def test():
start = Label()
end = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(ebx, ebp.addr + 8),
mov(eax, 1),
start,
cmp(ebx, 1),
jl(end),
mul(ebx),
dec(ebx),
jmp(start),
end,
pop(ebp),
ret(),
)
fun = prog.compile(restype=c_int, argtypes=[c_int])
for i in range(
13
): #factorial of 12 is the maximum integer fitting a 32 bit register
assert fun(i) == factorial(i)
def test():
label = Label('foo')
prog = Program(
mov(eax, 0),
label,
inc(eax),
cmp(eax, 100),
jl(label),
ret(),
)
fun = prog.compile(c_int)
assert fun() == 100
def test():
label = Label('foo')
prog = Program(
mov(eax, 0),
label,
inc(eax),
cmp(eax, 100),
jl(label),
ret(),
)
fun = prog.compile(c_int)
assert fun() == 100
def test():
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr+8),
cmp(eax, 100),
pushfd(),
pop(eax),
pop(ebp),
ret(),
)
fun = prog.compile(c_int, [c_int])
assert fun(99) & CF
assert fun(100) & ZF
assert not (fun(101) & CF or fun(101) & ZF)
def test():
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr + 8),
cmp(eax, 100),
pushfd(),
pop(eax),
pop(ebp),
ret(),
)
fun = prog.compile(c_int, [c_int])
assert fun(99) & CF
assert fun(100) & ZF
assert not (fun(101) & CF or fun(101) & ZF)
def test():
loop = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr + 8),
mov(ebx, ebp.addr + 12),
mov(ecx, ebp.addr + 16),
mov(edx, ebp.addr + 20),
loop,
fld(ebx.addr),
fmul(ecx.addr),
fstp(edx.addr),
add(ebx, 4),
add(ecx, 4),
add(edx, 4),
dec(eax),
cmp(eax, 0),
jg(loop),
pop(ebp),
ret(),
)
fun = prog.compile(argtypes=[
c_int,
POINTER(c_float),
POINTER(c_float),
POINTER(c_float),
])
size = 1000
a = (c_float * size)()
b = (c_float * size)()
c = (c_float * size)()
a[:] = range(0, size)
b[:] = range(size, size * 2)
c[:] = [0] * size
fun(size, a, b, c)
#weirdly off by 1 for higher result numbers
for i in range(size):
assert c[i] == a[i] * b[i]
def test():
loop = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr+8),
mov(ebx, ebp.addr+12),
mov(ecx, ebp.addr+16),
mov(edx, ebp.addr+20),
loop,
fld(ebx.addr),
fmul(ecx.addr),
fstp(edx.addr),
add(ebx, 4),
add(ecx, 4),
add(edx, 4),
dec(eax),
cmp(eax, 0),
jg(loop),
pop(ebp),
ret(),
)
fun = prog.compile(argtypes=[
c_int,
POINTER(c_float),
POINTER(c_float),
POINTER(c_float),
])
size = 1000
a = (c_float*size)()
b = (c_float*size)()
c = (c_float*size)()
a[:] = range(0, size)
b[:] = range(size, size*2)
c[:] = [0] * size
fun(size, a, b, c)
#weirdly off by 1 for higher result numbers
for i in range(size):
assert c[i] == a[i] * b[i]
def test():
start = Label()
end = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(ebx, ebp.addr+8),
mov(eax, 1),
start,
cmp(ebx, 1),
jl(end),
mul(ebx),
dec(ebx),
jmp(start),
end,
pop(ebp),
ret(),
)
fun = prog.compile(restype=c_int, argtypes=[c_int])
for i in range(13): #factorial of 12 is the maximum integer fitting a 32 bit register
assert fun(i) == factorial(i)
def compile():
outer_loop = Label()
inner_loop = Label()
distance_condition = Label()
distance_else = Label()
min_condition = Label()
size_condition = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr + 8), #size
cmp(eax, 2),
jl(size_condition),
mov(ecx, ebp.addr + 12), #array pointer
fld(ebp.addr + 16), #force factor
fld(ebp.addr + 20), #minimum distance
fld(ebp.addr + 24), #maximum distance
sub(eax, 1),
mov(ebx, 4 * 4),
mul(ebx),
add(ecx, eax),
mov(eax, ebp.addr + 12), #array pointer
outer_loop,
mov(ebx, eax),
add(ebx, 4 * 4),
inner_loop,
# x vector
fld(eax.addr + 0 * 4),
fsub(ebx.addr + 0 * 4),
# y vector
fld(eax.addr + 1 * 4),
fsub(ebx.addr + 1 * 4),
# distance
fld(st(1)),
fmul(st(0), st(0)),
fld(st(1)),
fmul(st(0), st(0)),
faddp(),
fsqrt(),
#make sure the distance is bigger then 10
fcomi(st(4)),
fcmovb(st(4)),
#only calculate if the distance is less then 200
fcomi(st(3)),
ja(distance_condition),
#cube distance
fld(st(0)),
fmul(st(1), st(0)),
fmulp(),
#compute force vector
fdiv(st(2), st(0)),
fdivp(),
fld(st(4)),
fmul(st(2), st(0)),
fmulp(),
#accumulate y component
fld(st(0)),
fadd(eax.addr + 3 * 4),
fstp(eax.addr + 3 * 4),
fsubr(ebx.addr + 3 * 4),
fstp(ebx.addr + 3 * 4),
#accumulate x component
fld(st(0)),
fadd(eax.addr + 2 * 4),
fstp(eax.addr + 2 * 4),
fsubr(ebx.addr + 2 * 4),
fstp(ebx.addr + 2 * 4),
jmp(distance_else),
distance_condition,
fstp(st(0)),
fstp(st(0)),
fstp(st(0)),
distance_else,
add(ebx, 4 * 4),
cmp(ebx, ecx),
jbe(inner_loop),
add(eax, 4 * 4),
cmp(eax, ecx),
jb(outer_loop),
#restore the fpu
fstp(st(0)),
fstp(st(0)),
fstp(st(0)),
size_condition,
pop(ebp),
ret(),
)
fun = prog.compile(argtypes=[
c_int,
POINTER(c_float),
c_float,
c_float,
c_float,
])
return fun
def compile():
outer_loop = Label()
inner_loop = Label()
distance_condition = Label()
distance_else = Label()
min_condition = Label()
size_condition = Label()
prog = Program(
push(ebp),
mov(ebp, esp),
mov(eax, ebp.addr+8), #size
cmp(eax, 2),
jl(size_condition),
mov(ecx, ebp.addr+12), #array pointer
fld(ebp.addr+16), #force factor
fld(ebp.addr+20), #minimum distance
fld(ebp.addr+24), #maximum distance
sub(eax, 1),
mov(ebx, 4*4),
mul(ebx),
add(ecx, eax),
mov(eax, ebp.addr+12), #array pointer
outer_loop,
mov(ebx, eax),
add(ebx, 4*4),
inner_loop,
# x vector
fld(eax.addr+0*4),
fsub(ebx.addr+0*4),
# y vector
fld(eax.addr+1*4),
fsub(ebx.addr+1*4),
# distance
fld(st(1)),
fmul(st(0), st(0)),
fld(st(1)),
fmul(st(0), st(0)),
faddp(),
fsqrt(),
#make sure the distance is bigger then 10
fcomi(st(4)),
fcmovb(st(4)),
#only calculate if the distance is less then 200
fcomi(st(3)),
ja(distance_condition),
#cube distance
fld(st(0)),
fmul(st(1), st(0)),
fmulp(),
#compute force vector
fdiv(st(2), st(0)),
fdivp(),
fld(st(4)),
fmul(st(2), st(0)),
fmulp(),
#accumulate y component
fld(st(0)),
fadd(eax.addr+3*4),
fstp(eax.addr+3*4),
fsubr(ebx.addr+3*4),
fstp(ebx.addr+3*4),
#accumulate x component
fld(st(0)),
fadd(eax.addr+2*4),
fstp(eax.addr+2*4),
fsubr(ebx.addr+2*4),
fstp(ebx.addr+2*4),
jmp(distance_else),
distance_condition,
fstp(st(0)),
fstp(st(0)),
fstp(st(0)),
distance_else,
add(ebx, 4*4),
cmp(ebx, ecx),
jbe(inner_loop),
add(eax, 4*4),
cmp(eax, ecx),
jb(outer_loop),
#restore the fpu
fstp(st(0)),
fstp(st(0)),
fstp(st(0)),
size_condition,
pop(ebp),
ret(),
)
fun = prog.compile(argtypes=[
c_int,
POINTER(c_float),
c_float,
c_float,
c_float,
])
return fun
from ctypes import POINTER, c_long
from pyasm import Program
from pyasm.base import Label
from pyasm.instructions import mov, mul, dec, cmp, jg, ret
from pyasm.registers import rax, rdi
def factorial(value):
accum = 1
while value > 1:
accum *= value
value -= 1
return accum
if __name__ == '__main__':
loop = Label()
prog = Program(
mov(rax, 1),
loop,
mul(rax, rdi),
dec(rdi),
cmp(rax, 1),
ja(loop),
ret()
)
fun = prog.compile(restype=c_long, argtypes=[c_long])
for i in xrange(20):
assert fun(i) == factorial(i)