prompts

SVD implementation

The Singular Value Decomposition (SVD) implementation was done with OpenAI's Codex, model "gpt-5.5". (Several interactions were made in order to get good enough sparse SVD results.)

First

Examine the file "./src/SparseMatrixFunctions.c" and implement the routine svd the does thin Singular Value Decomposition (SVD) over sparse matrices. In the input is an object CSRStruct, the results matrices are "u", "s", "v". Make sure:

• You use only data structures and operations implemented in this package; do not use other libraries

• The sparse SVD implementation is robust and standard

• The routine svd has an integer argument k and returns the decomposition that goes with the k largest singular values

After profiling

Your implementation is slow -- hundreds of times slower than expected. It seems you implemented a dense matrix SVD. Maybe, because I said "standard". Implement a "standard sparse matrix SVD".