代码

#pragma GCC optimize("Ofast")
#pragma GCC optimize(2)
unsigned int cnt1[256|1],cnt2[256|1],cnt3[256|1],cnt4[256|1];
unsigned tmp[(int)1e8+1];
void sort(unsigned *a,int n)
{
    --a;
    for(register int i=1;i<=n;++i)
        ++cnt1[a[i]&0xff],
        ++cnt2[a[i]>>8&0xff],
        ++cnt3[a[i]>>16&0xff],
        ++cnt4[a[i]>>24&0xff];
 
    for(register int i=1;i<256;++i)
        cnt1[i]+=cnt1[i-1],
        cnt2[i]+=cnt2[i-1],
        cnt3[i]+=cnt3[i-1],
        cnt4[i]+=cnt4[i-1];
 
    for(register int i=n;i;--i)tmp[cnt1[a[i]&0xff]--]=a[i];
    for(register int i=n;i;--i)a[cnt2[tmp[i]>>8&0xff]--]=tmp[i];
    for(register int i=n;i;--i)tmp[cnt3[a[i]>>16&0xff]--]=a[i];
    for(register int i=n;i;--i)a[cnt4[tmp[i]>>24&0xff]--]=tmp[i];
 
}

评测结果