Files
DucielandClaude Opus 5.5 797673cc37 Phase 2: performance module (decompression, MPQ file cache, Lua strings)
- Decompression: MPQ zlib blocks (DecompressData_WithOptions, 0x661A80)
  decoded with libdeflate 1.26 (vendored, decompression only), one
  decompressor per thread. Like the original, reports the produced size
  through outSize (verified in the client at +0x174).
- MPQ file-lookup cache for File_FindInArchive (0x6549a0), re-implemented
  from WeirdUtils with an SRW lock (lookups also happen on loading
  threads) and stricter archive re-validation on hits.
- Lua strings: table-driven matchclass (built lazily from the original on
  first use), a literal prefilter for string.find / gsub / the gfind
  iterator, and luaS_newlstr with a hash check before comparing bytes.
- Claims WeirdUtils' "weirdperformance" mutex so an old weirdutils.dll never
  runs its custom Lua GC alongside.

Prefilter fix over WeirdUtils: string.gfind does not treat a leading '^'
as an anchor (Lua 5.0 gfind_aux calls match() directly), so '^' is an
ordinary, quantifiable character there. Treating it as an anchor made
patterns like "^*abab" wrongly report no match.

Tests: tests/test_prefilter checks the prefilter against the real Lua
5.0.3 string library (vendored in tests/lua50): 4.7M fuzz checks plus
10M checks of 531 real addon patterns against combat-log and chat lines,
0 violations; the matcher is skipped in about half of real-world calls.
Decompression verified through the hook in the offline harness.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-14 23:51:31 +01:00

58 lines
2.1 KiB
C

#ifndef LIB_X86_DECOMPRESS_IMPL_H
#define LIB_X86_DECOMPRESS_IMPL_H
#include "cpu_features.h"
/*
* BMI2 optimized decompression function.
*
* With gcc and clang we just compile the whole function with
* __attribute__((target("bmi2"))), and the compiler uses bmi2 automatically.
*
* With MSVC, there is no target function attribute, but it's still possible to
* use bmi2 intrinsics explicitly. Currently we mostly don't, but there's a
* case in which we do (see below), so we at least take advantage of that.
* However, MSVC from VS2017 (toolset v141) apparently miscompiles the _bzhi_*()
* intrinsics. It seems to be fixed in VS2022. Hence, use MSVC_PREREQ(1930).
*/
#if defined(__GNUC__) || defined(__clang__) || MSVC_PREREQ(1930)
# define deflate_decompress_bmi2 deflate_decompress_bmi2
# define FUNCNAME deflate_decompress_bmi2
# define ATTRIBUTES _target_attribute("bmi2")
/*
* Even with __attribute__((target("bmi2"))), gcc doesn't reliably use the
* bzhi instruction for 'word & BITMASK(count)'. So use the bzhi intrinsic
* explicitly. EXTRACT_VARBITS() is equivalent to 'word & BITMASK(count)';
* EXTRACT_VARBITS8() is equivalent to 'word & BITMASK((u8)count)'.
* Nevertheless, their implementation using the bzhi intrinsic is identical,
* as the bzhi instruction truncates the count to 8 bits implicitly.
*/
# ifndef __clang__
# ifdef ARCH_X86_64
# define EXTRACT_VARBITS(word, count) _bzhi_u64((word), (count))
# define EXTRACT_VARBITS8(word, count) _bzhi_u64((word), (count))
# else
# define EXTRACT_VARBITS(word, count) _bzhi_u32((word), (count))
# define EXTRACT_VARBITS8(word, count) _bzhi_u32((word), (count))
# endif
# endif
# include "../decompress_template.h"
#endif
#if defined(deflate_decompress_bmi2) && HAVE_BMI2_NATIVE
#define DEFAULT_IMPL deflate_decompress_bmi2
#else
static inline decompress_func_t
arch_select_decompress_func(void)
{
#ifdef deflate_decompress_bmi2
if (HAVE_BMI2(get_x86_cpu_features()))
return deflate_decompress_bmi2;
#endif
return NULL;
}
#define arch_select_decompress_func arch_select_decompress_func
#endif
#endif /* LIB_X86_DECOMPRESS_IMPL_H */