Hi folks-
Yesterday I was in a meeting discussing what path we should take due to our decision (not mine, thankfully) to use an undersized FPGA that will host NIOS. NIOS is a 32 bit soft code machine that runs in Altera FPGAs. At that time it suddenly occurred to me that, since instructions are 32 bits, C code memory usage (not data) might not be as efficient as a 16 bit C code memory usage.
I realize the issue is complex to analyze. One must formemost consider (1) the compiler and linker and (2) the instruction set. In our case, NIOS code is built using GCC. I have my suspicions that my particular implementation of GCC for NIOS isn't so hot in terms of optimizing for code size. My frame of reference is the IAR toolset on the STM32, which I believe is tops in its class in terms of code footprint efficiency. It's plausible to me that even though 32 bit instruction storage is twice that of a 16 bit machine, that loss might be more than offset by gains in additional addressing modes and a greater number of other instructions, if indeed those advantages really exist.
What is the real world comparison of 16 bit vs. 32 bit code footprint among popular professional grade MCU cores and tools?
Thanks in advance for your discussion.
JJS
--- news://freenews.netfront.net/ - complaints: snipped-for-privacy@netfront.net ---