I recently discovered that using the base SVD function on a large matrix (109865 x 414) was causing RStudio 1.3.1073 to crash (running R 4.0.2 on macOS 10.15.6 with 32GB RAM). Running the same code in terminal provided the following error:
*** caught bus error *** address 0x1195ed000, cause 'non-existent physical address'
Traceback: 1: La.svd(x, nu, nv) 2: svd(problemMatrix) An irrecoverable exception occurred. R is aborting now ...
*** caught bus error *** address 0x11d604000, cause 'non-existent physical address'
Traceback: 1: La.svd(x, nu, nv) 2: svd(problemMatrix) An irrecoverable exception occurred. R is aborting now ...
*** caught segfault *** address 0x1297b9000, cause 'memory not mapped'
No error is given when performing SVD on a 109865 x 414 matrix of random values. Running the same code on the same original matrix on a different computer (running R 4.0.0 on macOS 10.15.4 with 8 GB RAM) performs the SVD as expected, without throwing an error. The newer R installation has otherwise been completely stable.
So, admittedly a sort of hyperspecific, potentially non-reproducible question, but any thoughts on what might be responsible for this error? Problematic matrix uploaded as an R data file here.
Thanks!