path: root/hdl/core/alu.py

from amaranth import *
from amaranth.sim import Simulator, Settle, Delay
from enum import Enum, unique

from hdl.utils import sim, e2s, cmd, rand_bits_mix
from hdl.lib.in_out_buff import InOutBuff
from hdl.config import NUM_RAND_TESTS

@unique
class AluOpCodes(Enum):
    add = 0
    addc = 1
    sub = 2
    subc = 3
    bit_and = 4
    bit_or = 5
    bit_xor = 6
    bit_nor = 7
    lleft = 8
    lright = 9
    aright = 10
    multul = 11     # low 32 bits of unsigned multiplication
    multuh = 12     # high 32 bits of unsigned multiplication
    multsl = 13     # low 32 bits of signed multiplication
    multsh = 14     # high 32 bits of signed multiplication

@unique
class ALUFlags(Enum):
    zero = 0
    carry = 1
    overflow = 2
    negative = 3

class ALU(Elaboratable):

    def __init__(self, **kargs):
        self.in1 = Signal(32, reset_less=True)
        self.in2 = Signal(32, reset_less=True)
        self.c_in = Signal(1)
        self.op = Signal(e2s(AluOpCodes), reset_less=True)

        self.tmp = Signal(33, reset_less=True)

        self.c_out = Signal(1, reset_less=True)
        self.overflow = Signal(1, reset_less=True)
        self.zero = Signal(1, reset_less=True)
        self.neg = Signal(1, reset_less=True)

        self.alu_flags = Signal(len(ALUFlags), reset_less=True) # alu flags is one hot
        self.out = Signal(32, reset_less=True)

        self.sim = kargs.get('sim', False)

        ports_in = [self.in1, self.in2, self.op, self.c_in]
        ports_out = [self.c_in, self.out, self.alu_flags]
        self.ports = {'in': ports_in, 'out': ports_out}

    def elaborate(self, platform=None):
        m = Module()

        with m.Switch(self.op):
            with m.Case(AluOpCodes.add.value):
                m.d.comb += self.tmp.eq(self.in1 + self.in2)

            with m.Case(AluOpCodes.addc.value):
                m.d.comb += self.tmp.eq(self.in1 + self.in2 + self.c_in)

            with m.Case(AluOpCodes.sub.value):
                m.d.comb += self.tmp.eq(self.in1 - self.in2)

            with m.Case(AluOpCodes.subc.value):
                m.d.comb += self.tmp.eq(self.in1 + ~self.in2 + self.c_in)

            with m.Case(AluOpCodes.bit_and.value):
                m.d.comb += self.tmp.eq(Cat(self.in1 & self.in2, 0))

            with m.Case(AluOpCodes.bit_or.value):
                m.d.comb += self.tmp.eq(Cat(self.in1 | self.in2, 0))

            with m.Case(AluOpCodes.bit_xor.value):
                m.d.comb += self.tmp.eq(Cat(self.in1 ^ self.in2, 0))
            
            with m.Case(AluOpCodes.bit_nor.value):
                m.d.comb += self.tmp.eq(Cat(~(self.in1 | self.in2), 0))

            with m.Case(AluOpCodes.lleft.value):
                m.d.comb += self.tmp.eq(Cat(self.in1, 0) << self.in2[0:5])

            with m.Case(AluOpCodes.lright.value):
                tmp2 = Signal(33)
                m.d.comb += tmp2.eq(Cat(0, self.in1) >> self.in2)
                m.d.comb += self.tmp.eq(Cat(tmp2[1:33], tmp2[0]))   # move shifted bit to carry bit

            with m.Case(AluOpCodes.aright.value):
                tmp2 = Signal(33)
                m.d.comb += tmp2.eq(Cat(0, self.in1).as_signed() >> self.in2)
                m.d.comb += self.tmp.eq(Cat(tmp2[1:33], tmp2[0]))   # move shifted bit to carry bit

            with m.Case(AluOpCodes.multul.value):
                m.d.comb += self.tmp.eq(Cat((self.in1 * self.in2)[:32], 0))

            with m.Case(AluOpCodes.multuh.value):
                m.d.comb += self.tmp.eq(Cat((self.in1 * self.in2)[32:], 0))
            
            with m.Case(AluOpCodes.multsl.value):
                m.d.comb += self.tmp.eq(Cat((self.in1.as_signed() * self.in2.as_signed())[:32], 0))

            with m.Case(AluOpCodes.multsh.value):
                m.d.comb += self.tmp.eq(Cat((self.in1.as_signed() * self.in2.as_signed())[32:], 0))
            

            # bad juju, 
            # TODO: come back and check this will work
            # with m.Case(AluOpCodes.udiv.value):
            #     m.d.comb += self.tmp.eq(Cat(self.in1 // self.in2, 0))
            
            # with m.Case(AluOpCodes.sdiv.value):
            #     m.d.comb += self.tmp.eq(self.in1.as_signed() // self.in2.as_signed())   # for some reason I have not confirmed, signed div can yield a 33 bit number, acording to amaranth

            with m.Case():
                m.d.comb += self.tmp.eq(0)
        
        m.d.comb += self.c_out.eq(self.tmp[32])
        m.d.comb += self.overflow.eq(self.tmp[32] ^ self.tmp[31])
        m.d.comb += self.neg.eq(self.out[31])
        m.d.comb += self.zero.eq(self.out == 0)

        m.d.comb += self.alu_flags[ALUFlags.zero.value].eq(self.zero)
        m.d.comb += self.alu_flags[ALUFlags.carry.value].eq(self.c_out)
        m.d.comb += self.alu_flags[ALUFlags.overflow.value].eq(self.overflow)
        m.d.comb += self.alu_flags[ALUFlags.negative.value].eq(self.neg)
        m.d.comb += self.out.eq(self.tmp[0:32])
        
        return m

def sub_proc(dut, val1, val2, c_in=0):
    yield dut.in1.eq(val1)
    yield dut.in2.eq(val2)
    yield dut.c_in.eq(c_in)
    yield Settle()

# test addition
def test_alu_add():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.add.value)
        yield from sub_proc(dut, 27, 13)
        out = yield dut.out
        assert 27 + 13 == (out), f'ERROR: {out} != {27 + 13}'
    sim(dut, proc, sync=False)

# test addition with carry
def test_alu_addc():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.addc.value)
        yield from sub_proc(dut, 11, 43, 1)
        out = yield dut.out.as_signed()
        assert 11 + 43 + 1 == out, f'ERROR: {out} != {11 + 43 + 1}'
    sim(dut, proc, sync=False)

# test subtraction
def test_alu_sub():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.sub.value)
        yield from sub_proc(dut, 25, 13)
        out = yield dut.out
        assert 25 - 13 == out, f'ERROR: {out} != {25 - 13}'
    sim(dut, proc, sync=False)
        
# test subtraction with carry
def test_alu_subc_0():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.subc.value)
        yield from sub_proc(dut, 25, -13, 0)
        out = yield dut.out.as_signed()
        assert 25 + 13 -1 +0 == out, f'ERROR: {out} != {25 + 13 -1 +0}'
    sim(dut, proc, sync=False)

# test subtraction with carry
def test_alu_subc_1():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.subc.value)
        yield from sub_proc(dut, 25, -13, 1)
        out = yield dut.out.as_signed()
        assert 25 + 13 -1 +1 == out, f'ERROR: {out} != {25 + 13 -1 +1}'
    sim(dut, proc, sync=False)

# test binary and
def test_alu_and():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.bit_and.value)
        yield from sub_proc(dut, 0b10101011, 0b01010101)
        out = yield dut.out
        assert 0b00000001 == out, f'ERROR: {out} != {0b00000001}'
    sim(dut, proc, sync=False)

# test binary or
def test_alu_or():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.bit_or.value)
        yield from sub_proc(dut, 0b10101011, 0b01000101)
        out = yield dut.out
        assert 0b11101111 == out, f'ERROR: {out} != {0b11101111}'
    sim(dut, proc, sync=False)

# test binary nor
def test_alu_nor():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.bit_nor.value)
        yield from sub_proc(dut, 0b10001011, 0b01000101)
        out = yield dut.out
        assert 0b11111111111111111111111100110000 == out, f'ERROR: {bin(out)} != {bin(0b11111111111111111111111100110000)}'
    sim(dut, proc, sync=False)

# test binary xor
def test_alu_xor():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.bit_xor.value)
        yield from sub_proc(dut, 0b10001011, 0b01000101)
        out = yield dut.out
        assert 0b11001110 == out, f'ERROR: {out} != {0b11001110}'
    sim(dut, proc, sync=False)

# test logical shift left
def test_alu_logic_shift_left():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.lleft.value)
        yield from sub_proc(dut, 0b10001011, 25) # shift left by 5
        out = yield dut.out
        assert 0b00010110000000000000000000000000 == out, f'ERROR: {bin(out)} != {bin(0b00010110000000000000000000000000)}'
        out = yield dut.c_out
        assert 1 == out, f'ERROR: {out} != {1}'
    sim(dut, proc, sync=False)

# test logical shift right
def test_alu_logic_shift_right():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.lright.value)
        yield from sub_proc(dut, 0b10001011, 4) # shift right by 5
        out = yield dut.out
        assert 0b1000 == out, f'ERROR: {bin(out)} != {bin(0b1000)}'
        out = yield dut.c_out
        assert 1 == out, f'ERROR: {out} != {1}'
    sim(dut, proc, sync=False)

# test arithmetic shift right
def test_alu_arith_shift_right():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.aright.value)
        yield from sub_proc(dut, 0x80001234, 4) # shift right by 4
        out = yield dut.out
        assert 0xF8000123 == out, f'ERROR: {out} != {0xF8000123}'
        out = yield dut.c_out
        assert 0 == out, f'ERROR: {out} != {0}'
    sim(dut, proc, sync=False)

# test low unsigned multiply
def test_alu_mul_low_u(tests=NUM_RAND_TESTS):
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.multul.value)
        yield dut.c_in.eq(0)
        
        for _ in range(tests):
            in1 = rand_bits_mix(32, sus='u')
            in2 = rand_bits_mix(32, sus='u')
            yield dut.in1.eq(in1)
            yield dut.in2.eq(in2)
            yield Settle()
            expected = (in1 * in2) & 0xFFFFFFFF
            assert (yield dut.out) == expected, f"mul_low_u failed: in1={hex(in1)}, in2={hex(in2)}, out={hex((yield dut.out))}, expected={hex(expected)}"

    sim(dut, proc, sync=False)

# test high unsigned multiply
def test_alu_mul_high_u(tests=NUM_RAND_TESTS):
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.multuh.value)
        yield dut.c_in.eq(0)
        
        for _ in range(tests):
            in1 = rand_bits_mix(32, sus='u')
            in2 = rand_bits_mix(32, sus='u')
            yield dut.in1.eq(in1)
            yield dut.in2.eq(in2)
            yield Settle()
            expected = ((in1 * in2) >> 32) & 0xFFFFFFFF
            assert (yield dut.out) == expected, f"mul_high_u failed: in1={hex(in1)}, in2={hex(in2)}, out={hex((yield dut.out))}, expected={hex(expected)}"

    sim(dut, proc, sync=False)

# test low signed multiply
def test_alu_mul_low_s(tests=NUM_RAND_TESTS):
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.multsl.value)
        yield dut.c_in.eq(0)
        
        for _ in range(tests):
            in1 = rand_bits_mix(32, sus='s')
            in2 = rand_bits_mix(32, sus='s')
            yield dut.in1.eq(in1)
            yield dut.in2.eq(in2)
            yield Settle()
            expected = (in1 * in2) & 0xFFFFFFFF
            assert (yield dut.out) == expected, f"mul_low_s failed: in1={hex(in1)}, in2={hex(in2)}, out={hex((yield dut.out))}, expected={hex(expected)}"

    sim(dut, proc, sync=False)

# test high signed multiply
def test_alu_mul_high_s(tests=NUM_RAND_TESTS):
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.multsh.value)
        yield dut.c_in.eq(0)
        
        for _ in range(tests):
            in1 = rand_bits_mix(32, sus='s')
            in2 = rand_bits_mix(32, sus='s')
            yield dut.in1.eq(in1)
            yield dut.in2.eq(in2)
            yield Settle()
            expected = ((in1 * in2) >> 32) & 0xFFFFFFFF
            assert (yield dut.out) == expected, f"mul_high_s failed: in1={hex(in1)}, in2={hex(in2)}, out={hex((yield dut.out))}, expected={hex(expected)}"

    sim(dut, proc, sync=False)


# test unsigned overflow
def test_alu_unsigned_overflow():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.add.value)
        yield from sub_proc(dut, 0xFFFFFFFF, 1) # add 1 to 0xFFFFFFFF
        out = yield dut.overflow
        assert out == 1, f'ERROR: {out} != {1}'
        out = yield dut.c_out
        assert out == 1, f'ERROR: {out} != {1}'
    sim(dut, proc, sync=False)

# test unsigned underflow
def test_alu_unsigned_underflow():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.add.value)
        yield from sub_proc(dut, 0, -1) # subtract 1 from 0
        out = yield dut.overflow
        assert out == 1, f'ERROR: {out} != {1}'
        out = yield dut.c_out
        assert out == 0, f'ERROR: {out} != {0}'
    sim(dut, proc, sync=False)

# test zero
def test_alu_zero_0():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.add.value)
        yield from sub_proc(dut, 0, 1) # add 0 to 0
        out = yield dut.zero
        assert out == 0, f'ERROR: {out} != {0}'
    sim(dut, proc, sync=False)

# test zero
def test_alu_zero_1():
    dut = ALU(sim=True)
    def proc():
        yield dut.op.eq(AluOpCodes.add.value)
        yield from sub_proc(dut, 0, 0) # add 0 to 0
        out = yield dut.zero
        assert out == 1, f'ERROR: {out} != {1}'
    sim(dut, proc, sync=False)

       


if __name__ == '__main__':
    hdl = InOutBuff(ALU())
    cmd(hdl)