Closed xper0418 closed 1 year ago
Hi there, no plans to support float32 as of now. I assume the accuracy of the result will suffer a lot when using float32. Have you tried to solve your system with scipy's spsolve to check if the result is acceptable in terms of accuracy and computation time?
Thanks for the response. I tested it with sicpy, but the calculation was not finished, so I stopped in the middle. Since I already store the results as a float32, I don't think there will be any accuracy issues.
Hi. As far as I know, PyPardiso only supports float64. I hope to use float32 to save memory. Any update plans?