[opt] Simplify modulo POT (Stage 1)

[xumingkuan]

Related issue = #944

Test case:

import taichi as ti

ti.init(print_ir=True)


@ti.kernel
def func(i: ti.u32):
    if i % 2 == 1:
        print(i)


func(1)

Before:

    <i32> $1 = const [1]
    <u32> $2 = arg[0]
    <u32> $3 = const [1]
    <u32> $4 = bit_sar $2 $3
    <u32> $5 = bit_shl $4 $3
    <u32> $6 = sub $2 $5
    <i32> $7 = cmp_eq $6 $3
    <i32> $8 = bit_and $7 $1
    $9 : if $8 {
      print $2, "\n"
    }

After:

    <i32> $1 = const [1]
    <u32> $2 = arg[0]
    <u32> $3 = const [1]
    <u32> $4 = bit_and $2 $3
    <u32> $5 = sub $2 $4
    <i32> $6 = cmp_eq $5 $3
    <i32> $7 = bit_and $6 $1
    $8 : if $7 {
      print $2, "\n"
    }

[Click here for the format server]

---

[k-ye]

Great LGTM!