I've been doing some benchmarks to compare APD's performance vs other approaches we're considering (cosmos/cosmos-sdk#7772).

It seems that Quo operations perform pretty poorly compare to Add, Mul, etc.:

Benchmark/1_Ops:_apd.Decimal128.Add
Benchmark/1_Ops:_apd.Decimal128.Add-16         	 5285344	       218 ns/op
Benchmark/1_Ops:_apd.Decimal128.Sub
Benchmark/1_Ops:_apd.Decimal128.Sub-16         	 5665453	       208 ns/op
Benchmark/1_Ops:_apd.Decimal128.Mul
Benchmark/1_Ops:_apd.Decimal128.Mul-16         	 9138602	       138 ns/op
Benchmark/1_Ops:_apd.Decimal128.Quo
Benchmark/1_Ops:_apd.Decimal128.Quo-16         	  162537	      7790 ns/op

This benchmark is basically just perform a single operation on two random floating point numbers (different ones each round). I'm using a Context configured with decimal128 parameters (i.e. Precision = 34, etc.)

Is Quo inherently expected to be 40x slower than other operations or could there be a performance bottleneck?