Description: Measures the maximum device memory bandwidth for different levels of the memory hierarchy and different access patterns. Results are reported in GB/s.

Problem Sizes: Problem size automatically scales up with device performance (designed to saturate bandwidth).

Precision: N/A

Includes PCIe Transfer Time: No

Specific Tests:

• ImageCacheAccess - Reports read bandwidth from texture/image memory
• ImageLinearAccess - Reports read bandwidth from texture/image memory using a linear access pattern
• ImageRandAccess - Reports read bandwidth from texture/image memory using a random access pattern
• readConstantMemoryCoalesced - Reports read bandwidth from constant memory using coalesced (per thread-block contiguous) accesses
• readGlobalMemoryCoalesced - Reports read bandwidth from global memory using coalesced (per thread-block contiguous) accesses
• readGlobalMemoryUnit - Reports read bandwidth from global memory using uncoalesced (per thread contiguous) accesses
• readLocalMemory - Reports read bandwidth from local memory
• writeGlobalMemoryCoalesced - Reports write bandwidth to global memory using coalesced (per thread-block contiguous) accesses
• writeGlobalMemoryUnit - Reports write bandwidth to global memory using uncoalesced (per thread contiguous) accesses
• writeLocalMemory - Reports write bandwidth to local memory