[math/big] Rename internal_int_shl_digit to _private_int_shl_leg.

Same for the SHR variant. These are pure implementation details to shift by a leg/word at a time.
Prevent accidental usage.
This commit is contained in:
Jeroen van Rijn
2021-12-11 15:22:24 +01:00
parent 84b84d9f7d
commit 938744b276
5 changed files with 116 additions and 149 deletions

View File

@@ -2648,7 +2648,7 @@ internal_int_shrmod :: proc(quotient, remainder, numerator: ^Int, bits: int, all
Shift by as many digits in the bit count.
*/
if bits >= _DIGIT_BITS {
internal_shr_digit(quotient, bits / _DIGIT_BITS) or_return
_private_int_shr_leg(quotient, bits / _DIGIT_BITS) or_return
}
/*
@@ -2687,37 +2687,6 @@ internal_int_shr :: proc(dest, source: ^Int, bits: int, allocator := context.all
}
internal_shr :: proc { internal_int_shr, }
/*
Shift right by `digits` * _DIGIT_BITS bits.
*/
internal_int_shr_digit :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
context.allocator = allocator
if digits <= 0 { return nil }
/*
If digits > used simply zero and return.
*/
if digits > quotient.used { return internal_zero(quotient) }
/*
Much like `int_shl_digit`, this is implemented using a sliding window,
except the window goes the other way around.
b-2 | b-1 | b0 | b1 | b2 | ... | bb | ---->
/\ | ---->
\-------------------/ ---->
*/
#no_bounds_check for x := 0; x < (quotient.used - digits); x += 1 {
quotient.digit[x] = quotient.digit[x + digits]
}
quotient.used -= digits
internal_zero_unused(quotient)
return internal_clamp(quotient)
}
internal_shr_digit :: proc { internal_int_shr_digit, }
/*
Shift right by a certain bit count with sign extension.
*/
@@ -2756,7 +2725,7 @@ internal_int_shl :: proc(dest, src: ^Int, bits: int, allocator := context.alloca
Shift by as many digits in the bit count as we have.
*/
if bits >= _DIGIT_BITS {
internal_shl_digit(dest, bits / _DIGIT_BITS) or_return
_private_int_shl_leg(dest, bits / _DIGIT_BITS) or_return
}
/*
@@ -2786,45 +2755,6 @@ internal_int_shl :: proc(dest, src: ^Int, bits: int, allocator := context.alloca
}
internal_shl :: proc { internal_int_shl, }
/*
Shift left by `digits` * _DIGIT_BITS bits.
*/
internal_int_shl_digit :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
context.allocator = allocator
if digits <= 0 { return nil }
/*
No need to shift a zero.
*/
if #force_inline internal_is_zero(quotient) {
return nil
}
/*
Resize `quotient` to accomodate extra digits.
*/
#force_inline internal_grow(quotient, quotient.used + digits) or_return
/*
Increment the used by the shift amount then copy upwards.
*/
/*
Much like `int_shr_digit`, this is implemented using a sliding window,
except the window goes the other way around.
*/
#no_bounds_check for x := quotient.used; x > 0; x -= 1 {
quotient.digit[x+digits-1] = quotient.digit[x-1]
}
quotient.used += digits
mem.zero_slice(quotient.digit[:digits])
return nil
}
internal_shl_digit :: proc { internal_int_shl_digit, }
/*
Count bits in an `Int`.
Assumes `a` not to be `nil` and to have been initialized.

View File

@@ -86,21 +86,6 @@ int_shr :: proc(dest, source: ^Int, bits: int, allocator := context.allocator) -
}
shr :: proc { int_shr, }
/*
Shift right by `digits` * _DIGIT_BITS bits.
*/
int_shr_digit :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
/*
Check that `quotient` is usable.
*/
assert_if_nil(quotient)
context.allocator = allocator
internal_clear_if_uninitialized(quotient) or_return
return #force_inline internal_int_shr_digit(quotient, digits)
}
shr_digit :: proc { int_shr_digit, }
/*
Shift right by a certain bit count with sign extension.
*/
@@ -124,20 +109,4 @@ int_shl :: proc(dest, src: ^Int, bits: int, allocator := context.allocator) -> (
internal_clear_if_uninitialized(dest, src) or_return
return #force_inline internal_int_shl(dest, src, bits)
}
shl :: proc { int_shl, }
/*
Shift left by `digits` * _DIGIT_BITS bits.
*/
int_shl_digit :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
/*
Check that `quotient` is usable.
*/
assert_if_nil(quotient)
context.allocator = allocator
internal_clear_if_uninitialized(quotient) or_return
return #force_inline internal_int_shl_digit(quotient, digits)
}
shl_digit :: proc { int_shl_digit, };
shl :: proc { int_shl, }

View File

@@ -211,12 +211,12 @@ _private_int_mul_toom :: proc(dest, a, b: ^Int, allocator := context.allocator)
/*
P = b1*x^4+ S2*x^3+ S1*x^2+ a1*x + a0;
*/
internal_shl_digit(b1, 4 * B) or_return
internal_shl_digit(S2, 3 * B) or_return
_private_int_shl_leg(b1, 4 * B) or_return
_private_int_shl_leg(S2, 3 * B) or_return
internal_add(b1, b1, S2) or_return
internal_shl_digit(S1, 2 * B) or_return
_private_int_shl_leg(S1, 2 * B) or_return
internal_add(b1, b1, S1) or_return
internal_shl_digit(a1, 1 * B) or_return
_private_int_shl_leg(a1, 1 * B) or_return
internal_add(b1, b1, a1) or_return
internal_add(dest, b1, a0) or_return
@@ -317,8 +317,8 @@ _private_int_mul_karatsuba :: proc(dest, a, b: ^Int, allocator := context.alloca
/*
shift by B.
*/
internal_shl_digit(t1, B) or_return /* t1 = (x0y0 + x1y1 - (x1-x0)*(y1-y0))<<B */
internal_shl_digit(x1y1, B * 2) or_return /* x1y1 = x1y1 << 2*B */
_private_int_shl_leg(t1, B) or_return /* t1 = (x0y0 + x1y1 - (x1-x0)*(y1-y0))<<B */
_private_int_shl_leg(x1y1, B * 2) or_return /* x1y1 = x1y1 << 2*B */
internal_add(t1, x0y0, t1) or_return /* t1 = x0y0 + t1 */
internal_add(dest, t1, x1y1) or_return /* t1 = x0y0 + t1 + x1y1 */
@@ -588,7 +588,7 @@ _private_int_mul_balance :: proc(dest, a, b: ^Int, allocator := context.allocato
/*
Shift `tmp` to the correct position.
*/
internal_shl_digit(tmp, b_size * i) or_return
_private_int_shl_leg(tmp, b_size * i) or_return
/*
Add to output. No carry needed.
@@ -606,7 +606,7 @@ _private_int_mul_balance :: proc(dest, a, b: ^Int, allocator := context.allocato
internal_clamp(a0)
internal_mul(tmp, a0, b) or_return
internal_shl_digit(tmp, b_size * i) or_return
_private_int_shl_leg(tmp, b_size * i) or_return
internal_add(r, r, tmp) or_return
}
@@ -840,8 +840,8 @@ _private_int_sqr_karatsuba :: proc(dest, src: ^Int, allocator := context.allocat
/*
Shift by B.
*/
internal_shl_digit(t1, B) or_return
internal_shl_digit(x1x1, B * 2) or_return
_private_int_shl_leg(t1, B) or_return
_private_int_shl_leg(x1x1, B * 2) or_return
internal_add(t1, t1, x0x0) or_return
internal_add(dest, t1, x1x1) or_return
@@ -942,10 +942,10 @@ _private_int_sqr_toom :: proc(dest, src: ^Int, allocator := context.allocator) -
internal_sub(dest, dest, S0) or_return
/** \\P = S4*x^4 + S3*x^3 + S2*x^2 + S1*x + S0; */
/** P = a2*x^4 + a1*x^3 + b*x^2 + a0*x + S0; */
internal_shl_digit( a2, 4 * B) or_return
internal_shl_digit( a1, 3 * B) or_return
internal_shl_digit(dest, 2 * B) or_return
internal_shl_digit( a0, 1 * B) or_return
_private_int_shl_leg( a2, 4 * B) or_return
_private_int_shl_leg( a1, 3 * B) or_return
_private_int_shl_leg(dest, 2 * B) or_return
_private_int_shl_leg( a0, 1 * B) or_return
internal_add(a2, a2, a1) or_return
internal_add(dest, dest, a2) or_return
@@ -1069,7 +1069,7 @@ _private_int_div_school :: proc(quotient, remainder, numerator, denominator: ^In
y = y*b**{n-t}
*/
internal_shl_digit(y, n - t) or_return
_private_int_shl_leg(y, n - t) or_return
gte := internal_gte(x, y)
for gte {
@@ -1081,7 +1081,7 @@ _private_int_div_school :: proc(quotient, remainder, numerator, denominator: ^In
/*
Reset y by shifting it back down.
*/
internal_shr_digit(y, n - t)
_private_int_shr_leg(y, n - t)
/*
Step 3. for i from n down to (t + 1).
@@ -1146,7 +1146,7 @@ _private_int_div_school :: proc(quotient, remainder, numerator, denominator: ^In
Step 3.3 x = x - q{i-t-1} * y * b**{i-t-1}
*/
int_mul_digit(t1, y, q.digit[(i - t) - 1]) or_return
internal_shl_digit(t1, (i - t) - 1) or_return
_private_int_shl_leg(t1, (i - t) - 1) or_return
internal_sub(x, x, t1) or_return
/*
@@ -1154,7 +1154,7 @@ _private_int_div_school :: proc(quotient, remainder, numerator, denominator: ^In
*/
if x.sign == .Negative {
internal_copy(t1, y) or_return
internal_shl_digit(t1, (i - t) - 1) or_return
_private_int_shl_leg(t1, (i - t) - 1) or_return
internal_add(x, x, t1) or_return
q.digit[(i - t) - 1] = (q.digit[(i - t) - 1] - 1) & _MASK
@@ -1220,7 +1220,7 @@ _private_div_recursion :: proc(quotient, remainder, a, b: ^Int, allocator := con
/*
A1 = (R1 * beta^(2k)) + (A % beta^(2k)) - (Q1 * B0 * beta^k)
*/
internal_shl_digit(R1, 2 * k) or_return
_private_int_shl_leg(R1, 2 * k) or_return
internal_add(A1, R1, t) or_return
internal_mul(t, Q1, B0) or_return
@@ -1246,7 +1246,7 @@ _private_div_recursion :: proc(quotient, remainder, a, b: ^Int, allocator := con
/*
A2 = (R0*beta^k) + (A1 % beta^k) - (Q0*B0)
*/
internal_shl_digit(R0, k) or_return
_private_int_shl_leg(R0, k) or_return
internal_add(A2, R0, t) or_return
internal_mul(t, Q0, B0) or_return
internal_sub(A2, A2, t) or_return
@@ -1262,7 +1262,7 @@ _private_div_recursion :: proc(quotient, remainder, a, b: ^Int, allocator := con
/*
Return q = (Q1*beta^k) + Q0, r = A2.
*/
internal_shl_digit(Q1, k) or_return
_private_int_shl_leg(Q1, k) or_return
internal_add(quotient, Q1, Q0) or_return
return internal_copy(remainder, A2)
@@ -1923,7 +1923,7 @@ _private_int_montgomery_reduce :: proc(x, n: ^Int, rho: DIGIT, allocator := cont
x = x/b**n.used.
*/
internal_clamp(x)
internal_shr_digit(x, n.used)
_private_int_shr_leg(x, n.used)
/*
if x >= n then x = x - n
@@ -2026,7 +2026,7 @@ _private_int_reduce :: proc(x, m, mu: ^Int, allocator := context.allocator) -> (
/*
q1 = x / b**(k-1)
*/
internal_shr_digit(q, um - 1)
_private_int_shr_leg(q, um - 1)
/*
According to HAC this optimization is ok.
@@ -2040,7 +2040,7 @@ _private_int_reduce :: proc(x, m, mu: ^Int, allocator := context.allocator) -> (
/*
q3 = q2 / b**(k+1)
*/
internal_shr_digit(q, um + 1)
_private_int_shr_leg(q, um + 1)
/*
x = x mod b**(k+1), quick (no division)
@@ -2062,7 +2062,7 @@ _private_int_reduce :: proc(x, m, mu: ^Int, allocator := context.allocator) -> (
*/
if internal_is_negative(x) {
internal_set(q, 1) or_return
internal_shl_digit(q, um + 1) or_return
_private_int_shl_leg(q, um + 1) or_return
internal_add(x, x, q) or_return
}
@@ -3192,6 +3192,74 @@ _private_copy_digits :: proc(dest, src: ^Int, digits: int, offset := int(0)) ->
return nil
}
/*
Shift left by `digits` * _DIGIT_BITS bits.
*/
_private_int_shl_leg :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
context.allocator = allocator
if digits <= 0 { return nil }
/*
No need to shift a zero.
*/
if #force_inline internal_is_zero(quotient) {
return nil
}
/*
Resize `quotient` to accomodate extra digits.
*/
#force_inline internal_grow(quotient, quotient.used + digits) or_return
/*
Increment the used by the shift amount then copy upwards.
*/
/*
Much like `_private_int_shr_leg`, this is implemented using a sliding window,
except the window goes the other way around.
*/
#no_bounds_check for x := quotient.used; x > 0; x -= 1 {
quotient.digit[x+digits-1] = quotient.digit[x-1]
}
quotient.used += digits
mem.zero_slice(quotient.digit[:digits])
return nil
}
/*
Shift right by `digits` * _DIGIT_BITS bits.
*/
_private_int_shr_leg :: proc(quotient: ^Int, digits: int, allocator := context.allocator) -> (err: Error) {
context.allocator = allocator
if digits <= 0 { return nil }
/*
If digits > used simply zero and return.
*/
if digits > quotient.used { return internal_zero(quotient) }
/*
Much like `int_shl_digit`, this is implemented using a sliding window,
except the window goes the other way around.
b-2 | b-1 | b0 | b1 | b2 | ... | bb | ---->
/\ | ---->
\-------------------/ ---->
*/
#no_bounds_check for x := 0; x < (quotient.used - digits); x += 1 {
quotient.digit[x] = quotient.digit[x + digits]
}
quotient.used -= digits
internal_zero_unused(quotient)
return internal_clamp(quotient)
}
/*
======================== End of private procedures =======================

View File

@@ -208,7 +208,7 @@ print_to_buffer :: proc(val: ^big.Int) -> cstring {
/*
dest = shr_digit(src, digits)
*/
@export test_shr_digit :: proc "c" (source: cstring, digits: int) -> (res: PyRes) {
@export test_shr_leg :: proc "c" (source: cstring, digits: int) -> (res: PyRes) {
context = runtime.default_context()
err: big.Error
@@ -216,7 +216,7 @@ print_to_buffer :: proc(val: ^big.Int) -> cstring {
defer big.internal_destroy(src)
if err = big.atoi(src, string(source), 16); err != nil { return PyRes{res=":shr_digit:atoi(src):", err=err} }
if err = #force_inline big.internal_shr_digit(src, digits); err != nil { return PyRes{res=":shr_digit:shr_digit(src):", err=err} }
if err = #force_inline big._private_int_shr_leg(src, digits); err != nil { return PyRes{res=":shr_digit:shr_digit(src):", err=err} }
r := print_to_buffer(src)
return PyRes{res = r, err = nil}
@@ -225,7 +225,7 @@ print_to_buffer :: proc(val: ^big.Int) -> cstring {
/*
dest = shl_digit(src, digits)
*/
@export test_shl_digit :: proc "c" (source: cstring, digits: int) -> (res: PyRes) {
@export test_shl_leg :: proc "c" (source: cstring, digits: int) -> (res: PyRes) {
context = runtime.default_context()
err: big.Error
@@ -233,7 +233,7 @@ print_to_buffer :: proc(val: ^big.Int) -> cstring {
defer big.internal_destroy(src)
if err = big.atoi(src, string(source), 16); err != nil { return PyRes{res=":shl_digit:atoi(src):", err=err} }
if err = #force_inline big.internal_shl_digit(src, digits); err != nil { return PyRes{res=":shl_digit:shr_digit(src):", err=err} }
if err = #force_inline big._private_int_shl_leg(src, digits); err != nil { return PyRes{res=":shl_digit:shr_digit(src):", err=err} }
r := print_to_buffer(src)
return PyRes{res = r, err = nil}

View File

@@ -187,8 +187,8 @@ int_sqrt = load(l.test_sqrt, [c_char_p ], Res)
int_root_n = load(l.test_root_n, [c_char_p, c_longlong], Res)
# Logical operations
int_shl_digit = load(l.test_shl_digit, [c_char_p, c_longlong], Res)
int_shr_digit = load(l.test_shr_digit, [c_char_p, c_longlong], Res)
int_shl_leg = load(l.test_shl_leg, [c_char_p, c_longlong], Res)
int_shr_leg = load(l.test_shr_leg, [c_char_p, c_longlong], Res)
int_shl = load(l.test_shl, [c_char_p, c_longlong], Res)
int_shr = load(l.test_shr, [c_char_p, c_longlong], Res)
int_shr_signed = load(l.test_shr_signed, [c_char_p, c_longlong], Res)
@@ -402,17 +402,17 @@ def test_root_n(number = 0, root = 0, expected_error = Error.Okay):
return test("test_root_n", res, [number, root], expected_error, expected_result)
def test_shl_digit(a = 0, digits = 0, expected_error = Error.Okay):
def test_shl_leg(a = 0, digits = 0, expected_error = Error.Okay):
args = [arg_to_odin(a), digits]
res = int_shl_digit(*args)
res = int_shl_leg(*args)
expected_result = None
if expected_error == Error.Okay:
expected_result = a << (digits * 60)
return test("test_shl_digit", res, [a, digits], expected_error, expected_result)
return test("test_shl_leg", res, [a, digits], expected_error, expected_result)
def test_shr_digit(a = 0, digits = 0, expected_error = Error.Okay):
def test_shr_leg(a = 0, digits = 0, expected_error = Error.Okay):
args = [arg_to_odin(a), digits]
res = int_shr_digit(*args)
res = int_shr_leg(*args)
expected_result = None
if expected_error == Error.Okay:
if a < 0:
@@ -421,7 +421,7 @@ def test_shr_digit(a = 0, digits = 0, expected_error = Error.Okay):
else:
expected_result = a >> (digits * 60)
return test("test_shr_digit", res, [a, digits], expected_error, expected_result)
return test("test_shr_leg", res, [a, digits], expected_error, expected_result)
def test_shl(a = 0, bits = 0, expected_error = Error.Okay):
args = [arg_to_odin(a), bits]
@@ -556,12 +556,12 @@ TESTS = {
test_root_n: [
[ 1298074214633706907132624082305024, 2, Error.Okay, ],
],
test_shl_digit: [
test_shl_leg: [
[ 3192, 1 ],
[ 1298074214633706907132624082305024, 2 ],
[ 1024, 3 ],
],
test_shr_digit: [
test_shr_leg: [
[ 3680125442705055547392, 1 ],
[ 1725436586697640946858688965569256363112777243042596638790631055949824, 2 ],
[ 219504133884436710204395031992179571, 2 ],
@@ -619,10 +619,10 @@ total_failures = 0
# test_shr_signed also tests shr, so we're not going to test shr randomly.
#
RANDOM_TESTS = [
test_add, test_sub, test_mul, test_sqr, test_div,
test_log, test_pow, test_sqrt, test_root_n,
test_shl_digit, test_shr_digit, test_shl, test_shr_signed,
test_gcd, test_lcm, test_is_square,
test_add, test_sub, test_mul, test_sqr,
test_log, test_pow, test_sqrt, test_root_n,
test_shl_leg, test_shr_leg, test_shl, test_shr_signed,
test_gcd, test_lcm, test_is_square, test_div,
]
SKIP_LARGE = [
test_pow, test_root_n, # test_gcd,
@@ -719,9 +719,9 @@ if __name__ == '__main__':
a = randint(1, 1 << BITS)
b = TEST_ROOT_N_PARAMS[index]
index = (index + 1) % len(TEST_ROOT_N_PARAMS)
elif test_proc == test_shl_digit:
elif test_proc == test_shl_leg:
b = randint(0, 10);
elif test_proc == test_shr_digit:
elif test_proc == test_shr_leg:
a = abs(a)
b = randint(0, 10);
elif test_proc == test_shl: