Because the entire encrypted bitstream is loaded into the FPGA’s configuration memory prior to decryption, the bitstream size is limited by the amount of memory in the device. In this case, what must be considered is the percent impact to overall configuration time. This impact depends on which external port is used for configuration. As mentioned previously, the internal RDW phase is always performed at 32 bits wide and at the rate of the specified clock frequency used for configuration. Therefore, if the external port is slower (the “bottleneck”), the overall configuration time might not be affected that much. Conversely, if a fast external port is used, there will be a larger overall percent increase in configuration time. Some examples are:
- JTAG and Master SPI (x1):
- Overall configuration time increased by ~8%
- Serial load phase followed by 32-bit parallel RDW phase
- Master SPI Quad (x4):
- Overall configuration time increased by 32%
- Quad SPI load phase followed by 32-bit parallel RDW phase
- Master SPI Dual Quad (x8):
- Overall configuration time increased by 63%
- Dual Quad SPI load phase followed by 32-bit parallel RDW phase
- Master (CCLK) asynchronous BPI (x16):
- Configuration time increased by ~125%
- 16-bit parallel load phase followed by a 32-bit parallel RDW phase
- Master (EMCCLK) synchronous BPI (x16):
- Configuration time increased by ~125%
- 16-bit parallel load phase followed by a 32-bit parallel RDW phas