cm_predivide;
-		ima->flag |= IMA_CM_PREDIVIDE;
-	}
-	else {
-		ibuf->flags &= ~IB_cm_predivide;
-		ima->flag &= ~IMA_CM_PREDIVIDE;
-	}
-
 	ima->ok = IMA_OK_LOADED;
 
 	return ibuf;
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);
 	}
 }
 
diff --git a/source/blender/blenkernel/intern/seqmodifier.c b/source/blender/blenkernel/intern/seqmodifier.c
index 5b2e9f2bf23..3d8a2f7cddf 100644
--- a/source/blender/blenkernel/intern/seqmodifier.c
+++ b/source/blender/blenkernel/intern/seqmodifier.c
@@ -226,24 +226,28 @@ static void curves_apply_threaded(int width, int height, unsigned char *rect, fl
 			}
 			if (rect) {
 				unsigned char *pixel = rect + pixel_index;
-				unsigned char result[3];
+				float result[3], tempc[4];
 
-				curvemapping_evaluate_premulRGB(curve_mapping, result, pixel);
+				straight_uchar_to_premul_float(tempc, pixel);
+
+				curvemapping_evaluate_premulRGBF(curve_mapping, result, tempc);
 
 				if (mask_rect) {
 					float t[3];
 
 					rgb_uchar_to_float(t, mask_rect + pixel_index);
 
-					pixel[0] = pixel[0] * (1.0f - t[0]) + result[0] * t[0];
-					pixel[1] = pixel[1] * (1.0f - t[1]) + result[1] * t[1];
-					pixel[2] = pixel[2] * (1.0f - t[2]) + result[2] * t[2];
+					tempc[0] = pixel[0] * (1.0f - t[0]) + result[0] * t[0];
+					tempc[1] = pixel[1] * (1.0f - t[1]) + result[1] * t[1];
+					tempc[2] = pixel[2] * (1.0f - t[2]) + result[2] * t[2];
 				}
 				else {
-					pixel[0] = result[0];
-					pixel[1] = result[1];
-					pixel[2] = result[2];
+					tempc[0] = result[0];
+					tempc[1] = result[1];
+					tempc[2] = result[2];
 				}
+
+				premul_float_to_straight_uchar(pixel, tempc);
 			}
 		}
 	}
diff --git a/source/blender/blenkernel/intern/sequencer.c b/source/blender/blenkernel/intern/sequencer.c
index a82704d9e38..69c125b0ece 100644
--- a/source/blender/blenkernel/intern/sequencer.c
+++ b/source/blender/blenkernel/intern/sequencer.c
@@ -324,7 +324,6 @@ void BKE_sequencer_imbuf_to_sequencer_space(Scene *scene, ImBuf *ibuf, int make_
 {
 	const char *from_colorspace = IMB_colormanagement_role_colorspace_name_get(COLOR_ROLE_SCENE_LINEAR);
 	const char *to_colorspace = scene->sequencer_colorspace_settings.name;
-	int predivide = ibuf->flags & IB_cm_predivide;
 
 	if (!ibuf->rect_float) {
 		if (make_float && ibuf->rect) {
@@ -354,7 +353,7 @@ void BKE_sequencer_imbuf_to_sequencer_space(Scene *scene, ImBuf *ibuf, int make_
 			imb_freerectImBuf(ibuf);
 
 		IMB_colormanagement_transform_threaded(ibuf->rect_float, ibuf->x, ibuf->y, ibuf->channels,
-		                                       from_colorspace, to_colorspace, predivide);
+		                                       from_colorspace, to_colorspace, TRUE);
 	}
 }
 
@@ -367,10 +366,8 @@ void BKE_sequencer_imbuf_from_sequencer_space(Scene *scene, ImBuf *ibuf)
 		return;
 
 	if (to_colorspace && to_colorspace[0] != '\0') {
-		int predivide = ibuf->flags & IB_cm_predivide;
-
 		IMB_colormanagement_transform_threaded(ibuf->rect_float, ibuf->x, ibuf->y, ibuf->channels,
-		                                       from_colorspace, to_colorspace, predivide);
+		                                       from_colorspace, to_colorspace, TRUE);
 	}
 }
 
@@ -1517,18 +1514,6 @@ MINLINE float color_balance_fl(float in, const float lift, const float gain, con
 	return powf(x, gamma) * mul;
 }
 
-static void make_cb_table_byte(float lift, float gain, float gamma,
-                               unsigned char *table, float mul)
-{
-	int y;
-
-	for (y = 0; y < 256; y++) {
-		float v = color_balance_fl((float)y * (1.0f / 255.0f), lift, gain, gamma, mul);
-
-		table[y] = FTOCHAR(v);
-	}
-}
-
 static void make_cb_table_float(float lift, float gain, float gamma,
                                 float *table, float mul)
 {
@@ -1543,35 +1528,33 @@ static void make_cb_table_float(float lift, float gain, float gamma,
 
 static void color_balance_byte_byte(StripColorBalance *cb_, unsigned char *rect, unsigned char *mask_rect, int width, int height, float mul)
 {
-	unsigned char cb_tab[3][256];
-	int c;
-	unsigned char *p = rect;
-	unsigned char *e = p + width * 4 * height;
+	//unsigned char cb_tab[3][256];
+	unsigned char *cp = rect;
+	unsigned char *e = cp + width * 4 * height;
 	unsigned char *m = mask_rect;
 
 	StripColorBalance cb = calc_cb(cb_);
 
-	for (c = 0; c < 3; c++) {
-		make_cb_table_byte(cb.lift[c], cb.gain[c], cb.gamma[c], cb_tab[c], mul);
-	}
+	while (cp < e) {
+		float p[4];
+		int c;
 
-	while (p < e) {
-		if (m) {
-			float t[3] = {m[0] / 255.0f, m[1] / 255.0f, m[2] / 255.0f};
+		straight_uchar_to_premul_float(p, cp);
 
-			p[0] = p[0] * (1.0f - t[0]) + t[0] * cb_tab[0][p[0]];
-			p[1] = p[1] * (1.0f - t[1]) + t[1] * cb_tab[1][p[1]];
-			p[2] = p[2] * (1.0f - t[2]) + t[2] * cb_tab[2][p[2]];
+		for (c = 0; c < 3; c++) {
+			float t = color_balance_fl(p[c], cb.lift[c], cb.gain[c], cb.gamma[c], mul);
 
-			m += 4;
-		}
-		else {
-			p[0] = cb_tab[0][p[0]];
-			p[1] = cb_tab[1][p[1]];
-			p[2] = cb_tab[2][p[2]];
+			if (m)
+				p[c] = p[c] * (1.0f - (float)m[c] / 255.0f) + t * m[c];
+			else
+				p[c] = t;
 		}
 		
-		p += 4;
+		premul_float_to_straight_uchar(cp, p);
+
+		cp += 4;
+		if (m)
+			m += 4;
 	}
 }
 
@@ -1795,7 +1778,7 @@ int BKE_sequencer_input_have_to_preprocess(SeqRenderData UNUSED(context), Sequen
 {
 	float mul;
 
-	if (seq->flag & (SEQ_FILTERY | SEQ_USE_CROP | SEQ_USE_TRANSFORM | SEQ_FLIPX | SEQ_FLIPY | SEQ_MAKE_PREMUL | SEQ_MAKE_FLOAT)) {
+	if (seq->flag & (SEQ_FILTERY | SEQ_USE_CROP | SEQ_USE_TRANSFORM | SEQ_FLIPX | SEQ_FLIPY | SEQ_MAKE_FLOAT)) {
 		return TRUE;
 	}
 
@@ -1892,7 +1875,8 @@ static ImBuf *input_preprocess(SeqRenderData context, Sequence *seq, float cfra,
 			ImBuf *i = IMB_allocImBuf(dx, dy, 32, ibuf->rect_float ? IB_rectfloat : IB_rect);
 
 			IMB_rectcpy(i, ibuf, t.xofs, t.yofs, c.left, c.bottom, sx, sy);
-			
+			sequencer_imbuf_assign_spaces(context.scene, i);
+
 			IMB_freeImBuf(ibuf);
 
 			ibuf = i;
@@ -1931,12 +1915,6 @@ static ImBuf *input_preprocess(SeqRenderData context, Sequence *seq, float cfra,
 		multibuf(ibuf, mul);
 	}
 
-	if (seq->flag & SEQ_MAKE_PREMUL) {
-		if (ibuf->planes == 32 && ibuf->zbuf == NULL) {
-			IMB_premultiply_alpha(ibuf);
-		}
-	}
-
 	if (ibuf->x != context.rectx || ibuf->y != context.recty) {
 		if (context.scene->r.mode & R_OSA) {
 			IMB_scaleImBuf(ibuf, (short)context.rectx, (short)context.recty);
@@ -2546,13 +2524,18 @@ static ImBuf *do_render_strip_uncached(SeqRenderData context, Sequence *seq, flo
 		case SEQ_TYPE_IMAGE:
 		{
 			StripElem *s_elem = BKE_sequencer_give_stripelem(seq, cfra);
+			int flag;
 
 			if (s_elem) {
 				BLI_join_dirfile(name, sizeof(name), seq->strip->dir, s_elem->name);
 				BLI_path_abs(name, G.main->name);
 			}
 
-			if (s_elem && (ibuf = IMB_loadiffname(name, IB_rect, seq->strip->colorspace_settings.name))) {
+			flag = IB_rect;
+			if (seq->alpha_mode == SEQ_ALPHA_PREMUL)
+				flag |= IB_alphamode_premul;
+
+			if (s_elem && (ibuf = IMB_loadiffname(name, flag, seq->strip->colorspace_settings.name))) {
 				/* we don't need both (speed reasons)! */
 				if (ibuf->rect_float && ibuf->rect)
 					imb_freerectImBuf(ibuf);
@@ -2641,7 +2624,7 @@ static ImBuf *do_render_strip_uncached(SeqRenderData context, Sequence *seq, flo
 static ImBuf *seq_render_strip(SeqRenderData context, Sequence *seq, float cfra)
 {
 	ImBuf *ibuf = NULL;
-	int use_preprocess = BKE_sequencer_input_have_to_preprocess(context, seq, cfra);
+	int use_preprocess = FALSE;
 	int is_proxy_image = FALSE;
 	float nr = give_stripelem_index(seq, cfra);
 	/* all effects are handled similarly with the exception of speed effect */
@@ -2650,30 +2633,36 @@ static ImBuf *seq_render_strip(SeqRenderData context, Sequence *seq, float cfra)
 
 	ibuf = BKE_sequencer_cache_get(context, seq, cfra, SEQ_STRIPELEM_IBUF);
 
-	/* currently, we cache preprocessed images in SEQ_STRIPELEM_IBUF,
-	 * but not(!) on SEQ_STRIPELEM_IBUF_ENDSTILL and ..._STARTSTILL */
-	if (ibuf)
-		use_preprocess = FALSE;
-
-	if (ibuf == NULL)
-		ibuf = copy_from_ibuf_still(context, seq, nr);
-
 	if (ibuf == NULL) {
-		ibuf = BKE_sequencer_preprocessed_cache_get(context, seq, cfra, SEQ_STRIPELEM_IBUF);
+		if (ibuf == NULL)
+			ibuf = copy_from_ibuf_still(context, seq, nr);
 
 		if (ibuf == NULL) {
-			/* MOVIECLIPs have their own proxy management */
-			if (ibuf == NULL && seq->type != SEQ_TYPE_MOVIECLIP) {
-				ibuf = seq_proxy_fetch(context, seq, cfra);
-				is_proxy_image = (ibuf != NULL);
-			}
+			ibuf = BKE_sequencer_preprocessed_cache_get(context, seq, cfra, SEQ_STRIPELEM_IBUF);
 
-			if (ibuf == NULL)
-				ibuf = do_render_strip_uncached(context, seq, cfra);
+			if (ibuf == NULL) {
+				/* MOVIECLIPs have their own proxy management */
+				if (ibuf == NULL && seq->type != SEQ_TYPE_MOVIECLIP) {
+					ibuf = seq_proxy_fetch(context, seq, cfra);
+					is_proxy_image = (ibuf != NULL);
+				}
 
-			if (ibuf)
-				BKE_sequencer_preprocessed_cache_put(context, seq, cfra, SEQ_STRIPELEM_IBUF, ibuf);
+				if (ibuf == NULL)
+					ibuf = do_render_strip_uncached(context, seq, cfra);
+
+				if (ibuf)
+					BKE_sequencer_preprocessed_cache_put(context, seq, cfra, SEQ_STRIPELEM_IBUF, ibuf);
+			}
 		}
+
+		if (ibuf)
+			use_preprocess = BKE_sequencer_input_have_to_preprocess(context, seq, cfra);
+	}
+	else {
+		/* currently, we cache preprocessed images in SEQ_STRIPELEM_IBUF,
+		 * but not(!) on SEQ_STRIPELEM_IBUF_ENDSTILL and ..._STARTSTILL
+		 * so, no need in check for preprocess here
+		 */
 	}
 
 	if (ibuf == NULL) {
@@ -3975,6 +3964,14 @@ void BKE_sequence_init_colorspace(Sequence *seq)
 		if (seq->type == SEQ_TYPE_IMAGE) {
 			ibuf = IMB_loadiffname(name, IB_rect, seq->strip->colorspace_settings.name);
 
+			/* byte images are default to straight alpha, however sequencer
+			 * works in premul space, so mark strip to be premultiplied first
+			 */
+			if (!ibuf->rect_float)
+				seq->alpha_mode = SEQ_ALPHA_STRAIGHT;
+			else
+				seq->alpha_mode = SEQ_ALPHA_PREMUL;
+
 			if (ibuf)
 				IMB_freeImBuf(ibuf);
 		}
diff --git a/source/blender/blenkernel/intern/texture.c b/source/blender/blenkernel/intern/texture.c
index 149842bc038..fbaf6f70fbc 100644
--- a/source/blender/blenkernel/intern/texture.c
+++ b/source/blender/blenkernel/intern/texture.c
@@ -440,7 +440,7 @@ void default_tex(Tex *tex)
 	tex->type = TEX_CLOUDS;
 	tex->stype = 0;
 	tex->flag = TEX_CHECKER_ODD;
-	tex->imaflag = TEX_INTERPOL | TEX_MIPMAP | TEX_USEALPHA;
+	tex->imaflag = TEX_INTERPOL | TEX_MIPMAP;
 	tex->extend = TEX_REPEAT;
 	tex->cropxmin = tex->cropymin = 0.0;
 	tex->cropxmax = tex->cropymax = 1.0;