06 / 08 · What we check

Get the design that fits, not just the one that compiles.

It's easy to write RTL that technically works but burns far more LUTs, flip-flops, or BRAM than it needs to. We look for patterns that inflate resource usage and suggest concrete ways to bring it down.

01
LUT usage patterns. Wide muxes, decoders, or comparators that could be restructured more efficiently.
02
Flip-flop count. Redundant pipeline stages or state that could be shared or reduced.
03
BRAM inference. Memory structures that fail to infer block RAM and fall back to LUT-based storage.
04
Replication risk. Fan-out driven register/logic duplication that inflates area unnecessarily.
05
Resource sharing. Arithmetic units that could be time-multiplexed instead of instantiated repeatedly.

Sample findingResource Optimization

// 256-entry lookup, written as a case statement
always @(*) case(index)
  8'd0: value = 8'h1A;
  8'd1: value = 8'h3F;
  // ... 254 more cases, not inferring BRAM
endcase

Large case-statement LUT can be replaced with an inferred block RAMSuggestion

As written, this table is likely to synthesize into distributed LUT logic rather than a single BRAM, using far more fabric than needed.

Why it matters