Agree with the above^^
A 20000x20000 matrix of double elements is 3.2GB, so one of them would fit in 8GB, at least. If it is a square matrix you could do an in-place transpose.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| CUDA Memory Transpose | 10 | 2135 | March 23, 2015 | |
| Batch matrix transposed | 0 | 2414 | August 21, 2009 | |
| Max matrix size for matrix transposition | 4 | 6492 | April 3, 2011 | |
| CUDA matrix transpose using shared memory | 0 | 535 | July 6, 2020 | |
| Could GPU computing speed up computation in data-larger-than-RAM cases? | 5 | 2066 | April 19, 2016 | |
| Fast matrix transpose of any size library/implementation? | 0 | 377 | July 3, 2020 | |
| Optimizing the transpose code | 0 | 2948 | August 23, 2011 | |
| Efficient in-place transpose of multiple square float matrices | 10 | 5947 | October 10, 2019 | |
| Transpose matrix like 8x1M in bytes by memcpy2d | 10 | 221 | November 13, 2024 | |
| Transpose bitwise x:y matrix (x!=y) | 7 | 1172 | October 4, 2021 |