Hello dear all,
I want to copy a big image from host to device using cudamemcopy and I have problem with memory. What is the best way to do this for an image of 10000*10000 pixels, knowing that I have to do other processing operations on the image? I have one GPU core of ability 5.0.
Thank you very much
Please tell us:
1. Is it a color image?
1.2. Is the color indexed or RGB?
2. How much VRAM do your GPU have?
3. What kind of analysis are you going to do?
Thank you very much for your replies. It is a grayscale color, I have 2GB of DDR3 memory in my GPU. I want to compute a wavelet transform of this image.
Thank you very much
What color depth?
Assuming 32 bit that's 10000*10000*4byte = 381MByte.
So Generally the image should fit into the VRAM. Probably the memory gets duplicated by your implementation of the desired transform. Therefore you should design you code in a way, that it will not duplicate the data where possible.
Also reducing the color depth to only 1byte, (256 shades) the memory consumption is reduced to 25%.
Thank you very much. Effectively, I reduced the color depth to 1 byte and I have solved the problem
- Badges
- Science topic
- Similar topics
- Computer Science
- Images