Previously, we aligned every cstring to 16 bytes as a temporary hack to
deal with https://github.com/llvm/llvm-project/issues/50135. However, it
was highly wasteful in terms of binary size.
To recap, in contrast to ELF, which puts strings that need different
alignments into different sections, `clang`'s Mach-O backend puts them
all in one section. Strings that need to be aligned have the .p2align
directive emitted before them, which simply translates into zero padding
in the object file. In other words, we have to infer the alignment of
the cstrings from their addresses.
We differ slightly from ld64 in how we've chosen to align these
cstrings. Both LLD and ld64 preserve the number of trailing zeros in
each cstring's address in the input object files. When deduplicating
identical cstrings, both linkers pick the cstring whose address has more
trailing zeros, and preserve the alignment of that address in the final
binary. However, ld64 goes a step further and also preserves the offset
of the cstring from the last section-aligned address. I.e. if a cstring
is at offset 18 in the input, with a section alignment of 16, then both
LLD and ld64 will ensure the final address is 2-byte aligned (since
`18 == 16 + 2`). But ld64 will also ensure that the final address is of
the form 16 * k + 2 for some k (which implies 2-byte alignment).
Note that ld64's heuristic means that a dedup'ed cstring's final address is
dependent on the order of the input object files. E.g. if in addition to the
cstring at offset 18 above, we have a duplicate one in another file with a
`.cstring` section alignment of 2 and an offset of zero, then ld64 will pick
the cstring from the object file earlier on the command line (since both have
the same number of trailing zeros in their address). So the final cstring may
either be at some address `16 * k + 2` or at some address `2 * k`.
I've opted not to follow this behavior primarily for implementation
simplicity, and secondarily to save a few more bytes. It's not clear to me
that preserving the section alignment + offset is ever necessary, and there
are many cases that are clearly redundant. In particular, if an x86_64 object
file contains some strings that are accessed via SIMD instructions, then the
.cstring section in the object file will be 16-byte-aligned (since SIMD
requires its operand addresses to be 16-byte aligned). However, there will
typically also be other cstrings in the same file that aren't used via SIMD
and don't need this alignment. They will be emitted at some arbitrary address
`A`, but ld64 will treat them as being 16-byte aligned with an offset of
`16 % A`.
I have verified that the two repros in https://github.com/llvm/llvm-project/issues/50135
work well with the new alignment behavior.
Fixes https://github.com/llvm/llvm-project/issues/54036.
Reviewed By: #lld-macho, oontvoo
Differential Revision: https://reviews.llvm.org/D121342
Our implementation draws heavily from LLD-ELF's, which in turn delegates
its string deduplication to llvm-mc's StringTableBuilder. The messiness of
this diff is largely due to the fact that we've previously assumed that
all InputSections get concatenated together to form the output. This is
no longer true with CStringInputSections, which split their contents into
StringPieces. StringPieces are much more lightweight than InputSections,
which is important as we create a lot of them. They may also overlap in
the output, which makes it possible for strings to be tail-merged. In
fact, the initial version of this diff implemented tail merging, but
I've dropped it for reasons I'll explain later.
**Alignment Issues**
Mergeable cstring literals are found under the `__TEXT,__cstring`
section. In contrast to ELF, which puts strings that need different
alignments into different sections, clang's Mach-O backend puts them all
in one section. Strings that need to be aligned have the `.p2align`
directive emitted before them, which simply translates into zero padding
in the object file.
I *think* ld64 extracts the desired per-string alignment from this data
by preserving each string's offset from the last section-aligned
address. I'm not entirely certain since it doesn't seem consistent about
doing this; but perhaps this can be chalked up to cases where ld64 has
to deduplicate strings with different offset/alignment combos -- it
seems to pick one of their alignments to preserve. This doesn't seem
correct in general; we can in fact can induce ld64 to produce a crashing
binary just by linking in an additional object file that only contains
cstrings and no code. See PR50563 for details.
Moreover, this scheme seems rather inefficient: since unaligned and
aligned strings are all put in the same section, which has a single
alignment value, it doesn't seem possible to tell whether a given string
doesn't have any alignment requirements. Preserving offset+alignments
for strings that don't need it is wasteful.
In practice, the crashes seen so far seem to stem from x86_64 SIMD
operations on cstrings. X86_64 requires SIMD accesses to be
16-byte-aligned. So for now, I'm thinking of just aligning all strings
to 16 bytes on x86_64. This is indeed wasteful, but implementation-wise
it's simpler than preserving per-string alignment+offsets. It also
avoids the aforementioned crash after deduplication of
differently-aligned strings. Finally, the overhead is not huge: using
16-byte alignment (vs no alignment) is only a 0.5% size overhead when
linking chromium_framework.
With these alignment requirements, it doesn't make sense to attempt tail
merging -- most strings will not be eligible since their overlaps aren't
likely to start at a 16-byte boundary. Tail-merging (with alignment) for
chromium_framework only improves size by 0.3%.
It's worth noting that LLD-ELF only does tail merging at `-O2`. By
default (at `-O1`), it just deduplicates w/o tail merging. @thakis has
also mentioned that they saw it regress compressed size in some cases
and therefore turned it off. `ld64` does not seem to do tail merging at
all.
**Performance Numbers**
CString deduplication reduces chromium_framework from 250MB to 242MB, or
about a 3.2% reduction.
Numbers for linking chromium_framework on my 3.2 GHz 16-Core Intel Xeon W:
N Min Max Median Avg Stddev
x 20 3.91 4.03 3.935 3.95 0.034641016
+ 20 3.99 4.14 4.015 4.0365 0.0492336
Difference at 95.0% confidence
0.0865 +/- 0.027245
2.18987% +/- 0.689746%
(Student's t, pooled s = 0.0425673)
As expected, cstring merging incurs some non-trivial overhead.
When passing `--no-literal-merge`, it seems that performance is the
same, i.e. the refactoring in this diff didn't cost us.
N Min Max Median Avg Stddev
x 20 3.91 4.03 3.935 3.95 0.034641016
+ 20 3.89 4.02 3.935 3.9435 0.043197831
No difference proven at 95.0% confidence
Reviewed By: #lld-macho, gkm
Differential Revision: https://reviews.llvm.org/D102964
Jez Ng