Re: [PATCH v2 0/2] MR8970: kernel32/kernelbase: Implemented NUMA functions and improved affinity support

[Bernhard Kölbl (＠besentv)]

Bernhard Kölbl (@besentv) commented about dlls/kernelbase/memory.c:
> -
>        numa_nodes[0].cpu_mask = 1;
> -
>    numa_nodes[0].valid = TRUE; max_node = 0;
> -
>    TRACE_(numa)("NUMA fallback single: mask=0x%llx procs=%u\n", (unsigned long long)numa_nodes[0].cpu_mask, (unsigned)popcount_ulongptr(numa_nodes[0].cpu_mask));
> - }
> - else if (numa_contig && max_node > 0)
> - {
> -
>    /* Remap each node to a contiguous block of bits in ascending order */
> -
>    ULONG_PTR new_masks[64] = {0}; unsigned int bit_offset = 0; BOOL ok = TRUE;
> -
>    for (i = 0; i <= max_node; i++) if (numa_nodes[i].valid)
> -
>    {
> -
>        unsigned int cnt = popcount_ulongptr(numa_nodes[i].cpu_mask);
> -
>        if (!cnt || bit_offset + cnt > sizeof(ULONG_PTR)*8) { ok = FALSE; break; }
> -
>        new_masks[i] = (((ULONG_PTR)1 << cnt) - 1) << bit_offset;
> -
>        TRACE_(numa)("NUMA remap: node=%lu raw=0x%llx cnt=%u -> contig=0x%llx base=%u\n",
> -
>              (unsigned long)i, (unsigned long long)numa_nodes[i].cpu_mask, cnt,
> -
>              (unsigned long long)new_masks[i], bit_offset);

Why all these casts, instead of using the correct format?