Please indicate in this forum under quadmath if there is any interest in expanding lapack to quadmath. Alternatively indicate if this topic is undesirable; so it can be expunged.
I have very mechanically converted lapack to gcc quadmath (gcc-4.7.0). I am not an expert in linear algebra and I am experiencing problems with the extensive lapack test packages.
For my own purpose I have only done test with matrix inversion. Using a 500x500 matrix filled with random numbers I obtain, after suitable multiplication, an approximate identity matrix
with terms correct to plus 28 decimal places. Including the gcc matrix multiplication to obtain the approximate identiy matrix the whole process is about 150 times slower than double precision.
I am not claiming any originality and all converted source code is freely available under whatever restrictions stated in lapack-3.3.1.
Best Regards Ray Malitzke