Skip to content

Commit

Permalink
Rollup merge of rust-lang#100823 - WaffleLapkin:less_offsets, r=scottmcm
Browse files Browse the repository at this point in the history 
Refactor some `std` code that works with pointer offstes

This PR replaces `pointer::offset` in standard library with `pointer::add` and `pointer::sub`, [re]moving some casts and using `.addr()` while we are at it.

This is a more complicated refactor than all other sibling PRs, so take a closer look when reviewing, please 😃  (though I've checked this multiple times and it looks fine).

r? ````@scottmcm````

_split off from rust-lang#100746, continuation of #100822_
  • Loading branch information
@matthiaskrgr
matthiaskrgr authored Sep 24, 2022
2 parents bb5a016 + 98a3230 commit 1b1596c
Show file tree
Hide file tree
Showing 3 changed files with 62 additions and 62 deletions.
92 changes: 46 additions & 46 deletions library/core/src/fmt/num.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -211,7 +211,7 @@ macro_rules! impl_Display {
fn $name(mut n: $u, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::Result {
// 2^128 is about 3*10^38, so 39 gives an extra byte of space
let mut buf = [MaybeUninit::<u8>::uninit(); 39];
let mut curr = buf.len() as isize;
let mut curr = buf.len();
let buf_ptr = MaybeUninit::slice_as_mut_ptr(&mut buf);
let lut_ptr = DEC_DIGITS_LUT.as_ptr();

Expand All @@ -228,7 +228,7 @@ macro_rules! impl_Display {

// eagerly decode 4 characters at a time
while n >= 10000 {
let rem = (n % 10000) as isize;
let rem = (n % 10000) as usize;
n /= 10000;

let d1 = (rem / 100) << 1;
Expand All @@ -238,37 +238,37 @@ macro_rules! impl_Display {
// We are allowed to copy to `buf_ptr[curr..curr + 3]` here since
// otherwise `curr < 0`. But then `n` was originally at least `10000^10`
// which is `10^40 > 2^128 > n`.
ptr::copy_nonoverlapping(lut_ptr.offset(d1), buf_ptr.offset(curr), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d2), buf_ptr.offset(curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1), buf_ptr.add(curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d2), buf_ptr.add(curr + 2), 2);
}

// if we reach here numbers are <= 9999, so at most 4 chars long
let mut n = n as isize; // possibly reduce 64bit math
let mut n = n as usize; // possibly reduce 64bit math

// decode 2 more chars, if > 2 chars
if n >= 100 {
let d1 = (n % 100) << 1;
n /= 100;
curr -= 2;
ptr::copy_nonoverlapping(lut_ptr.offset(d1), buf_ptr.offset(curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1), buf_ptr.add(curr), 2);
}

// decode last 1 or 2 chars
if n < 10 {
curr -= 1;
*buf_ptr.offset(curr) = (n as u8) + b'0';
*buf_ptr.add(curr) = (n as u8) + b'0';
} else {
let d1 = n << 1;
curr -= 2;
ptr::copy_nonoverlapping(lut_ptr.offset(d1), buf_ptr.offset(curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1), buf_ptr.add(curr), 2);
}
}

// SAFETY: `curr` > 0 (since we made `buf` large enough), and all the chars are valid
// UTF-8 since `DEC_DIGITS_LUT` is
let buf_slice = unsafe {
str::from_utf8_unchecked(
slice::from_raw_parts(buf_ptr.offset(curr), buf.len() - curr as usize))
slice::from_raw_parts(buf_ptr.add(curr), buf.len() - curr))
};
f.pad_integral(is_nonnegative, "", buf_slice)
}
Expand Down Expand Up @@ -339,18 +339,18 @@ macro_rules! impl_Exp {
// Since `curr` always decreases by the number of digits copied, this means
// that `curr >= 0`.
let mut buf = [MaybeUninit::<u8>::uninit(); 40];
let mut curr = buf.len() as isize; //index for buf
let mut curr = buf.len(); //index for buf
let buf_ptr = MaybeUninit::slice_as_mut_ptr(&mut buf);
let lut_ptr = DEC_DIGITS_LUT.as_ptr();

// decode 2 chars at a time
while n >= 100 {
let d1 = ((n % 100) as isize) << 1;
let d1 = ((n % 100) as usize) << 1;
curr -= 2;
// SAFETY: `d1 <= 198`, so we can copy from `lut_ptr[d1..d1 + 2]` since
// `DEC_DIGITS_LUT` has a length of 200.
unsafe {
ptr::copy_nonoverlapping(lut_ptr.offset(d1), buf_ptr.offset(curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1), buf_ptr.add(curr), 2);
}
n /= 100;
exponent += 2;
Expand All @@ -362,7 +362,7 @@ macro_rules! impl_Exp {
curr -= 1;
// SAFETY: Safe since `40 > curr >= 0` (see comment)
unsafe {
*buf_ptr.offset(curr) = (n as u8 % 10_u8) + b'0';
*buf_ptr.add(curr) = (n as u8 % 10_u8) + b'0';
}
n /= 10;
exponent += 1;
Expand All @@ -372,18 +372,18 @@ macro_rules! impl_Exp {
curr -= 1;
// SAFETY: Safe since `40 > curr >= 0`
unsafe {
*buf_ptr.offset(curr) = b'.';
*buf_ptr.add(curr) = b'.';
}
}

// SAFETY: Safe since `40 > curr >= 0`
let buf_slice = unsafe {
// decode last character
curr -= 1;
*buf_ptr.offset(curr) = (n as u8) + b'0';
*buf_ptr.add(curr) = (n as u8) + b'0';

let len = buf.len() - curr as usize;
slice::from_raw_parts(buf_ptr.offset(curr), len)
slice::from_raw_parts(buf_ptr.add(curr), len)
};

// stores 'e' (or 'E') and the up to 2-digit exponent
Expand All @@ -392,13 +392,13 @@ macro_rules! impl_Exp {
// SAFETY: In either case, `exp_buf` is written within bounds and `exp_ptr[..len]`
// is contained within `exp_buf` since `len <= 3`.
let exp_slice = unsafe {
*exp_ptr.offset(0) = if upper { b'E' } else { b'e' };
*exp_ptr.add(0) = if upper { b'E' } else { b'e' };
let len = if exponent < 10 {
*exp_ptr.offset(1) = (exponent as u8) + b'0';
*exp_ptr.add(1) = (exponent as u8) + b'0';
2
} else {
let off = exponent << 1;
ptr::copy_nonoverlapping(lut_ptr.offset(off), exp_ptr.offset(1), 2);
ptr::copy_nonoverlapping(lut_ptr.add(off), exp_ptr.add(1), 2);
3
};
slice::from_raw_parts(exp_ptr, len)
Expand Down Expand Up @@ -479,7 +479,7 @@ mod imp {
impl_Exp!(i128, u128 as u128 via to_u128 named exp_u128);

/// Helper function for writing a u64 into `buf` going from last to first, with `curr`.
fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], curr: &mut isize) {
fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], curr: &mut usize) {
let buf_ptr = MaybeUninit::slice_as_mut_ptr(buf);
let lut_ptr = DEC_DIGITS_LUT.as_ptr();
assert!(*curr > 19);
Expand All @@ -505,14 +505,14 @@ fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], cu

*curr -= 16;

ptr::copy_nonoverlapping(lut_ptr.offset(d1 as isize), buf_ptr.offset(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d2 as isize), buf_ptr.offset(*curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d3 as isize), buf_ptr.offset(*curr + 4), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d4 as isize), buf_ptr.offset(*curr + 6), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d5 as isize), buf_ptr.offset(*curr + 8), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d6 as isize), buf_ptr.offset(*curr + 10), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d7 as isize), buf_ptr.offset(*curr + 12), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d8 as isize), buf_ptr.offset(*curr + 14), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1 as usize), buf_ptr.add(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d2 as usize), buf_ptr.add(*curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d3 as usize), buf_ptr.add(*curr + 4), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d4 as usize), buf_ptr.add(*curr + 6), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d5 as usize), buf_ptr.add(*curr + 8), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d6 as usize), buf_ptr.add(*curr + 10), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d7 as usize), buf_ptr.add(*curr + 12), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d8 as usize), buf_ptr.add(*curr + 14), 2);
}
if n >= 1e8 as u64 {
let to_parse = n % 1e8 as u64;
Expand All @@ -525,10 +525,10 @@ fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], cu
let d4 = ((to_parse / 1e0 as u64) % 100) << 1;
*curr -= 8;

ptr::copy_nonoverlapping(lut_ptr.offset(d1 as isize), buf_ptr.offset(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d2 as isize), buf_ptr.offset(*curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d3 as isize), buf_ptr.offset(*curr + 4), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d4 as isize), buf_ptr.offset(*curr + 6), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1 as usize), buf_ptr.add(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d2 as usize), buf_ptr.add(*curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d3 as usize), buf_ptr.add(*curr + 4), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d4 as usize), buf_ptr.add(*curr + 6), 2);
}
// `n` < 1e8 < (1 << 32)
let mut n = n as u32;
Expand All @@ -540,8 +540,8 @@ fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], cu
let d2 = (to_parse % 100) << 1;
*curr -= 4;

ptr::copy_nonoverlapping(lut_ptr.offset(d1 as isize), buf_ptr.offset(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.offset(d2 as isize), buf_ptr.offset(*curr + 2), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1 as usize), buf_ptr.add(*curr + 0), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d2 as usize), buf_ptr.add(*curr + 2), 2);
}

// `n` < 1e4 < (1 << 16)
Expand All @@ -550,17 +550,17 @@ fn parse_u64_into<const N: usize>(mut n: u64, buf: &mut [MaybeUninit<u8>; N], cu
let d1 = (n % 100) << 1;
n /= 100;
*curr -= 2;
ptr::copy_nonoverlapping(lut_ptr.offset(d1 as isize), buf_ptr.offset(*curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1 as usize), buf_ptr.add(*curr), 2);
}

// decode last 1 or 2 chars
if n < 10 {
*curr -= 1;
*buf_ptr.offset(*curr) = (n as u8) + b'0';
*buf_ptr.add(*curr) = (n as u8) + b'0';
} else {
let d1 = n << 1;
*curr -= 2;
ptr::copy_nonoverlapping(lut_ptr.offset(d1 as isize), buf_ptr.offset(*curr), 2);
ptr::copy_nonoverlapping(lut_ptr.add(d1 as usize), buf_ptr.add(*curr), 2);
}
}
}
Expand Down Expand Up @@ -593,21 +593,21 @@ impl fmt::Display for i128 {
fn fmt_u128(n: u128, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::Result {
// 2^128 is about 3*10^38, so 39 gives an extra byte of space
let mut buf = [MaybeUninit::<u8>::uninit(); 39];
let mut curr = buf.len() as isize;
let mut curr = buf.len();

let (n, rem) = udiv_1e19(n);
parse_u64_into(rem, &mut buf, &mut curr);

if n != 0 {
// 0 pad up to point
let target = (buf.len() - 19) as isize;
let target = buf.len() - 19;
// SAFETY: Guaranteed that we wrote at most 19 bytes, and there must be space
// remaining since it has length 39
unsafe {
ptr::write_bytes(
MaybeUninit::slice_as_mut_ptr(&mut buf).offset(target),
MaybeUninit::slice_as_mut_ptr(&mut buf).add(target),
b'0',
(curr - target) as usize,
curr - target,
);
}
curr = target;
Expand All @@ -616,16 +616,16 @@ fn fmt_u128(n: u128, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::R
parse_u64_into(rem, &mut buf, &mut curr);
// Should this following branch be annotated with unlikely?
if n != 0 {
let target = (buf.len() - 38) as isize;
let target = buf.len() - 38;
// The raw `buf_ptr` pointer is only valid until `buf` is used the next time,
// buf `buf` is not used in this scope so we are good.
let buf_ptr = MaybeUninit::slice_as_mut_ptr(&mut buf);
// SAFETY: At this point we wrote at most 38 bytes, pad up to that point,
// There can only be at most 1 digit remaining.
unsafe {
ptr::write_bytes(buf_ptr.offset(target), b'0', (curr - target) as usize);
ptr::write_bytes(buf_ptr.add(target), b'0', curr - target);
curr = target - 1;
*buf_ptr.offset(curr) = (n as u8) + b'0';
*buf_ptr.add(curr) = (n as u8) + b'0';
}
}
}
Expand All @@ -634,8 +634,8 @@ fn fmt_u128(n: u128, is_nonnegative: bool, f: &mut fmt::Formatter<'_>) -> fmt::R
// UTF-8 since `DEC_DIGITS_LUT` is
let buf_slice = unsafe {
str::from_utf8_unchecked(slice::from_raw_parts(
MaybeUninit::slice_as_mut_ptr(&mut buf).offset(curr),
buf.len() - curr as usize,
MaybeUninit::slice_as_mut_ptr(&mut buf).add(curr),
buf.len() - curr,
))
};
f.pad_integral(is_nonnegative, "", buf_slice)
Expand Down
4 changes: 2 additions & 2 deletions library/core/src/slice/memchr.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -141,8 +141,8 @@ pub fn memrchr(x: u8, text: &[u8]) -> Option<usize> {
// SAFETY: offset starts at len - suffix.len(), as long as it is greater than
// min_aligned_offset (prefix.len()) the remaining distance is at least 2 * chunk_bytes.
unsafe {
let u = *(ptr.offset(offset as isize - 2 * chunk_bytes as isize) as *const Chunk);
let v = *(ptr.offset(offset as isize - chunk_bytes as isize) as *const Chunk);
let u = *(ptr.add(offset - 2 * chunk_bytes) as *const Chunk);
let v = *(ptr.add(offset - chunk_bytes) as *const Chunk);

// Break if there is a matching byte.
let zu = contains_zero_byte(u ^ repeated_x);
Expand Down
28 changes: 14 additions & 14 deletions library/std/src/sys/sgx/abi/usercalls/alloc.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -316,9 +316,9 @@ where
// | small1 | Chunk smaller than 8 bytes
// +--------+
fn region_as_aligned_chunks(ptr: *const u8, len: usize) -> (usize, usize, usize) {
let small0_size = if ptr as usize % 8 == 0 { 0 } else { 8 - ptr as usize % 8 };
let small1_size = (len - small0_size as usize) % 8;
let big_size = len - small0_size as usize - small1_size as usize;
let small0_size = if ptr.is_aligned_to(8) { 0 } else { 8 - ptr.addr() % 8 };
let small1_size = (len - small0_size) % 8;
let big_size = len - small0_size - small1_size;

(small0_size, big_size, small1_size)
}
Expand Down Expand Up @@ -364,8 +364,8 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize)
mfence
lfence
",
val = in(reg_byte) *src.offset(off as isize),
dst = in(reg) dst.offset(off as isize),
val = in(reg_byte) *src.add(off),
dst = in(reg) dst.add(off),
seg_sel = in(reg) &mut seg_sel,
options(nostack, att_syntax)
);
Expand All @@ -378,8 +378,8 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize)
assert!(is_enclave_range(src, len));
assert!(is_user_range(dst, len));
assert!(len < isize::MAX as usize);
assert!(!(src as usize).overflowing_add(len).1);
assert!(!(dst as usize).overflowing_add(len).1);
assert!(!src.addr().overflowing_add(len).1);
assert!(!dst.addr().overflowing_add(len).1);

if len < 8 {
// Can't align on 8 byte boundary: copy safely byte per byte
Expand All @@ -404,17 +404,17 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize)

unsafe {
// Copy small0
copy_bytewise_to_userspace(src, dst, small0_size as _);
copy_bytewise_to_userspace(src, dst, small0_size);

// Copy big
let big_src = src.offset(small0_size as _);
let big_dst = dst.offset(small0_size as _);
copy_quadwords(big_src as _, big_dst, big_size);
let big_src = src.add(small0_size);
let big_dst = dst.add(small0_size);
copy_quadwords(big_src, big_dst, big_size);

// Copy small1
let small1_src = src.offset(big_size as isize + small0_size as isize);
let small1_dst = dst.offset(big_size as isize + small0_size as isize);
copy_bytewise_to_userspace(small1_src, small1_dst, small1_size as _);
let small1_src = src.add(big_size + small0_size);
let small1_dst = dst.add(big_size + small0_size);
copy_bytewise_to_userspace(small1_src, small1_dst, small1_size);
}
}
}
Expand Down

0 comments on commit 1b1596c

Please sign in to comment.