D627: Memory usage optimization for the compositor.

The compostor used a fixed size of 4 floats to hold pixel data. this
patch will select size of a pixel based on its type.
It uses 1 float for Value, 3 float for vector and 4 floats for color
data types.

When benchmarking on shots (opening shot of caminandes) we get a
reduction of memory of 30% and a tiny speedup as less data
transformations needs to take place (but these are negligable.

More information of the patch can be found on
https://developer.blender.org/D627 and
http://wiki.blender.org/index.php/Dev:Ref/Proposals/Compositor2014_p1.1_TD

Developers: jbakker & mdewanchand
Thanks for Sergey for his indept review.

1 files changed, 12 insertions, 13 deletions

diff --git a/source/blender/compositor/operations/COM_DilateErodeOperation.cpp b/source/blender/compositor/operations/COM_DilateErodeOperation.cpp
index cbf4ce693d9..cee39994aed 100644
--- a/source/blender/compositor/operations/COM_DilateErodeOperation.cpp
+++ b/source/blender/compositor/operations/COM_DilateErodeOperation.cpp
@@ -72,7 +72,7 @@ void DilateErodeThresholdOperation::executePixel(float output[4], int x, int y,
 	const float inset = this->m_inset;
 	float mindist = rd * 2;
 
-	MemoryBuffer *inputBuffer = (MemoryBuffer *)data;
+	MemoryBuffer *inputBuffer = (MemoryBuffer*)data;
 	float *buffer = inputBuffer->getBuffer();
 	rcti *rect = inputBuffer->getRect();
 	const int minx = max(x - this->m_scope, rect->xmin);
@@ -82,18 +82,18 @@ void DilateErodeThresholdOperation::executePixel(float output[4], int x, int y,
 	const int bufferWidth = BLI_rcti_size_x(rect);
 	int offset;
 
-	this->m_inputProgram->read(inputValue, x, y, NULL);
+	inputBuffer->read(inputValue, x, y);
 	if (inputValue[0] > sw) {
 		for (int yi = miny; yi < maxy; yi++) {
 			const float dy = yi - y;
-			offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin)) * 4;
+			offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin));
 			for (int xi = minx; xi < maxx; xi++) {
 				if (buffer[offset] < sw) {
 					const float dx = xi - x;
 					const float dis = dx * dx + dy * dy;
 					mindist = min(mindist, dis);
 				}
-				offset += 4;
+				offset ++;
 			}
 		}
 		pixelvalue = -sqrtf(mindist);
@@ -101,15 +101,14 @@ void DilateErodeThresholdOperation::executePixel(float output[4], int x, int y,
 	else {
 		for (int yi = miny; yi < maxy; yi++) {
 			const float dy = yi - y;
-			offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin)) * 4;
+			offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin));
 			for (int xi = minx; xi < maxx; xi++) {
 				if (buffer[offset] > sw) {
 					const float dx = xi - x;
 					const float dis = dx * dx + dy * dy;
 					mindist = min(mindist, dis);
 				}
-				offset += 4;
-
+				offset ++;
 			}
 		}
 		pixelvalue = sqrtf(mindist);
@@ -206,14 +205,14 @@ void DilateDistanceOperation::executePixel(float output[4], int x, int y, void *
 
 	for (int yi = miny; yi < maxy; yi++) {
 		const float dy = yi - y;
-		offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin)) * 4;
+		offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin));
 		for (int xi = minx; xi < maxx; xi++) {
 			const float dx = xi - x;
 			const float dis = dx * dx + dy * dy;
 			if (dis <= mindist) {
 				value = max(buffer[offset], value);
 			}
-			offset += 4;
+			offset ++;
 		}
 	}
 	output[0] = value;
@@ -280,14 +279,14 @@ void ErodeDistanceOperation::executePixel(float output[4], int x, int y, void *d
 
 	for (int yi = miny; yi < maxy; yi++) {
 		const float dy = yi - y;
-		offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin)) * 4;
+		offset = ((yi - rect->ymin) * bufferWidth + (minx - rect->xmin));
 		for (int xi = minx; xi < maxx; xi++) {
 			const float dx = xi - x;
 			const float dis = dx * dx + dy * dy;
 			if (dis <= mindist) {
 				value = min(buffer[offset], value);
 			}
-			offset += 4;
+			offset ++;
 		}
 	}
 	output[0] = value;
@@ -383,7 +382,7 @@ void *DilateStepOperation::initializeTileData(rcti *rect)
 			buf[x] = -FLT_MAX;
 		}
 		for (x = xmin; x < xmax; ++x) {
-			buf[x - rect->xmin + window - 1] = buffer[4 * (y * width + x)];
+			buf[x - rect->xmin + window - 1] = buffer[(y * width + x)];
 		}
 
 		for (i = 0; i < (bwidth + 3 * half_window) / window; i++) {
@@ -510,7 +509,7 @@ void *ErodeStepOperation::initializeTileData(rcti *rect)
 			buf[x] = FLT_MAX;
 		}
 		for (x = xmin; x < xmax; ++x) {
-			buf[x - rect->xmin + window - 1] = buffer[4 * (y * width + x)];
+			buf[x - rect->xmin + window - 1] = buffer[(y * width + x)];
 		}
 
 		for (i = 0; i < (bwidth + 3 * half_window) / window; i++) {