Before this commit, GeneralPurposeAllocator could run into incredibly degraded performance in scenarios where the bucket count for a particular size class grew to be large. For example, if exactly `slot_count` allocations of a single size class were performed and then all of them were freed except one, then the bucket for those allocations would have to be kept around indefinitely. If that pattern of allocation were done over and over, then the bucket list for that size class could grow incredibly large.
This allocation pattern has been seen in the wild: https://github.com/Vexu/arocc/issues/508#issuecomment-1738275688
In that case, the length of the bucket list for the `128` size class would grow to tens of thousands of buckets and cause Debug runtime to balloon to ~8 minutes whereas with the c_allocator the Debug runtime would be ~3 seconds.
To address this, there are three different changes happening here:
1. std.Treap is used instead of a doubly linked list for the lists of buckets. This takes the time complexity of searchBucket [used in resize and free] from O(n) to O(log n), but increases the time complexity of insert from O(1) to O(log n) [before, all new buckets would get added to the head of the list]. Note: Any data structure with O(log n) or better search/insert/delete would also work for this use-case.
2. If the 'current' bucket for a size class is full, the list of buckets is never traversed and instead a new bucket is allocated. Previously, traversing the bucket list could only find a non-full bucket in specific circumstances, and only because of a separate optimization that is no longer needed (before, after any resize/free, the affected bucket would be moved to the head of the bucket list to allow searchBucket to perform better on average). Now, the current_bucket for each size class only changes when either (1) the current bucket is emptied/freed, or (2) a new bucket is allocated (due to the current bucket being full or null). Because each bucket's alloc_cursor only moves forward (i.e. slots within a bucket are never re-used), we can therefore always know that any bucket besides the current_bucket will be full, so traversing the list in the hopes of finding an existing non-full bucket is entirely pointless.
3. Size + alignment information for small allocations has been moved into the Bucket data instead of keeping it in a separate HashMap. This offers an improvement over the HashMap since whenever we need to get/modify the length/alignment of an allocation it's extremely likely we will already have calculated any bucket-related information necessary to get the data.
The first change is the most relevant and accounts for most of the benefit here. Also note that the overall functionality of GeneralPurposeAllocator is unchanged.
In the degraded `arocc` case, these changes bring Debug performance from ~8 minutes to ~20 seconds.
Benchmark 1: test-master.bat
Time (mean ± σ): 481.263 s ± 5.440 s [User: 479.159 s, System: 1.937 s]
Range (min … max): 477.416 s … 485.109 s 2 runs
Benchmark 2: test-optim-treap.bat
Time (mean ± σ): 19.639 s ± 0.037 s [User: 18.183 s, System: 1.452 s]
Range (min … max): 19.613 s … 19.665 s 2 runs
Summary
'test-optim-treap.bat' ran
24.51 ± 0.28 times faster than 'test-master.bat'
Note: Much of the time taken on Windows in this particular case is related to gathering stack traces. With `.stack_trace_frames = 0` the runtime goes down to 6.7 seconds, which is a little more than 2.5x slower compared to when the c_allocator is used.
These changes may or mat not introduce a slight performance regression in the average case:
Here's the standard library tests on Windows in Debug mode:
Benchmark 1 (10 runs): std-tests-master.exe
measurement mean ± σ min … max outliers delta
wall_time 16.0s ± 30.8ms 15.9s … 16.1s 1 (10%) 0%
peak_rss 42.8MB ± 8.24KB 42.8MB … 42.8MB 0 ( 0%) 0%
Benchmark 2 (10 runs): std-tests-optim-treap.exe
measurement mean ± σ min … max outliers delta
wall_time 16.2s ± 37.6ms 16.1s … 16.3s 0 ( 0%) 💩+ 1.3% ± 0.2%
peak_rss 42.8MB ± 5.18KB 42.8MB … 42.8MB 0 ( 0%) + 0.1% ± 0.0%
And on Linux:
Benchmark 1: ./test-master
Time (mean ± σ): 16.091 s ± 0.088 s [User: 15.856 s, System: 0.453 s]
Range (min … max): 15.870 s … 16.166 s 10 runs
Benchmark 2: ./test-optim-treap
Time (mean ± σ): 16.028 s ± 0.325 s [User: 15.755 s, System: 0.492 s]
Range (min … max): 15.735 s … 16.709 s 10 runs
Summary
'./test-optim-treap' ran
1.00 ± 0.02 times faster than './test-master'
* add nested packed struct/union behavior tests
* use ptr_info.packed_offset rather than trying to duplicate the logic from Sema.structFieldPtrByIndex()
* use the container_ptr_info.packed_offset to account for non-aligned nested structs.
* dedup type.packedStructFieldBitOffset() and module.structPackedFieldBitOffset()
zig fetch [options] <url>
zig fetch [options] <path>
Fetches a package which is found at <url> or <path> into the global
cache directory, printing the package hash to stdout.
Closes#16972
Related to #14280
Additionally, this commit:
* Adds uncompressed .tar support to package fetching
* Introduces symlink support to package fetching
The 5.11 in uname is not something that is ever updated. There is no
versioning of the illumos system in general. Illumos prefers to rely
on feature detection.
I can't say what Solaris does these days as I do not work at Oracle;
so I left it alone.
- Adds `illumos` to the `Target.Os.Tag` enum. A new function,
`isSolarish` has been added that returns true if the tag is either
Solaris or Illumos. This matches the naming convention found in Rust's
`libc` crate[1].
- Add the tag wherever `.solaris` is being checked against.
- Check for the C pre-processor macro `__illumos__` in CMake to set the
proper target tuple. Illumos distros patch their compilers to have
this in the "built-in" set (verified with `echo | cc -dM -E -`).
Alternatively you could check the output of `uname -o`.
Right now, both Solaris and Illumos import from `c/solaris.zig`. In the
future it may be worth putting the shared ABI bits in a base file, and
mixing that in with specific `c/solaris.zig`/`c/illumos.zig` files.
[1]: 6e02a329a2/src/unix/solarish
My previous change for reading / writing to unions at comptime did not handle
union field read/writes correctly in all cases. Previously, if a field was
written to a union, it would overwrite the entire value. This is problematic
when a field of a larger size is subsequently read, because the value would not
be long enough, causing a panic.
Additionally, the writing behaviour itself was incorrect. Writing to a field of
a packed or extern union should only overwrite the bits corresponding to that
field, allowing for memory reintepretation via field writes / reads.
I addressed these problems as follows:
Add the concept of a "backing type" for extern / packed unions
(`Type.unionBackingType`). For extern unions, this is a `u8` array, for packed
unions it's an integer matching the `bitSize` of the union. Whenever union
memory is read at comptime, it's read as this type.
When union memory is written at comptime, the tag may still be known. If so, the
memory is written using the tagged type. If the tag is unknown (because this
union had previously been read from memory), it's simply written back out as the
backing type.
I added `write_packed` to the `reinterpret` field of
`ComptimePtrMutationKit`. This causes writes of the operand to be packed - which
is necessary when writing to a field of a packed union. Without this, writing a
value to a `u1` field would overwrite the entire byte it occupied.
The final case to address was reading a different (potentially larger) field
from a union when it was written with a known tag. To handle this, a new kind of
bitcast was introduced (`bitCastUnionFieldVal`) which supports reading a larger
field by using a backing buffer that has the unwritten bits set to
undefined. The reason to support this (vs always just writing the union as it's
backing type), is that no reads to larger fields ever occur at comptime, it
would be strictly worse to have spent time writing the full backing type.
Closes#14298
This commit adds support for fetching dependencies over git+http(s)
using a minimal implementation of the Git protocols and formats relevant
to fetching repository data.
Git URLs can be specified in `build.zig.zon` as follows:
```zig
.xml = .{
.url = "git+https://github.com/ianprime0509/zig-xml#7380d59d50f1cd8460fd748b5f6f179306679e2f",
.hash = "122085c1e4045fa9cb69632ff771c56acdb6760f34ca5177e80f70b0b92cd80da3e9",
},
```
The fragment part of the URL may specify a commit ID (SHA1 hash), branch
name, or tag. It is an error to omit the fragment: if this happens, the
compiler will prompt the user to add it, using the commit ID of the HEAD
commit of the repository (that is, the latest commit of the default
branch):
```
Fetch Packages... xml... /var/home/ian/src/zig-gobject/build.zig.zon:6:20: error: url field is missing an explicit ref
.url = "git+https://github.com/ianprime0509/zig-xml",
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
note: try .url = "git+https://github.com/ianprime0509/zig-xml#dfdc044f3271641c7d428dc8ec8cd46423d8b8b6",
```
This implementation currently supports only version 2 of Git's wire
protocol (documented in
[protocol-v2](https://git-scm.com/docs/protocol-v2)), which was first
introduced in Git 2.19 (2018) and made the default in 2.26 (2020).
The wire protocol behaves similarly when used over other transports,
such as SSH and the "Git protocol" (git:// URLs), so it should be
reasonably straightforward to support fetching dependencies from such
URLs if the necessary transports are implemented (e.g. #14295).
(Unmanaged)ArrayList.insert has the same inefficiency as the old insertSlice. With the new addManyAt, the solution is trivial.
Also improves the test "growing memory preserves contents". In the previous implementation, if any changes were made to the ArrayList memory growth policy (function growMemory), the list could end up with enough capacity to not trigger a memory growth, defeating the purpose of the test. The new implementation more robustly triggers a memory growth.
