Open cartazio opened 10 years ago
could adapt the very dumb lacking test program https://github.com/wellposed/hblas/blob/master/testing/Test.hs to test this
Small mistake : the function signature should be :
sger ::( PrimMonad s) =>
Float -> Mvector (PrimState s) Float -> MVector (PrimState s) Float
-> MutDenseMatrix (PrimState m) orient Float -> m ()
good catch! As specced in http://software.intel.com/sites/products/documentation/hpc/mkl/mklman/GUID-BD2E87B3-5FA7-4E0C-88E2-1982AB0773A2.htm
it should be
A := alpha*x*y'+ A,
note that this may not be implemented correctly in terms of certain corner cases, reopening for now and this ticket + associated audit needs to be done before i do the next release
sger and dger
http://software.intel.com/sites/products/documentation/hpc/mkl/mklman/GUID-BD2E87B3-5FA7-4E0C-88E2-1982AB0773A2.htm
https://github.com/wellposed/hblas/blob/master/src/Numerical/HBLAS/BLAS/FFI.hs#L247-L250
the haskelly high level version would look like (assuming you punt on dealing with modelling the vectors properly because i've not added strided vector support to hblas yet, and you assume inx and incy are set to 1)
note this would actually be done slightly differently once I add strided vector support
a mini explanation of some of the API stuff I do currently is on this other ticket https://github.com/wellposed/hblas/issues/13
don't hesitate to ask for help