ed/Odin
1
0
Fork 0
mirror of https://github.com/Ed94/Odin.git synced 2026-08-05 15:18:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1249 from Yawning/feature/math-bits-improvements

Browse Source 
core:math/bits: Improve add/sub/mul_u64
This commit is contained in:
gingerBill
2021-10-28 00:19:57 +01:00
committed by GitHub
parent 90d587df13 ee724f183a
commit fa07e45eda
1 changed files with 30 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files
core/math/bits/bits.odin
+30 -42
View File
@@ -147,55 +147,53 @@ len :: proc{len_u8, len_u16, len_u32, len_u64, len_uint}
add_u32 :: proc(x, y, carry: u32) -> (sum, carry_out: u32) { add_u32 :: proc(x, y, carry: u32) -> (sum, carry_out: u32) {
yc := y + carry tmp_carry, tmp_carry2: bool
sum = x + yc sum, tmp_carry = intrinsics.overflow_add(x, y)
if sum < x || yc < y { sum, tmp_carry2 = intrinsics.overflow_add(sum, carry)
carry_out = 1 carry_out = u32(tmp_carry | tmp_carry2)
}
return return
} }
add_u64 :: proc(x, y, carry: u64) -> (sum, carry_out: u64) { add_u64 :: proc(x, y, carry: u64) -> (sum, carry_out: u64) {
yc := y + carry tmp_carry, tmp_carry2: bool
sum = x + yc sum, tmp_carry = intrinsics.overflow_add(x, y)
if sum < x || yc < y { sum, tmp_carry2 = intrinsics.overflow_add(sum, carry)
carry_out = 1 carry_out = u64(tmp_carry | tmp_carry2)
}
return return
} }
add_uint :: proc(x, y, carry: uint) -> (sum, carry_out: uint) { add_uint :: proc(x, y, carry: uint) -> (sum, carry_out: uint) {
yc := y + carry when size_of(uint) == size_of(u64) {
sum = x + yc a, b := add_u64(u64(x), u64(y), u64(carry))
if sum < x || yc < y { } else {
carry_out = 1 #assert(size_of(uint) == size_of(u32))
a, b := add_u32(u32(x), u32(y), u32(carry))
} }
return return uint(a), uint(b)
} }
add :: proc{add_u32, add_u64, add_uint} add :: proc{add_u32, add_u64, add_uint}
sub_u32 :: proc(x, y, borrow: u32) -> (diff, borrow_out: u32) { sub_u32 :: proc(x, y, borrow: u32) -> (diff, borrow_out: u32) {
yb := y + borrow tmp_borrow, tmp_borrow2: bool
diff = x - yb diff, tmp_borrow = intrinsics.overflow_sub(x, y)
if diff > x || yb < y { diff, tmp_borrow2 = intrinsics.overflow_sub(diff, borrow)
borrow_out = 1 borrow_out = u32(tmp_borrow | tmp_borrow2)
}
return return
} }
sub_u64 :: proc(x, y, borrow: u64) -> (diff, borrow_out: u64) { sub_u64 :: proc(x, y, borrow: u64) -> (diff, borrow_out: u64) {
yb := y + borrow tmp_borrow, tmp_borrow2: bool
diff = x - yb diff, tmp_borrow = intrinsics.overflow_sub(x, y)
if diff > x || yb < y { diff, tmp_borrow2 = intrinsics.overflow_sub(diff, borrow)
borrow_out = 1 borrow_out = u64(tmp_borrow | tmp_borrow2)
}
return return
} }
sub_uint :: proc(x, y, borrow: uint) -> (diff, borrow_out: uint) { sub_uint :: proc(x, y, borrow: uint) -> (diff, borrow_out: uint) {
yb := y + borrow when size_of(uint) == size_of(u64) {
diff = x - yb a, b := sub_u64(u64(x), u64(y), u64(borrow))
if diff > x || yb < y { } else {
borrow_out = 1 #assert(size_of(uint) == size_of(u32))
a, b := sub_u32(u32(x), u32(y), u32(borrow))
} }
return return uint(a), uint(b)
} }
sub :: proc{sub_u32, sub_u64, sub_uint} sub :: proc{sub_u32, sub_u64, sub_uint}
@@ -206,18 +204,8 @@ mul_u32 :: proc(x, y: u32) -> (hi, lo: u32) {
return return
} }
mul_u64 :: proc(x, y: u64) -> (hi, lo: u64) { mul_u64 :: proc(x, y: u64) -> (hi, lo: u64) {
mask :: 1<<32 - 1 prod_wide := u128(x) * u128(y)
hi, lo = u64(prod_wide>>64), u64(prod_wide)
x0, x1 := x & mask, x >> 32
y0, y1 := y & mask, y >> 32
w0 := x0 * y0
t := x1*y0 + w0>>32
w1, w2 := t & mask, t >> 32
w1 += x0 * y1
hi = x1*y1 + w2 + w1>>32
lo = x * y
return return
} }