This brings io_uring helper methods to Zig for kernels >= 5.4.
We follow liburing's design decisions so that anyone who is comfortable with
liburing (https://unixism.net/loti/ref-liburing/index.html) will feel at home.
Thanks to @daurnimator for the first draft.
Refs: https://github.com/ziglang/zig/pull/3083
Signed-off-by: Joran Dirk Greef <joran@coil.com>
Speed up a little the slicing-by-8 code path by replacing the
(load+shift+xor)*4 sequence with a single u32 load plus a xor.
Before:
```
iterative: 1018 MiB/s [000000006c3b110d]
small keys: 1075 MiB/s [0035bf3dcac00000]
```
After:
```
iterative: 1114 MiB/s [000000006c3b110d]
small keys: 1324 MiB/s [0035bf3dcac00000]
```
It turns out that the kernel won't read or write more than 0x7fffffff
bytes in a single call, failing with EINVAL when trying to do so.
Adjust the limit and curse whoever is responsible for this.
Closes#6332
A profiler run showed that the main bottleneck was the naive decoding of
the Huffman codes, replacing it with a nice trick borrowed by Zlib gave
a substantial speedup.
Replacing a `%` with a `and (mask-1)` gave another significant
improvement (yay for low hanging fruits).
A few numbers obtained by decompressing a 22M file:
Before:
```
./decompress 2,39s user 0,00s system 99% cpu 2,400 total
```
After:
```
./decompress 0,79s user 0,00s system 99% cpu 0,798 total
````
This commit changes the behavior of stage1 to emit libfoo.so instead
of libfoo.so.0.0.0 when none of the --ver-major, --ver-minor, or
--ver-patch flags are set.
It also makes it possible to create unversioned shared libraries
using the zig build system, changing the version parameter of
addSharedLibrary() to a tagged union.
std.event.Loop does not yet work in single threaded builds. However,
using evented io on a single thread can be very convenient. This commit
allows settind @import("root").event_loop_mode to .single_threaded
in order to allow this without reimplementing the startup code in
start.zig
