Ive seen great speedups in applications that have switch to mimalloc for memory allocation.
I humbly ask the qspice dev team check it out and consider it for qspice.
Considering malloc is only used once at the very first Cblock call, and later at the very end to free the memory. I don’t think a more efficient malloc implementation may affect the overall compute speed.
And 2nd, understanding that digital mars is ancient compiler, there is some possibility more modern library can’t get compiled. Unless you are using another compiler as @RDunn has demonstrated in his github.