*(uint64_t*)(p + 0+3*stride)= v;
*(uint64_t*)(p + 8+3*stride)= v;
#else
- const uint32_t v= size==2 ? val*0x00010001 : val*0x01010101;
+ const uint32_t v= size==2 ? val*0x00010001 : val;
*(uint32_t*)(p + 0+0*stride)= v;
*(uint32_t*)(p + 4+0*stride)= v;
if(h==1) return;