$NetBSD$ * Part of patchset to build electron on NetBSD * Based on OpenBSD's chromium patches, and FreeBSD's electron patches --- third_party/lzma_sdk/C/Bra86.c.orig 2025-02-24 19:59:26.000000000 +0000 +++ third_party/lzma_sdk/C/Bra86.c @@ -1,82 +1,187 @@ -/* Bra86.c -- Converter for x86 code (BCJ) -2021-02-09 : Igor Pavlov : Public domain */ +/* Bra86.c -- Branch converter for X86 code (BCJ) +2023-04-02 : Igor Pavlov : Public domain */ #include "Precomp.h" #include "Bra.h" +#include "CpuArch.h" -#define Test86MSByte(b) ((((b) + 1) & 0xFE) == 0) -SizeT x86_Convert(Byte *data, SizeT size, UInt32 ip, UInt32 *state, int encoding) +#if defined(MY_CPU_SIZEOF_POINTER) \ + && ( MY_CPU_SIZEOF_POINTER == 4 \ + || MY_CPU_SIZEOF_POINTER == 8) + #define BR_CONV_USE_OPT_PC_PTR +#endif + +#ifdef BR_CONV_USE_OPT_PC_PTR +#define BR_PC_INIT pc -= (UInt32)(SizeT)p; // (MY_uintptr_t) +#define BR_PC_GET (pc + (UInt32)(SizeT)p) +#else +#define BR_PC_INIT pc += (UInt32)size; +#define BR_PC_GET (pc - (UInt32)(SizeT)(lim - p)) +// #define BR_PC_INIT +// #define BR_PC_GET (pc + (UInt32)(SizeT)(p - data)) +#endif + +#define BR_CONVERT_VAL(v, c) if (encoding) v += c; else v -= c; +// #define BR_CONVERT_VAL(v, c) if (!encoding) c = (UInt32)0 - c; v += c; + +#define Z7_BRANCH_CONV_ST(name) z7_BranchConvSt_ ## name + +#define BR86_NEED_CONV_FOR_MS_BYTE(b) ((((b) + 1) & 0xfe) == 0) + +#ifdef MY_CPU_LE_UNALIGN + #define BR86_PREPARE_BCJ_SCAN const UInt32 v = GetUi32(p) ^ 0xe8e8e8e8; + #define BR86_IS_BCJ_BYTE(n) ((v & ((UInt32)0xfe << (n) * 8)) == 0) +#else + #define BR86_PREPARE_BCJ_SCAN + // bad for MSVC X86 (partial write to byte reg): + #define BR86_IS_BCJ_BYTE(n) ((p[n - 4] & 0xfe) == 0xe8) + // bad for old MSVC (partial write to byte reg): + // #define BR86_IS_BCJ_BYTE(n) (((*p ^ 0xe8) & 0xfe) == 0) +#endif + +static +Z7_FORCE_INLINE +Z7_ATTRIB_NO_VECTOR +Byte *Z7_BRANCH_CONV_ST(X86)(Byte *p, SizeT size, UInt32 pc, UInt32 *state, int encoding) { - SizeT pos = 0; - UInt32 mask = *state & 7; if (size < 5) - return 0; - size -= 4; - ip += 5; + return p; + { + // Byte *p = data; + const Byte *lim = p + size - 4; + unsigned mask = (unsigned)*state; // & 7; +#ifdef BR_CONV_USE_OPT_PC_PTR + /* if BR_CONV_USE_OPT_PC_PTR is defined: we need to adjust (pc) for (+4), + because call/jump offset is relative to the next instruction. + if BR_CONV_USE_OPT_PC_PTR is not defined : we don't need to adjust (pc) for (+4), + because BR_PC_GET uses (pc - (lim - p)), and lim was adjusted for (-4) before. + */ + pc += 4; +#endif + BR_PC_INIT + goto start; - for (;;) + for (;; mask |= 4) { - Byte *p = data + pos; - const Byte *limit = data + size; - for (; p < limit; p++) - if ((*p & 0xFE) == 0xE8) - break; - + // cont: mask |= 4; + start: + if (p >= lim) + goto fin; { - SizeT d = (SizeT)(p - data) - pos; - pos = (SizeT)(p - data); - if (p >= limit) - { - *state = (d > 2 ? 0 : mask >> (unsigned)d); - return pos; - } - if (d > 2) - mask = 0; - else - { - mask >>= (unsigned)d; - if (mask != 0 && (mask > 4 || mask == 3 || Test86MSByte(p[(size_t)(mask >> 1) + 1]))) - { - mask = (mask >> 1) | 4; - pos++; - continue; - } - } + BR86_PREPARE_BCJ_SCAN + p += 4; + if (BR86_IS_BCJ_BYTE(0)) { goto m0; } mask >>= 1; + if (BR86_IS_BCJ_BYTE(1)) { goto m1; } mask >>= 1; + if (BR86_IS_BCJ_BYTE(2)) { goto m2; } mask = 0; + if (BR86_IS_BCJ_BYTE(3)) { goto a3; } } + goto main_loop; + + m0: p--; + m1: p--; + m2: p--; + if (mask == 0) + goto a3; + if (p > lim) + goto fin_p; - if (Test86MSByte(p[4])) + // if (((0x17u >> mask) & 1) == 0) + if (mask > 4 || mask == 3) { - UInt32 v = ((UInt32)p[4] << 24) | ((UInt32)p[3] << 16) | ((UInt32)p[2] << 8) | ((UInt32)p[1]); - UInt32 cur = ip + (UInt32)pos; - pos += 5; - if (encoding) - v += cur; - else - v -= cur; - if (mask != 0) + mask >>= 1; + continue; // goto cont; + } + mask >>= 1; + if (BR86_NEED_CONV_FOR_MS_BYTE(p[mask])) + continue; // goto cont; + // if (!BR86_NEED_CONV_FOR_MS_BYTE(p[3])) continue; // goto cont; + { + UInt32 v = GetUi32(p); + UInt32 c; + v += (1 << 24); if (v & 0xfe000000) continue; // goto cont; + c = BR_PC_GET; + BR_CONVERT_VAL(v, c) { - unsigned sh = (mask & 6) << 2; - if (Test86MSByte((Byte)(v >> sh))) + mask <<= 3; + if (BR86_NEED_CONV_FOR_MS_BYTE(v >> mask)) { - v ^= (((UInt32)0x100 << sh) - 1); - if (encoding) - v += cur; - else - v -= cur; + v ^= (((UInt32)0x100 << mask) - 1); + #ifdef MY_CPU_X86 + // for X86 : we can recalculate (c) to reduce register pressure + c = BR_PC_GET; + #endif + BR_CONVERT_VAL(v, c) } mask = 0; } - p[1] = (Byte)v; - p[2] = (Byte)(v >> 8); - p[3] = (Byte)(v >> 16); - p[4] = (Byte)(0 - ((v >> 24) & 1)); + // v = (v & ((1 << 24) - 1)) - (v & (1 << 24)); + v &= (1 << 25) - 1; v -= (1 << 24); + SetUi32(p, v) + p += 4; + goto main_loop; } - else + + main_loop: + if (p >= lim) + goto fin; + for (;;) { - mask = (mask >> 1) | 4; - pos++; + BR86_PREPARE_BCJ_SCAN + p += 4; + if (BR86_IS_BCJ_BYTE(0)) { goto a0; } + if (BR86_IS_BCJ_BYTE(1)) { goto a1; } + if (BR86_IS_BCJ_BYTE(2)) { goto a2; } + if (BR86_IS_BCJ_BYTE(3)) { goto a3; } + if (p >= lim) + goto fin; + } + + a0: p--; + a1: p--; + a2: p--; + a3: + if (p > lim) + goto fin_p; + // if (!BR86_NEED_CONV_FOR_MS_BYTE(p[3])) continue; // goto cont; + { + UInt32 v = GetUi32(p); + UInt32 c; + v += (1 << 24); if (v & 0xfe000000) continue; // goto cont; + c = BR_PC_GET; + BR_CONVERT_VAL(v, c) + // v = (v & ((1 << 24) - 1)) - (v & (1 << 24)); + v &= (1 << 25) - 1; v -= (1 << 24); + SetUi32(p, v) + p += 4; + goto main_loop; } } + +fin_p: + p--; +fin: + // the following processing for tail is optional and can be commented + /* + lim += 4; + for (; p < lim; p++, mask >>= 1) + if ((*p & 0xfe) == 0xe8) + break; + */ + *state = (UInt32)mask; + return p; + } } + + +#define Z7_BRANCH_CONV_ST_FUNC_IMP(name, m, encoding) \ +Z7_NO_INLINE \ +Z7_ATTRIB_NO_VECTOR \ +Byte *m(name)(Byte *data, SizeT size, UInt32 pc, UInt32 *state) \ + { return Z7_BRANCH_CONV_ST(name)(data, size, pc, state, encoding); } + +Z7_BRANCH_CONV_ST_FUNC_IMP(X86, Z7_BRANCH_CONV_ST_DEC, 0) +#ifndef Z7_EXTRACT_ONLY +Z7_BRANCH_CONV_ST_FUNC_IMP(X86, Z7_BRANCH_CONV_ST_ENC, 1) +#endif