Franck HOUSSEN
Franck HOUSSEN
>What should we be doing here? @caliarim may answer. I never had specific problems with initialization (but a lot on computations! :))
> just using a rand initialized v0 is sufficient to not trigger the error any more Would say make sense to me (in difficult cases). AFAIR, this way, one may...
BTW, in corner cases where `arpack` would still fail, you can restart `arpack` using the previous "not-good-enough" solution as `v0`: for difficult cases, you may run `arpack` twice, but, the...
Sounds good! Looks more like a discussion than an issue to me.
> and the final result differes. BTW, how much in norm for instance? 1.e-6 in double? 1.e-3 in float? more?
If you bootstrapped before installing `pkg-config 0.29.2`, try ``` >> make distclean >> ./bootstrap >> ./configure ```
>Does arpack-ng require Eigen? No, `arpack-ng` needs `eigen` only when asking for `exmm` option (matrix market examples using C/C++ bindings allowing direct use of fortran). Can you print with `AC_MSG_RESULT`...
>I added a line AC_MSG_RESULT($enable_icb_exmm), then ran ./configure again. Modifying `./configure` will not help: you need to regenerate it. Follow this steps, what are your traces? ``` >> git diff...
Put the `AC_MSG_RESULT` before https://github.com/opencollab/arpack-ng/blob/db55a7ff4180bb440a980d618e302eded0db52f3/configure.ac#L92
This patch is becoming huge which is no good: mixing-up things is the best way to reg! :( I thought at first it was only about using `find_dependency` in `./cmake/arpackng-config.cmake.in`...