Hi
Sorry if this is a really obvious/stupid question. If all threads in a warp read a float2 or float4 from the same global memory address does it count as coalesced, or does each thread read the same address independently? I should probably mention I’m talking about raw global memory access, not texture memory access.