Replace many foreign llvm calls with intrinsics

This commit is contained in:
gingerBill
2021-04-25 20:22:26 +01:00
parent cb2e6ea31d
commit 72aa0e6e38
12 changed files with 78 additions and 247 deletions

View File

@@ -1,5 +1,6 @@
package math_bits
import "intrinsics"
import "core:runtime"
U8_MIN :: 0;
@@ -22,32 +23,10 @@ I16_MAX :: 1 << 15 - 1;
I32_MAX :: 1 << 31 - 1;
I64_MAX :: 1 << 63 - 1;
@(default_calling_convention="none")
foreign {
@(link_name="llvm.ctpop.i8") count_ones8 :: proc(i: u8) -> u8 ---
@(link_name="llvm.ctpop.i16") count_ones16 :: proc(i: u16) -> u16 ---
@(link_name="llvm.ctpop.i32") count_ones32 :: proc(i: u32) -> u32 ---
@(link_name="llvm.ctpop.i64") count_ones64 :: proc(i: u64) -> u64 ---
@(link_name="llvm.cttz.i8") trailing_zeros8 :: proc(i: u8, is_zero_undef := false) -> u8 ---
@(link_name="llvm.cttz.i16") trailing_zeros16 :: proc(i: u16, is_zero_undef := false) -> u16 ---
@(link_name="llvm.cttz.i32") trailing_zeros32 :: proc(i: u32, is_zero_undef := false) -> u32 ---
@(link_name="llvm.cttz.i64") trailing_zeros64 :: proc(i: u64, is_zero_undef := false) -> u64 ---
@(link_name="llvm.bitreverse.i8") reverse_bits8 :: proc(i: u8) -> u8 ---
@(link_name="llvm.bitreverse.i16") reverse_bits16 :: proc(i: u16) -> u16 ---
@(link_name="llvm.bitreverse.i32") reverse_bits32 :: proc(i: u32) -> u32 ---
@(link_name="llvm.bitreverse.i64") reverse_bits64 :: proc(i: u64) -> u64 ---
}
trailing_zeros_uint :: proc(i: uint) -> uint {
when size_of(uint) == size_of(u64) {
return uint(trailing_zeros64(u64(i)));
} else {
return uint(trailing_zeros32(u32(i)));
}
}
count_ones :: intrinsics.count_ones;
trailing_zeros :: intrinsics.trailing_zeros;
reverse_bits :: intrinsics.reverse_bits;
leading_zeros_u8 :: proc(i: u8) -> int {
@@ -117,10 +96,17 @@ byte_swap :: proc{
byte_swap_int,
};
count_zeros8 :: proc(i: u8) -> u8 { return 8 - count_ones8(i); }
count_zeros16 :: proc(i: u16) -> u16 { return 16 - count_ones16(i); }
count_zeros32 :: proc(i: u32) -> u32 { return 32 - count_ones32(i); }
count_zeros64 :: proc(i: u64) -> u64 { return 64 - count_ones64(i); }
count_zeros8 :: proc(i: u8) -> u8 { return 8 - count_ones(i); }
count_zeros16 :: proc(i: u16) -> u16 { return 16 - count_ones(i); }
count_zeros32 :: proc(i: u32) -> u32 { return 32 - count_ones(i); }
count_zeros64 :: proc(i: u64) -> u64 { return 64 - count_ones(i); }
count_zeros :: proc{
count_zeros8,
count_zeros16,
count_zeros32,
count_zeros64,
};
rotate_left8 :: proc(x: u8, k: int) -> u8 {
@@ -176,120 +162,10 @@ to_le_u64 :: proc(i: u64) -> u64 { when ODIN_ENDIAN == "little" { return i; }
to_le_uint :: proc(i: uint) -> uint { when ODIN_ENDIAN == "little" { return i; } else { return byte_swap(i); } }
@(default_calling_convention="none")
foreign {
@(link_name="llvm.uadd.with.overflow.i8") overflowing_add_u8 :: proc(lhs, rhs: u8) -> (u8, bool) ---
@(link_name="llvm.sadd.with.overflow.i8") overflowing_add_i8 :: proc(lhs, rhs: i8) -> (i8, bool) ---
@(link_name="llvm.uadd.with.overflow.i16") overflowing_add_u16 :: proc(lhs, rhs: u16) -> (u16, bool) ---
@(link_name="llvm.sadd.with.overflow.i16") overflowing_add_i16 :: proc(lhs, rhs: i16) -> (i16, bool) ---
@(link_name="llvm.uadd.with.overflow.i32") overflowing_add_u32 :: proc(lhs, rhs: u32) -> (u32, bool) ---
@(link_name="llvm.sadd.with.overflow.i32") overflowing_add_i32 :: proc(lhs, rhs: i32) -> (i32, bool) ---
@(link_name="llvm.uadd.with.overflow.i64") overflowing_add_u64 :: proc(lhs, rhs: u64) -> (u64, bool) ---
@(link_name="llvm.sadd.with.overflow.i64") overflowing_add_i64 :: proc(lhs, rhs: i64) -> (i64, bool) ---
}
overflowing_add_uint :: proc(lhs, rhs: uint) -> (uint, bool) {
when size_of(uint) == size_of(u32) {
x, ok := overflowing_add_u32(u32(lhs), u32(rhs));
return uint(x), ok;
} else {
x, ok := overflowing_add_u64(u64(lhs), u64(rhs));
return uint(x), ok;
}
}
overflowing_add_int :: proc(lhs, rhs: int) -> (int, bool) {
when size_of(int) == size_of(i32) {
x, ok := overflowing_add_i32(i32(lhs), i32(rhs));
return int(x), ok;
} else {
x, ok := overflowing_add_i64(i64(lhs), i64(rhs));
return int(x), ok;
}
}
overflowing_add :: proc{
overflowing_add_u8, overflowing_add_i8,
overflowing_add_u16, overflowing_add_i16,
overflowing_add_u32, overflowing_add_i32,
overflowing_add_u64, overflowing_add_i64,
overflowing_add_uint, overflowing_add_int,
};
@(default_calling_convention="none")
foreign {
@(link_name="llvm.usub.with.overflow.i8") overflowing_sub_u8 :: proc(lhs, rhs: u8) -> (u8, bool) ---
@(link_name="llvm.ssub.with.overflow.i8") overflowing_sub_i8 :: proc(lhs, rhs: i8) -> (i8, bool) ---
@(link_name="llvm.usub.with.overflow.i16") overflowing_sub_u16 :: proc(lhs, rhs: u16) -> (u16, bool) ---
@(link_name="llvm.ssub.with.overflow.i16") overflowing_sub_i16 :: proc(lhs, rhs: i16) -> (i16, bool) ---
@(link_name="llvm.usub.with.overflow.i32") overflowing_sub_u32 :: proc(lhs, rhs: u32) -> (u32, bool) ---
@(link_name="llvm.ssub.with.overflow.i32") overflowing_sub_i32 :: proc(lhs, rhs: i32) -> (i32, bool) ---
@(link_name="llvm.usub.with.overflow.i64") overflowing_sub_u64 :: proc(lhs, rhs: u64) -> (u64, bool) ---
@(link_name="llvm.ssub.with.overflow.i64") overflowing_sub_i64 :: proc(lhs, rhs: i64) -> (i64, bool) ---
}
overflowing_sub_uint :: proc(lhs, rhs: uint) -> (uint, bool) {
when size_of(uint) == size_of(u32) {
x, ok := overflowing_sub_u32(u32(lhs), u32(rhs));
return uint(x), ok;
} else {
x, ok := overflowing_sub_u64(u64(lhs), u64(rhs));
return uint(x), ok;
}
}
overflowing_sub_int :: proc(lhs, rhs: int) -> (int, bool) {
when size_of(int) == size_of(i32) {
x, ok := overflowing_sub_i32(i32(lhs), i32(rhs));
return int(x), ok;
} else {
x, ok := overflowing_sub_i64(i64(lhs), i64(rhs));
return int(x), ok;
}
}
overflowing_sub :: proc{
overflowing_sub_u8, overflowing_sub_i8,
overflowing_sub_u16, overflowing_sub_i16,
overflowing_sub_u32, overflowing_sub_i32,
overflowing_sub_u64, overflowing_sub_i64,
overflowing_sub_uint, overflowing_sub_int,
};
@(default_calling_convention="none")
foreign {
@(link_name="llvm.umul.with.overflow.i8") overflowing_mul_u8 :: proc(lhs, rhs: u8) -> (u8, bool) ---
@(link_name="llvm.smul.with.overflow.i8") overflowing_mul_i8 :: proc(lhs, rhs: i8) -> (i8, bool) ---
@(link_name="llvm.umul.with.overflow.i16") overflowing_mul_u16 :: proc(lhs, rhs: u16) -> (u16, bool) ---
@(link_name="llvm.smul.with.overflow.i16") overflowing_mul_i16 :: proc(lhs, rhs: i16) -> (i16, bool) ---
@(link_name="llvm.umul.with.overflow.i32") overflowing_mul_u32 :: proc(lhs, rhs: u32) -> (u32, bool) ---
@(link_name="llvm.smul.with.overflow.i32") overflowing_mul_i32 :: proc(lhs, rhs: i32) -> (i32, bool) ---
@(link_name="llvm.umul.with.overflow.i64") overflowing_mul_u64 :: proc(lhs, rhs: u64) -> (u64, bool) ---
@(link_name="llvm.smul.with.overflow.i64") overflowing_mul_i64 :: proc(lhs, rhs: i64) -> (i64, bool) ---
}
overflowing_mul_uint :: proc(lhs, rhs: uint) -> (uint, bool) {
when size_of(uint) == size_of(u32) {
x, ok := overflowing_mul_u32(u32(lhs), u32(rhs));
return uint(x), ok;
} else {
x, ok := overflowing_mul_u64(u64(lhs), u64(rhs));
return uint(x), ok;
}
}
overflowing_mul_int :: proc(lhs, rhs: int) -> (int, bool) {
when size_of(int) == size_of(i32) {
x, ok := overflowing_mul_i32(i32(lhs), i32(rhs));
return int(x), ok;
} else {
x, ok := overflowing_mul_i64(i64(lhs), i64(rhs));
return int(x), ok;
}
}
overflowing_mul :: proc{
overflowing_mul_u8, overflowing_mul_i8,
overflowing_mul_u16, overflowing_mul_i16,
overflowing_mul_u32, overflowing_mul_i32,
overflowing_mul_u64, overflowing_mul_i64,
overflowing_mul_uint, overflowing_mul_int,
};
overflowing_add :: intrinsics.overflow_add;
overflowing_sub :: intrinsics.overflow_sub;
overflowing_mul :: intrinsics.overflow_mul;
len_u8 :: proc(x: u8) -> int {