Merge branch 'master' of https://github.com/weshardee/Odin

16 files changed, 295 insertions, 235 deletions

diff --git a/core/bytes/bytes.odin b/core/bytes/bytes.odin
index cbc1e2506..1e83b93c8 100644
--- a/core/bytes/bytes.odin
+++ b/core/bytes/bytes.odin
@@ -1143,7 +1143,7 @@ fields_proc :: proc(s: []byte, f: proc(rune) -> bool, allocator := context.alloc
 	}
 
 	if start >= 0 {
-		append(&subslices, s[start : end])
+		append(&subslices, s[start : len(s)])
 	}
 
 	return subslices[:]
diff --git a/core/math/big/common.odin b/core/math/big/common.odin
index 5b7d162bc..31ad54b14 100644
--- a/core/math/big/common.odin
+++ b/core/math/big/common.odin
@@ -158,13 +158,14 @@ Error :: enum int {
 	Invalid_Pointer         = 2,
 	Invalid_Argument        = 3,
 
-	Assignment_To_Immutable = 4,
-	Max_Iterations_Reached  = 5,
-	Buffer_Overflow         = 6,
-	Integer_Overflow        = 7,
+	Assignment_To_Immutable = 10,
+	Max_Iterations_Reached  = 11,
+	Buffer_Overflow         = 12,
+	Integer_Overflow        = 13,
+	Integer_Underflow       = 14,
 
-	Division_by_Zero        = 8,
-	Math_Domain_Error       = 9,
+	Division_by_Zero        = 30,
+	Math_Domain_Error       = 31,
 
 	Cannot_Open_File        = 50,
 	Cannot_Read_File        = 51,
@@ -215,7 +216,7 @@ _MIN_DIGIT_COUNT :: max(3, ((size_of(u128) + _DIGIT_BITS) - 1) / _DIGIT_BITS)
 /*
 	Maximum number of digits.
 	- Must be small enough such that `_bit_count` does not overflow.
- 	- Must be small enough such that `_radix_size` for base 2 does not overflow.
+	- Must be small enough such that `_radix_size` for base 2 does not overflow.
 	`_radix_size` needs two additional bytes for zero termination and sign.
 */
 _MAX_BIT_COUNT   :: (max(int) - 2)
@@ -251,7 +252,7 @@ Order :: enum i8 {
 }
 
 Endianness :: enum i8 {
-   Little   = -1,
-   Platform =  0,
-   Big      =  1,
-};
\ No newline at end of file
+	Little   = -1,
+	Platform =  0,
+	Big      =  1,
+}
\ No newline at end of file
diff --git a/core/math/big/internal.odin b/core/math/big/internal.odin
index 4702e76a3..437f6e5fc 100644
--- a/core/math/big/internal.odin
+++ b/core/math/big/internal.odin
@@ -34,6 +34,7 @@ package math_big
 import "core:mem"
 import "core:intrinsics"
 import rnd "core:math/rand"
+import "core:builtin"
 
 /*
 	Low-level addition, unsigned. Handbook of Applied Cryptography, algorithm 14.7.
@@ -1880,8 +1881,6 @@ internal_int_set_from_integer :: proc(dest: ^Int, src: $T, minimize := false, al
 	where intrinsics.type_is_integer(T) {
 	context.allocator = allocator
 
-	src := src
-
 	internal_error_if_immutable(dest) or_return
 	/*
 		Most internal procs asssume an Int to have already been initialize,
@@ -1892,13 +1891,27 @@ internal_int_set_from_integer :: proc(dest: ^Int, src: $T, minimize := false, al
 	dest.flags = {} // We're not -Inf, Inf, NaN or Immutable.
 
 	dest.used  = 0
-	dest.sign = .Zero_or_Positive if src >= 0 else .Negative
-	src = internal_abs(src)
+	dest.sign = .Negative if src < 0 else .Zero_or_Positive
+
+	temp := src
+
+	is_maximally_negative := src == min(T)
+	if is_maximally_negative {
+		/*
+			Prevent overflow on abs()
+		*/
+		temp += 1
+	}
+	temp = -temp if temp < 0 else temp
 
-	#no_bounds_check for src != 0 {
-		dest.digit[dest.used] = DIGIT(src) & _MASK
+	#no_bounds_check for temp != 0 {
+		dest.digit[dest.used] = DIGIT(temp) & _MASK
 		dest.used += 1
-		src >>= _DIGIT_BITS
+		temp >>= _DIGIT_BITS
+	}
+
+	if is_maximally_negative {
+		return internal_sub(dest, dest, 1)
 	}
 	internal_zero_unused(dest)
 	return nil
@@ -2307,29 +2320,70 @@ internal_int_get_i32 :: proc(a: ^Int) -> (res: i32, err: Error) {
 }
 internal_get_i32 :: proc { internal_int_get_i32, }
 
+internal_get_low_u32 :: proc(a: ^Int) -> u32 #no_bounds_check {
+	if a == nil {
+		return 0
+	}
+	
+	if a.used == 0 {
+		return 0
+	}
+	
+	return u32(a.digit[0])
+}
+internal_get_low_u64 :: proc(a: ^Int) -> u64 #no_bounds_check {
+	if a == nil {
+		return 0
+	}
+	
+	if a.used == 0 {
+		return 0
+	}
+	
+	v := u64(a.digit[0])
+	when size_of(DIGIT) == 4 {
+		if a.used > 1 {
+			return u64(a.digit[1])<<32 | v
+		}
+	}
+	return v
+}
+
 /*
 	TODO: Think about using `count_bits` to check if the value could be returned completely,
 	and maybe return max(T), .Integer_Overflow if not?
 */
 internal_int_get :: proc(a: ^Int, $T: typeid) -> (res: T, err: Error) where intrinsics.type_is_integer(T) {
-	size_in_bits := int(size_of(T) * 8)
-	i := int((size_in_bits + _DIGIT_BITS - 1) / _DIGIT_BITS)
-	i  = min(int(a.used), i)
-
-	#no_bounds_check for ; i >= 0; i -= 1 {
-		res <<= uint(0) if size_in_bits <= _DIGIT_BITS else _DIGIT_BITS
-		res |= T(a.digit[i])
-		if size_in_bits <= _DIGIT_BITS {
-			break
+	/*
+		Calculate target bit size.
+	*/
+	target_bit_size := int(size_of(T) * 8)
+	when !intrinsics.type_is_unsigned(T) {
+		if a.sign == .Zero_or_Positive {
+			target_bit_size -= 1
+		}
+	} else {
+		if a.sign == .Negative {
+			return 0, .Integer_Underflow
+		}
+	}
+
+	bits_used := internal_count_bits(a)
+
+	if bits_used > target_bit_size {
+		if a.sign == .Negative {
+			return min(T), .Integer_Underflow
 		}
+		return max(T), .Integer_Overflow
+	}
+
+	for i := a.used; i > 0; i -= 1 {
+		res <<= _DIGIT_BITS
+		res |=  T(a.digit[i - 1])
 	}
 
 	when !intrinsics.type_is_unsigned(T) {
 		/*
-			Mask off sign bit.
-		*/
-		res ~= 1 << uint(size_in_bits - 1)
-		/*
 			Set the sign.
 		*/
 		if a.sign == .Negative { res = -res }
@@ -2594,7 +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
 	}
 
 	/*
@@ -2634,37 +2688,6 @@ 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.
 */
 internal_int_shr_signed :: proc(dest, src: ^Int, bits: int, allocator := context.allocator) -> (err: Error) {
@@ -2702,7 +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
 	}
 
 	/*
@@ -2732,45 +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.
diff --git a/core/math/big/logical.odin b/core/math/big/logical.odin
index dbcf566c8..e7e55cc47 100644
--- a/core/math/big/logical.odin
+++ b/core/math/big/logical.odin
@@ -87,21 +87,6 @@ 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.
 */
 int_shr_signed :: proc(dest, src: ^Int, bits: int, allocator := context.allocator) -> (err: Error) {
@@ -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, };
\ No newline at end of file
+shl :: proc { int_shl, }
\ No newline at end of file
diff --git a/core/math/big/private.odin b/core/math/big/private.odin
index 14a27f600..9989a208a 100644
--- a/core/math/big/private.odin
+++ b/core/math/big/private.odin
@@ -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    =======================

 

diff --git a/core/math/big/rat.odin b/core/math/big/rat.odin
index 837af6fd3..c3efc30aa 100644
--- a/core/math/big/rat.odin
+++ b/core/math/big/rat.odin
@@ -42,9 +42,9 @@ rat_set_f64 :: proc(dst: ^Rat, f: f64, allocator := context.allocator) -> (err:
 	dst.a.sign = .Negative if f < 0 else .Zero_or_Positive
 	
 	if shift > 0 {
-		internal_int_shl_digit(&dst.b, shift) or_return
+		internal_int_shl(&dst.b, &dst.b, shift) or_return
 	} else {
-		internal_int_shl_digit(&dst.a, -shift) or_return
+		internal_int_shl(&dst.a, &dst.a, -shift) or_return
 	}
 	
 	return internal_rat_norm(dst)
@@ -389,9 +389,9 @@ internal_rat_to_float :: proc($T: typeid, z: ^Rat, allocator := context.allocato
 	internal_int_abs(b2, b) or_return
 	
 	if shift := MSIZE2 - exp; shift > 0 {
-		internal_int_shl_digit(a2, shift) or_return
-	} else {
-		internal_int_shl_digit(b2, -shift) or_return
+		internal_int_shl(a2, a2, shift) or_return
+	} else if shift < 0 {
+		internal_int_shl(b2, b2, -shift) or_return
 	}
 	
 	q, r := &Int{}, &Int{}
diff --git a/core/math/math.odin b/core/math/math.odin
index caaa6f51b..b81598da9 100644
--- a/core/math/math.odin
+++ b/core/math/math.odin
@@ -1196,6 +1196,7 @@ sum :: proc "contextless" (x: $T/[]$E) -> (res: E)
 
 prod :: proc "contextless" (x: $T/[]$E) -> (res: E)
 	where intrinsics.type_is_numeric(E) {
+	res = 1
 	for i in x {
 		res *= i
 	}
diff --git a/core/runtime/internal.odin b/core/runtime/internal.odin
index ed58f4318..6498c4db7 100644
--- a/core/runtime/internal.odin
+++ b/core/runtime/internal.odin
@@ -2,13 +2,15 @@ package runtime
 
 import "core:intrinsics"
 
+@(private="file")
+IS_WASM :: ODIN_ARCH == "wasm32" || ODIN_ARCH == "wasm64"
+
 @(private)
 RUNTIME_LINKAGE :: "strong" when (
 	(ODIN_USE_SEPARATE_MODULES || 
 	ODIN_BUILD_MODE == "dynamic" || 
 	!ODIN_NO_CRT) &&
-	!(ODIN_ARCH == "wasm32" || 
-	  ODIN_ARCH == "wasm64")) else "internal"
+	!IS_WASM) else "internal"
 RUNTIME_REQUIRE :: true
 
 
@@ -752,6 +754,9 @@ extendhfsf2 :: proc "c" (value: u16) -> f32 {
 
 @(link_name="__floattidf", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
 floattidf :: proc "c" (a: i128) -> f64 {
+when IS_WASM {
+	return 0
+} else {
 	DBL_MANT_DIG :: 53
 	if a == 0 {
 		return 0.0
@@ -791,10 +796,14 @@ floattidf :: proc "c" (a: i128) -> f64 {
 	fb[0] = u32(a)                           // mantissa-low
 	return transmute(f64)fb
 }
+}
 
 
 @(link_name="__floattidf_unsigned", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
 floattidf_unsigned :: proc "c" (a: u128) -> f64 {
+when IS_WASM {
+	return 0
+} else {
 	DBL_MANT_DIG :: 53
 	if a == 0 {
 		return 0.0
@@ -832,6 +841,7 @@ floattidf_unsigned :: proc "c" (a: u128) -> f64 {
 	fb[0] = u32(a)                           // mantissa-low
 	return transmute(f64)fb
 }
+}
 
 
 
diff --git a/core/runtime/procs_wasm32.odin b/core/runtime/procs_wasm32.odin
index 5caf6d2f8..2a4210c1e 100644
--- a/core/runtime/procs_wasm32.odin
+++ b/core/runtime/procs_wasm32.odin
@@ -1,23 +1,40 @@
 //+build wasm32
 package runtime
 
+@(private="file")
+ti_int :: struct #raw_union {
+	using s: struct { lo, hi: u64 },
+	all: i128,
+}
+
 @(link_name="__ashlti3", linkage="strong")
-__ashlti3 :: proc "c" (a: i64, b_: i32) -> i64 {
-	/*
+__ashlti3 :: proc "c" (a: i128, b_: u32) -> i128 {
+	bits_in_dword :: size_of(u32)*8
 	b := u32(b_)
-	input := transmute([2]i32)a
-	result: [2]i32
-	if b & 32 != 0 {
-		result[0] = 0
-		result[1] = input[0] << (b - 32)
+	
+	input, result: ti_int
+	input.all = a
+	if b & bits_in_dword != 0 {
+		result.lo = 0
+		result.hi = input.lo << (b-bits_in_dword)
 	} else {
 		if b == 0 {
 			return a
 		}
-		result[0] = input[0]<<b
-		result[1] = (input[1]<<b) | (input[0]>>(32-b))
+		result.lo = input.lo<<b
+		result.hi = (input.hi<<b) | (input.lo>>(bits_in_dword-b))
 	}
-	return transmute(i64)result
-	*/
-	return 0
+	return result.all
+}
+
+
+@(link_name="__multi3", linkage="strong")
+__multi3 :: proc "c" (a, b: i128) -> i128 {
+	x, y, r: ti_int
+	
+	x.all = a
+	y.all = b
+	r.all = i128(x.lo * y.lo) // TODO this is incorrect
+	r.hi += x.hi*y.lo + x.lo*y.hi
+	return r.all
 }
\ No newline at end of file
diff --git a/core/strconv/strconv.odin b/core/strconv/strconv.odin
index 6b3a91b4c..6ea8b39e6 100644
--- a/core/strconv/strconv.odin
+++ b/core/strconv/strconv.odin
@@ -882,7 +882,9 @@ unquote_string :: proc(lit: string, allocator := context.allocator) -> (res: str
 		return -1
 	}
 
-	assert(len(lit) >= 2)
+	if len(lit) < 2 {
+		return
+	}
 	if lit[0] == '`' {
 		return lit[1:len(lit)-1], false, true
 	}
diff --git a/core/strings/strings.odin b/core/strings/strings.odin
index a8199e0cf..3f703372f 100644
--- a/core/strings/strings.odin
+++ b/core/strings/strings.odin
@@ -504,8 +504,8 @@ index_any :: proc(s, chars: string) -> int {
 		}
 	}
 
-	for c in chars {
-		if i := index_rune(s, c); i >= 0 {
+	for c, i in s {
+		if index_rune(chars, c) >= 0 {
 			return i
 		}
 	}
@@ -1288,7 +1288,7 @@ fields_proc :: proc(s: string, f: proc(rune) -> bool, allocator := context.alloc
 	}
 
 	if start >= 0 {
-		append(&substrings, s[start : end])
+		append(&substrings, s[start : len(s)])
 	}
 
 	return substrings[:]
diff --git a/src/check_stmt.cpp b/src/check_stmt.cpp
index 1a424240c..396388629 100644
--- a/src/check_stmt.cpp
+++ b/src/check_stmt.cpp
@@ -2243,6 +2243,9 @@ void check_stmt_internal(CheckerContext *ctx, Ast *node, u32 flags) {
 						error(e->token, "The 'static' attribute is not allowed to be applied to '_'");
 					} else {
 						e->flags |= EntityFlag_Static;
+						if (ctx->in_defer) {
+							error(e->token, "'static' variables cannot be declared within a defer statement");
+						}
 					}
 				}
 				if (ac.thread_local_model != "") {
@@ -2251,9 +2254,13 @@ void check_stmt_internal(CheckerContext *ctx, Ast *node, u32 flags) {
 						error(e->token, "The 'thread_local' attribute is not allowed to be applied to '_'");
 					} else {
 						e->flags |= EntityFlag_Static;
+						if (ctx->in_defer) {
+							error(e->token, "'thread_local' variables cannot be declared within a defer statement");
+						}
 					}
 					e->Variable.thread_local_model = ac.thread_local_model;
 				}
+				
 
 				if (ac.is_static && ac.thread_local_model != "") {
 					error(e->token, "The 'static' attribute is not needed if 'thread_local' is applied");
diff --git a/src/llvm_backend_opt.cpp b/src/llvm_backend_opt.cpp
index de925655f..5b8468799 100644
--- a/src/llvm_backend_opt.cpp
+++ b/src/llvm_backend_opt.cpp
@@ -48,12 +48,6 @@ LLVMBool lb_must_preserve_predicate_callback(LLVMValueRef value, void *user_data
 	return LLVMIsAAllocaInst(value) != nullptr;
 }
 
-void lb_add_must_preserve_predicate_pass(lbModule *m, LLVMPassManagerRef fpm, i32 optimization_level) {
-	if (false && optimization_level == 0 && m->debug_builder) {
-		// LLVMAddInternalizePassWithMustPreservePredicate(fpm, m, lb_must_preserve_predicate_callback);
-	}
-}
-
 
 #if LLVM_VERSION_MAJOR < 12
 #define LLVM_ADD_CONSTANT_VALUE_PASS(fpm) LLVMAddConstantPropagationPass(fpm)
@@ -61,7 +55,10 @@ void lb_add_must_preserve_predicate_pass(lbModule *m, LLVMPassManagerRef fpm, i3
 #define LLVM_ADD_CONSTANT_VALUE_PASS(fpm) 
 #endif
 
-void lb_basic_populate_function_pass_manager(LLVMPassManagerRef fpm) {
+void lb_basic_populate_function_pass_manager(LLVMPassManagerRef fpm, i32 optimization_level) {
+	if (optimization_level == 0 && build_context.ODIN_DEBUG) {
+		return;
+	}
 	LLVMAddPromoteMemoryToRegisterPass(fpm);
 	LLVMAddMergedLoadStoreMotionPass(fpm);
 	LLVM_ADD_CONSTANT_VALUE_PASS(fpm);
@@ -78,14 +75,12 @@ void lb_populate_function_pass_manager(lbModule *m, LLVMPassManagerRef fpm, bool
 	// TODO(bill): Determine which opt definitions should exist in the first place
 	optimization_level = gb_clamp(optimization_level, 0, 2);
 
-	lb_add_must_preserve_predicate_pass(m, fpm, optimization_level);
-
 	if (ignore_memcpy_pass) {
-		lb_basic_populate_function_pass_manager(fpm);
+		lb_basic_populate_function_pass_manager(fpm, optimization_level);
 		return;
 	} else if (optimization_level == 0) {
 		LLVMAddMemCpyOptPass(fpm);
-		lb_basic_populate_function_pass_manager(fpm);
+		lb_basic_populate_function_pass_manager(fpm, optimization_level);
 		return;
 	}
 
@@ -96,7 +91,7 @@ void lb_populate_function_pass_manager(lbModule *m, LLVMPassManagerRef fpm, bool
 	LLVMPassManagerBuilderPopulateFunctionPassManager(pmb, fpm);
 #else
 	LLVMAddMemCpyOptPass(fpm);
-	lb_basic_populate_function_pass_manager(fpm);
+	lb_basic_populate_function_pass_manager(fpm, optimization_level);
 
 	LLVMAddSCCPPass(fpm);
 
@@ -114,11 +109,9 @@ void lb_populate_function_pass_manager_specific(lbModule *m, LLVMPassManagerRef
 	// TODO(bill): Determine which opt definitions should exist in the first place
 	optimization_level = gb_clamp(optimization_level, 0, 2);
 
-	lb_add_must_preserve_predicate_pass(m, fpm, optimization_level);
-
 	if (optimization_level == 0) {
 		LLVMAddMemCpyOptPass(fpm);
-		lb_basic_populate_function_pass_manager(fpm);
+		lb_basic_populate_function_pass_manager(fpm, optimization_level);
 		return;
 	}
 
@@ -191,6 +184,9 @@ void lb_populate_module_pass_manager(LLVMTargetMachineRef target_machine, LLVMPa
 	// NOTE(bill): Treat -opt:3 as if it was -opt:2
 	// TODO(bill): Determine which opt definitions should exist in the first place
 	optimization_level = gb_clamp(optimization_level, 0, 2);
+	if (optimization_level == 0 && build_context.ODIN_DEBUG) {
+		return;
+	}
 
 	LLVMAddAlwaysInlinerPass(mpm);
 	LLVMAddStripDeadPrototypesPass(mpm);
diff --git a/tests/core/math/big/test.odin b/tests/core/math/big/test.odin
index 07fa0364b..81f1956dc 100644
--- a/tests/core/math/big/test.odin
+++ b/tests/core/math/big/test.odin
@@ -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}
diff --git a/tests/core/math/big/test.py b/tests/core/math/big/test.py
index 6b17336bc..d292a3ff4 100644
--- a/tests/core/math/big/test.py
+++ b/tests/core/math/big/test.py
@@ -127,17 +127,22 @@ def we_iterate():
 # Error enum values

 #

 class Error(Enum):

-	Okay                   = 0

-	Out_Of_Memory          = 1

-	Invalid_Pointer        = 2

-	Invalid_Argument       = 3

-	Unknown_Error          = 4

-	Max_Iterations_Reached = 5

-	Buffer_Overflow        = 6

-	Integer_Overflow       = 7

-	Division_by_Zero       = 8

-	Math_Domain_Error      = 9

-	Unimplemented          = 127

+	Okay                    = 0

+	Out_Of_Memory           = 1

+	Invalid_Pointer         = 2

+	Invalid_Argument        = 3

+	Unknown_Error           = 4

+	Assignment_To_Immutable = 10

+	Max_Iterations_Reached  = 11

+	Buffer_Overflow         = 12

+	Integer_Overflow        = 13

+	Integer_Underflow       = 14

+	Division_by_Zero        = 30

+	Math_Domain_Error       = 31

+	Cannot_Open_File        = 50

+	Cannot_Read_File        = 51

+	Cannot_Write_File       = 52

+	Unimplemented           = 127

 

 #

 # Disable garbage collection

@@ -182,8 +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)

@@ -397,17 +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:

@@ -416,7 +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]

@@ -551,12 +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 ],

@@ -614,10 +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,

@@ -714,9 +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:

diff --git a/vendor/wasm/WebGL/runtime.js b/vendor/wasm/WebGL/runtime.js
index b6994a8ec..18b540b5c 100644
--- a/vendor/wasm/WebGL/runtime.js
+++ b/vendor/wasm/WebGL/runtime.js
@@ -416,7 +416,7 @@ class WebGLInterface {
 					log = log.substring(0, n);
 					this.mem.loadBytes(buf_ptr, buf_len).set(new TextEncoder("utf-8").encode(log))
 					
-					storeInt(length_ptr, n);
+					this.mem.storeInt(length_ptr, n);
 				}
 			},
 			GetShaderInfoLog: (shader, buf_ptr, buf_len, length_ptr) => {
@@ -429,7 +429,7 @@ class WebGLInterface {
 					log = log.substring(0, n);
 					this.mem.loadBytes(buf_ptr, buf_len).set(new TextEncoder("utf-8").encode(log))
 					
-					storeInt(length_ptr, n);
+					this.mem.storeInt(length_ptr, n);
 				}
 			},
 			GetShaderiv: (shader, pname, p) => {
@@ -439,11 +439,11 @@ class WebGLInterface {
 						if (log === null) {
 							log = "(unknown error)";
 						}
-						storeInt(p, log.length+1);
+						this.mem.storeInt(p, log.length+1);
 					} else if (pname == 35720) {
 						let source = this.ctx.getShaderSource(this.shaders[shader]);
 						let sourceLength = (source === null || source.length == 0) ? 0 : source.length+1;
-						storeInt(p, sourceLength);
+						this.mem.storeInt(p, sourceLength);
 					} else {
 						let param = this.ctx.getShaderParameter(this.shaders[shader], pname);
 						this.mem.storeI32(p, param);
@@ -994,7 +994,7 @@ class WebGLInterface {
 				let n = Math.min(buf_len, name.length);
 				name = name.substring(0, n);
 				this.mem.loadBytes(buf_ptr, buf_len).set(new TextEncoder("utf-8").encode(name))
-				storeInt(length_ptr, n);
+				this.mem.storeInt(length_ptr, n);
 			},
 			UniformBlockBinding: (program, uniformBlockIndex, uniformBlockBinding) => {
 				this.assertWebGL2();