pwyq / bme351 Goto Github PK

course chap. 6

possible optimization techniques

Replication and Sharing

• mux pushing
• common subexpression elimination
• computation replication
• input trimming

Stuff to fix

• "Blue area is large, means you are having edges that shouldn't"
  • see picture below (generated by diff_ted_to_bmp)

Stuff to improve

• "Bit width is way too large for the last stage." (He meant those 16 downto 0 ones
  • see issue #3

Eric Commits

Another comments Eric made were that

• we're not having any skew (see the blue image above), which is a good thing
• he has never use resize() before, and suggest we use leading 0 and & to do resizing.
• don't worry about direction right now, try to do the computation manually for the first few rows to see what your problem(s) is

@ali-abdullah @mosobase @sepehrafla

• Draw a new DFD based on following hand-written DFD
  • suggested tool: draw.io
• Requirements:
  • same quality as our dfd-prelim (dfd-name.pdf under kirsch_proj/)
  • label stages with distinct color
  • label each operator/component with its name (eg max1, r1)
  • clear data flow (eg arrow, lining)
  • Overlapping stages must have same alignment to clearly show clock cycles
  • same box size for each operator (max, add, sub, shift comp)
    

we introduced a new clock cycle to compensate the final output in #7 ... need to fix it

spec picture

our picture

• 3-page report