Fix T44541 aka gigapixel image render support in blender.

Moral of the story: Make sure that size_t is used whenever pointer
arithmetic is involved. For images, that basically means whenever any
squared dimensions are involved. Casting an operand to size_t early in
the operation is usually sufficient to force the entire operation to
size_t.

There might still be places lurking where we don't support this
correctly. This has been tested with render pipeline, quite a few image
functions (meaning we can paint on such images now, albeit somewhat
slowly ;) ) and export to jpeg. Too many places in code to check so I
guess we'll be handling cases as they come.

Don't try this at home unless you have an immense ammount of RAM.
First GPixel render of suzanne in the multiverse can be found here:

http://download.blender.org/demo/test/suzanne-billion-pixel.jpg

Can be viewed from blender (takes about 3.3 GB after loading but may
take more during loading so 8GB might be more safe to try this).

1 files changed, 10 insertions, 10 deletions

diff --git a/source/blender/imbuf/intern/colormanagement.c b/source/blender/imbuf/intern/colormanagement.c
index 20baa1614c9..d6fdcad0819 100644
--- a/source/blender/imbuf/intern/colormanagement.c
+++ b/source/blender/imbuf/intern/colormanagement.c
@@ -1306,8 +1306,8 @@ static void display_buffer_init_handle(void *handle_v, int start_line, int tot_l
 	float dither = ibuf->dither;
 	bool is_data = (ibuf->colormanage_flag & IMB_COLORMANAGE_IS_DATA) != 0;
 
-	int offset = channels * start_line * ibuf->x;
-	int display_buffer_byte_offset = DISPLAY_BUFFER_CHANNELS * start_line * ibuf->x;
+	size_t offset = ((size_t)channels) * start_line * ibuf->x;
+	size_t display_buffer_byte_offset = ((size_t)DISPLAY_BUFFER_CHANNELS) * start_line * ibuf->x;
 
 	memset(handle, 0, sizeof(DisplayBufferThread));
 
@@ -1344,7 +1344,7 @@ static void display_buffer_apply_get_linear_buffer(DisplayBufferThread *handle,
 	int channels = handle->channels;
 	int width = handle->width;
 
-	int buffer_size = channels * width * height;
+	size_t buffer_size = ((size_t)channels) * width * height;
 
 	bool is_data = handle->is_data;
 	bool is_data_display = handle->cm_processor->is_data_result;
@@ -1357,11 +1357,11 @@ static void display_buffer_apply_get_linear_buffer(DisplayBufferThread *handle,
 
 		float *fp;
 		unsigned char *cp;
-		int i;
+		size_t i;
 
 		/* first convert byte buffer to float, keep in image space */
 		for (i = 0, fp = linear_buffer, cp = byte_buffer;
-		     i < width * height;
+		     i < ((size_t)width) * height;
 		     i++, fp += channels, cp += channels)
 		{
 			if (channels == 3) {
@@ -1440,7 +1440,7 @@ static void *do_display_buffer_apply_thread(void *handle_v)
 	}
 	else {
 		bool is_straight_alpha, predivide;
-		float *linear_buffer = MEM_mallocN(channels * width * height * sizeof(float),
+		float *linear_buffer = MEM_mallocN(((size_t)channels) * width * height * sizeof(float),
 		                                   "color conversion linear buffer");
 
 		display_buffer_apply_get_linear_buffer(handle, height, linear_buffer, &is_straight_alpha);
@@ -1467,14 +1467,14 @@ static void *do_display_buffer_apply_thread(void *handle_v)
 		}
 
 		if (display_buffer) {
-			memcpy(display_buffer, linear_buffer, width * height * channels * sizeof(float));
+			memcpy(display_buffer, linear_buffer, ((size_t)width) * height * channels * sizeof(float));
 
 			if (is_straight_alpha && channels == 4) {
-				int i;
+				size_t i;
 				float *fp;
 
 				for (i = 0, fp = display_buffer;
-				     i < width * height;
+				     i < ((size_t)width) * height;
 				     i++, fp += channels)
 				{
 					straight_to_premul_v4(fp);
@@ -2959,7 +2959,7 @@ void IMB_colormanagement_processor_apply(ColormanageProcessor *cm_processor, flo
 
 		for (y = 0; y < height; y++) {
 			for (x = 0; x < width; x++) {
-				float *pixel = buffer + channels * (y * width + x);
+				float *pixel = buffer + channels * (((size_t)y) * width + x);
 
 				curve_mapping_apply_pixel(cm_processor->curve_mapping, pixel, channels);
 			}