2 CoreMark

CoreMark is a benchmarking suite that is intended for the performance of the CPU, when running with specific workloads. The internal CPU architecture is independent of the on-chip memory and peripherals. CoreMark focuses on processor core and cache memory read/write by utilizing the CPU pipeline architecture. The advantage of CoreMark is its size, which allows CoreMark to easily fit in a processor's memory and makes it suitable to get the performance outcome. To evaluate processor performance, CoreMark uses about 20k size of code, which includes I/O computations. All the computations are made at run time. CoreMark mainly focuses on basic read/write operations and integer operations. CoreMark uses realistic workloads. The CoreMark workload comprises of several commonly used algorithms, including:

• Matrix manipulation, to exercise common math operations.
• Linked list manipulation, to exercise the common use of pointers.
• State machine operation, to exercise data-dependent branches.
• Cyclic Redundancy Check (CRC) is a common function in the embedded systems.

The outcome of the CoreMark execution is as follows:

• The four bits of workload tested are matrix manipulation, linked lists, state machines, and CRCs. The output of each stage of the compilers is intended to get the best outcome of the performance results.
• PolarFire SoC device can load the entire test in the cache, DDR, and Flash memory.

The GCC on the RISC-V gives a good performance with -O2 optimization compiler flag and without a compressed extension (RVC).

For more information about CoreMark, see EEMBC.