Currently furnish_kernels is download kernels one at a time even though filecache is capable of massive parallelism. This is an enormous performance problem. The issue appears to be at spicedb/__init__.py line 1523:
abspaths = pfx.retrieve(filepaths, exception_on_fail=False)
Even though this looks like it's retrieving multiple kernels, it's really not because the outer loop is the one looping through the available kernels.
Maybe this isn't worth fixing if spyceman is coming soon.
Currently
furnish_kernelsis download kernels one at a time even though filecache is capable of massive parallelism. This is an enormous performance problem. The issue appears to be atspicedb/__init__.pyline 1523:Even though this looks like it's retrieving multiple kernels, it's really not because the outer loop is the one looping through the available kernels.
Maybe this isn't worth fixing if
spycemanis coming soon.