[CombToSynth] Add a lowering for mod/div where rhs is constant

[okekayode]

Concerns #10569

In this PR we add lowerings for comb.divu, comb.modu, comb.divs, and comb.mods in the case that the RHS is a constant value. For the division operators , we utilise UnsignedDivisionByConstantInfo and SignedDivisionByConstantInfo to replace the operation with magic multiplication, shifts, and adds. For modulo, we reuse the division lowering via x % k = x - ((x / k) * k) = remainder

Assisted-by: claude: sonnet 4.6

[circt-bot]

Results of circt-tests run for 759682f compared to results for 7aa41cc: no change to test results.

[circt-bot]

Results of circt-tests run for 1767468 compared to results for 7aa41cc: no change to test results.

[circt-bot]

Results of circt-tests run for 4bb59c4 compared to results for 7aa41cc: no change to test results.

[uenoku]

This fails LEC. I enumerated all possible inputs up to i8, but mods %lhs, -1 was only failure, so it's mostly correct. I think this mods %lhs, -1 is always zero, but apparently the result is different.

hw.module @const_divmod_mods_neg1_i3(in %lhs: i3, out out: i3) {
  %c_neg1_i3 = hw.constant -1 : i3
 
  %0 = comb.mods %lhs, %c_neg1_i3 : i3
  hw.output %0 : i3
}

[okekayode]

This fails LEC. I enumerated all possible inputs up to i8, but mods %lhs, -1 was only failure, so it's mostly correct. I think this mods %lhs, -1 is always zero, but apparently the result is different.

hw.module @const_divmod_mods_neg1_i3(in %lhs: i3, out out: i3) {
  %c_neg1_i3 = hw.constant -1 : i3
 
  %0 = comb.mods %lhs, %c_neg1_i3 : i3
  hw.output %0 : i3
}

oops, let me add that case right now, yes it is always zero, seems i only considered the 0 case in mods

[uenoku]

It's great that there is only one failure 👍 Great work!

[circt-bot]

Results of circt-tests run for 5a47792 compared to results for 7aa41cc: no change to test results.