Skip to content

Latest commit

 

History

History
34 lines (25 loc) · 1.52 KB

0138-2023-08-31.md

File metadata and controls

34 lines (25 loc) · 1.52 KB

31 Aug 2023

Previous journal: Next journal:
0137-2023-08-30.md 0139-2023-09-01.md

raybox-zero

Accomplishments

  • Add a 5th harden_test combo: PL_TARGET_DENSITY=0.65, clock is 25MHz.
  • Made harden_test able to run selective combos:
    • Combos are named 4x2:1 8x2:1 4x2:2, etc.
    • With no argument, it will run them all (but beware: it clears out all other ?x?-combo-?.txt files)
    • Otherwise, the argument is an 'egrep'-compatible regular expression. If it matches any of the combo names, it will run that combo and skip the rest.
    • Examples:
      • ./harden_test.sh -- do all 10.
      • ./harden_test.sh 4x2:5 -- do just that one.
      • ./harden_test.sh .x2:[3-5] -- run combos 3, 4, and 5 in both 4x2 and 8x2 sizes.
  • I tried disabling the tex logic. It made almost no difference to area (though interestingly there was some difference). My hypothesis was that tex and the chain leading to it would be optimised out anyway because it's not actually used.

Next steps

Run harden_test after each of these changes.

  • Make rcp_in a register instead of mux.
  • Cut back states.
  • Try combining TraceStep and TraceTest.
  • Reduce vector ranges. This could be tricky because we'll be working with mixing different sizes.
    • Either that or just make sure the bits we don't care about are set to 0: Optimiser could work out the rest. Don't forget sign bit! Sign-extend instead of 0.

Other:

  • Run on FPGA