Am I right in concluding, from some testing, that this works only for out-of-place operations, i.e., gemm(res, a, b) where a≠res≠b?
When I try something like gemm(c, c, a), the result is full of zeros.
If this is the case, it might be worth mentioning on the README. Are there any workarounds to enabling in-place multiplication, or would that ruin performance?
Am I right in concluding, from some testing, that this works only for out-of-place operations, i.e.,
gemm(res, a, b)
wherea
≠res
≠b
?When I try something like
gemm(c, c, a)
, the result is full of zeros.If this is the case, it might be worth mentioning on the README. Are there any workarounds to enabling in-place multiplication, or would that ruin performance?