--- Quote Start ---
Altera makes 3th way -- from embedded ARMv7 in FPGA, booting from SD-card and running with Linux.
If you may compile kernels from Win/Lin and write its to SD-flash, then may insert big PCIe board to Mac and send control manipulations through ARMs through PCIe, write data arrays to shared memory and read back results from it.
However, I recommend NVIDIA GPUs -- more faster for image processing.
--- Quote End ---
That would only work for SoC FPGAs that have an ARM core. The majority of high-end FPGA boards only have a non-SoC FPGAs and will need to be connected to a standard x86 host to be usable with OpenCL.
And of course, if you are comparing high-end Nvidia GPUs to low-end SoC FPGAs, the GPU will be much faster for any workload, not just image processing.
Honored Contributor
