+1 -1

src/external/tracy/COMMIT

··· 1 1 - v0.9 1 1 + v0.9.1

+1

src/external/tracy/TracyClient.cpp

··· 28 28 #include "client/tracy_rpmalloc.cpp" 29 29 #include "client/TracyDxt1.cpp" 30 30 #include "client/TracyAlloc.cpp" 31 31 + #include "client/TracyOverride.cpp" 31 32 32 33 #if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6 33 34 # include "libbacktrace/alloc.cpp"

+4

src/external/tracy/client/TracyCallstack.cpp

··· 227 227 const auto res = GetModuleFileNameA( mod[i], name, 1021 ); 228 228 if( res > 0 ) 229 229 { 230 230 + // This may be a new module loaded since our call to SymInitialize. 231 231 + // Just in case, force DbgHelp to load its pdb ! 232 232 + SymLoadModuleEx(proc, NULL, name, NULL, (DWORD64)info.lpBaseOfDll, info.SizeOfImage, NULL, 0); 233 233 + 230 234 auto ptr = name + res; 231 235 while( ptr > name && *ptr != '\\' && *ptr != '/' ) ptr--; 232 236 if( ptr > name ) ptr++;

+9 -2

src/external/tracy/client/TracyCallstack.hpp

··· 2 2 #define __TRACYCALLSTACK_HPP__ 3 3 4 4 #include "../common/TracyApi.h" 5 5 + #include "../common/TracyForceInline.hpp" 5 6 #include "TracyCallstack.h" 6 7 7 8 #if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 5 ··· 11 12 #endif 12 13 13 14 14 14 - #ifdef TRACY_HAS_CALLSTACK 15 15 + #ifndef TRACY_HAS_CALLSTACK 16 16 + 17 17 + namespace tracy 18 18 + { 19 19 + static tracy_force_inline void* Callstack( int depth ) { return nullptr; } 20 20 + } 21 21 + 22 22 + #else 15 23 16 24 #ifdef TRACY_DEBUGINFOD 17 25 # include <elfutils/debuginfod.h> ··· 21 29 #include <stdint.h> 22 30 23 31 #include "../common/TracyAlloc.hpp" 24 24 - #include "../common/TracyForceInline.hpp" 25 32 26 33 namespace tracy 27 34 {

+26

src/external/tracy/client/TracyOverride.cpp

··· 1 1 + #ifdef TRACY_ENABLE 2 2 + # ifdef __linux__ 3 3 + # include "TracyDebug.hpp" 4 4 + # ifdef TRACY_VERBOSE 5 5 + # include <dlfcn.h> 6 6 + # include <link.h> 7 7 + # endif 8 8 + 9 9 + extern "C" int dlclose( void* hnd ) 10 10 + { 11 11 + #ifdef TRACY_VERBOSE 12 12 + struct link_map* lm; 13 13 + if( dlinfo( hnd, RTLD_DI_LINKMAP, &lm ) == 0 ) 14 14 + { 15 15 + TracyDebug( "Overriding dlclose for %s\n", lm->l_name ); 16 16 + } 17 17 + else 18 18 + { 19 19 + TracyDebug( "Overriding dlclose for unknown object (%s)\n", dlerror() ); 20 20 + } 21 21 + #endif 22 22 + return 0; 23 23 + } 24 24 + 25 25 + # endif 26 26 + #endif

+13 -11

src/external/tracy/client/TracyProfiler.cpp

··· 1415 1415 1416 1416 #if defined _WIN32 && !defined TRACY_UWP && !defined TRACY_NO_CRASH_HANDLER 1417 1417 s_profilerThreadId = GetThreadId( s_thread->Handle() ); 1418 1418 + # ifdef TRACY_HAS_CALLSTACK 1418 1419 s_symbolThreadId = GetThreadId( s_symbolThread->Handle() ); 1420 1420 + # endif 1419 1421 m_exceptionHandler = AddVectoredExceptionHandler( 1, CrashFilter ); 1420 1422 #endif 1421 1423 ··· 1454 1456 if( m_crashHandlerInstalled ) RemoveVectoredExceptionHandler( m_exceptionHandler ); 1455 1457 #endif 1456 1458 1457 1457 - #ifdef __linux__ 1459 1459 + #if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER 1458 1460 if( m_crashHandlerInstalled ) 1459 1461 { 1460 1462 sigaction( TRACY_CRASH_SIGNAL, &m_prevSignal.pwr, nullptr ); ··· 1520 1522 1521 1523 void Profiler::Worker() 1522 1524 { 1523 1523 - #ifdef __linux__ 1525 1525 + #if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER 1524 1526 s_profilerTid = syscall( SYS_gettid ); 1525 1527 #endif 1526 1528 ··· 3024 3026 MemWrite( &item.srcloc.file, (uint64_t)srcloc->file ); 3025 3027 MemWrite( &item.srcloc.function, (uint64_t)srcloc->function ); 3026 3028 MemWrite( &item.srcloc.line, srcloc->line ); 3027 3027 - MemWrite( &item.srcloc.r, uint8_t( ( srcloc->color ) & 0xFF ) ); 3029 3029 + MemWrite( &item.srcloc.b, uint8_t( ( srcloc->color ) & 0xFF ) ); 3028 3030 MemWrite( &item.srcloc.g, uint8_t( ( srcloc->color >> 8 ) & 0xFF ) ); 3029 3029 - MemWrite( &item.srcloc.b, uint8_t( ( srcloc->color >> 16 ) & 0xFF ) ); 3031 3031 + MemWrite( &item.srcloc.r, uint8_t( ( srcloc->color >> 16 ) & 0xFF ) ); 3030 3032 AppendData( &item, QueueDataSize[(int)QueueType::SourceLocation] ); 3031 3033 } 3032 3034 ··· 3329 3331 3330 3332 uint8_t type; 3331 3333 uint64_t ptr; 3332 3332 - uint32_t extra; 3333 3334 memcpy( &type, &payload.type, sizeof( payload.type ) ); 3334 3335 memcpy( &ptr, &payload.ptr, sizeof( payload.ptr ) ); 3335 3335 - memcpy( &extra, &payload.extra, sizeof( payload.extra ) ); 3336 3336 3337 3337 switch( type ) 3338 3338 { ··· 3379 3379 break; 3380 3380 #ifndef TRACY_NO_CODE_TRANSFER 3381 3381 case ServerQuerySymbolCode: 3382 3382 - HandleSymbolCodeQuery( ptr, extra ); 3382 3382 + HandleSymbolCodeQuery( ptr, payload.extra ); 3383 3383 break; 3384 3384 #endif 3385 3385 case ServerQuerySourceCode: ··· 3396 3396 break; 3397 3397 case ServerQueryDataTransferPart: 3398 3398 memcpy( m_queryDataPtr, &ptr, 8 ); 3399 3399 - memcpy( m_queryDataPtr+8, &extra, 4 ); 3399 3399 + memcpy( m_queryDataPtr+8, &payload.extra, 4 ); 3400 3400 m_queryDataPtr += 12; 3401 3401 AckServerQuery(); 3402 3402 break; ··· 3546 3546 constexpr int Events = Iterations * 2; // start + end 3547 3547 static_assert( Events < QueuePrealloc, "Delay calibration loop will allocate memory in queue" ); 3548 3548 3549 3549 - static const tracy::SourceLocationData __tracy_source_location { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; 3549 3549 + static const tracy::SourceLocationData __tracy_source_location { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; 3550 3550 const auto t0 = GetTime(); 3551 3551 for( int i=0; i<Iterations; i++ ) 3552 3552 { ··· 4110 4110 #endif 4111 4111 { 4112 4112 TracyQueuePrepareC( tracy::QueueType::ZoneColor ); 4113 4113 - tracy::MemWrite( &item->zoneColor.r, uint8_t( ( color ) & 0xFF ) ); 4113 4113 + tracy::MemWrite( &item->zoneColor.b, uint8_t( ( color ) & 0xFF ) ); 4114 4114 tracy::MemWrite( &item->zoneColor.g, uint8_t( ( color >> 8 ) & 0xFF ) ); 4115 4115 - tracy::MemWrite( &item->zoneColor.b, uint8_t( ( color >> 16 ) & 0xFF ) ); 4115 4115 + tracy::MemWrite( &item->zoneColor.r, uint8_t( ( color >> 16 ) & 0xFF ) ); 4116 4116 TracyQueueCommitC( zoneColorThread ); 4117 4117 } 4118 4118 } ··· 4147 4147 TRACY_API void ___tracy_emit_frame_mark_end( const char* name ) { tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgEnd ); } 4148 4148 TRACY_API void ___tracy_emit_frame_image( const void* image, uint16_t w, uint16_t h, uint8_t offset, int flip ) { tracy::Profiler::SendFrameImage( image, w, h, offset, flip ); } 4149 4149 TRACY_API void ___tracy_emit_plot( const char* name, double val ) { tracy::Profiler::PlotData( name, val ); } 4150 4150 + TRACY_API void ___tracy_emit_plot_float( const char* name, float val ) { tracy::Profiler::PlotData( name, val ); } 4151 4151 + TRACY_API void ___tracy_emit_plot_int( const char* name, int64_t val ) { tracy::Profiler::PlotData( name, val ); } 4150 4152 TRACY_API void ___tracy_emit_message( const char* txt, size_t size, int callstack ) { tracy::Profiler::Message( txt, size, callstack ); } 4151 4153 TRACY_API void ___tracy_emit_messageL( const char* txt, int callstack ) { tracy::Profiler::Message( txt, callstack ); } 4152 4154 TRACY_API void ___tracy_emit_messageC( const char* txt, size_t size, uint32_t color, int callstack ) { tracy::Profiler::MessageColor( txt, size, color, callstack ); }

+4 -4

src/external/tracy/client/TracyProfiler.hpp

··· 414 414 TracyQueuePrepare( callstack == 0 ? QueueType::MessageColor : QueueType::MessageColorCallstack ); 415 415 MemWrite( &item->messageColorFat.time, GetTime() ); 416 416 MemWrite( &item->messageColorFat.text, (uint64_t)ptr ); 417 417 - MemWrite( &item->messageColorFat.r, uint8_t( ( color ) & 0xFF ) ); 417 417 + MemWrite( &item->messageColorFat.b, uint8_t( ( color ) & 0xFF ) ); 418 418 MemWrite( &item->messageColorFat.g, uint8_t( ( color >> 8 ) & 0xFF ) ); 419 419 - MemWrite( &item->messageColorFat.b, uint8_t( ( color >> 16 ) & 0xFF ) ); 419 419 + MemWrite( &item->messageColorFat.r, uint8_t( ( color >> 16 ) & 0xFF ) ); 420 420 MemWrite( &item->messageColorFat.size, (uint16_t)size ); 421 421 TracyQueueCommit( messageColorFatThread ); 422 422 } ··· 434 434 TracyQueuePrepare( callstack == 0 ? QueueType::MessageLiteralColor : QueueType::MessageLiteralColorCallstack ); 435 435 MemWrite( &item->messageColorLiteral.time, GetTime() ); 436 436 MemWrite( &item->messageColorLiteral.text, (uint64_t)txt ); 437 437 - MemWrite( &item->messageColorLiteral.r, uint8_t( ( color ) & 0xFF ) ); 437 437 + MemWrite( &item->messageColorLiteral.b, uint8_t( ( color ) & 0xFF ) ); 438 438 MemWrite( &item->messageColorLiteral.g, uint8_t( ( color >> 8 ) & 0xFF ) ); 439 439 - MemWrite( &item->messageColorLiteral.b, uint8_t( ( color >> 16 ) & 0xFF ) ); 439 439 + MemWrite( &item->messageColorLiteral.r, uint8_t( ( color >> 16 ) & 0xFF ) ); 440 440 TracyQueueCommit( messageColorLiteralThread ); 441 441 } 442 442

+3 -3

src/external/tracy/client/TracyScoped.hpp

··· 143 143 if( GetProfiler().ConnectionId() != m_connectionId ) return; 144 144 #endif 145 145 TracyQueuePrepare( QueueType::ZoneColor ); 146 146 - MemWrite( &item->zoneColor.r, uint8_t( ( color ) & 0xFF ) ); 146 146 + MemWrite( &item->zoneColor.b, uint8_t( ( color ) & 0xFF ) ); 147 147 MemWrite( &item->zoneColor.g, uint8_t( ( color >> 8 ) & 0xFF ) ); 148 148 - MemWrite( &item->zoneColor.b, uint8_t( ( color >> 16 ) & 0xFF ) ); 148 148 + MemWrite( &item->zoneColor.r, uint8_t( ( color >> 16 ) & 0xFF ) ); 149 149 TracyQueueCommit( zoneColorThread ); 150 150 } 151 151 ··· 166 166 const bool m_active; 167 167 168 168 #ifdef TRACY_ON_DEMAND 169 169 - uint64_t m_connectionId; 169 169 + uint64_t m_connectionId = 0; 170 170 #endif 171 171 }; 172 172

+3 -2

src/external/tracy/client/TracySysTrace.cpp

··· 409 409 return false; 410 410 } 411 411 412 412 + #ifndef TRACY_NO_SAMPLING 412 413 if( isOs64Bit ) 413 414 { 414 415 CLASSIC_EVENT_ID stackId[2] = {}; ··· 423 424 return false; 424 425 } 425 426 } 427 427 + #endif 426 428 427 429 #ifdef UNICODE 428 430 WCHAR KernelLoggerName[sizeof( KERNEL_LOGGER_NAME )]; ··· 1263 1265 type = QueueType::HwSampleBranchMiss; 1264 1266 break; 1265 1267 default: 1266 1266 - assert( false ); 1267 1267 - break; 1268 1268 + abort(); 1268 1269 } 1269 1270 1270 1271 TracyLfqPrepare( type );

+3 -8

src/external/tracy/client/tracy_concurrentqueue.h

··· 210 210 } 211 211 }; 212 212 213 213 - #ifdef _MSC_VER 214 214 - #pragma warning(push) 215 215 - #pragma warning(disable: 4554) 216 216 - #endif 217 213 template<typename T> 218 214 static inline bool circular_less_than(T a, T b) 219 215 { 220 216 static_assert(std::is_integral<T>::value && !std::numeric_limits<T>::is_signed, "circular_less_than is intended to be used only with unsigned integer types"); 221 221 - return static_cast<T>(a - b) > (static_cast<T>(static_cast<T>(1) << static_cast<T>(sizeof(T) * CHAR_BIT - 1))); 217 217 + return static_cast<T>(a - b) > static_cast<T>(static_cast<T>(1) << (static_cast<T>(sizeof(T) * CHAR_BIT - 1))); 218 218 + // Note: extra parens around rhs of operator<< is MSVC bug: https://developercommunity2.visualstudio.com/t/C4554-triggers-when-both-lhs-and-rhs-is/10034931 219 219 + // silencing the bug requires #pragma warning(disable: 4554) around the calling code and has no effect when done here. 222 220 } 223 223 - #ifdef _MSC_VER 224 224 - #pragma warning(pop) 225 225 - #endif 226 221 227 222 template<typename U> 228 223 static inline char* align_for(char* ptr)

+6 -6

src/external/tracy/common/TracyQueue.hpp

··· 166 166 167 167 struct QueueZoneColor 168 168 { 169 169 - uint8_t r; 170 170 - uint8_t g; 171 169 uint8_t b; 170 170 + uint8_t g; 171 171 + uint8_t r; 172 172 }; 173 173 174 174 struct QueueZoneColorThread : public QueueZoneColor ··· 222 222 uint64_t function; // ptr 223 223 uint64_t file; // ptr 224 224 uint32_t line; 225 225 - uint8_t r; 226 226 - uint8_t g; 227 225 uint8_t b; 226 226 + uint8_t g; 227 227 + uint8_t r; 228 228 }; 229 229 230 230 struct QueueZoneTextFat ··· 342 342 343 343 struct QueueMessageColor : public QueueMessage 344 344 { 345 345 - uint8_t r; 346 346 - uint8_t g; 347 345 uint8_t b; 346 346 + uint8_t g; 347 347 + uint8_t r; 348 348 }; 349 349 350 350 struct QueueMessageLiteral : public QueueMessage

+50 -54

src/external/tracy/common/TracySystem.cpp

··· 205 205 } 206 206 ptr = ptr->next; 207 207 } 208 208 - #else 209 209 - # if defined _WIN32 210 210 - # ifdef TRACY_UWP 211 211 - static auto _GetThreadDescription = &::GetThreadDescription; 212 212 - # else 213 213 - static auto _GetThreadDescription = (t_GetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetThreadDescription" ); 214 214 - # endif 215 215 - if( _GetThreadDescription ) 216 216 - { 217 217 - auto hnd = OpenThread( THREAD_QUERY_LIMITED_INFORMATION, FALSE, (DWORD)id ); 218 218 - if( hnd != 0 ) 219 219 - { 220 220 - PWSTR tmp; 221 221 - _GetThreadDescription( hnd, &tmp ); 222 222 - auto ret = wcstombs( buf, tmp, 256 ); 223 223 - CloseHandle( hnd ); 224 224 - if( ret != 0 ) 225 225 - { 226 226 - return buf; 227 227 - } 228 228 - } 229 229 - } 230 230 - # elif defined __linux__ 231 231 - int cs, fd; 232 232 - char path[32]; 233 233 - # ifdef __ANDROID__ 234 234 - int tid = gettid(); 235 235 - # else 236 236 - int tid = (int) syscall( SYS_gettid ); 237 237 - # endif 238 238 - snprintf( path, sizeof( path ), "/proc/self/task/%d/comm", tid ); 239 239 - sprintf( buf, "%" PRIu32, id ); 240 240 - # ifndef __ANDROID__ 241 241 - pthread_setcancelstate( PTHREAD_CANCEL_DISABLE, &cs ); 242 242 - # endif 243 243 - if ( ( fd = open( path, O_RDONLY ) ) > 0) { 244 244 - int len = read( fd, buf, 255 ); 245 245 - if( len > 0 ) 246 246 - { 247 247 - buf[len] = 0; 248 248 - if( len > 1 && buf[len-1] == '\n' ) 249 249 - { 250 250 - buf[len-1] = 0; 251 251 - } 252 252 - } 253 253 - close( fd ); 254 254 - } 255 255 - # ifndef __ANDROID__ 256 256 - pthread_setcancelstate( cs, 0 ); 257 257 - # endif 258 258 - return buf; 259 259 - # endif 260 208 #endif 261 261 - sprintf( buf, "%" PRIu32, id ); 262 262 - return buf; 209 209 + 210 210 + #if defined _WIN32 211 211 + # ifdef TRACY_UWP 212 212 + static auto _GetThreadDescription = &::GetThreadDescription; 213 213 + # else 214 214 + static auto _GetThreadDescription = (t_GetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetThreadDescription" ); 215 215 + # endif 216 216 + if( _GetThreadDescription ) 217 217 + { 218 218 + auto hnd = OpenThread( THREAD_QUERY_LIMITED_INFORMATION, FALSE, (DWORD)id ); 219 219 + if( hnd != 0 ) 220 220 + { 221 221 + PWSTR tmp; 222 222 + _GetThreadDescription( hnd, &tmp ); 223 223 + auto ret = wcstombs( buf, tmp, 256 ); 224 224 + CloseHandle( hnd ); 225 225 + if( ret != 0 ) 226 226 + { 227 227 + return buf; 228 228 + } 229 229 + } 230 230 + } 231 231 + #elif defined __linux__ 232 232 + int cs, fd; 233 233 + char path[32]; 234 234 + snprintf( path, sizeof( path ), "/proc/self/task/%d/comm", id ); 235 235 + sprintf( buf, "%" PRIu32, id ); 236 236 + # ifndef __ANDROID__ 237 237 + pthread_setcancelstate( PTHREAD_CANCEL_DISABLE, &cs ); 238 238 + # endif 239 239 + if ( ( fd = open( path, O_RDONLY ) ) > 0) { 240 240 + int len = read( fd, buf, 255 ); 241 241 + if( len > 0 ) 242 242 + { 243 243 + buf[len] = 0; 244 244 + if( len > 1 && buf[len-1] == '\n' ) 245 245 + { 246 246 + buf[len-1] = 0; 247 247 + } 248 248 + } 249 249 + close( fd ); 250 250 + } 251 251 + # ifndef __ANDROID__ 252 252 + pthread_setcancelstate( cs, 0 ); 253 253 + # endif 254 254 + return buf; 255 255 + #endif 256 256 + 257 257 + sprintf( buf, "%" PRIu32, id ); 258 258 + return buf; 263 259 } 264 260 265 261 TRACY_API const char* GetEnvVar( const char* name )

+1 -1

src/external/tracy/common/TracyVersion.hpp

··· 7 7 { 8 8 enum { Major = 0 }; 9 9 enum { Minor = 9 }; 10 10 - enum { Patch = 0 }; 10 10 + enum { Patch = 1 }; 11 11 } 12 12 } 13 13

+2589 -70

src/external/tracy/libbacktrace/elf.cpp

··· 193 193 #undef STT_FUNC 194 194 #undef NT_GNU_BUILD_ID 195 195 #undef ELFCOMPRESS_ZLIB 196 196 + #undef ELFCOMPRESS_ZSTD 196 197 197 198 /* Basic types. */ 198 199 ··· 350 351 #endif /* BACKTRACE_ELF_SIZE != 32 */ 351 352 352 353 #define ELFCOMPRESS_ZLIB 1 354 354 + #define ELFCOMPRESS_ZSTD 2 353 355 354 356 /* Names of sections, indexed by enum dwarf_section in internal.h. */ 355 357 ··· 1130 1132 on error. */ 1131 1133 1132 1134 static int 1133 1133 - elf_zlib_fetch (const unsigned char **ppin, const unsigned char *pinend, 1135 1135 + elf_fetch_bits (const unsigned char **ppin, const unsigned char *pinend, 1134 1136 uint64_t *pval, unsigned int *pbits) 1135 1137 { 1136 1138 unsigned int bits; ··· 1177 1179 return 1; 1178 1180 } 1179 1181 1182 1182 + /* This is like elf_fetch_bits, but it fetchs the bits backward, and ensures at 1183 1183 + least 16 bits. This is for zstd. */ 1184 1184 + 1185 1185 + static int 1186 1186 + elf_fetch_bits_backward (const unsigned char **ppin, 1187 1187 + const unsigned char *pinend, 1188 1188 + uint64_t *pval, unsigned int *pbits) 1189 1189 + { 1190 1190 + unsigned int bits; 1191 1191 + const unsigned char *pin; 1192 1192 + uint64_t val; 1193 1193 + uint32_t next; 1194 1194 + 1195 1195 + bits = *pbits; 1196 1196 + if (bits >= 16) 1197 1197 + return 1; 1198 1198 + pin = *ppin; 1199 1199 + val = *pval; 1200 1200 + 1201 1201 + if (unlikely (pin <= pinend)) 1202 1202 + { 1203 1203 + if (bits == 0) 1204 1204 + { 1205 1205 + elf_uncompress_failed (); 1206 1206 + return 0; 1207 1207 + } 1208 1208 + return 1; 1209 1209 + } 1210 1210 + 1211 1211 + pin -= 4; 1212 1212 + 1213 1213 + #if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) \ 1214 1214 + && defined(__ORDER_BIG_ENDIAN__) \ 1215 1215 + && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ \ 1216 1216 + || __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) 1217 1217 + /* We've ensured that PIN is aligned. */ 1218 1218 + next = *(const uint32_t *)pin; 1219 1219 + 1220 1220 + #if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ 1221 1221 + next = __builtin_bswap32 (next); 1222 1222 + #endif 1223 1223 + #else 1224 1224 + next = pin[0] | (pin[1] << 8) | (pin[2] << 16) | (pin[3] << 24); 1225 1225 + #endif 1226 1226 + 1227 1227 + val <<= 32; 1228 1228 + val |= next; 1229 1229 + bits += 32; 1230 1230 + 1231 1231 + if (unlikely (pin < pinend)) 1232 1232 + { 1233 1233 + val >>= (pinend - pin) * 8; 1234 1234 + bits -= (pinend - pin) * 8; 1235 1235 + } 1236 1236 + 1237 1237 + *ppin = pin; 1238 1238 + *pval = val; 1239 1239 + *pbits = bits; 1240 1240 + return 1; 1241 1241 + } 1242 1242 + 1243 1243 + /* Initialize backward fetching when the bitstream starts with a 1 bit in the 1244 1244 + last byte in memory (which is the first one that we read). This is used by 1245 1245 + zstd decompression. Returns 1 on success, 0 on error. */ 1246 1246 + 1247 1247 + static int 1248 1248 + elf_fetch_backward_init (const unsigned char **ppin, 1249 1249 + const unsigned char *pinend, 1250 1250 + uint64_t *pval, unsigned int *pbits) 1251 1251 + { 1252 1252 + const unsigned char *pin; 1253 1253 + unsigned int stream_start; 1254 1254 + uint64_t val; 1255 1255 + unsigned int bits; 1256 1256 + 1257 1257 + pin = *ppin; 1258 1258 + stream_start = (unsigned int)*pin; 1259 1259 + if (unlikely (stream_start == 0)) 1260 1260 + { 1261 1261 + elf_uncompress_failed (); 1262 1262 + return 0; 1263 1263 + } 1264 1264 + val = 0; 1265 1265 + bits = 0; 1266 1266 + 1267 1267 + /* Align to a 32-bit boundary. */ 1268 1268 + while ((((uintptr_t)pin) & 3) != 0) 1269 1269 + { 1270 1270 + val <<= 8; 1271 1271 + val |= (uint64_t)*pin; 1272 1272 + bits += 8; 1273 1273 + --pin; 1274 1274 + } 1275 1275 + 1276 1276 + val <<= 8; 1277 1277 + val |= (uint64_t)*pin; 1278 1278 + bits += 8; 1279 1279 + 1280 1280 + *ppin = pin; 1281 1281 + *pval = val; 1282 1282 + *pbits = bits; 1283 1283 + if (!elf_fetch_bits_backward (ppin, pinend, pval, pbits)) 1284 1284 + return 0; 1285 1285 + 1286 1286 + *pbits -= __builtin_clz (stream_start) - (sizeof (unsigned int) - 1) * 8 + 1; 1287 1287 + 1288 1288 + if (!elf_fetch_bits_backward (ppin, pinend, pval, pbits)) 1289 1289 + return 0; 1290 1290 + 1291 1291 + return 1; 1292 1292 + } 1293 1293 + 1180 1294 /* Huffman code tables, like the rest of the zlib format, are defined 1181 1295 by RFC 1951. We store a Huffman code table as a series of tables 1182 1296 stored sequentially in memory. Each entry in a table is 16 bits. ··· 1211 1325 /* Number of entries we allocate to for one code table. We get a page 1212 1326 for the two code tables we need. */ 1213 1327 1214 1214 - #define HUFFMAN_TABLE_SIZE (1024) 1328 1328 + #define ZLIB_HUFFMAN_TABLE_SIZE (1024) 1215 1329 1216 1330 /* Bit masks and shifts for the values in the table. */ 1217 1331 1218 1218 - #define HUFFMAN_VALUE_MASK 0x01ff 1219 1219 - #define HUFFMAN_BITS_SHIFT 9 1220 1220 - #define HUFFMAN_BITS_MASK 0x7 1221 1221 - #define HUFFMAN_SECONDARY_SHIFT 12 1332 1332 + #define ZLIB_HUFFMAN_VALUE_MASK 0x01ff 1333 1333 + #define ZLIB_HUFFMAN_BITS_SHIFT 9 1334 1334 + #define ZLIB_HUFFMAN_BITS_MASK 0x7 1335 1335 + #define ZLIB_HUFFMAN_SECONDARY_SHIFT 12 1222 1336 1223 1337 /* For working memory while inflating we need two code tables, we need 1224 1338 an array of code lengths (max value 15, so we use unsigned char), ··· 1226 1340 latter two arrays must be large enough to hold the maximum number 1227 1341 of code lengths, which RFC 1951 defines as 286 + 30. */ 1228 1342 1229 1229 - #define ZDEBUG_TABLE_SIZE \ 1230 1230 - (2 * HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ 1343 1343 + #define ZLIB_TABLE_SIZE \ 1344 1344 + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ 1231 1345 + (286 + 30) * sizeof (uint16_t) \ 1232 1346 + (286 + 30) * sizeof (unsigned char)) 1233 1347 1234 1234 - #define ZDEBUG_TABLE_CODELEN_OFFSET \ 1235 1235 - (2 * HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ 1348 1348 + #define ZLIB_TABLE_CODELEN_OFFSET \ 1349 1349 + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ 1236 1350 + (286 + 30) * sizeof (uint16_t)) 1237 1351 1238 1238 - #define ZDEBUG_TABLE_WORK_OFFSET \ 1239 1239 - (2 * HUFFMAN_TABLE_SIZE * sizeof (uint16_t)) 1352 1352 + #define ZLIB_TABLE_WORK_OFFSET \ 1353 1353 + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t)) 1240 1354 1241 1355 #ifdef BACKTRACE_GENERATE_FIXED_HUFFMAN_TABLE 1242 1356 ··· 1269 1383 next value after VAL with the same bit length. */ 1270 1384 1271 1385 next = (uint16_t *) (((unsigned char *) zdebug_table) 1272 1272 - + ZDEBUG_TABLE_WORK_OFFSET); 1386 1386 + + ZLIB_TABLE_WORK_OFFSET); 1273 1387 1274 1388 memset (&count[0], 0, 16 * sizeof (uint16_t)); 1275 1389 for (i = 0; i < codes_len; ++i) ··· 1297 1411 /* For each length, fill in the table for the codes of that 1298 1412 length. */ 1299 1413 1300 1300 - memset (table, 0, HUFFMAN_TABLE_SIZE * sizeof (uint16_t)); 1414 1414 + memset (table, 0, ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t)); 1301 1415 1302 1416 /* Handle the values that do not require a secondary table. */ 1303 1417 ··· 1331 1445 /* In the compressed bit stream, the value VAL is encoded as 1332 1446 J bits with the value C. */ 1333 1447 1334 1334 - if (unlikely ((val & ~HUFFMAN_VALUE_MASK) != 0)) 1448 1448 + if (unlikely ((val & ~ZLIB_HUFFMAN_VALUE_MASK) != 0)) 1335 1449 { 1336 1450 elf_uncompress_failed (); 1337 1451 return 0; 1338 1452 } 1339 1453 1340 1340 - tval = val | ((j - 1) << HUFFMAN_BITS_SHIFT); 1454 1454 + tval = val | ((j - 1) << ZLIB_HUFFMAN_BITS_SHIFT); 1341 1455 1342 1456 /* The table lookup uses 8 bits. If J is less than 8, we 1343 1457 don't know what the other bits will be. We need to fill ··· 1487 1601 { 1488 1602 /* Start a new secondary table. */ 1489 1603 1490 1490 - if (unlikely ((next_secondary & HUFFMAN_VALUE_MASK) 1604 1604 + if (unlikely ((next_secondary & ZLIB_HUFFMAN_VALUE_MASK) 1491 1605 != next_secondary)) 1492 1606 { 1493 1607 elf_uncompress_failed (); ··· 1498 1612 secondary_bits = j - 8; 1499 1613 next_secondary += 1 << secondary_bits; 1500 1614 table[primary] = (secondary 1501 1501 - + ((j - 8) << HUFFMAN_BITS_SHIFT) 1502 1502 - + (1U << HUFFMAN_SECONDARY_SHIFT)); 1615 1615 + + ((j - 8) << ZLIB_HUFFMAN_BITS_SHIFT) 1616 1616 + + (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)); 1503 1617 } 1504 1618 else 1505 1619 { 1506 1620 /* There is an existing entry. It had better be a 1507 1621 secondary table with enough bits. */ 1508 1508 - if (unlikely ((tprimary & (1U << HUFFMAN_SECONDARY_SHIFT)) 1622 1622 + if (unlikely ((tprimary 1623 1623 + & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) 1509 1624 == 0)) 1510 1625 { 1511 1626 elf_uncompress_failed (); 1512 1627 return 0; 1513 1628 } 1514 1514 - secondary = tprimary & HUFFMAN_VALUE_MASK; 1515 1515 - secondary_bits = ((tprimary >> HUFFMAN_BITS_SHIFT) 1516 1516 - & HUFFMAN_BITS_MASK); 1629 1629 + secondary = tprimary & ZLIB_HUFFMAN_VALUE_MASK; 1630 1630 + secondary_bits = ((tprimary >> ZLIB_HUFFMAN_BITS_SHIFT) 1631 1631 + & ZLIB_HUFFMAN_BITS_MASK); 1517 1632 if (unlikely (secondary_bits < j - 8)) 1518 1633 { 1519 1634 elf_uncompress_failed (); ··· 1524 1639 1525 1640 /* Fill in secondary table entries. */ 1526 1641 1527 1527 - tval = val | ((j - 8) << HUFFMAN_BITS_SHIFT); 1642 1642 + tval = val | ((j - 8) << ZLIB_HUFFMAN_BITS_SHIFT); 1528 1643 1529 1644 for (ind = code >> 8; 1530 1645 ind < (1U << secondary_bits); ··· 1567 1682 1568 1683 #include <stdio.h> 1569 1684 1570 1570 - static uint16_t table[ZDEBUG_TABLE_SIZE]; 1685 1685 + static uint16_t table[ZLIB_TABLE_SIZE]; 1571 1686 static unsigned char codes[288]; 1572 1687 1573 1688 int ··· 1795 1910 const uint16_t *tlit; 1796 1911 const uint16_t *tdist; 1797 1912 1798 1798 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 1913 1913 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 1799 1914 return 0; 1800 1915 1801 1916 last = val & 1; ··· 1883 1998 /* Read a Huffman encoding table. The various magic 1884 1999 numbers here are from RFC 1951. */ 1885 2000 1886 1886 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2001 2001 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 1887 2002 return 0; 1888 2003 1889 2004 nlit = (val & 0x1f) + 257; ··· 1908 2023 /* There are always at least 4 elements in the 1909 2024 table. */ 1910 2025 1911 1911 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2026 2026 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 1912 2027 return 0; 1913 2028 1914 2029 codebits[16] = val & 7; ··· 1928 2043 if (nclen == 5) 1929 2044 goto codebitsdone; 1930 2045 1931 1931 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2046 2046 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 1932 2047 return 0; 1933 2048 1934 2049 codebits[7] = val & 7; ··· 1966 2081 if (nclen == 10) 1967 2082 goto codebitsdone; 1968 2083 1969 1969 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2084 2084 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 1970 2085 return 0; 1971 2086 1972 2087 codebits[11] = val & 7; ··· 2004 2119 if (nclen == 15) 2005 2120 goto codebitsdone; 2006 2121 2007 2007 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2122 2122 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2008 2123 return 0; 2009 2124 2010 2125 codebits[2] = val & 7; ··· 2043 2158 at the end of zdebug_table to hold them. */ 2044 2159 2045 2160 plenbase = (((unsigned char *) zdebug_table) 2046 2046 - + ZDEBUG_TABLE_CODELEN_OFFSET); 2161 2161 + + ZLIB_TABLE_CODELEN_OFFSET); 2047 2162 plen = plenbase; 2048 2163 plenend = plen + nlit + ndist; 2049 2164 while (plen < plenend) ··· 2052 2167 unsigned int b; 2053 2168 uint16_t v; 2054 2169 2055 2055 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2170 2170 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2056 2171 return 0; 2057 2172 2058 2173 t = zdebug_table[val & 0xff]; 2059 2174 2060 2175 /* The compression here uses bit lengths up to 7, so 2061 2176 a secondary table is never necessary. */ 2062 2062 - if (unlikely ((t & (1U << HUFFMAN_SECONDARY_SHIFT)) != 0)) 2177 2177 + if (unlikely ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) 2178 2178 + != 0)) 2063 2179 { 2064 2180 elf_uncompress_failed (); 2065 2181 return 0; 2066 2182 } 2067 2183 2068 2068 - b = (t >> HUFFMAN_BITS_SHIFT) & HUFFMAN_BITS_MASK; 2184 2184 + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; 2069 2185 val >>= b + 1; 2070 2186 bits -= b + 1; 2071 2187 2072 2072 - v = t & HUFFMAN_VALUE_MASK; 2188 2188 + v = t & ZLIB_HUFFMAN_VALUE_MASK; 2073 2189 if (v < 16) 2074 2190 *plen++ = v; 2075 2191 else if (v == 16) ··· 2086 2202 } 2087 2203 2088 2204 /* We used up to 7 bits since the last 2089 2089 - elf_zlib_fetch, so we have at least 8 bits 2205 2205 + elf_fetch_bits, so we have at least 8 bits 2090 2206 available here. */ 2091 2207 2092 2208 c = 3 + (val & 0x3); ··· 2121 2237 /* Store zero 3 to 10 times. */ 2122 2238 2123 2239 /* We used up to 7 bits since the last 2124 2124 - elf_zlib_fetch, so we have at least 8 bits 2240 2240 + elf_fetch_bits, so we have at least 8 bits 2125 2241 available here. */ 2126 2242 2127 2243 c = 3 + (val & 0x7); ··· 2167 2283 /* Store zero 11 to 138 times. */ 2168 2284 2169 2285 /* We used up to 7 bits since the last 2170 2170 - elf_zlib_fetch, so we have at least 8 bits 2286 2286 + elf_fetch_bits, so we have at least 8 bits 2171 2287 available here. */ 2172 2288 2173 2289 c = 11 + (val & 0x7f); ··· 2204 2320 zdebug_table)) 2205 2321 return 0; 2206 2322 if (!elf_zlib_inflate_table (plen + nlit, ndist, zdebug_table, 2207 2207 - zdebug_table + HUFFMAN_TABLE_SIZE)) 2323 2323 + (zdebug_table 2324 2324 + + ZLIB_HUFFMAN_TABLE_SIZE))) 2208 2325 return 0; 2209 2326 tlit = zdebug_table; 2210 2210 - tdist = zdebug_table + HUFFMAN_TABLE_SIZE; 2327 2327 + tdist = zdebug_table + ZLIB_HUFFMAN_TABLE_SIZE; 2211 2328 } 2212 2329 2213 2330 /* Inflate values until the end of the block. This is the ··· 2220 2337 uint16_t v; 2221 2338 unsigned int lit; 2222 2339 2223 2223 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2340 2340 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2224 2341 return 0; 2225 2342 2226 2343 t = tlit[val & 0xff]; 2227 2227 - b = (t >> HUFFMAN_BITS_SHIFT) & HUFFMAN_BITS_MASK; 2228 2228 - v = t & HUFFMAN_VALUE_MASK; 2344 2344 + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; 2345 2345 + v = t & ZLIB_HUFFMAN_VALUE_MASK; 2229 2346 2230 2230 - if ((t & (1U << HUFFMAN_SECONDARY_SHIFT)) == 0) 2347 2347 + if ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) == 0) 2231 2348 { 2232 2349 lit = v; 2233 2350 val >>= b + 1; ··· 2236 2353 else 2237 2354 { 2238 2355 t = tlit[v + 0x100 + ((val >> 8) & ((1U << b) - 1))]; 2239 2239 - b = (t >> HUFFMAN_BITS_SHIFT) & HUFFMAN_BITS_MASK; 2240 2240 - lit = t & HUFFMAN_VALUE_MASK; 2356 2356 + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; 2357 2357 + lit = t & ZLIB_HUFFMAN_VALUE_MASK; 2241 2358 val >>= b + 8; 2242 2359 bits -= b + 8; 2243 2360 } ··· 2282 2399 { 2283 2400 unsigned int extra; 2284 2401 2285 2285 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2402 2402 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2286 2403 return 0; 2287 2404 2288 2405 /* This is an expression for the table of length ··· 2297 2414 bits -= extra; 2298 2415 } 2299 2416 2300 2300 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2417 2417 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2301 2418 return 0; 2302 2419 2303 2420 t = tdist[val & 0xff]; 2304 2304 - b = (t >> HUFFMAN_BITS_SHIFT) & HUFFMAN_BITS_MASK; 2305 2305 - v = t & HUFFMAN_VALUE_MASK; 2421 2421 + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; 2422 2422 + v = t & ZLIB_HUFFMAN_VALUE_MASK; 2306 2423 2307 2307 - if ((t & (1U << HUFFMAN_SECONDARY_SHIFT)) == 0) 2424 2424 + if ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) == 0) 2308 2425 { 2309 2426 dist = v; 2310 2427 val >>= b + 1; ··· 2313 2430 else 2314 2431 { 2315 2432 t = tdist[v + 0x100 + ((val >> 8) & ((1U << b) - 1))]; 2316 2316 - b = (t >> HUFFMAN_BITS_SHIFT) & HUFFMAN_BITS_MASK; 2317 2317 - dist = t & HUFFMAN_VALUE_MASK; 2433 2433 + b = ((t >> ZLIB_HUFFMAN_BITS_SHIFT) 2434 2434 + & ZLIB_HUFFMAN_BITS_MASK); 2435 2435 + dist = t & ZLIB_HUFFMAN_VALUE_MASK; 2318 2436 val >>= b + 8; 2319 2437 bits -= b + 8; 2320 2438 } ··· 2354 2472 { 2355 2473 unsigned int extra; 2356 2474 2357 2357 - if (!elf_zlib_fetch (&pin, pinend, &val, &bits)) 2475 2475 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2358 2476 return 0; 2359 2477 2360 2478 /* This is an expression for the table of ··· 2559 2677 return 1; 2560 2678 } 2561 2679 2680 2680 + /* For working memory during zstd compression, we need 2681 2681 + - a literal length FSE table: 512 64-bit values == 4096 bytes 2682 2682 + - a match length FSE table: 512 64-bit values == 4096 bytes 2683 2683 + - a offset FSE table: 256 64-bit values == 2048 bytes 2684 2684 + - a Huffman tree: 2048 uint16_t values == 4096 bytes 2685 2685 + - scratch space, one of 2686 2686 + - to build an FSE table: 512 uint16_t values == 1024 bytes 2687 2687 + - to build a Huffman tree: 512 uint16_t + 256 uint32_t == 2048 bytes 2688 2688 + */ 2689 2689 + 2690 2690 + #define ZSTD_TABLE_SIZE \ 2691 2691 + (2 * 512 * sizeof (struct elf_zstd_fse_baseline_entry) \ 2692 2692 + + 256 * sizeof (struct elf_zstd_fse_baseline_entry) \ 2693 2693 + + 2048 * sizeof (uint16_t) \ 2694 2694 + + 512 * sizeof (uint16_t) + 256 * sizeof (uint32_t)) 2695 2695 + 2696 2696 + #define ZSTD_TABLE_LITERAL_FSE_OFFSET (0) 2697 2697 + 2698 2698 + #define ZSTD_TABLE_MATCH_FSE_OFFSET \ 2699 2699 + (512 * sizeof (struct elf_zstd_fse_baseline_entry)) 2700 2700 + 2701 2701 + #define ZSTD_TABLE_OFFSET_FSE_OFFSET \ 2702 2702 + (ZSTD_TABLE_MATCH_FSE_OFFSET \ 2703 2703 + + 512 * sizeof (struct elf_zstd_fse_baseline_entry)) 2704 2704 + 2705 2705 + #define ZSTD_TABLE_HUFFMAN_OFFSET \ 2706 2706 + (ZSTD_TABLE_OFFSET_FSE_OFFSET \ 2707 2707 + + 256 * sizeof (struct elf_zstd_fse_baseline_entry)) 2708 2708 + 2709 2709 + #define ZSTD_TABLE_WORK_OFFSET \ 2710 2710 + (ZSTD_TABLE_HUFFMAN_OFFSET + 2048 * sizeof (uint16_t)) 2711 2711 + 2712 2712 + /* An entry in a zstd FSE table. */ 2713 2713 + 2714 2714 + struct elf_zstd_fse_entry 2715 2715 + { 2716 2716 + /* The value that this FSE entry represents. */ 2717 2717 + unsigned char symbol; 2718 2718 + /* The number of bits to read to determine the next state. */ 2719 2719 + unsigned char bits; 2720 2720 + /* Add the bits to this base to get the next state. */ 2721 2721 + uint16_t base; 2722 2722 + }; 2723 2723 + 2724 2724 + static int 2725 2725 + elf_zstd_build_fse (const int16_t *, int, uint16_t *, int, 2726 2726 + struct elf_zstd_fse_entry *); 2727 2727 + 2728 2728 + /* Read a zstd FSE table and build the decoding table in *TABLE, updating *PPIN 2729 2729 + as it reads. ZDEBUG_TABLE is scratch space; it must be enough for 512 2730 2730 + uint16_t values (1024 bytes). MAXIDX is the maximum number of symbols 2731 2731 + permitted. *TABLE_BITS is the maximum number of bits for symbols in the 2732 2732 + table: the size of *TABLE is at least 1 << *TABLE_BITS. This updates 2733 2733 + *TABLE_BITS to the actual number of bits. Returns 1 on success, 0 on 2734 2734 + error. */ 2735 2735 + 2736 2736 + static int 2737 2737 + elf_zstd_read_fse (const unsigned char **ppin, const unsigned char *pinend, 2738 2738 + uint16_t *zdebug_table, int maxidx, 2739 2739 + struct elf_zstd_fse_entry *table, int *table_bits) 2740 2740 + { 2741 2741 + const unsigned char *pin; 2742 2742 + int16_t *norm; 2743 2743 + uint16_t *next; 2744 2744 + uint64_t val; 2745 2745 + unsigned int bits; 2746 2746 + int accuracy_log; 2747 2747 + uint32_t remaining; 2748 2748 + uint32_t threshold; 2749 2749 + int bits_needed; 2750 2750 + int idx; 2751 2751 + int prev0; 2752 2752 + 2753 2753 + pin = *ppin; 2754 2754 + 2755 2755 + norm = (int16_t *) zdebug_table; 2756 2756 + next = zdebug_table + 256; 2757 2757 + 2758 2758 + if (unlikely (pin + 3 >= pinend)) 2759 2759 + { 2760 2760 + elf_uncompress_failed (); 2761 2761 + return 0; 2762 2762 + } 2763 2763 + 2764 2764 + /* Align PIN to a 32-bit boundary. */ 2765 2765 + 2766 2766 + val = 0; 2767 2767 + bits = 0; 2768 2768 + while ((((uintptr_t) pin) & 3) != 0) 2769 2769 + { 2770 2770 + val |= (uint64_t)*pin << bits; 2771 2771 + bits += 8; 2772 2772 + ++pin; 2773 2773 + } 2774 2774 + 2775 2775 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2776 2776 + return 0; 2777 2777 + 2778 2778 + accuracy_log = (val & 0xf) + 5; 2779 2779 + if (accuracy_log > *table_bits) 2780 2780 + { 2781 2781 + elf_uncompress_failed (); 2782 2782 + return 0; 2783 2783 + } 2784 2784 + *table_bits = accuracy_log; 2785 2785 + val >>= 4; 2786 2786 + bits -= 4; 2787 2787 + 2788 2788 + /* This code is mostly copied from the reference implementation. */ 2789 2789 + 2790 2790 + /* The number of remaining probabilities, plus 1. This sets the number of 2791 2791 + bits that need to be read for the next value. */ 2792 2792 + remaining = (1 << accuracy_log) + 1; 2793 2793 + 2794 2794 + /* The current difference between small and large values, which depends on 2795 2795 + the number of remaining values. Small values use one less bit. */ 2796 2796 + threshold = 1 << accuracy_log; 2797 2797 + 2798 2798 + /* The number of bits used to compute threshold. */ 2799 2799 + bits_needed = accuracy_log + 1; 2800 2800 + 2801 2801 + /* The next character value. */ 2802 2802 + idx = 0; 2803 2803 + 2804 2804 + /* Whether the last count was 0. */ 2805 2805 + prev0 = 0; 2806 2806 + 2807 2807 + while (remaining > 1 && idx <= maxidx) 2808 2808 + { 2809 2809 + uint32_t max; 2810 2810 + int32_t count; 2811 2811 + 2812 2812 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2813 2813 + return 0; 2814 2814 + 2815 2815 + if (prev0) 2816 2816 + { 2817 2817 + int zidx; 2818 2818 + 2819 2819 + /* Previous count was 0, so there is a 2-bit repeat flag. If the 2820 2820 + 2-bit flag is 0b11, it adds 3 and then there is another repeat 2821 2821 + flag. */ 2822 2822 + zidx = idx; 2823 2823 + while ((val & 0xfff) == 0xfff) 2824 2824 + { 2825 2825 + zidx += 3 * 6; 2826 2826 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2827 2827 + return 0; 2828 2828 + val >>= 12; 2829 2829 + bits -= 12; 2830 2830 + } 2831 2831 + while ((val & 3) == 3) 2832 2832 + { 2833 2833 + zidx += 3; 2834 2834 + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) 2835 2835 + return 0; 2836 2836 + val >>= 2; 2837 2837 + bits -= 2; 2838 2838 + } 2839 2839 + /* We have at least 13 bits here, don't need to fetch. */ 2840 2840 + zidx += val & 3; 2841 2841 + val >>= 2; 2842 2842 + bits -= 2; 2843 2843 + 2844 2844 + if (unlikely (zidx > maxidx)) 2845 2845 + { 2846 2846 + elf_uncompress_failed (); 2847 2847 + return 0; 2848 2848 + } 2849 2849 + 2850 2850 + for (; idx < zidx; idx++) 2851 2851 + norm[idx] = 0; 2852 2852 + 2853 2853 + prev0 = 0; 2854 2854 + continue; 2855 2855 + } 2856 2856 + 2857 2857 + max = (2 * threshold - 1) - remaining; 2858 2858 + if ((val & (threshold - 1)) < max) 2859 2859 + { 2860 2860 + /* A small value. */ 2861 2861 + count = (int32_t) ((uint32_t) val & (threshold - 1)); 2862 2862 + val >>= bits_needed - 1; 2863 2863 + bits -= bits_needed - 1; 2864 2864 + } 2865 2865 + else 2866 2866 + { 2867 2867 + /* A large value. */ 2868 2868 + count = (int32_t) ((uint32_t) val & (2 * threshold - 1)); 2869 2869 + if (count >= (int32_t) threshold) 2870 2870 + count -= (int32_t) max; 2871 2871 + val >>= bits_needed; 2872 2872 + bits -= bits_needed; 2873 2873 + } 2874 2874 + 2875 2875 + count--; 2876 2876 + if (count >= 0) 2877 2877 + remaining -= count; 2878 2878 + else 2879 2879 + remaining--; 2880 2880 + if (unlikely (idx >= 256)) 2881 2881 + { 2882 2882 + elf_uncompress_failed (); 2883 2883 + return 0; 2884 2884 + } 2885 2885 + norm[idx] = (int16_t) count; 2886 2886 + ++idx; 2887 2887 + 2888 2888 + prev0 = count == 0; 2889 2889 + 2890 2890 + while (remaining < threshold) 2891 2891 + { 2892 2892 + bits_needed--; 2893 2893 + threshold >>= 1; 2894 2894 + } 2895 2895 + } 2896 2896 + 2897 2897 + if (unlikely (remaining != 1)) 2898 2898 + { 2899 2899 + elf_uncompress_failed (); 2900 2900 + return 0; 2901 2901 + } 2902 2902 + 2903 2903 + /* If we've read ahead more than a byte, back up. */ 2904 2904 + while (bits >= 8) 2905 2905 + { 2906 2906 + --pin; 2907 2907 + bits -= 8; 2908 2908 + } 2909 2909 + 2910 2910 + *ppin = pin; 2911 2911 + 2912 2912 + for (; idx <= maxidx; idx++) 2913 2913 + norm[idx] = 0; 2914 2914 + 2915 2915 + return elf_zstd_build_fse (norm, idx, next, *table_bits, table); 2916 2916 + } 2917 2917 + 2918 2918 + /* Build the FSE decoding table from a list of probabilities. This reads from 2919 2919 + NORM of length IDX, uses NEXT as scratch space, and writes to *TABLE, whose 2920 2920 + size is TABLE_BITS. */ 2921 2921 + 2922 2922 + static int 2923 2923 + elf_zstd_build_fse (const int16_t *norm, int idx, uint16_t *next, 2924 2924 + int table_bits, struct elf_zstd_fse_entry *table) 2925 2925 + { 2926 2926 + int table_size; 2927 2927 + int high_threshold; 2928 2928 + int i; 2929 2929 + int pos; 2930 2930 + int step; 2931 2931 + int mask; 2932 2932 + 2933 2933 + table_size = 1 << table_bits; 2934 2934 + high_threshold = table_size - 1; 2935 2935 + for (i = 0; i < idx; i++) 2936 2936 + { 2937 2937 + int16_t n; 2938 2938 + 2939 2939 + n = norm[i]; 2940 2940 + if (n >= 0) 2941 2941 + next[i] = (uint16_t) n; 2942 2942 + else 2943 2943 + { 2944 2944 + table[high_threshold].symbol = (unsigned char) i; 2945 2945 + high_threshold--; 2946 2946 + next[i] = 1; 2947 2947 + } 2948 2948 + } 2949 2949 + 2950 2950 + pos = 0; 2951 2951 + step = (table_size >> 1) + (table_size >> 3) + 3; 2952 2952 + mask = table_size - 1; 2953 2953 + for (i = 0; i < idx; i++) 2954 2954 + { 2955 2955 + int n; 2956 2956 + int j; 2957 2957 + 2958 2958 + n = (int) norm[i]; 2959 2959 + for (j = 0; j < n; j++) 2960 2960 + { 2961 2961 + table[pos].symbol = (unsigned char) i; 2962 2962 + pos = (pos + step) & mask; 2963 2963 + while (unlikely (pos > high_threshold)) 2964 2964 + pos = (pos + step) & mask; 2965 2965 + } 2966 2966 + } 2967 2967 + if (pos != 0) 2968 2968 + { 2969 2969 + elf_uncompress_failed (); 2970 2970 + return 0; 2971 2971 + } 2972 2972 + 2973 2973 + for (i = 0; i < table_size; i++) 2974 2974 + { 2975 2975 + unsigned char sym; 2976 2976 + uint16_t next_state; 2977 2977 + int high_bit; 2978 2978 + int bits; 2979 2979 + 2980 2980 + sym = table[i].symbol; 2981 2981 + next_state = next[sym]; 2982 2982 + ++next[sym]; 2983 2983 + 2984 2984 + if (next_state == 0) 2985 2985 + { 2986 2986 + elf_uncompress_failed (); 2987 2987 + return 0; 2988 2988 + } 2989 2989 + high_bit = 31 - __builtin_clz (next_state); 2990 2990 + 2991 2991 + bits = table_bits - high_bit; 2992 2992 + table[i].bits = (unsigned char) bits; 2993 2993 + table[i].base = (uint16_t) ((next_state << bits) - table_size); 2994 2994 + } 2995 2995 + 2996 2996 + return 1; 2997 2997 + } 2998 2998 + 2999 2999 + /* Encode the baseline and bits into a single 32-bit value. */ 3000 3000 + 3001 3001 + #define ZSTD_ENCODE_BASELINE_BITS(baseline, basebits) \ 3002 3002 + ((uint32_t)(baseline) | ((uint32_t)(basebits) << 24)) 3003 3003 + 3004 3004 + #define ZSTD_DECODE_BASELINE(baseline_basebits) \ 3005 3005 + ((uint32_t)(baseline_basebits) & 0xffffff) 3006 3006 + 3007 3007 + #define ZSTD_DECODE_BASEBITS(baseline_basebits) \ 3008 3008 + ((uint32_t)(baseline_basebits) >> 24) 3009 3009 + 3010 3010 + /* Given a literal length code, we need to read a number of bits and add that 3011 3011 + to a baseline. For states 0 to 15 the baseline is the state and the number 3012 3012 + of bits is zero. */ 3013 3013 + 3014 3014 + #define ZSTD_LITERAL_LENGTH_BASELINE_OFFSET (16) 3015 3015 + 3016 3016 + static const uint32_t elf_zstd_literal_length_base[] = 3017 3017 + { 3018 3018 + ZSTD_ENCODE_BASELINE_BITS(16, 1), 3019 3019 + ZSTD_ENCODE_BASELINE_BITS(18, 1), 3020 3020 + ZSTD_ENCODE_BASELINE_BITS(20, 1), 3021 3021 + ZSTD_ENCODE_BASELINE_BITS(22, 1), 3022 3022 + ZSTD_ENCODE_BASELINE_BITS(24, 2), 3023 3023 + ZSTD_ENCODE_BASELINE_BITS(28, 2), 3024 3024 + ZSTD_ENCODE_BASELINE_BITS(32, 3), 3025 3025 + ZSTD_ENCODE_BASELINE_BITS(40, 3), 3026 3026 + ZSTD_ENCODE_BASELINE_BITS(48, 4), 3027 3027 + ZSTD_ENCODE_BASELINE_BITS(64, 6), 3028 3028 + ZSTD_ENCODE_BASELINE_BITS(128, 7), 3029 3029 + ZSTD_ENCODE_BASELINE_BITS(256, 8), 3030 3030 + ZSTD_ENCODE_BASELINE_BITS(512, 9), 3031 3031 + ZSTD_ENCODE_BASELINE_BITS(1024, 10), 3032 3032 + ZSTD_ENCODE_BASELINE_BITS(2048, 11), 3033 3033 + ZSTD_ENCODE_BASELINE_BITS(4096, 12), 3034 3034 + ZSTD_ENCODE_BASELINE_BITS(8192, 13), 3035 3035 + ZSTD_ENCODE_BASELINE_BITS(16384, 14), 3036 3036 + ZSTD_ENCODE_BASELINE_BITS(32768, 15), 3037 3037 + ZSTD_ENCODE_BASELINE_BITS(65536, 16) 3038 3038 + }; 3039 3039 + 3040 3040 + /* The same applies to match length codes. For states 0 to 31 the baseline is 3041 3041 + the state + 3 and the number of bits is zero. */ 3042 3042 + 3043 3043 + #define ZSTD_MATCH_LENGTH_BASELINE_OFFSET (32) 3044 3044 + 3045 3045 + static const uint32_t elf_zstd_match_length_base[] = 3046 3046 + { 3047 3047 + ZSTD_ENCODE_BASELINE_BITS(35, 1), 3048 3048 + ZSTD_ENCODE_BASELINE_BITS(37, 1), 3049 3049 + ZSTD_ENCODE_BASELINE_BITS(39, 1), 3050 3050 + ZSTD_ENCODE_BASELINE_BITS(41, 1), 3051 3051 + ZSTD_ENCODE_BASELINE_BITS(43, 2), 3052 3052 + ZSTD_ENCODE_BASELINE_BITS(47, 2), 3053 3053 + ZSTD_ENCODE_BASELINE_BITS(51, 3), 3054 3054 + ZSTD_ENCODE_BASELINE_BITS(59, 3), 3055 3055 + ZSTD_ENCODE_BASELINE_BITS(67, 4), 3056 3056 + ZSTD_ENCODE_BASELINE_BITS(83, 4), 3057 3057 + ZSTD_ENCODE_BASELINE_BITS(99, 5), 3058 3058 + ZSTD_ENCODE_BASELINE_BITS(131, 7), 3059 3059 + ZSTD_ENCODE_BASELINE_BITS(259, 8), 3060 3060 + ZSTD_ENCODE_BASELINE_BITS(515, 9), 3061 3061 + ZSTD_ENCODE_BASELINE_BITS(1027, 10), 3062 3062 + ZSTD_ENCODE_BASELINE_BITS(2051, 11), 3063 3063 + ZSTD_ENCODE_BASELINE_BITS(4099, 12), 3064 3064 + ZSTD_ENCODE_BASELINE_BITS(8195, 13), 3065 3065 + ZSTD_ENCODE_BASELINE_BITS(16387, 14), 3066 3066 + ZSTD_ENCODE_BASELINE_BITS(32771, 15), 3067 3067 + ZSTD_ENCODE_BASELINE_BITS(65539, 16) 3068 3068 + }; 3069 3069 + 3070 3070 + /* An entry in an FSE table used for literal/match/length values. For these we 3071 3071 + have to map the symbol to a baseline value, and we have to read zero or more 3072 3072 + bits and add that value to the baseline value. Rather than look the values 3073 3073 + up in a separate table, we grow the FSE table so that we get better memory 3074 3074 + caching. */ 3075 3075 + 3076 3076 + struct elf_zstd_fse_baseline_entry 3077 3077 + { 3078 3078 + /* The baseline for the value that this FSE entry represents.. */ 3079 3079 + uint32_t baseline; 3080 3080 + /* The number of bits to read to add to the baseline. */ 3081 3081 + unsigned char basebits; 3082 3082 + /* The number of bits to read to determine the next state. */ 3083 3083 + unsigned char bits; 3084 3084 + /* Add the bits to this base to get the next state. */ 3085 3085 + uint16_t base; 3086 3086 + }; 3087 3087 + 3088 3088 + /* Convert the literal length FSE table FSE_TABLE to an FSE baseline table at 3089 3089 + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ 3090 3090 + 3091 3091 + static int 3092 3092 + elf_zstd_make_literal_baseline_fse ( 3093 3093 + const struct elf_zstd_fse_entry *fse_table, 3094 3094 + int table_bits, 3095 3095 + struct elf_zstd_fse_baseline_entry *baseline_table) 3096 3096 + { 3097 3097 + size_t count; 3098 3098 + const struct elf_zstd_fse_entry *pfse; 3099 3099 + struct elf_zstd_fse_baseline_entry *pbaseline; 3100 3100 + 3101 3101 + /* Convert backward to avoid overlap. */ 3102 3102 + 3103 3103 + count = 1U << table_bits; 3104 3104 + pfse = fse_table + count; 3105 3105 + pbaseline = baseline_table + count; 3106 3106 + while (pfse > fse_table) 3107 3107 + { 3108 3108 + unsigned char symbol; 3109 3109 + unsigned char bits; 3110 3110 + uint16_t base; 3111 3111 + 3112 3112 + --pfse; 3113 3113 + --pbaseline; 3114 3114 + symbol = pfse->symbol; 3115 3115 + bits = pfse->bits; 3116 3116 + base = pfse->base; 3117 3117 + if (symbol < ZSTD_LITERAL_LENGTH_BASELINE_OFFSET) 3118 3118 + { 3119 3119 + pbaseline->baseline = (uint32_t)symbol; 3120 3120 + pbaseline->basebits = 0; 3121 3121 + } 3122 3122 + else 3123 3123 + { 3124 3124 + unsigned int idx; 3125 3125 + uint32_t basebits; 3126 3126 + 3127 3127 + if (unlikely (symbol > 35)) 3128 3128 + { 3129 3129 + elf_uncompress_failed (); 3130 3130 + return 0; 3131 3131 + } 3132 3132 + idx = symbol - ZSTD_LITERAL_LENGTH_BASELINE_OFFSET; 3133 3133 + basebits = elf_zstd_literal_length_base[idx]; 3134 3134 + pbaseline->baseline = ZSTD_DECODE_BASELINE(basebits); 3135 3135 + pbaseline->basebits = ZSTD_DECODE_BASEBITS(basebits); 3136 3136 + } 3137 3137 + pbaseline->bits = bits; 3138 3138 + pbaseline->base = base; 3139 3139 + } 3140 3140 + 3141 3141 + return 1; 3142 3142 + } 3143 3143 + 3144 3144 + /* Convert the offset length FSE table FSE_TABLE to an FSE baseline table at 3145 3145 + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ 3146 3146 + 3147 3147 + static int 3148 3148 + elf_zstd_make_offset_baseline_fse ( 3149 3149 + const struct elf_zstd_fse_entry *fse_table, 3150 3150 + int table_bits, 3151 3151 + struct elf_zstd_fse_baseline_entry *baseline_table) 3152 3152 + { 3153 3153 + size_t count; 3154 3154 + const struct elf_zstd_fse_entry *pfse; 3155 3155 + struct elf_zstd_fse_baseline_entry *pbaseline; 3156 3156 + 3157 3157 + /* Convert backward to avoid overlap. */ 3158 3158 + 3159 3159 + count = 1U << table_bits; 3160 3160 + pfse = fse_table + count; 3161 3161 + pbaseline = baseline_table + count; 3162 3162 + while (pfse > fse_table) 3163 3163 + { 3164 3164 + unsigned char symbol; 3165 3165 + unsigned char bits; 3166 3166 + uint16_t base; 3167 3167 + 3168 3168 + --pfse; 3169 3169 + --pbaseline; 3170 3170 + symbol = pfse->symbol; 3171 3171 + bits = pfse->bits; 3172 3172 + base = pfse->base; 3173 3173 + if (unlikely (symbol > 31)) 3174 3174 + { 3175 3175 + elf_uncompress_failed (); 3176 3176 + return 0; 3177 3177 + } 3178 3178 + 3179 3179 + /* The simple way to write this is 3180 3180 + 3181 3181 + pbaseline->baseline = (uint32_t)1 << symbol; 3182 3182 + pbaseline->basebits = symbol; 3183 3183 + 3184 3184 + That will give us an offset value that corresponds to the one 3185 3185 + described in the RFC. However, for offset values > 3, we have to 3186 3186 + subtract 3. And for offset values 1, 2, 3 we use a repeated offset. 3187 3187 + The baseline is always a power of 2, and is never 0, so for these low 3188 3188 + values we will see one entry that is baseline 1, basebits 0, and one 3189 3189 + entry that is baseline 2, basebits 1. All other entries will have 3190 3190 + baseline >= 4 and basebits >= 2. 3191 3191 + 3192 3192 + So we can check for RFC offset <= 3 by checking for basebits <= 1. 3193 3193 + And that means that we can subtract 3 here and not worry about doing 3194 3194 + it in the hot loop. */ 3195 3195 + 3196 3196 + pbaseline->baseline = (uint32_t)1 << symbol; 3197 3197 + if (symbol >= 2) 3198 3198 + pbaseline->baseline -= 3; 3199 3199 + pbaseline->basebits = symbol; 3200 3200 + pbaseline->bits = bits; 3201 3201 + pbaseline->base = base; 3202 3202 + } 3203 3203 + 3204 3204 + return 1; 3205 3205 + } 3206 3206 + 3207 3207 + /* Convert the match length FSE table FSE_TABLE to an FSE baseline table at 3208 3208 + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ 3209 3209 + 3210 3210 + static int 3211 3211 + elf_zstd_make_match_baseline_fse ( 3212 3212 + const struct elf_zstd_fse_entry *fse_table, 3213 3213 + int table_bits, 3214 3214 + struct elf_zstd_fse_baseline_entry *baseline_table) 3215 3215 + { 3216 3216 + size_t count; 3217 3217 + const struct elf_zstd_fse_entry *pfse; 3218 3218 + struct elf_zstd_fse_baseline_entry *pbaseline; 3219 3219 + 3220 3220 + /* Convert backward to avoid overlap. */ 3221 3221 + 3222 3222 + count = 1U << table_bits; 3223 3223 + pfse = fse_table + count; 3224 3224 + pbaseline = baseline_table + count; 3225 3225 + while (pfse > fse_table) 3226 3226 + { 3227 3227 + unsigned char symbol; 3228 3228 + unsigned char bits; 3229 3229 + uint16_t base; 3230 3230 + 3231 3231 + --pfse; 3232 3232 + --pbaseline; 3233 3233 + symbol = pfse->symbol; 3234 3234 + bits = pfse->bits; 3235 3235 + base = pfse->base; 3236 3236 + if (symbol < ZSTD_MATCH_LENGTH_BASELINE_OFFSET) 3237 3237 + { 3238 3238 + pbaseline->baseline = (uint32_t)symbol + 3; 3239 3239 + pbaseline->basebits = 0; 3240 3240 + } 3241 3241 + else 3242 3242 + { 3243 3243 + unsigned int idx; 3244 3244 + uint32_t basebits; 3245 3245 + 3246 3246 + if (unlikely (symbol > 52)) 3247 3247 + { 3248 3248 + elf_uncompress_failed (); 3249 3249 + return 0; 3250 3250 + } 3251 3251 + idx = symbol - ZSTD_MATCH_LENGTH_BASELINE_OFFSET; 3252 3252 + basebits = elf_zstd_match_length_base[idx]; 3253 3253 + pbaseline->baseline = ZSTD_DECODE_BASELINE(basebits); 3254 3254 + pbaseline->basebits = ZSTD_DECODE_BASEBITS(basebits); 3255 3255 + } 3256 3256 + pbaseline->bits = bits; 3257 3257 + pbaseline->base = base; 3258 3258 + } 3259 3259 + 3260 3260 + return 1; 3261 3261 + } 3262 3262 + 3263 3263 + #ifdef BACKTRACE_GENERATE_ZSTD_FSE_TABLES 3264 3264 + 3265 3265 + /* Used to generate the predefined FSE decoding tables for zstd. */ 3266 3266 + 3267 3267 + #include <stdio.h> 3268 3268 + 3269 3269 + /* These values are straight from RFC 8878. */ 3270 3270 + 3271 3271 + static int16_t lit[36] = 3272 3272 + { 3273 3273 + 4, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 3274 3274 + 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 2, 1, 1, 1, 1, 1, 3275 3275 + -1,-1,-1,-1 3276 3276 + }; 3277 3277 + 3278 3278 + static int16_t match[53] = 3279 3279 + { 3280 3280 + 1, 4, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 3281 3281 + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3282 3282 + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,-1,-1, 3283 3283 + -1,-1,-1,-1,-1 3284 3284 + }; 3285 3285 + 3286 3286 + static int16_t offset[29] = 3287 3287 + { 3288 3288 + 1, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 3289 3289 + 1, 1, 1, 1, 1, 1, 1, 1,-1,-1,-1,-1,-1 3290 3290 + }; 3291 3291 + 3292 3292 + static uint16_t next[256]; 3293 3293 + 3294 3294 + static void 3295 3295 + print_table (const struct elf_zstd_fse_baseline_entry *table, size_t size) 3296 3296 + { 3297 3297 + size_t i; 3298 3298 + 3299 3299 + printf ("{\n"); 3300 3300 + for (i = 0; i < size; i += 3) 3301 3301 + { 3302 3302 + int j; 3303 3303 + 3304 3304 + printf (" "); 3305 3305 + for (j = 0; j < 3 && i + j < size; ++j) 3306 3306 + printf (" { %u, %d, %d, %d },", table[i + j].baseline, 3307 3307 + table[i + j].basebits, table[i + j].bits, 3308 3308 + table[i + j].base); 3309 3309 + printf ("\n"); 3310 3310 + } 3311 3311 + printf ("};\n"); 3312 3312 + } 3313 3313 + 3314 3314 + int 3315 3315 + main () 3316 3316 + { 3317 3317 + struct elf_zstd_fse_entry lit_table[64]; 3318 3318 + struct elf_zstd_fse_baseline_entry lit_baseline[64]; 3319 3319 + struct elf_zstd_fse_entry match_table[64]; 3320 3320 + struct elf_zstd_fse_baseline_entry match_baseline[64]; 3321 3321 + struct elf_zstd_fse_entry offset_table[32]; 3322 3322 + struct elf_zstd_fse_baseline_entry offset_baseline[32]; 3323 3323 + 3324 3324 + if (!elf_zstd_build_fse (lit, sizeof lit / sizeof lit[0], next, 3325 3325 + 6, lit_table)) 3326 3326 + { 3327 3327 + fprintf (stderr, "elf_zstd_build_fse failed\n"); 3328 3328 + exit (EXIT_FAILURE); 3329 3329 + } 3330 3330 + 3331 3331 + if (!elf_zstd_make_literal_baseline_fse (lit_table, 6, lit_baseline)) 3332 3332 + { 3333 3333 + fprintf (stderr, "elf_zstd_make_literal_baseline_fse failed\n"); 3334 3334 + exit (EXIT_FAILURE); 3335 3335 + } 3336 3336 + 3337 3337 + printf ("static const struct elf_zstd_fse_baseline_entry " 3338 3338 + "elf_zstd_lit_table[64] =\n"); 3339 3339 + print_table (lit_baseline, 3340 3340 + sizeof lit_baseline / sizeof lit_baseline[0]); 3341 3341 + printf ("\n"); 3342 3342 + 3343 3343 + if (!elf_zstd_build_fse (match, sizeof match / sizeof match[0], next, 3344 3344 + 6, match_table)) 3345 3345 + { 3346 3346 + fprintf (stderr, "elf_zstd_build_fse failed\n"); 3347 3347 + exit (EXIT_FAILURE); 3348 3348 + } 3349 3349 + 3350 3350 + if (!elf_zstd_make_match_baseline_fse (match_table, 6, match_baseline)) 3351 3351 + { 3352 3352 + fprintf (stderr, "elf_zstd_make_match_baseline_fse failed\n"); 3353 3353 + exit (EXIT_FAILURE); 3354 3354 + } 3355 3355 + 3356 3356 + printf ("static const struct elf_zstd_fse_baseline_entry " 3357 3357 + "elf_zstd_match_table[64] =\n"); 3358 3358 + print_table (match_baseline, 3359 3359 + sizeof match_baseline / sizeof match_baseline[0]); 3360 3360 + printf ("\n"); 3361 3361 + 3362 3362 + if (!elf_zstd_build_fse (offset, sizeof offset / sizeof offset[0], next, 3363 3363 + 5, offset_table)) 3364 3364 + { 3365 3365 + fprintf (stderr, "elf_zstd_build_fse failed\n"); 3366 3366 + exit (EXIT_FAILURE); 3367 3367 + } 3368 3368 + 3369 3369 + if (!elf_zstd_make_offset_baseline_fse (offset_table, 5, offset_baseline)) 3370 3370 + { 3371 3371 + fprintf (stderr, "elf_zstd_make_offset_baseline_fse failed\n"); 3372 3372 + exit (EXIT_FAILURE); 3373 3373 + } 3374 3374 + 3375 3375 + printf ("static const struct elf_zstd_fse_baseline_entry " 3376 3376 + "elf_zstd_offset_table[32] =\n"); 3377 3377 + print_table (offset_baseline, 3378 3378 + sizeof offset_baseline / sizeof offset_baseline[0]); 3379 3379 + printf ("\n"); 3380 3380 + 3381 3381 + return 0; 3382 3382 + } 3383 3383 + 3384 3384 + #endif 3385 3385 + 3386 3386 + /* The fixed tables generated by the #ifdef'ed out main function 3387 3387 + above. */ 3388 3388 + 3389 3389 + static const struct elf_zstd_fse_baseline_entry elf_zstd_lit_table[64] = 3390 3390 + { 3391 3391 + { 0, 0, 4, 0 }, { 0, 0, 4, 16 }, { 1, 0, 5, 32 }, 3392 3392 + { 3, 0, 5, 0 }, { 4, 0, 5, 0 }, { 6, 0, 5, 0 }, 3393 3393 + { 7, 0, 5, 0 }, { 9, 0, 5, 0 }, { 10, 0, 5, 0 }, 3394 3394 + { 12, 0, 5, 0 }, { 14, 0, 6, 0 }, { 16, 1, 5, 0 }, 3395 3395 + { 20, 1, 5, 0 }, { 22, 1, 5, 0 }, { 28, 2, 5, 0 }, 3396 3396 + { 32, 3, 5, 0 }, { 48, 4, 5, 0 }, { 64, 6, 5, 32 }, 3397 3397 + { 128, 7, 5, 0 }, { 256, 8, 6, 0 }, { 1024, 10, 6, 0 }, 3398 3398 + { 4096, 12, 6, 0 }, { 0, 0, 4, 32 }, { 1, 0, 4, 0 }, 3399 3399 + { 2, 0, 5, 0 }, { 4, 0, 5, 32 }, { 5, 0, 5, 0 }, 3400 3400 + { 7, 0, 5, 32 }, { 8, 0, 5, 0 }, { 10, 0, 5, 32 }, 3401 3401 + { 11, 0, 5, 0 }, { 13, 0, 6, 0 }, { 16, 1, 5, 32 }, 3402 3402 + { 18, 1, 5, 0 }, { 22, 1, 5, 32 }, { 24, 2, 5, 0 }, 3403 3403 + { 32, 3, 5, 32 }, { 40, 3, 5, 0 }, { 64, 6, 4, 0 }, 3404 3404 + { 64, 6, 4, 16 }, { 128, 7, 5, 32 }, { 512, 9, 6, 0 }, 3405 3405 + { 2048, 11, 6, 0 }, { 0, 0, 4, 48 }, { 1, 0, 4, 16 }, 3406 3406 + { 2, 0, 5, 32 }, { 3, 0, 5, 32 }, { 5, 0, 5, 32 }, 3407 3407 + { 6, 0, 5, 32 }, { 8, 0, 5, 32 }, { 9, 0, 5, 32 }, 3408 3408 + { 11, 0, 5, 32 }, { 12, 0, 5, 32 }, { 15, 0, 6, 0 }, 3409 3409 + { 18, 1, 5, 32 }, { 20, 1, 5, 32 }, { 24, 2, 5, 32 }, 3410 3410 + { 28, 2, 5, 32 }, { 40, 3, 5, 32 }, { 48, 4, 5, 32 }, 3411 3411 + { 65536, 16, 6, 0 }, { 32768, 15, 6, 0 }, { 16384, 14, 6, 0 }, 3412 3412 + { 8192, 13, 6, 0 }, 3413 3413 + }; 3414 3414 + 3415 3415 + static const struct elf_zstd_fse_baseline_entry elf_zstd_match_table[64] = 3416 3416 + { 3417 3417 + { 3, 0, 6, 0 }, { 4, 0, 4, 0 }, { 5, 0, 5, 32 }, 3418 3418 + { 6, 0, 5, 0 }, { 8, 0, 5, 0 }, { 9, 0, 5, 0 }, 3419 3419 + { 11, 0, 5, 0 }, { 13, 0, 6, 0 }, { 16, 0, 6, 0 }, 3420 3420 + { 19, 0, 6, 0 }, { 22, 0, 6, 0 }, { 25, 0, 6, 0 }, 3421 3421 + { 28, 0, 6, 0 }, { 31, 0, 6, 0 }, { 34, 0, 6, 0 }, 3422 3422 + { 37, 1, 6, 0 }, { 41, 1, 6, 0 }, { 47, 2, 6, 0 }, 3423 3423 + { 59, 3, 6, 0 }, { 83, 4, 6, 0 }, { 131, 7, 6, 0 }, 3424 3424 + { 515, 9, 6, 0 }, { 4, 0, 4, 16 }, { 5, 0, 4, 0 }, 3425 3425 + { 6, 0, 5, 32 }, { 7, 0, 5, 0 }, { 9, 0, 5, 32 }, 3426 3426 + { 10, 0, 5, 0 }, { 12, 0, 6, 0 }, { 15, 0, 6, 0 }, 3427 3427 + { 18, 0, 6, 0 }, { 21, 0, 6, 0 }, { 24, 0, 6, 0 }, 3428 3428 + { 27, 0, 6, 0 }, { 30, 0, 6, 0 }, { 33, 0, 6, 0 }, 3429 3429 + { 35, 1, 6, 0 }, { 39, 1, 6, 0 }, { 43, 2, 6, 0 }, 3430 3430 + { 51, 3, 6, 0 }, { 67, 4, 6, 0 }, { 99, 5, 6, 0 }, 3431 3431 + { 259, 8, 6, 0 }, { 4, 0, 4, 32 }, { 4, 0, 4, 48 }, 3432 3432 + { 5, 0, 4, 16 }, { 7, 0, 5, 32 }, { 8, 0, 5, 32 }, 3433 3433 + { 10, 0, 5, 32 }, { 11, 0, 5, 32 }, { 14, 0, 6, 0 }, 3434 3434 + { 17, 0, 6, 0 }, { 20, 0, 6, 0 }, { 23, 0, 6, 0 }, 3435 3435 + { 26, 0, 6, 0 }, { 29, 0, 6, 0 }, { 32, 0, 6, 0 }, 3436 3436 + { 65539, 16, 6, 0 }, { 32771, 15, 6, 0 }, { 16387, 14, 6, 0 }, 3437 3437 + { 8195, 13, 6, 0 }, { 4099, 12, 6, 0 }, { 2051, 11, 6, 0 }, 3438 3438 + { 1027, 10, 6, 0 }, 3439 3439 + }; 3440 3440 + 3441 3441 + static const struct elf_zstd_fse_baseline_entry elf_zstd_offset_table[32] = 3442 3442 + { 3443 3443 + { 1, 0, 5, 0 }, { 64, 6, 4, 0 }, { 512, 9, 5, 0 }, 3444 3444 + { 32768, 15, 5, 0 }, { 2097152, 21, 5, 0 }, { 8, 3, 5, 0 }, 3445 3445 + { 128, 7, 4, 0 }, { 4096, 12, 5, 0 }, { 262144, 18, 5, 0 }, 3446 3446 + { 8388608, 23, 5, 0 }, { 32, 5, 5, 0 }, { 256, 8, 4, 0 }, 3447 3447 + { 16384, 14, 5, 0 }, { 1048576, 20, 5, 0 }, { 4, 2, 5, 0 }, 3448 3448 + { 128, 7, 4, 16 }, { 2048, 11, 5, 0 }, { 131072, 17, 5, 0 }, 3449 3449 + { 4194304, 22, 5, 0 }, { 16, 4, 5, 0 }, { 256, 8, 4, 16 }, 3450 3450 + { 8192, 13, 5, 0 }, { 524288, 19, 5, 0 }, { 2, 1, 5, 0 }, 3451 3451 + { 64, 6, 4, 16 }, { 1024, 10, 5, 0 }, { 65536, 16, 5, 0 }, 3452 3452 + { 268435456, 28, 5, 0 }, { 134217728, 27, 5, 0 }, { 67108864, 26, 5, 0 }, 3453 3453 + { 33554432, 25, 5, 0 }, { 16777216, 24, 5, 0 }, 3454 3454 + }; 3455 3455 + 3456 3456 + /* Read a zstd Huffman table and build the decoding table in *TABLE, reading 3457 3457 + and updating *PPIN. This sets *PTABLE_BITS to the number of bits of the 3458 3458 + table, such that the table length is 1 << *TABLE_BITS. ZDEBUG_TABLE is 3459 3459 + scratch space; it must be enough for 512 uint16_t values + 256 32-bit values 3460 3460 + (2048 bytes). Returns 1 on success, 0 on error. */ 3461 3461 + 3462 3462 + static int 3463 3463 + elf_zstd_read_huff (const unsigned char **ppin, const unsigned char *pinend, 3464 3464 + uint16_t *zdebug_table, uint16_t *table, int *ptable_bits) 3465 3465 + { 3466 3466 + const unsigned char *pin; 3467 3467 + unsigned char hdr; 3468 3468 + unsigned char *weights; 3469 3469 + size_t count; 3470 3470 + uint32_t *weight_mark; 3471 3471 + size_t i; 3472 3472 + uint32_t weight_mask; 3473 3473 + size_t table_bits; 3474 3474 + 3475 3475 + pin = *ppin; 3476 3476 + if (unlikely (pin >= pinend)) 3477 3477 + { 3478 3478 + elf_uncompress_failed (); 3479 3479 + return 0; 3480 3480 + } 3481 3481 + hdr = *pin; 3482 3482 + ++pin; 3483 3483 + 3484 3484 + weights = (unsigned char *) zdebug_table; 3485 3485 + 3486 3486 + if (hdr < 128) 3487 3487 + { 3488 3488 + /* Table is compressed using FSE. */ 3489 3489 + 3490 3490 + struct elf_zstd_fse_entry *fse_table; 3491 3491 + int fse_table_bits; 3492 3492 + uint16_t *scratch; 3493 3493 + const unsigned char *pfse; 3494 3494 + const unsigned char *pback; 3495 3495 + uint64_t val; 3496 3496 + unsigned int bits; 3497 3497 + unsigned int state1, state2; 3498 3498 + 3499 3499 + /* SCRATCH is used temporarily by elf_zstd_read_fse. It overlaps 3500 3500 + WEIGHTS. */ 3501 3501 + scratch = zdebug_table; 3502 3502 + fse_table = (struct elf_zstd_fse_entry *) (scratch + 512); 3503 3503 + fse_table_bits = 6; 3504 3504 + 3505 3505 + pfse = pin; 3506 3506 + if (!elf_zstd_read_fse (&pfse, pinend, scratch, 255, fse_table, 3507 3507 + &fse_table_bits)) 3508 3508 + return 0; 3509 3509 + 3510 3510 + if (unlikely (pin + hdr > pinend)) 3511 3511 + { 3512 3512 + elf_uncompress_failed (); 3513 3513 + return 0; 3514 3514 + } 3515 3515 + 3516 3516 + /* We no longer need SCRATCH. Start recording weights. We need up to 3517 3517 + 256 bytes of weights and 64 bytes of rank counts, so it won't overlap 3518 3518 + FSE_TABLE. */ 3519 3519 + 3520 3520 + pback = pin + hdr - 1; 3521 3521 + 3522 3522 + if (!elf_fetch_backward_init (&pback, pfse, &val, &bits)) 3523 3523 + return 0; 3524 3524 + 3525 3525 + bits -= fse_table_bits; 3526 3526 + state1 = (val >> bits) & ((1U << fse_table_bits) - 1); 3527 3527 + bits -= fse_table_bits; 3528 3528 + state2 = (val >> bits) & ((1U << fse_table_bits) - 1); 3529 3529 + 3530 3530 + /* There are two independent FSE streams, tracked by STATE1 and STATE2. 3531 3531 + We decode them alternately. */ 3532 3532 + 3533 3533 + count = 0; 3534 3534 + while (1) 3535 3535 + { 3536 3536 + struct elf_zstd_fse_entry *pt; 3537 3537 + uint64_t v; 3538 3538 + 3539 3539 + pt = &fse_table[state1]; 3540 3540 + 3541 3541 + if (unlikely (pin < pinend) && bits < pt->bits) 3542 3542 + { 3543 3543 + if (unlikely (count >= 254)) 3544 3544 + { 3545 3545 + elf_uncompress_failed (); 3546 3546 + return 0; 3547 3547 + } 3548 3548 + weights[count] = (unsigned char) pt->symbol; 3549 3549 + weights[count + 1] = (unsigned char) fse_table[state2].symbol; 3550 3550 + count += 2; 3551 3551 + break; 3552 3552 + } 3553 3553 + 3554 3554 + if (unlikely (pt->bits == 0)) 3555 3555 + v = 0; 3556 3556 + else 3557 3557 + { 3558 3558 + if (!elf_fetch_bits_backward (&pback, pfse, &val, &bits)) 3559 3559 + return 0; 3560 3560 + 3561 3561 + bits -= pt->bits; 3562 3562 + v = (val >> bits) & (((uint64_t)1 << pt->bits) - 1); 3563 3563 + } 3564 3564 + 3565 3565 + state1 = pt->base + v; 3566 3566 + 3567 3567 + if (unlikely (count >= 255)) 3568 3568 + { 3569 3569 + elf_uncompress_failed (); 3570 3570 + return 0; 3571 3571 + } 3572 3572 + 3573 3573 + weights[count] = pt->symbol; 3574 3574 + ++count; 3575 3575 + 3576 3576 + pt = &fse_table[state2]; 3577 3577 + 3578 3578 + if (unlikely (pin < pinend && bits < pt->bits)) 3579 3579 + { 3580 3580 + if (unlikely (count >= 254)) 3581 3581 + { 3582 3582 + elf_uncompress_failed (); 3583 3583 + return 0; 3584 3584 + } 3585 3585 + weights[count] = (unsigned char) pt->symbol; 3586 3586 + weights[count + 1] = (unsigned char) fse_table[state1].symbol; 3587 3587 + count += 2; 3588 3588 + break; 3589 3589 + } 3590 3590 + 3591 3591 + if (unlikely (pt->bits == 0)) 3592 3592 + v = 0; 3593 3593 + else 3594 3594 + { 3595 3595 + if (!elf_fetch_bits_backward (&pback, pfse, &val, &bits)) 3596 3596 + return 0; 3597 3597 + 3598 3598 + bits -= pt->bits; 3599 3599 + v = (val >> bits) & (((uint64_t)1 << pt->bits) - 1); 3600 3600 + } 3601 3601 + 3602 3602 + state2 = pt->base + v; 3603 3603 + 3604 3604 + if (unlikely (count >= 255)) 3605 3605 + { 3606 3606 + elf_uncompress_failed (); 3607 3607 + return 0; 3608 3608 + } 3609 3609 + 3610 3610 + weights[count] = pt->symbol; 3611 3611 + ++count; 3612 3612 + } 3613 3613 + 3614 3614 + pin += hdr; 3615 3615 + } 3616 3616 + else 3617 3617 + { 3618 3618 + /* Table is not compressed. Each weight is 4 bits. */ 3619 3619 + 3620 3620 + count = hdr - 127; 3621 3621 + if (unlikely (pin + ((count + 1) / 2) >= pinend)) 3622 3622 + { 3623 3623 + elf_uncompress_failed (); 3624 3624 + return 0; 3625 3625 + } 3626 3626 + for (i = 0; i < count; i += 2) 3627 3627 + { 3628 3628 + unsigned char b; 3629 3629 + 3630 3630 + b = *pin; 3631 3631 + ++pin; 3632 3632 + weights[i] = b >> 4; 3633 3633 + weights[i + 1] = b & 0xf; 3634 3634 + } 3635 3635 + } 3636 3636 + 3637 3637 + weight_mark = (uint32_t *) (weights + 256); 3638 3638 + memset (weight_mark, 0, 12 * sizeof (uint32_t)); 3639 3639 + weight_mask = 0; 3640 3640 + for (i = 0; i < count; ++i) 3641 3641 + { 3642 3642 + unsigned char w; 3643 3643 + 3644 3644 + w = weights[i]; 3645 3645 + if (unlikely (w > 12)) 3646 3646 + { 3647 3647 + elf_uncompress_failed (); 3648 3648 + return 0; 3649 3649 + } 3650 3650 + ++weight_mark[w]; 3651 3651 + if (w > 0) 3652 3652 + weight_mask += 1U << (w - 1); 3653 3653 + } 3654 3654 + if (unlikely (weight_mask == 0)) 3655 3655 + { 3656 3656 + elf_uncompress_failed (); 3657 3657 + return 0; 3658 3658 + } 3659 3659 + 3660 3660 + table_bits = 32 - __builtin_clz (weight_mask); 3661 3661 + if (unlikely (table_bits > 11)) 3662 3662 + { 3663 3663 + elf_uncompress_failed (); 3664 3664 + return 0; 3665 3665 + } 3666 3666 + 3667 3667 + /* Work out the last weight value, which is omitted because the weights must 3668 3668 + sum to a power of two. */ 3669 3669 + { 3670 3670 + uint32_t left; 3671 3671 + uint32_t high_bit; 3672 3672 + 3673 3673 + left = ((uint32_t)1 << table_bits) - weight_mask; 3674 3674 + if (left == 0) 3675 3675 + { 3676 3676 + elf_uncompress_failed (); 3677 3677 + return 0; 3678 3678 + } 3679 3679 + high_bit = 31 - __builtin_clz (left); 3680 3680 + if (((uint32_t)1 << high_bit) != left) 3681 3681 + { 3682 3682 + elf_uncompress_failed (); 3683 3683 + return 0; 3684 3684 + } 3685 3685 + 3686 3686 + if (unlikely (count >= 256)) 3687 3687 + { 3688 3688 + elf_uncompress_failed (); 3689 3689 + return 0; 3690 3690 + } 3691 3691 + 3692 3692 + weights[count] = high_bit + 1; 3693 3693 + ++count; 3694 3694 + ++weight_mark[high_bit + 1]; 3695 3695 + } 3696 3696 + 3697 3697 + if (weight_mark[1] < 2 || (weight_mark[1] & 1) != 0) 3698 3698 + { 3699 3699 + elf_uncompress_failed (); 3700 3700 + return 0; 3701 3701 + } 3702 3702 + 3703 3703 + /* Change WEIGHT_MARK from a count of weights to the index of the first 3704 3704 + symbol for that weight. We shift the indexes to also store how many we 3705 3705 + hae seen so far, below. */ 3706 3706 + { 3707 3707 + uint32_t next; 3708 3708 + 3709 3709 + next = 0; 3710 3710 + for (i = 0; i < table_bits; ++i) 3711 3711 + { 3712 3712 + uint32_t cur; 3713 3713 + 3714 3714 + cur = next; 3715 3715 + next += weight_mark[i + 1] << i; 3716 3716 + weight_mark[i + 1] = cur; 3717 3717 + } 3718 3718 + } 3719 3719 + 3720 3720 + for (i = 0; i < count; ++i) 3721 3721 + { 3722 3722 + unsigned char weight; 3723 3723 + uint32_t length; 3724 3724 + uint16_t tval; 3725 3725 + size_t start; 3726 3726 + uint32_t j; 3727 3727 + 3728 3728 + weight = weights[i]; 3729 3729 + if (weight == 0) 3730 3730 + continue; 3731 3731 + 3732 3732 + length = 1U << (weight - 1); 3733 3733 + tval = (i << 8) | (table_bits + 1 - weight); 3734 3734 + start = weight_mark[weight]; 3735 3735 + for (j = 0; j < length; ++j) 3736 3736 + table[start + j] = tval; 3737 3737 + weight_mark[weight] += length; 3738 3738 + } 3739 3739 + 3740 3740 + *ppin = pin; 3741 3741 + *ptable_bits = (int)table_bits; 3742 3742 + 3743 3743 + return 1; 3744 3744 + } 3745 3745 + 3746 3746 + /* Read and decompress the literals and store them ending at POUTEND. This 3747 3747 + works because we are going to use all the literals in the output, so they 3748 3748 + must fit into the output buffer. HUFFMAN_TABLE, and PHUFFMAN_TABLE_BITS 3749 3749 + store the Huffman table across calls. SCRATCH is used to read a Huffman 3750 3750 + table. Store the start of the decompressed literals in *PPLIT. Update 3751 3751 + *PPIN. Return 1 on success, 0 on error. */ 3752 3752 + 3753 3753 + static int 3754 3754 + elf_zstd_read_literals (const unsigned char **ppin, 3755 3755 + const unsigned char *pinend, 3756 3756 + unsigned char *pout, 3757 3757 + unsigned char *poutend, 3758 3758 + uint16_t *scratch, 3759 3759 + uint16_t *huffman_table, 3760 3760 + int *phuffman_table_bits, 3761 3761 + unsigned char **pplit) 3762 3762 + { 3763 3763 + const unsigned char *pin; 3764 3764 + unsigned char *plit; 3765 3765 + unsigned char hdr; 3766 3766 + uint32_t regenerated_size; 3767 3767 + uint32_t compressed_size; 3768 3768 + int streams; 3769 3769 + uint32_t total_streams_size; 3770 3770 + unsigned int huffman_table_bits; 3771 3771 + uint64_t huffman_mask; 3772 3772 + 3773 3773 + pin = *ppin; 3774 3774 + if (unlikely (pin >= pinend)) 3775 3775 + { 3776 3776 + elf_uncompress_failed (); 3777 3777 + return 0; 3778 3778 + } 3779 3779 + hdr = *pin; 3780 3780 + ++pin; 3781 3781 + 3782 3782 + if ((hdr & 3) == 0 || (hdr & 3) == 1) 3783 3783 + { 3784 3784 + int raw; 3785 3785 + 3786 3786 + /* Raw_literals_Block or RLE_Literals_Block */ 3787 3787 + 3788 3788 + raw = (hdr & 3) == 0; 3789 3789 + 3790 3790 + switch ((hdr >> 2) & 3) 3791 3791 + { 3792 3792 + case 0: case 2: 3793 3793 + regenerated_size = hdr >> 3; 3794 3794 + break; 3795 3795 + case 1: 3796 3796 + if (unlikely (pin >= pinend)) 3797 3797 + { 3798 3798 + elf_uncompress_failed (); 3799 3799 + return 0; 3800 3800 + } 3801 3801 + regenerated_size = (hdr >> 4) + ((uint32_t)(*pin) << 4); 3802 3802 + ++pin; 3803 3803 + break; 3804 3804 + case 3: 3805 3805 + if (unlikely (pin + 1 >= pinend)) 3806 3806 + { 3807 3807 + elf_uncompress_failed (); 3808 3808 + return 0; 3809 3809 + } 3810 3810 + regenerated_size = ((hdr >> 4) 3811 3811 + + ((uint32_t)*pin << 4) 3812 3812 + + ((uint32_t)pin[1] << 12)); 3813 3813 + pin += 2; 3814 3814 + break; 3815 3815 + default: 3816 3816 + elf_uncompress_failed (); 3817 3817 + return 0; 3818 3818 + } 3819 3819 + 3820 3820 + if (unlikely ((size_t)(poutend - pout) < regenerated_size)) 3821 3821 + { 3822 3822 + elf_uncompress_failed (); 3823 3823 + return 0; 3824 3824 + } 3825 3825 + 3826 3826 + plit = poutend - regenerated_size; 3827 3827 + 3828 3828 + if (raw) 3829 3829 + { 3830 3830 + if (unlikely (pin + regenerated_size >= pinend)) 3831 3831 + { 3832 3832 + elf_uncompress_failed (); 3833 3833 + return 0; 3834 3834 + } 3835 3835 + memcpy (plit, pin, regenerated_size); 3836 3836 + pin += regenerated_size; 3837 3837 + } 3838 3838 + else 3839 3839 + { 3840 3840 + if (pin >= pinend) 3841 3841 + { 3842 3842 + elf_uncompress_failed (); 3843 3843 + return 0; 3844 3844 + } 3845 3845 + memset (plit, *pin, regenerated_size); 3846 3846 + ++pin; 3847 3847 + } 3848 3848 + 3849 3849 + *ppin = pin; 3850 3850 + *pplit = plit; 3851 3851 + 3852 3852 + return 1; 3853 3853 + } 3854 3854 + 3855 3855 + /* Compressed_Literals_Block or Treeless_Literals_Block */ 3856 3856 + 3857 3857 + switch ((hdr >> 2) & 3) 3858 3858 + { 3859 3859 + case 0: case 1: 3860 3860 + if (unlikely (pin + 1 >= pinend)) 3861 3861 + { 3862 3862 + elf_uncompress_failed (); 3863 3863 + return 0; 3864 3864 + } 3865 3865 + regenerated_size = (hdr >> 4) | ((uint32_t)(*pin & 0x3f) << 4); 3866 3866 + compressed_size = (uint32_t)*pin >> 6 | ((uint32_t)pin[1] << 2); 3867 3867 + pin += 2; 3868 3868 + streams = ((hdr >> 2) & 3) == 0 ? 1 : 4; 3869 3869 + break; 3870 3870 + case 2: 3871 3871 + if (unlikely (pin + 2 >= pinend)) 3872 3872 + { 3873 3873 + elf_uncompress_failed (); 3874 3874 + return 0; 3875 3875 + } 3876 3876 + regenerated_size = (((uint32_t)hdr >> 4) 3877 3877 + | ((uint32_t)*pin << 4) 3878 3878 + | (((uint32_t)pin[1] & 3) << 12)); 3879 3879 + compressed_size = (((uint32_t)pin[1] >> 2) 3880 3880 + | ((uint32_t)pin[2] << 6)); 3881 3881 + pin += 3; 3882 3882 + streams = 4; 3883 3883 + break; 3884 3884 + case 3: 3885 3885 + if (unlikely (pin + 3 >= pinend)) 3886 3886 + { 3887 3887 + elf_uncompress_failed (); 3888 3888 + return 0; 3889 3889 + } 3890 3890 + regenerated_size = (((uint32_t)hdr >> 4) 3891 3891 + | ((uint32_t)*pin << 4) 3892 3892 + | (((uint32_t)pin[1] & 0x3f) << 12)); 3893 3893 + compressed_size = (((uint32_t)pin[1] >> 6) 3894 3894 + | ((uint32_t)pin[2] << 2) 3895 3895 + | ((uint32_t)pin[3] << 10)); 3896 3896 + pin += 4; 3897 3897 + streams = 4; 3898 3898 + break; 3899 3899 + default: 3900 3900 + elf_uncompress_failed (); 3901 3901 + return 0; 3902 3902 + } 3903 3903 + 3904 3904 + if (unlikely (pin + compressed_size > pinend)) 3905 3905 + { 3906 3906 + elf_uncompress_failed (); 3907 3907 + return 0; 3908 3908 + } 3909 3909 + 3910 3910 + pinend = pin + compressed_size; 3911 3911 + *ppin = pinend; 3912 3912 + 3913 3913 + if (unlikely ((size_t)(poutend - pout) < regenerated_size)) 3914 3914 + { 3915 3915 + elf_uncompress_failed (); 3916 3916 + return 0; 3917 3917 + } 3918 3918 + 3919 3919 + plit = poutend - regenerated_size; 3920 3920 + 3921 3921 + *pplit = plit; 3922 3922 + 3923 3923 + total_streams_size = compressed_size; 3924 3924 + if ((hdr & 3) == 2) 3925 3925 + { 3926 3926 + const unsigned char *ptable; 3927 3927 + 3928 3928 + /* Compressed_Literals_Block. Read Huffman tree. */ 3929 3929 + 3930 3930 + ptable = pin; 3931 3931 + if (!elf_zstd_read_huff (&ptable, pinend, scratch, huffman_table, 3932 3932 + phuffman_table_bits)) 3933 3933 + return 0; 3934 3934 + 3935 3935 + if (unlikely (total_streams_size < (size_t)(ptable - pin))) 3936 3936 + { 3937 3937 + elf_uncompress_failed (); 3938 3938 + return 0; 3939 3939 + } 3940 3940 + 3941 3941 + total_streams_size -= ptable - pin; 3942 3942 + pin = ptable; 3943 3943 + } 3944 3944 + else 3945 3945 + { 3946 3946 + /* Treeless_Literals_Block. Reuse previous Huffman tree. */ 3947 3947 + if (unlikely (*phuffman_table_bits == 0)) 3948 3948 + { 3949 3949 + elf_uncompress_failed (); 3950 3950 + return 0; 3951 3951 + } 3952 3952 + } 3953 3953 + 3954 3954 + /* Decompress COMPRESSED_SIZE bytes of data at PIN using the huffman table, 3955 3955 + storing REGENERATED_SIZE bytes of decompressed data at PLIT. */ 3956 3956 + 3957 3957 + huffman_table_bits = (unsigned int)*phuffman_table_bits; 3958 3958 + huffman_mask = ((uint64_t)1 << huffman_table_bits) - 1; 3959 3959 + 3960 3960 + if (streams == 1) 3961 3961 + { 3962 3962 + const unsigned char *pback; 3963 3963 + const unsigned char *pbackend; 3964 3964 + uint64_t val; 3965 3965 + unsigned int bits; 3966 3966 + uint32_t i; 3967 3967 + 3968 3968 + pback = pin + compressed_size - 1; 3969 3969 + pbackend = pin; 3970 3970 + if (!elf_fetch_backward_init (&pback, pbackend, &val, &bits)) 3971 3971 + return 0; 3972 3972 + 3973 3973 + /* This is one of the inner loops of the decompression algorithm, so we 3974 3974 + put some effort into optimization. We can't get more than 64 bytes 3975 3975 + from a single call to elf_fetch_bits_backward, and we can't subtract 3976 3976 + more than 11 bits at a time. */ 3977 3977 + 3978 3978 + if (regenerated_size >= 64) 3979 3979 + { 3980 3980 + unsigned char *plitstart; 3981 3981 + unsigned char *plitstop; 3982 3982 + 3983 3983 + plitstart = plit; 3984 3984 + plitstop = plit + regenerated_size - 64; 3985 3985 + while (plit < plitstop) 3986 3986 + { 3987 3987 + uint16_t t; 3988 3988 + 3989 3989 + if (!elf_fetch_bits_backward (&pback, pbackend, &val, &bits)) 3990 3990 + return 0; 3991 3991 + 3992 3992 + if (bits < 16) 3993 3993 + break; 3994 3994 + 3995 3995 + while (bits >= 33) 3996 3996 + { 3997 3997 + t = huffman_table[(val >> (bits - huffman_table_bits)) 3998 3998 + & huffman_mask]; 3999 3999 + *plit = t >> 8; 4000 4000 + ++plit; 4001 4001 + bits -= t & 0xff; 4002 4002 + 4003 4003 + t = huffman_table[(val >> (bits - huffman_table_bits)) 4004 4004 + & huffman_mask]; 4005 4005 + *plit = t >> 8; 4006 4006 + ++plit; 4007 4007 + bits -= t & 0xff; 4008 4008 + 4009 4009 + t = huffman_table[(val >> (bits - huffman_table_bits)) 4010 4010 + & huffman_mask]; 4011 4011 + *plit = t >> 8; 4012 4012 + ++plit; 4013 4013 + bits -= t & 0xff; 4014 4014 + } 4015 4015 + 4016 4016 + while (bits > 11) 4017 4017 + { 4018 4018 + t = huffman_table[(val >> (bits - huffman_table_bits)) 4019 4019 + & huffman_mask]; 4020 4020 + *plit = t >> 8; 4021 4021 + ++plit; 4022 4022 + bits -= t & 0xff; 4023 4023 + } 4024 4024 + } 4025 4025 + 4026 4026 + regenerated_size -= plit - plitstart; 4027 4027 + } 4028 4028 + 4029 4029 + for (i = 0; i < regenerated_size; ++i) 4030 4030 + { 4031 4031 + uint16_t t; 4032 4032 + 4033 4033 + if (!elf_fetch_bits_backward (&pback, pbackend, &val, &bits)) 4034 4034 + return 0; 4035 4035 + 4036 4036 + if (unlikely (bits < huffman_table_bits)) 4037 4037 + { 4038 4038 + t = huffman_table[(val << (huffman_table_bits - bits)) 4039 4039 + & huffman_mask]; 4040 4040 + if (unlikely (bits < (t & 0xff))) 4041 4041 + { 4042 4042 + elf_uncompress_failed (); 4043 4043 + return 0; 4044 4044 + } 4045 4045 + } 4046 4046 + else 4047 4047 + t = huffman_table[(val >> (bits - huffman_table_bits)) 4048 4048 + & huffman_mask]; 4049 4049 + 4050 4050 + *plit = t >> 8; 4051 4051 + ++plit; 4052 4052 + bits -= t & 0xff; 4053 4053 + } 4054 4054 + 4055 4055 + return 1; 4056 4056 + } 4057 4057 + 4058 4058 + { 4059 4059 + uint32_t stream_size1, stream_size2, stream_size3, stream_size4; 4060 4060 + uint32_t tot; 4061 4061 + const unsigned char *pback1, *pback2, *pback3, *pback4; 4062 4062 + const unsigned char *pbackend1, *pbackend2, *pbackend3, *pbackend4; 4063 4063 + uint64_t val1, val2, val3, val4; 4064 4064 + unsigned int bits1, bits2, bits3, bits4; 4065 4065 + unsigned char *plit1, *plit2, *plit3, *plit4; 4066 4066 + uint32_t regenerated_stream_size; 4067 4067 + uint32_t regenerated_stream_size4; 4068 4068 + uint16_t t1, t2, t3, t4; 4069 4069 + uint32_t i; 4070 4070 + uint32_t limit; 4071 4071 + 4072 4072 + /* Read jump table. */ 4073 4073 + if (unlikely (pin + 5 >= pinend)) 4074 4074 + { 4075 4075 + elf_uncompress_failed (); 4076 4076 + return 0; 4077 4077 + } 4078 4078 + stream_size1 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); 4079 4079 + pin += 2; 4080 4080 + stream_size2 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); 4081 4081 + pin += 2; 4082 4082 + stream_size3 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); 4083 4083 + pin += 2; 4084 4084 + tot = stream_size1 + stream_size2 + stream_size3; 4085 4085 + if (unlikely (tot > total_streams_size - 6)) 4086 4086 + { 4087 4087 + elf_uncompress_failed (); 4088 4088 + return 0; 4089 4089 + } 4090 4090 + stream_size4 = total_streams_size - 6 - tot; 4091 4091 + 4092 4092 + pback1 = pin + stream_size1 - 1; 4093 4093 + pbackend1 = pin; 4094 4094 + 4095 4095 + pback2 = pback1 + stream_size2; 4096 4096 + pbackend2 = pback1 + 1; 4097 4097 + 4098 4098 + pback3 = pback2 + stream_size3; 4099 4099 + pbackend3 = pback2 + 1; 4100 4100 + 4101 4101 + pback4 = pback3 + stream_size4; 4102 4102 + pbackend4 = pback3 + 1; 4103 4103 + 4104 4104 + if (!elf_fetch_backward_init (&pback1, pbackend1, &val1, &bits1)) 4105 4105 + return 0; 4106 4106 + if (!elf_fetch_backward_init (&pback2, pbackend2, &val2, &bits2)) 4107 4107 + return 0; 4108 4108 + if (!elf_fetch_backward_init (&pback3, pbackend3, &val3, &bits3)) 4109 4109 + return 0; 4110 4110 + if (!elf_fetch_backward_init (&pback4, pbackend4, &val4, &bits4)) 4111 4111 + return 0; 4112 4112 + 4113 4113 + regenerated_stream_size = (regenerated_size + 3) / 4; 4114 4114 + 4115 4115 + plit1 = plit; 4116 4116 + plit2 = plit1 + regenerated_stream_size; 4117 4117 + plit3 = plit2 + regenerated_stream_size; 4118 4118 + plit4 = plit3 + regenerated_stream_size; 4119 4119 + 4120 4120 + regenerated_stream_size4 = regenerated_size - regenerated_stream_size * 3; 4121 4121 + 4122 4122 + /* We can't get more than 64 literal bytes from a single call to 4123 4123 + elf_fetch_bits_backward. The fourth stream can be up to 3 bytes less, 4124 4124 + so use as the limit. */ 4125 4125 + 4126 4126 + limit = regenerated_stream_size4 <= 64 ? 0 : regenerated_stream_size4 - 64; 4127 4127 + i = 0; 4128 4128 + while (i < limit) 4129 4129 + { 4130 4130 + if (!elf_fetch_bits_backward (&pback1, pbackend1, &val1, &bits1)) 4131 4131 + return 0; 4132 4132 + if (!elf_fetch_bits_backward (&pback2, pbackend2, &val2, &bits2)) 4133 4133 + return 0; 4134 4134 + if (!elf_fetch_bits_backward (&pback3, pbackend3, &val3, &bits3)) 4135 4135 + return 0; 4136 4136 + if (!elf_fetch_bits_backward (&pback4, pbackend4, &val4, &bits4)) 4137 4137 + return 0; 4138 4138 + 4139 4139 + /* We can't subtract more than 11 bits at a time. */ 4140 4140 + 4141 4141 + do 4142 4142 + { 4143 4143 + t1 = huffman_table[(val1 >> (bits1 - huffman_table_bits)) 4144 4144 + & huffman_mask]; 4145 4145 + t2 = huffman_table[(val2 >> (bits2 - huffman_table_bits)) 4146 4146 + & huffman_mask]; 4147 4147 + t3 = huffman_table[(val3 >> (bits3 - huffman_table_bits)) 4148 4148 + & huffman_mask]; 4149 4149 + t4 = huffman_table[(val4 >> (bits4 - huffman_table_bits)) 4150 4150 + & huffman_mask]; 4151 4151 + 4152 4152 + *plit1 = t1 >> 8; 4153 4153 + ++plit1; 4154 4154 + bits1 -= t1 & 0xff; 4155 4155 + 4156 4156 + *plit2 = t2 >> 8; 4157 4157 + ++plit2; 4158 4158 + bits2 -= t2 & 0xff; 4159 4159 + 4160 4160 + *plit3 = t3 >> 8; 4161 4161 + ++plit3; 4162 4162 + bits3 -= t3 & 0xff; 4163 4163 + 4164 4164 + *plit4 = t4 >> 8; 4165 4165 + ++plit4; 4166 4166 + bits4 -= t4 & 0xff; 4167 4167 + 4168 4168 + ++i; 4169 4169 + } 4170 4170 + while (bits1 > 11 && bits2 > 11 && bits3 > 11 && bits4 > 11); 4171 4171 + } 4172 4172 + 4173 4173 + while (i < regenerated_stream_size) 4174 4174 + { 4175 4175 + int use4; 4176 4176 + 4177 4177 + use4 = i < regenerated_stream_size4; 4178 4178 + 4179 4179 + if (!elf_fetch_bits_backward (&pback1, pbackend1, &val1, &bits1)) 4180 4180 + return 0; 4181 4181 + if (!elf_fetch_bits_backward (&pback2, pbackend2, &val2, &bits2)) 4182 4182 + return 0; 4183 4183 + if (!elf_fetch_bits_backward (&pback3, pbackend3, &val3, &bits3)) 4184 4184 + return 0; 4185 4185 + if (use4) 4186 4186 + { 4187 4187 + if (!elf_fetch_bits_backward (&pback4, pbackend4, &val4, &bits4)) 4188 4188 + return 0; 4189 4189 + } 4190 4190 + 4191 4191 + if (unlikely (bits1 < huffman_table_bits)) 4192 4192 + { 4193 4193 + t1 = huffman_table[(val1 << (huffman_table_bits - bits1)) 4194 4194 + & huffman_mask]; 4195 4195 + if (unlikely (bits1 < (t1 & 0xff))) 4196 4196 + { 4197 4197 + elf_uncompress_failed (); 4198 4198 + return 0; 4199 4199 + } 4200 4200 + } 4201 4201 + else 4202 4202 + t1 = huffman_table[(val1 >> (bits1 - huffman_table_bits)) 4203 4203 + & huffman_mask]; 4204 4204 + 4205 4205 + if (unlikely (bits2 < huffman_table_bits)) 4206 4206 + { 4207 4207 + t2 = huffman_table[(val2 << (huffman_table_bits - bits2)) 4208 4208 + & huffman_mask]; 4209 4209 + if (unlikely (bits2 < (t2 & 0xff))) 4210 4210 + { 4211 4211 + elf_uncompress_failed (); 4212 4212 + return 0; 4213 4213 + } 4214 4214 + } 4215 4215 + else 4216 4216 + t2 = huffman_table[(val2 >> (bits2 - huffman_table_bits)) 4217 4217 + & huffman_mask]; 4218 4218 + 4219 4219 + if (unlikely (bits3 < huffman_table_bits)) 4220 4220 + { 4221 4221 + t3 = huffman_table[(val3 << (huffman_table_bits - bits3)) 4222 4222 + & huffman_mask]; 4223 4223 + if (unlikely (bits3 < (t3 & 0xff))) 4224 4224 + { 4225 4225 + elf_uncompress_failed (); 4226 4226 + return 0; 4227 4227 + } 4228 4228 + } 4229 4229 + else 4230 4230 + t3 = huffman_table[(val3 >> (bits3 - huffman_table_bits)) 4231 4231 + & huffman_mask]; 4232 4232 + 4233 4233 + if (use4) 4234 4234 + { 4235 4235 + if (unlikely (bits4 < huffman_table_bits)) 4236 4236 + { 4237 4237 + t4 = huffman_table[(val4 << (huffman_table_bits - bits4)) 4238 4238 + & huffman_mask]; 4239 4239 + if (unlikely (bits4 < (t4 & 0xff))) 4240 4240 + { 4241 4241 + elf_uncompress_failed (); 4242 4242 + return 0; 4243 4243 + } 4244 4244 + } 4245 4245 + else 4246 4246 + t4 = huffman_table[(val4 >> (bits4 - huffman_table_bits)) 4247 4247 + & huffman_mask]; 4248 4248 + 4249 4249 + *plit4 = t4 >> 8; 4250 4250 + ++plit4; 4251 4251 + bits4 -= t4 & 0xff; 4252 4252 + } 4253 4253 + 4254 4254 + *plit1 = t1 >> 8; 4255 4255 + ++plit1; 4256 4256 + bits1 -= t1 & 0xff; 4257 4257 + 4258 4258 + *plit2 = t2 >> 8; 4259 4259 + ++plit2; 4260 4260 + bits2 -= t2 & 0xff; 4261 4261 + 4262 4262 + *plit3 = t3 >> 8; 4263 4263 + ++plit3; 4264 4264 + bits3 -= t3 & 0xff; 4265 4265 + 4266 4266 + ++i; 4267 4267 + } 4268 4268 + } 4269 4269 + 4270 4270 + return 1; 4271 4271 + } 4272 4272 + 4273 4273 + /* The information used to decompress a sequence code, which can be a literal 4274 4274 + length, an offset, or a match length. */ 4275 4275 + 4276 4276 + struct elf_zstd_seq_decode 4277 4277 + { 4278 4278 + const struct elf_zstd_fse_baseline_entry *table; 4279 4279 + int table_bits; 4280 4280 + }; 4281 4281 + 4282 4282 + /* Unpack a sequence code compression mode. */ 4283 4283 + 4284 4284 + static int 4285 4285 + elf_zstd_unpack_seq_decode (int mode, 4286 4286 + const unsigned char **ppin, 4287 4287 + const unsigned char *pinend, 4288 4288 + const struct elf_zstd_fse_baseline_entry *predef, 4289 4289 + int predef_bits, 4290 4290 + uint16_t *scratch, 4291 4291 + int maxidx, 4292 4292 + struct elf_zstd_fse_baseline_entry *table, 4293 4293 + int table_bits, 4294 4294 + int (*conv)(const struct elf_zstd_fse_entry *, 4295 4295 + int, 4296 4296 + struct elf_zstd_fse_baseline_entry *), 4297 4297 + struct elf_zstd_seq_decode *decode) 4298 4298 + { 4299 4299 + switch (mode) 4300 4300 + { 4301 4301 + case 0: 4302 4302 + decode->table = predef; 4303 4303 + decode->table_bits = predef_bits; 4304 4304 + break; 4305 4305 + 4306 4306 + case 1: 4307 4307 + { 4308 4308 + struct elf_zstd_fse_entry entry; 4309 4309 + 4310 4310 + if (unlikely (*ppin >= pinend)) 4311 4311 + { 4312 4312 + elf_uncompress_failed (); 4313 4313 + return 0; 4314 4314 + } 4315 4315 + entry.symbol = **ppin; 4316 4316 + ++*ppin; 4317 4317 + entry.bits = 0; 4318 4318 + entry.base = 0; 4319 4319 + decode->table_bits = 0; 4320 4320 + if (!conv (&entry, 0, table)) 4321 4321 + return 0; 4322 4322 + } 4323 4323 + break; 4324 4324 + 4325 4325 + case 2: 4326 4326 + { 4327 4327 + struct elf_zstd_fse_entry *fse_table; 4328 4328 + 4329 4329 + /* We use the same space for the simple FSE table and the baseline 4330 4330 + table. */ 4331 4331 + fse_table = (struct elf_zstd_fse_entry *)table; 4332 4332 + decode->table_bits = table_bits; 4333 4333 + if (!elf_zstd_read_fse (ppin, pinend, scratch, maxidx, fse_table, 4334 4334 + &decode->table_bits)) 4335 4335 + return 0; 4336 4336 + if (!conv (fse_table, decode->table_bits, table)) 4337 4337 + return 0; 4338 4338 + decode->table = table; 4339 4339 + } 4340 4340 + break; 4341 4341 + 4342 4342 + case 3: 4343 4343 + if (unlikely (decode->table_bits == -1)) 4344 4344 + { 4345 4345 + elf_uncompress_failed (); 4346 4346 + return 0; 4347 4347 + } 4348 4348 + break; 4349 4349 + 4350 4350 + default: 4351 4351 + elf_uncompress_failed (); 4352 4352 + return 0; 4353 4353 + } 4354 4354 + 4355 4355 + return 1; 4356 4356 + } 4357 4357 + 4358 4358 + /* Decompress a zstd stream from PIN/SIN to POUT/SOUT. Code based on RFC 8878. 4359 4359 + Return 1 on success, 0 on error. */ 4360 4360 + 4361 4361 + static int 4362 4362 + elf_zstd_decompress (const unsigned char *pin, size_t sin, 4363 4363 + unsigned char *zdebug_table, unsigned char *pout, 4364 4364 + size_t sout) 4365 4365 + { 4366 4366 + const unsigned char *pinend; 4367 4367 + unsigned char *poutstart; 4368 4368 + unsigned char *poutend; 4369 4369 + struct elf_zstd_seq_decode literal_decode; 4370 4370 + struct elf_zstd_fse_baseline_entry *literal_fse_table; 4371 4371 + struct elf_zstd_seq_decode match_decode; 4372 4372 + struct elf_zstd_fse_baseline_entry *match_fse_table; 4373 4373 + struct elf_zstd_seq_decode offset_decode; 4374 4374 + struct elf_zstd_fse_baseline_entry *offset_fse_table; 4375 4375 + uint16_t *huffman_table; 4376 4376 + int huffman_table_bits; 4377 4377 + uint32_t repeated_offset1; 4378 4378 + uint32_t repeated_offset2; 4379 4379 + uint32_t repeated_offset3; 4380 4380 + uint16_t *scratch; 4381 4381 + unsigned char hdr; 4382 4382 + int has_checksum; 4383 4383 + uint64_t content_size; 4384 4384 + int last_block; 4385 4385 + 4386 4386 + pinend = pin + sin; 4387 4387 + poutstart = pout; 4388 4388 + poutend = pout + sout; 4389 4389 + 4390 4390 + literal_decode.table = NULL; 4391 4391 + literal_decode.table_bits = -1; 4392 4392 + literal_fse_table = ((struct elf_zstd_fse_baseline_entry *) 4393 4393 + (zdebug_table + ZSTD_TABLE_LITERAL_FSE_OFFSET)); 4394 4394 + 4395 4395 + match_decode.table = NULL; 4396 4396 + match_decode.table_bits = -1; 4397 4397 + match_fse_table = ((struct elf_zstd_fse_baseline_entry *) 4398 4398 + (zdebug_table + ZSTD_TABLE_MATCH_FSE_OFFSET)); 4399 4399 + 4400 4400 + offset_decode.table = NULL; 4401 4401 + offset_decode.table_bits = -1; 4402 4402 + offset_fse_table = ((struct elf_zstd_fse_baseline_entry *) 4403 4403 + (zdebug_table + ZSTD_TABLE_OFFSET_FSE_OFFSET)); 4404 4404 + huffman_table = ((uint16_t *) 4405 4405 + (zdebug_table + ZSTD_TABLE_HUFFMAN_OFFSET)); 4406 4406 + huffman_table_bits = 0; 4407 4407 + scratch = ((uint16_t *) 4408 4408 + (zdebug_table + ZSTD_TABLE_WORK_OFFSET)); 4409 4409 + 4410 4410 + repeated_offset1 = 1; 4411 4411 + repeated_offset2 = 4; 4412 4412 + repeated_offset3 = 8; 4413 4413 + 4414 4414 + if (unlikely (sin < 4)) 4415 4415 + { 4416 4416 + elf_uncompress_failed (); 4417 4417 + return 0; 4418 4418 + } 4419 4419 + 4420 4420 + /* These values are the zstd magic number. */ 4421 4421 + if (unlikely (pin[0] != 0x28 4422 4422 + || pin[1] != 0xb5 4423 4423 + || pin[2] != 0x2f 4424 4424 + || pin[3] != 0xfd)) 4425 4425 + { 4426 4426 + elf_uncompress_failed (); 4427 4427 + return 0; 4428 4428 + } 4429 4429 + 4430 4430 + pin += 4; 4431 4431 + 4432 4432 + if (unlikely (pin >= pinend)) 4433 4433 + { 4434 4434 + elf_uncompress_failed (); 4435 4435 + return 0; 4436 4436 + } 4437 4437 + 4438 4438 + hdr = *pin++; 4439 4439 + 4440 4440 + /* We expect a single frame. */ 4441 4441 + if (unlikely ((hdr & (1 << 5)) == 0)) 4442 4442 + { 4443 4443 + elf_uncompress_failed (); 4444 4444 + return 0; 4445 4445 + } 4446 4446 + /* Reserved bit must be zero. */ 4447 4447 + if (unlikely ((hdr & (1 << 3)) != 0)) 4448 4448 + { 4449 4449 + elf_uncompress_failed (); 4450 4450 + return 0; 4451 4451 + } 4452 4452 + /* We do not expect a dictionary. */ 4453 4453 + if (unlikely ((hdr & 3) != 0)) 4454 4454 + { 4455 4455 + elf_uncompress_failed (); 4456 4456 + return 0; 4457 4457 + } 4458 4458 + has_checksum = (hdr & (1 << 2)) != 0; 4459 4459 + switch (hdr >> 6) 4460 4460 + { 4461 4461 + case 0: 4462 4462 + if (unlikely (pin >= pinend)) 4463 4463 + { 4464 4464 + elf_uncompress_failed (); 4465 4465 + return 0; 4466 4466 + } 4467 4467 + content_size = (uint64_t) *pin++; 4468 4468 + break; 4469 4469 + case 1: 4470 4470 + if (unlikely (pin + 1 >= pinend)) 4471 4471 + { 4472 4472 + elf_uncompress_failed (); 4473 4473 + return 0; 4474 4474 + } 4475 4475 + content_size = (((uint64_t) pin[0]) | (((uint64_t) pin[1]) << 8)) + 256; 4476 4476 + pin += 2; 4477 4477 + break; 4478 4478 + case 2: 4479 4479 + if (unlikely (pin + 3 >= pinend)) 4480 4480 + { 4481 4481 + elf_uncompress_failed (); 4482 4482 + return 0; 4483 4483 + } 4484 4484 + content_size = ((uint64_t) pin[0] 4485 4485 + | (((uint64_t) pin[1]) << 8) 4486 4486 + | (((uint64_t) pin[2]) << 16) 4487 4487 + | (((uint64_t) pin[3]) << 24)); 4488 4488 + pin += 4; 4489 4489 + break; 4490 4490 + case 3: 4491 4491 + if (unlikely (pin + 7 >= pinend)) 4492 4492 + { 4493 4493 + elf_uncompress_failed (); 4494 4494 + return 0; 4495 4495 + } 4496 4496 + content_size = ((uint64_t) pin[0] 4497 4497 + | (((uint64_t) pin[1]) << 8) 4498 4498 + | (((uint64_t) pin[2]) << 16) 4499 4499 + | (((uint64_t) pin[3]) << 24) 4500 4500 + | (((uint64_t) pin[4]) << 32) 4501 4501 + | (((uint64_t) pin[5]) << 40) 4502 4502 + | (((uint64_t) pin[6]) << 48) 4503 4503 + | (((uint64_t) pin[7]) << 56)); 4504 4504 + pin += 8; 4505 4505 + break; 4506 4506 + default: 4507 4507 + elf_uncompress_failed (); 4508 4508 + return 0; 4509 4509 + } 4510 4510 + 4511 4511 + if (unlikely (content_size != (size_t) content_size 4512 4512 + || (size_t) content_size != sout)) 4513 4513 + { 4514 4514 + elf_uncompress_failed (); 4515 4515 + return 0; 4516 4516 + } 4517 4517 + 4518 4518 + last_block = 0; 4519 4519 + while (!last_block) 4520 4520 + { 4521 4521 + uint32_t block_hdr; 4522 4522 + int block_type; 4523 4523 + uint32_t block_size; 4524 4524 + 4525 4525 + if (unlikely (pin + 2 >= pinend)) 4526 4526 + { 4527 4527 + elf_uncompress_failed (); 4528 4528 + return 0; 4529 4529 + } 4530 4530 + block_hdr = ((uint32_t) pin[0] 4531 4531 + | (((uint32_t) pin[1]) << 8) 4532 4532 + | (((uint32_t) pin[2]) << 16)); 4533 4533 + pin += 3; 4534 4534 + 4535 4535 + last_block = block_hdr & 1; 4536 4536 + block_type = (block_hdr >> 1) & 3; 4537 4537 + block_size = block_hdr >> 3; 4538 4538 + 4539 4539 + switch (block_type) 4540 4540 + { 4541 4541 + case 0: 4542 4542 + /* Raw_Block */ 4543 4543 + if (unlikely ((size_t) block_size > (size_t) (pinend - pin))) 4544 4544 + { 4545 4545 + elf_uncompress_failed (); 4546 4546 + return 0; 4547 4547 + } 4548 4548 + if (unlikely ((size_t) block_size > (size_t) (poutend - pout))) 4549 4549 + { 4550 4550 + elf_uncompress_failed (); 4551 4551 + return 0; 4552 4552 + } 4553 4553 + memcpy (pout, pin, block_size); 4554 4554 + pout += block_size; 4555 4555 + pin += block_size; 4556 4556 + break; 4557 4557 + 4558 4558 + case 1: 4559 4559 + /* RLE_Block */ 4560 4560 + if (unlikely (pin >= pinend)) 4561 4561 + { 4562 4562 + elf_uncompress_failed (); 4563 4563 + return 0; 4564 4564 + } 4565 4565 + if (unlikely ((size_t) block_size > (size_t) (poutend - pout))) 4566 4566 + { 4567 4567 + elf_uncompress_failed (); 4568 4568 + return 0; 4569 4569 + } 4570 4570 + memset (pout, *pin, block_size); 4571 4571 + pout += block_size; 4572 4572 + pin++; 4573 4573 + break; 4574 4574 + 4575 4575 + case 2: 4576 4576 + { 4577 4577 + const unsigned char *pblockend; 4578 4578 + unsigned char *plitstack; 4579 4579 + unsigned char *plit; 4580 4580 + uint32_t literal_count; 4581 4581 + unsigned char seq_hdr; 4582 4582 + size_t seq_count; 4583 4583 + size_t seq; 4584 4584 + const unsigned char *pback; 4585 4585 + uint64_t val; 4586 4586 + unsigned int bits; 4587 4587 + unsigned int literal_state; 4588 4588 + unsigned int offset_state; 4589 4589 + unsigned int match_state; 4590 4590 + 4591 4591 + /* Compressed_Block */ 4592 4592 + if (unlikely ((size_t) block_size > (size_t) (pinend - pin))) 4593 4593 + { 4594 4594 + elf_uncompress_failed (); 4595 4595 + return 0; 4596 4596 + } 4597 4597 + 4598 4598 + pblockend = pin + block_size; 4599 4599 + 4600 4600 + /* Read the literals into the end of the output space, and leave 4601 4601 + PLIT pointing at them. */ 4602 4602 + 4603 4603 + if (!elf_zstd_read_literals (&pin, pblockend, pout, poutend, 4604 4604 + scratch, huffman_table, 4605 4605 + &huffman_table_bits, 4606 4606 + &plitstack)) 4607 4607 + return 0; 4608 4608 + plit = plitstack; 4609 4609 + literal_count = poutend - plit; 4610 4610 + 4611 4611 + seq_hdr = *pin; 4612 4612 + pin++; 4613 4613 + if (seq_hdr < 128) 4614 4614 + seq_count = seq_hdr; 4615 4615 + else if (seq_hdr < 255) 4616 4616 + { 4617 4617 + if (unlikely (pin >= pinend)) 4618 4618 + { 4619 4619 + elf_uncompress_failed (); 4620 4620 + return 0; 4621 4621 + } 4622 4622 + seq_count = ((seq_hdr - 128) << 8) + *pin; 4623 4623 + pin++; 4624 4624 + } 4625 4625 + else 4626 4626 + { 4627 4627 + if (unlikely (pin + 1 >= pinend)) 4628 4628 + { 4629 4629 + elf_uncompress_failed (); 4630 4630 + return 0; 4631 4631 + } 4632 4632 + seq_count = *pin + (pin[1] << 8) + 0x7f00; 4633 4633 + pin += 2; 4634 4634 + } 4635 4635 + 4636 4636 + if (seq_count > 0) 4637 4637 + { 4638 4638 + int (*pfn)(const struct elf_zstd_fse_entry *, 4639 4639 + int, struct elf_zstd_fse_baseline_entry *); 4640 4640 + 4641 4641 + if (unlikely (pin >= pinend)) 4642 4642 + { 4643 4643 + elf_uncompress_failed (); 4644 4644 + return 0; 4645 4645 + } 4646 4646 + seq_hdr = *pin; 4647 4647 + ++pin; 4648 4648 + 4649 4649 + pfn = elf_zstd_make_literal_baseline_fse; 4650 4650 + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 6) & 3, 4651 4651 + &pin, pinend, 4652 4652 + &elf_zstd_lit_table[0], 6, 4653 4653 + scratch, 35, 4654 4654 + literal_fse_table, 9, pfn, 4655 4655 + &literal_decode)) 4656 4656 + return 0; 4657 4657 + 4658 4658 + pfn = elf_zstd_make_offset_baseline_fse; 4659 4659 + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 4) & 3, 4660 4660 + &pin, pinend, 4661 4661 + &elf_zstd_offset_table[0], 5, 4662 4662 + scratch, 31, 4663 4663 + offset_fse_table, 8, pfn, 4664 4664 + &offset_decode)) 4665 4665 + return 0; 4666 4666 + 4667 4667 + pfn = elf_zstd_make_match_baseline_fse; 4668 4668 + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 2) & 3, 4669 4669 + &pin, pinend, 4670 4670 + &elf_zstd_match_table[0], 6, 4671 4671 + scratch, 52, 4672 4672 + match_fse_table, 9, pfn, 4673 4673 + &match_decode)) 4674 4674 + return 0; 4675 4675 + } 4676 4676 + 4677 4677 + pback = pblockend - 1; 4678 4678 + if (!elf_fetch_backward_init (&pback, pin, &val, &bits)) 4679 4679 + return 0; 4680 4680 + 4681 4681 + bits -= literal_decode.table_bits; 4682 4682 + literal_state = ((val >> bits) 4683 4683 + & ((1U << literal_decode.table_bits) - 1)); 4684 4684 + 4685 4685 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4686 4686 + return 0; 4687 4687 + bits -= offset_decode.table_bits; 4688 4688 + offset_state = ((val >> bits) 4689 4689 + & ((1U << offset_decode.table_bits) - 1)); 4690 4690 + 4691 4691 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4692 4692 + return 0; 4693 4693 + bits -= match_decode.table_bits; 4694 4694 + match_state = ((val >> bits) 4695 4695 + & ((1U << match_decode.table_bits) - 1)); 4696 4696 + 4697 4697 + seq = 0; 4698 4698 + while (1) 4699 4699 + { 4700 4700 + const struct elf_zstd_fse_baseline_entry *pt; 4701 4701 + uint32_t offset_basebits; 4702 4702 + uint32_t offset_baseline; 4703 4703 + uint32_t offset_bits; 4704 4704 + uint32_t offset_base; 4705 4705 + uint32_t offset; 4706 4706 + uint32_t match_baseline; 4707 4707 + uint32_t match_bits; 4708 4708 + uint32_t match_base; 4709 4709 + uint32_t match; 4710 4710 + uint32_t literal_baseline; 4711 4711 + uint32_t literal_bits; 4712 4712 + uint32_t literal_base; 4713 4713 + uint32_t literal; 4714 4714 + uint32_t need; 4715 4715 + uint32_t add; 4716 4716 + 4717 4717 + pt = &offset_decode.table[offset_state]; 4718 4718 + offset_basebits = pt->basebits; 4719 4719 + offset_baseline = pt->baseline; 4720 4720 + offset_bits = pt->bits; 4721 4721 + offset_base = pt->base; 4722 4722 + 4723 4723 + /* This case can be more than 16 bits, which is all that 4724 4724 + elf_fetch_bits_backward promises. */ 4725 4725 + need = offset_basebits; 4726 4726 + add = 0; 4727 4727 + if (unlikely (need > 16)) 4728 4728 + { 4729 4729 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4730 4730 + return 0; 4731 4731 + bits -= 16; 4732 4732 + add = (val >> bits) & ((1U << 16) - 1); 4733 4733 + need -= 16; 4734 4734 + add <<= need; 4735 4735 + } 4736 4736 + if (need > 0) 4737 4737 + { 4738 4738 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4739 4739 + return 0; 4740 4740 + bits -= need; 4741 4741 + add += (val >> bits) & ((1U << need) - 1); 4742 4742 + } 4743 4743 + 4744 4744 + offset = offset_baseline + add; 4745 4745 + 4746 4746 + pt = &match_decode.table[match_state]; 4747 4747 + need = pt->basebits; 4748 4748 + match_baseline = pt->baseline; 4749 4749 + match_bits = pt->bits; 4750 4750 + match_base = pt->base; 4751 4751 + 4752 4752 + add = 0; 4753 4753 + if (need > 0) 4754 4754 + { 4755 4755 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4756 4756 + return 0; 4757 4757 + bits -= need; 4758 4758 + add = (val >> bits) & ((1U << need) - 1); 4759 4759 + } 4760 4760 + 4761 4761 + match = match_baseline + add; 4762 4762 + 4763 4763 + pt = &literal_decode.table[literal_state]; 4764 4764 + need = pt->basebits; 4765 4765 + literal_baseline = pt->baseline; 4766 4766 + literal_bits = pt->bits; 4767 4767 + literal_base = pt->base; 4768 4768 + 4769 4769 + add = 0; 4770 4770 + if (need > 0) 4771 4771 + { 4772 4772 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4773 4773 + return 0; 4774 4774 + bits -= need; 4775 4775 + add = (val >> bits) & ((1U << need) - 1); 4776 4776 + } 4777 4777 + 4778 4778 + literal = literal_baseline + add; 4779 4779 + 4780 4780 + /* See the comment in elf_zstd_make_offset_baseline_fse. */ 4781 4781 + if (offset_basebits > 1) 4782 4782 + { 4783 4783 + repeated_offset3 = repeated_offset2; 4784 4784 + repeated_offset2 = repeated_offset1; 4785 4785 + repeated_offset1 = offset; 4786 4786 + } 4787 4787 + else 4788 4788 + { 4789 4789 + if (unlikely (literal == 0)) 4790 4790 + ++offset; 4791 4791 + switch (offset) 4792 4792 + { 4793 4793 + case 1: 4794 4794 + offset = repeated_offset1; 4795 4795 + break; 4796 4796 + case 2: 4797 4797 + offset = repeated_offset2; 4798 4798 + repeated_offset2 = repeated_offset1; 4799 4799 + repeated_offset1 = offset; 4800 4800 + break; 4801 4801 + case 3: 4802 4802 + offset = repeated_offset3; 4803 4803 + repeated_offset3 = repeated_offset2; 4804 4804 + repeated_offset2 = repeated_offset1; 4805 4805 + repeated_offset1 = offset; 4806 4806 + break; 4807 4807 + case 4: 4808 4808 + offset = repeated_offset1 - 1; 4809 4809 + repeated_offset3 = repeated_offset2; 4810 4810 + repeated_offset2 = repeated_offset1; 4811 4811 + repeated_offset1 = offset; 4812 4812 + break; 4813 4813 + } 4814 4814 + } 4815 4815 + 4816 4816 + ++seq; 4817 4817 + if (seq < seq_count) 4818 4818 + { 4819 4819 + uint32_t v; 4820 4820 + 4821 4821 + /* Update the three states. */ 4822 4822 + 4823 4823 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4824 4824 + return 0; 4825 4825 + 4826 4826 + need = literal_bits; 4827 4827 + bits -= need; 4828 4828 + v = (val >> bits) & (((uint32_t)1 << need) - 1); 4829 4829 + 4830 4830 + literal_state = literal_base + v; 4831 4831 + 4832 4832 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4833 4833 + return 0; 4834 4834 + 4835 4835 + need = match_bits; 4836 4836 + bits -= need; 4837 4837 + v = (val >> bits) & (((uint32_t)1 << need) - 1); 4838 4838 + 4839 4839 + match_state = match_base + v; 4840 4840 + 4841 4841 + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) 4842 4842 + return 0; 4843 4843 + 4844 4844 + need = offset_bits; 4845 4845 + bits -= need; 4846 4846 + v = (val >> bits) & (((uint32_t)1 << need) - 1); 4847 4847 + 4848 4848 + offset_state = offset_base + v; 4849 4849 + } 4850 4850 + 4851 4851 + /* The next sequence is now in LITERAL, OFFSET, MATCH. */ 4852 4852 + 4853 4853 + /* Copy LITERAL bytes from the literals. */ 4854 4854 + 4855 4855 + if (unlikely ((size_t)(poutend - pout) < literal)) 4856 4856 + { 4857 4857 + elf_uncompress_failed (); 4858 4858 + return 0; 4859 4859 + } 4860 4860 + 4861 4861 + if (unlikely (literal_count < literal)) 4862 4862 + { 4863 4863 + elf_uncompress_failed (); 4864 4864 + return 0; 4865 4865 + } 4866 4866 + 4867 4867 + literal_count -= literal; 4868 4868 + 4869 4869 + /* Often LITERAL is small, so handle small cases quickly. */ 4870 4870 + switch (literal) 4871 4871 + { 4872 4872 + case 8: 4873 4873 + *pout++ = *plit++; 4874 4874 + /* FALLTHROUGH */ 4875 4875 + case 7: 4876 4876 + *pout++ = *plit++; 4877 4877 + /* FALLTHROUGH */ 4878 4878 + case 6: 4879 4879 + *pout++ = *plit++; 4880 4880 + /* FALLTHROUGH */ 4881 4881 + case 5: 4882 4882 + *pout++ = *plit++; 4883 4883 + /* FALLTHROUGH */ 4884 4884 + case 4: 4885 4885 + *pout++ = *plit++; 4886 4886 + /* FALLTHROUGH */ 4887 4887 + case 3: 4888 4888 + *pout++ = *plit++; 4889 4889 + /* FALLTHROUGH */ 4890 4890 + case 2: 4891 4891 + *pout++ = *plit++; 4892 4892 + /* FALLTHROUGH */ 4893 4893 + case 1: 4894 4894 + *pout++ = *plit++; 4895 4895 + break; 4896 4896 + 4897 4897 + case 0: 4898 4898 + break; 4899 4899 + 4900 4900 + default: 4901 4901 + if (unlikely ((size_t)(plit - pout) < literal)) 4902 4902 + { 4903 4903 + uint32_t move; 4904 4904 + 4905 4905 + move = plit - pout; 4906 4906 + while (literal > move) 4907 4907 + { 4908 4908 + memcpy (pout, plit, move); 4909 4909 + pout += move; 4910 4910 + plit += move; 4911 4911 + literal -= move; 4912 4912 + } 4913 4913 + } 4914 4914 + 4915 4915 + memcpy (pout, plit, literal); 4916 4916 + pout += literal; 4917 4917 + plit += literal; 4918 4918 + } 4919 4919 + 4920 4920 + if (match > 0) 4921 4921 + { 4922 4922 + /* Copy MATCH bytes from the decoded output at OFFSET. */ 4923 4923 + 4924 4924 + if (unlikely ((size_t)(poutend - pout) < match)) 4925 4925 + { 4926 4926 + elf_uncompress_failed (); 4927 4927 + return 0; 4928 4928 + } 4929 4929 + 4930 4930 + if (unlikely ((size_t)(pout - poutstart) < offset)) 4931 4931 + { 4932 4932 + elf_uncompress_failed (); 4933 4933 + return 0; 4934 4934 + } 4935 4935 + 4936 4936 + if (offset >= match) 4937 4937 + { 4938 4938 + memcpy (pout, pout - offset, match); 4939 4939 + pout += match; 4940 4940 + } 4941 4941 + else 4942 4942 + { 4943 4943 + while (match > 0) 4944 4944 + { 4945 4945 + uint32_t copy; 4946 4946 + 4947 4947 + copy = match < offset ? match : offset; 4948 4948 + memcpy (pout, pout - offset, copy); 4949 4949 + match -= copy; 4950 4950 + pout += copy; 4951 4951 + } 4952 4952 + } 4953 4953 + } 4954 4954 + 4955 4955 + if (unlikely (seq >= seq_count)) 4956 4956 + { 4957 4957 + /* Copy remaining literals. */ 4958 4958 + if (literal_count > 0 && plit != pout) 4959 4959 + { 4960 4960 + if (unlikely ((size_t)(poutend - pout) 4961 4961 + < literal_count)) 4962 4962 + { 4963 4963 + elf_uncompress_failed (); 4964 4964 + return 0; 4965 4965 + } 4966 4966 + 4967 4967 + if ((size_t)(plit - pout) < literal_count) 4968 4968 + { 4969 4969 + uint32_t move; 4970 4970 + 4971 4971 + move = plit - pout; 4972 4972 + while (literal_count > move) 4973 4973 + { 4974 4974 + memcpy (pout, plit, move); 4975 4975 + pout += move; 4976 4976 + plit += move; 4977 4977 + literal_count -= move; 4978 4978 + } 4979 4979 + } 4980 4980 + 4981 4981 + memcpy (pout, plit, literal_count); 4982 4982 + } 4983 4983 + 4984 4984 + pout += literal_count; 4985 4985 + 4986 4986 + break; 4987 4987 + } 4988 4988 + } 4989 4989 + 4990 4990 + pin = pblockend; 4991 4991 + } 4992 4992 + break; 4993 4993 + 4994 4994 + case 3: 4995 4995 + default: 4996 4996 + elf_uncompress_failed (); 4997 4997 + return 0; 4998 4998 + } 4999 4999 + } 5000 5000 + 5001 5001 + if (has_checksum) 5002 5002 + { 5003 5003 + if (unlikely (pin + 4 > pinend)) 5004 5004 + { 5005 5005 + elf_uncompress_failed (); 5006 5006 + return 0; 5007 5007 + } 5008 5008 + 5009 5009 + /* We don't currently verify the checksum. Currently running GNU ld with 5010 5010 + --compress-debug-sections=zstd does not seem to generate a 5011 5011 + checksum. */ 5012 5012 + 5013 5013 + pin += 4; 5014 5014 + } 5015 5015 + 5016 5016 + if (pin != pinend) 5017 5017 + { 5018 5018 + elf_uncompress_failed (); 5019 5019 + return 0; 5020 5020 + } 5021 5021 + 5022 5022 + return 1; 5023 5023 + } 5024 5024 + 5025 5025 + #define ZDEBUG_TABLE_SIZE \ 5026 5026 + (ZLIB_TABLE_SIZE > ZSTD_TABLE_SIZE ? ZLIB_TABLE_SIZE : ZSTD_TABLE_SIZE) 5027 5027 + 2562 5028 /* Uncompress the old compressed debug format, the one emitted by 2563 5029 --compress-debug-sections=zlib-gnu. The compressed data is in 2564 5030 COMPRESSED / COMPRESSED_SIZE, and the function writes to ··· 2628 5094 unsigned char **uncompressed, size_t *uncompressed_size) 2629 5095 { 2630 5096 const b_elf_chdr *chdr; 5097 5097 + char *alc; 5098 5098 + size_t alc_len; 2631 5099 unsigned char *po; 2632 5100 2633 5101 *uncompressed = NULL; ··· 2639 5107 2640 5108 chdr = (const b_elf_chdr *) compressed; 2641 5109 2642 2642 - if (chdr->ch_type != ELFCOMPRESS_ZLIB) 2643 2643 - { 2644 2644 - /* Unsupported compression algorithm. */ 2645 2645 - return 1; 2646 2646 - } 2647 2647 - 5110 5110 + alc = NULL; 5111 5111 + alc_len = 0; 2648 5112 if (*uncompressed != NULL && *uncompressed_size >= chdr->ch_size) 2649 5113 po = *uncompressed; 2650 5114 else 2651 5115 { 2652 2652 - po = (unsigned char *) backtrace_alloc (state, chdr->ch_size, 2653 2653 - error_callback, data); 2654 2654 - if (po == NULL) 5116 5116 + alc_len = chdr->ch_size; 5117 5117 + alc = (char*)backtrace_alloc (state, alc_len, error_callback, data); 5118 5118 + if (alc == NULL) 2655 5119 return 0; 5120 5120 + po = (unsigned char *) alc; 2656 5121 } 2657 5122 2658 2658 - if (!elf_zlib_inflate_and_verify (compressed + sizeof (b_elf_chdr), 2659 2659 - compressed_size - sizeof (b_elf_chdr), 2660 2660 - zdebug_table, po, chdr->ch_size)) 2661 2661 - return 1; 5123 5123 + switch (chdr->ch_type) 5124 5124 + { 5125 5125 + case ELFCOMPRESS_ZLIB: 5126 5126 + if (!elf_zlib_inflate_and_verify (compressed + sizeof (b_elf_chdr), 5127 5127 + compressed_size - sizeof (b_elf_chdr), 5128 5128 + zdebug_table, po, chdr->ch_size)) 5129 5129 + goto skip; 5130 5130 + break; 5131 5131 + 5132 5132 + case ELFCOMPRESS_ZSTD: 5133 5133 + if (!elf_zstd_decompress (compressed + sizeof (b_elf_chdr), 5134 5134 + compressed_size - sizeof (b_elf_chdr), 5135 5135 + (unsigned char *)zdebug_table, po, 5136 5136 + chdr->ch_size)) 5137 5137 + goto skip; 5138 5138 + break; 5139 5139 + 5140 5140 + default: 5141 5141 + /* Unsupported compression algorithm. */ 5142 5142 + goto skip; 5143 5143 + } 2662 5144 2663 5145 *uncompressed = po; 2664 5146 *uncompressed_size = chdr->ch_size; 2665 5147 5148 5148 + return 1; 5149 5149 + 5150 5150 + skip: 5151 5151 + if (alc != NULL && alc_len > 0) 5152 5152 + backtrace_free (state, alc, alc_len, error_callback, data); 2666 5153 return 1; 2667 5154 } 2668 5155 ··· 2687 5174 ret = elf_uncompress_zdebug (state, compressed, compressed_size, 2688 5175 zdebug_table, error_callback, data, 2689 5176 uncompressed, uncompressed_size); 5177 5177 + backtrace_free (state, zdebug_table, ZDEBUG_TABLE_SIZE, 5178 5178 + error_callback, data); 5179 5179 + return ret; 5180 5180 + } 5181 5181 + 5182 5182 + /* This function is a hook for testing the zstd support. It is only used by 5183 5183 + tests. */ 5184 5184 + 5185 5185 + int 5186 5186 + backtrace_uncompress_zstd (struct backtrace_state *state, 5187 5187 + const unsigned char *compressed, 5188 5188 + size_t compressed_size, 5189 5189 + backtrace_error_callback error_callback, 5190 5190 + void *data, unsigned char *uncompressed, 5191 5191 + size_t uncompressed_size) 5192 5192 + { 5193 5193 + unsigned char *zdebug_table; 5194 5194 + int ret; 5195 5195 + 5196 5196 + zdebug_table = ((unsigned char *) backtrace_alloc (state, ZDEBUG_TABLE_SIZE, 5197 5197 + error_callback, data)); 5198 5198 + if (zdebug_table == NULL) 5199 5199 + return 0; 5200 5200 + ret = elf_zstd_decompress (compressed, compressed_size, 5201 5201 + zdebug_table, uncompressed, uncompressed_size); 2690 5202 backtrace_free (state, zdebug_table, ZDEBUG_TABLE_SIZE, 2691 5203 error_callback, data); 2692 5204 return ret; ··· 4688 7200 if (zdebug_table == NULL) 4689 7201 { 4690 7202 zdebug_table = ((uint16_t *) 4691 4691 - backtrace_alloc (state, ZDEBUG_TABLE_SIZE, 7203 7203 + backtrace_alloc (state, ZLIB_TABLE_SIZE, 4692 7204 error_callback, data)); 4693 7205 if (zdebug_table == NULL) 4694 7206 goto fail; ··· 4714 7226 } 4715 7227 } 4716 7228 7229 7229 + if (zdebug_table != NULL) 7230 7230 + { 7231 7231 + backtrace_free (state, zdebug_table, ZLIB_TABLE_SIZE, 7232 7232 + error_callback, data); 7233 7233 + zdebug_table = NULL; 7234 7234 + } 7235 7235 + 4717 7236 /* Uncompress the official ELF format 4718 4718 - (--compress-debug-sections=zlib-gabi). */ 7237 7237 + (--compress-debug-sections=zlib-gabi, --compress-debug-sections=zstd). */ 4719 7238 for (i = 0; i < (int) DEBUG_MAX; ++i) 4720 7239 { 4721 7240 unsigned char *uncompressed_data;

+9

src/external/tracy/libbacktrace/internal.hpp

··· 371 371 unsigned char **uncompressed, 372 372 size_t *uncompressed_size); 373 373 374 374 + /* A test-only hook for elf_zstd_decompress. */ 375 375 + 376 376 + extern int backtrace_uncompress_zstd (struct backtrace_state *, 377 377 + const unsigned char *compressed, 378 378 + size_t compressed_size, 379 379 + backtrace_error_callback, void *data, 380 380 + unsigned char *uncompressed, 381 381 + size_t uncompressed_size); 382 382 + 374 383 /* A test-only hook for elf_uncompress_lzma. */ 375 384 376 385 extern int backtrace_uncompress_lzma (struct backtrace_state *,

+41 -29

src/external/tracy/tracy/Tracy.hpp

··· 4 4 #include "../common/TracyColor.hpp" 5 5 #include "../common/TracySystem.hpp" 6 6 7 7 + #ifndef TracyFunction 8 8 + # define TracyFunction __FUNCTION__ 9 9 + #endif 10 10 + 11 11 + #ifndef TracyFile 12 12 + # define TracyFile __FILE__ 13 13 + #endif 14 14 + 15 15 + #ifndef TracyLine 16 16 + # define TracyLine __LINE__ 17 17 + #endif 18 18 + 7 19 #ifndef TRACY_ENABLE 8 20 9 21 #define ZoneNamed(x,y) ··· 37 49 38 50 #define FrameImage(x,y,z,w,a) 39 51 40 40 - #define TracyLockable( type, varname ) type varname; 41 41 - #define TracyLockableN( type, varname, desc ) type varname; 42 42 - #define TracySharedLockable( type, varname ) type varname; 43 43 - #define TracySharedLockableN( type, varname, desc ) type varname; 52 52 + #define TracyLockable( type, varname ) type varname 53 53 + #define TracyLockableN( type, varname, desc ) type varname 54 54 + #define TracySharedLockable( type, varname ) type varname 55 55 + #define TracySharedLockableN( type, varname, desc ) type varname 44 56 #define LockableBase( type ) type 45 57 #define SharedLockableBase( type ) type 46 46 - #define LockMark(x) (void)x; 47 47 - #define LockableName(x,y,z); 58 58 + #define LockMark(x) (void)x 59 59 + #define LockableName(x,y,z) 48 60 49 61 #define TracyPlot(x,y) 50 62 #define TracyPlotConfig(x,y,z,w,a) ··· 110 122 #include "../client/TracyScoped.hpp" 111 123 112 124 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 113 113 - # define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ) 114 114 - # define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ) 115 115 - # define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ) 116 116 - # define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ) 125 125 + # define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) 126 126 + # define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) 127 127 + # define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) 128 128 + # define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) 117 129 118 118 - # define ZoneTransient( varname, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, TRACY_CALLSTACK, active ) 119 119 - # define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), TRACY_CALLSTACK, active ) 130 130 + # define ZoneTransient( varname, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, TRACY_CALLSTACK, active ) 131 131 + # define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ) 120 132 #else 121 121 - # define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active ) 122 122 - # define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active ) 123 123 - # define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active ) 124 124 - # define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active ) 133 133 + # define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) 134 134 + # define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) 135 135 + # define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) 136 136 + # define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) 125 137 126 126 - # define ZoneTransient( varname, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, active ) 127 127 - # define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), active ) 138 138 + # define ZoneTransient( varname, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, active ) 139 139 + # define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ) 128 140 #endif 129 141 130 142 #define ZoneScoped ZoneNamed( ___tracy_scoped_zone, true ) ··· 150 162 151 163 #define FrameImage( image, width, height, offset, flip ) tracy::Profiler::SendFrameImage( image, width, height, offset, flip ) 152 164 153 153 - #define TracyLockable( type, varname ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() } 154 154 - #define TracyLockableN( type, varname, desc ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() } 155 155 - #define TracySharedLockable( type, varname ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() } 156 156 - #define TracySharedLockableN( type, varname, desc ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() } 165 165 + #define TracyLockable( type, varname ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, TracyFile, TracyLine, 0 }; return &srcloc; }() } 166 166 + #define TracyLockableN( type, varname, desc ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, TracyFile, TracyLine, 0 }; return &srcloc; }() } 167 167 + #define TracySharedLockable( type, varname ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, TracyFile, TracyLine, 0 }; return &srcloc; }() } 168 168 + #define TracySharedLockableN( type, varname, desc ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, TracyFile, TracyLine, 0 }; return &srcloc; }() } 157 169 #define LockableBase( type ) tracy::Lockable<type> 158 170 #define SharedLockableBase( type ) tracy::SharedLockable<type> 159 159 - #define LockMark( varname ) static constexpr tracy::SourceLocationData __tracy_lock_location_##varname { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; varname.Mark( &__tracy_lock_location_##varname ) 171 171 + #define LockMark( varname ) static constexpr tracy::SourceLocationData __tracy_lock_location_##varname { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; varname.Mark( &__tracy_lock_location_##varname ) 160 172 #define LockableName( varname, txt, size ) varname.CustomName( txt, size ) 161 173 162 174 #define TracyPlot( name, val ) tracy::Profiler::PlotData( name, val ) ··· 197 209 #endif 198 210 199 211 #ifdef TRACY_HAS_CALLSTACK 200 200 - # define ZoneNamedS( varname, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active ) 201 201 - # define ZoneNamedNS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active ) 202 202 - # define ZoneNamedCS( varname, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active ) 203 203 - # define ZoneNamedNCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active ) 212 212 + # define ZoneNamedS( varname, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) 213 213 + # define ZoneNamedNS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) 214 214 + # define ZoneNamedCS( varname, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) 215 215 + # define ZoneNamedNCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) 204 216 205 205 - # define ZoneTransientS( varname, depth, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, depth, active ) 206 206 - # define ZoneTransientNS( varname, name, depth, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), depth, active ) 217 217 + # define ZoneTransientS( varname, depth, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, depth, active ) 218 218 + # define ZoneTransientNS( varname, name, depth, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ) 207 219 208 220 # define ZoneScopedS( depth ) ZoneNamedS( ___tracy_scoped_zone, depth, true ) 209 221 # define ZoneScopedNS( name, depth ) ZoneNamedNS( ___tracy_scoped_zone, name, depth, true )

+30 -13

src/external/tracy/tracy/TracyC.h

··· 15 15 16 16 #define TracyCSetThreadName( name ) ___tracy_set_thread_name( name ); 17 17 18 18 + #ifndef TracyFunction 19 19 + # define TracyFunction __FUNCTION__ 20 20 + #endif 21 21 + 22 22 + #ifndef TracyFile 23 23 + # define TracyFile __FILE__ 24 24 + #endif 25 25 + 26 26 + #ifndef TracyLine 27 27 + # define TracyLine __LINE__ 28 28 + #endif 18 29 19 30 #ifndef TRACY_ENABLE 20 31 ··· 47 58 #define TracyCFrameImage(x,y,z,w,a) 48 59 49 60 #define TracyCPlot(x,y) 61 61 + #define TracyCPlotF(x,y) 62 62 + #define TracyCPlotI(x,y) 50 63 #define TracyCMessage(x,y) 51 64 #define TracyCMessageL(x) 52 65 #define TracyCMessageC(x,y,z) ··· 195 208 TRACY_API int ___tracy_connected(void); 196 209 197 210 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 198 198 - # define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ); 199 199 - # define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ); 200 200 - # define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ); 201 201 - # define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active ); 211 211 + # define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); 212 212 + # define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); 213 213 + # define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); 214 214 + # define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); 202 215 #else 203 203 - # define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,__LINE__), active ); 204 204 - # define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,__LINE__), active ); 205 205 - # define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,__LINE__), active ); 206 206 - # define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,__LINE__), active ); 216 216 + # define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); 217 217 + # define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); 218 218 + # define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); 219 219 + # define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); 207 220 #endif 208 221 209 222 #define TracyCZoneEnd( ctx ) ___tracy_emit_zone_end( ctx ); ··· 274 287 275 288 276 289 TRACY_API void ___tracy_emit_plot( const char* name, double val ); 290 290 + TRACY_API void ___tracy_emit_plot_float( const char* name, float val ); 291 291 + TRACY_API void ___tracy_emit_plot_int( const char* name, int64_t val ); 277 292 TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size ); 278 293 279 294 #define TracyCPlot( name, val ) ___tracy_emit_plot( name, val ); 295 295 + #define TracyCPlotF( name, val ) ___tracy_emit_plot_float( name, val ); 296 296 + #define TracyCPlotI( name, val ) ___tracy_emit_plot_int( name, val ); 280 297 #define TracyCAppInfo( txt, size ) ___tracy_emit_message_appinfo( txt, size ); 281 298 282 299 283 300 #ifdef TRACY_HAS_CALLSTACK 284 284 - # define TracyCZoneS( ctx, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active ); 285 285 - # define TracyCZoneNS( ctx, name, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active ); 286 286 - # define TracyCZoneCS( ctx, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active ); 287 287 - # define TracyCZoneNCS( ctx, name, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __func__, __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active ); 301 301 + # define TracyCZoneS( ctx, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); 302 302 + # define TracyCZoneNS( ctx, name, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); 303 303 + # define TracyCZoneCS( ctx, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); 304 304 + # define TracyCZoneNCS( ctx, name, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); 288 305 289 306 # define TracyCAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth, 0 ) 290 307 # define TracyCFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth, 0 ) ··· 324 341 325 342 #define TracyCIsConnected ___tracy_connected() 326 343 344 344 + #ifdef TRACY_FIBERS 327 345 TRACY_API void ___tracy_fiber_enter( const char* fiber ); 328 346 TRACY_API void ___tracy_fiber_leave( void ); 329 347 330 330 - #ifdef TRACY_FIBERS 331 348 # define TracyCFiberEnter( fiber ) ___tracy_fiber_enter( fiber ); 332 349 # define TracyCFiberLeave ___tracy_fiber_leave(); 333 350 #endif

+10 -8

src/external/tracy/tracy/TracyD3D11.hpp

··· 42 42 #include "../common/TracyAlign.hpp" 43 43 #include "../common/TracyAlloc.hpp" 44 44 45 45 + #include <d3d11.h> 46 46 + 45 47 namespace tracy 46 48 { 47 49 ··· 410 412 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 411 413 # define TracyD3D11Zone( ctx, name ) TracyD3D11NamedZoneS( ctx, ___tracy_gpu_zone, name, TRACY_CALLSTACK, true ) 412 414 # define TracyD3D11ZoneC( ctx, name, color ) TracyD3D11NamedZoneCS( ctx, ___tracy_gpu_zone, name, color, TRACY_CALLSTACK, true ) 413 413 - # define TracyD3D11NamedZone( ctx, varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 414 414 - # define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 415 415 + # define TracyD3D11NamedZone( ctx, varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 416 416 + # define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 415 417 # define TracyD3D11ZoneTransient(ctx, varname, name, active) TracyD3D11ZoneTransientS(ctx, varname, cmdList, name, TRACY_CALLSTACK, active) 416 418 #else 417 419 # define TracyD3D11Zone( ctx, name ) TracyD3D11NamedZone( ctx, ___tracy_gpu_zone, name, true ) 418 420 # define TracyD3D11ZoneC( ctx, name, color ) TracyD3D11NamedZoneC( ctx, ___tracy_gpu_zone, name, color, true ) 419 419 - # define TracyD3D11NamedZone( ctx, varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), active ); 420 420 - # define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), active ); 421 421 - # define TracyD3D11ZoneTransient(ctx, varname, name, active) tracy::D3D11ZoneScope varname{ ctx, __LINE__, __FILE__, strlen(__FILE__), __FUNCTION__, strlen(__FUNCTION__), name, strlen(name), active }; 421 421 + # define TracyD3D11NamedZone( ctx, varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); 422 422 + # define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); 423 423 + # define TracyD3D11ZoneTransient(ctx, varname, name, active) tracy::D3D11ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), active }; 422 424 #endif 423 425 424 426 #ifdef TRACY_HAS_CALLSTACK 425 427 # define TracyD3D11ZoneS( ctx, name, depth ) TracyD3D11NamedZoneS( ctx, ___tracy_gpu_zone, name, depth, true ) 426 428 # define TracyD3D11ZoneCS( ctx, name, color, depth ) TracyD3D11NamedZoneCS( ctx, ___tracy_gpu_zone, name, color, depth, true ) 427 427 - # define TracyD3D11NamedZoneS( ctx, varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active ); 428 428 - # define TracyD3D11NamedZoneCS( ctx, varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active ); 429 429 - # define TracyD3D11ZoneTransientS(ctx, varname, name, depth, active) tracy::D3D11ZoneScope varname{ ctx, __LINE__, __FILE__, strlen(__FILE__), __FUNCTION__, strlen(__FUNCTION__), name, strlen(name), depth, active }; 429 429 + # define TracyD3D11NamedZoneS( ctx, varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); 430 430 + # define TracyD3D11NamedZoneCS( ctx, varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D11ZoneScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); 431 431 + # define TracyD3D11ZoneTransientS(ctx, varname, name, depth, active) tracy::D3D11ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), depth, active }; 430 432 #else 431 433 # define TracyD3D11ZoneS( ctx, name, depth, active ) TracyD3D11Zone( ctx, name ) 432 434 # define TracyD3D11ZoneCS( ctx, name, color, depth, active ) TracyD3D11ZoneC( name, color )

+8 -8

src/external/tracy/tracy/TracyD3D12.hpp

··· 474 474 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 475 475 # define TracyD3D12Zone(ctx, cmdList, name) TracyD3D12NamedZoneS(ctx, ___tracy_gpu_zone, cmdList, name, TRACY_CALLSTACK, true) 476 476 # define TracyD3D12ZoneC(ctx, cmdList, name, color) TracyD3D12NamedZoneCS(ctx, ___tracy_gpu_zone, cmdList, name, color, TRACY_CALLSTACK, true) 477 477 - # define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), TRACY_CALLSTACK, active }; 478 478 - # define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), TRACY_CALLSTACK, active }; 477 477 + # define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), TRACY_CALLSTACK, active }; 478 478 + # define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), TRACY_CALLSTACK, active }; 479 479 # define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, TRACY_CALLSTACK, active) 480 480 #else 481 481 # define TracyD3D12Zone(ctx, cmdList, name) TracyD3D12NamedZone(ctx, ___tracy_gpu_zone, cmdList, name, true) 482 482 # define TracyD3D12ZoneC(ctx, cmdList, name, color) TracyD3D12NamedZoneC(ctx, ___tracy_gpu_zone, cmdList, name, color, true) 483 483 - # define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), active }; 484 484 - # define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), active }; 485 485 - # define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) tracy::D3D12ZoneScope varname{ ctx, __LINE__, __FILE__, strlen(__FILE__), __FUNCTION__, strlen(__FUNCTION__), name, strlen(name), cmdList, active }; 483 483 + # define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), active }; 484 484 + # define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), active }; 485 485 + # define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) tracy::D3D12ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), cmdList, active }; 486 486 #endif 487 487 488 488 #ifdef TRACY_HAS_CALLSTACK 489 489 # define TracyD3D12ZoneS(ctx, cmdList, name, depth) TracyD3D12NamedZoneS(ctx, ___tracy_gpu_zone, cmdList, name, depth, true) 490 490 # define TracyD3D12ZoneCS(ctx, cmdList, name, color, depth) TracyD3D12NamedZoneCS(ctx, ___tracy_gpu_zone, cmdList, name, color, depth, true) 491 491 - # define TracyD3D12NamedZoneS(ctx, varname, cmdList, name, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), depth, active }; 492 492 - # define TracyD3D12NamedZoneCS(ctx, varname, cmdList, name, color, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, __LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, __LINE__), depth, active }; 493 493 - # define TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, depth, active) tracy::D3D12ZoneScope varname{ ctx, __LINE__, __FILE__, strlen(__FILE__), __FUNCTION__, strlen(__FUNCTION__), name, strlen(name), cmdList, depth, active }; 491 491 + # define TracyD3D12NamedZoneS(ctx, varname, cmdList, name, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), depth, active }; 492 492 + # define TracyD3D12NamedZoneCS(ctx, varname, cmdList, name, color, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location, TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyConcat(__tracy_gpu_source_location, TracyLine), depth, active }; 493 493 + # define TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, depth, active) tracy::D3D12ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), cmdList, depth, active }; 494 494 #else 495 495 # define TracyD3D12ZoneS(ctx, cmdList, name, depth) TracyD3D12Zone(ctx, cmdList, name) 496 496 # define TracyD3D12ZoneCS(ctx, cmdList, name, color, depth) TracyD3D12Zone(ctx, cmdList, name, color)

+11 -11

src/external/tracy/tracy/TracyOpenCL.hpp

··· 47 47 #define TRACY_CL_TO_STRING_INDIRECT(T) #T 48 48 #define TRACY_CL_TO_STRING(T) TRACY_CL_TO_STRING_INDIRECT(T) 49 49 #define TRACY_CL_ASSERT(p) if(!(p)) { \ 50 50 - TracyMessageL( "TRACY_CL_ASSERT failed on " __FILE__ ":" TRACY_CL_TO_STRING(__LINE__) ); \ 50 50 + TracyMessageL( "TRACY_CL_ASSERT failed on " TracyFile ":" TRACY_CL_TO_STRING(TracyLine) ); \ 51 51 assert(false && "TRACY_CL_ASSERT failed"); \ 52 52 } 53 53 #define TRACY_CL_CHECK_ERROR(err) if(err != CL_SUCCESS) { \ 54 54 std::ostringstream oss; \ 55 55 - oss << "TRACY_CL_CHECK_ERROR failed on " << __FILE__ << ":" << __LINE__ \ 55 55 + oss << "TRACY_CL_CHECK_ERROR failed on " << TracyFile << ":" << TracyLine \ 56 56 << ": error code " << err; \ 57 57 auto msg = oss.str(); \ 58 58 TracyMessage(msg.data(), msg.size()); \ ··· 377 377 #define TracyCLDestroy(ctx) tracy::DestroyCLContext(ctx); 378 378 #define TracyCLContextName(context, name, size) ctx->Name(name, size); 379 379 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 380 380 - # define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 381 381 - # define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 380 380 + # define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 381 381 + # define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 382 382 # define TracyCLZone(ctx, name) TracyCLNamedZoneS(ctx, __tracy_gpu_zone, name, TRACY_CALLSTACK, true) 383 383 # define TracyCLZoneC(ctx, name, color) TracyCLNamedZoneCS(ctx, __tracy_gpu_zone, name, color, TRACY_CALLSTACK, true) 384 384 - # define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), TRACY_CALLSTACK, active ); 384 384 + # define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ); 385 385 #else 386 386 - # define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__){ name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), active); 387 387 - # define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__){ name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), active); 386 386 + # define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active); 387 387 + # define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active); 388 388 # define TracyCLZone(ctx, name) TracyCLNamedZone(ctx, __tracy_gpu_zone, name, true) 389 389 # define TracyCLZoneC(ctx, name, color) TracyCLNamedZoneC(ctx, __tracy_gpu_zone, name, color, true ) 390 390 - # define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), active ); 390 390 + # define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ); 391 391 #endif 392 392 393 393 #ifdef TRACY_HAS_CALLSTACK 394 394 - # define TracyCLNamedZoneS(ctx, varname, name, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__){ name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active); 395 395 - # define TracyCLNamedZoneCS(ctx, varname, name, color, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__){ name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active); 394 394 + # define TracyCLNamedZoneS(ctx, varname, name, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active); 395 395 + # define TracyCLNamedZoneCS(ctx, varname, name, color, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active); 396 396 # define TracyCLZoneS(ctx, name, depth) TracyCLNamedZoneS(ctx, __tracy_gpu_zone, name, depth, true) 397 397 # define TracyCLZoneCS(ctx, name, color, depth) TracyCLNamedZoneCS(ctx, __tracy_gpu_zone, name, color, depth, true) 398 398 - # define TracyCLZoneTransientS( ctx, varname, name, depth, active ) tracy::OpenCLCtxScope varname( ctx, __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), depth, active ); 398 398 + # define TracyCLZoneTransientS( ctx, varname, name, depth, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ); 399 399 #else 400 400 # define TracyCLNamedZoneS(ctx, varname, name, depth, active) TracyCLNamedZone(ctx, varname, name, active) 401 401 # define TracyCLNamedZoneCS(ctx, varname, name, color, depth, active) TracyCLNamedZoneC(ctx, varname, name, color, active)

+9 -9

src/external/tracy/tracy/TracyOpenGL.hpp

··· 52 52 #define TracyGpuContext tracy::GetGpuCtx().ptr = (tracy::GpuCtx*)tracy::tracy_malloc( sizeof( tracy::GpuCtx ) ); new(tracy::GetGpuCtx().ptr) tracy::GpuCtx; 53 53 #define TracyGpuContextName( name, size ) tracy::GetGpuCtx().ptr->Name( name, size ); 54 54 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 55 55 - # define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 56 56 - # define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active ); 55 55 + # define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 56 56 + # define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); 57 57 # define TracyGpuZone( name ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, TRACY_CALLSTACK, true ) 58 58 # define TracyGpuZoneC( name, color ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, TRACY_CALLSTACK, true ) 59 59 - # define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), TRACY_CALLSTACK, active ); 59 59 + # define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ); 60 60 #else 61 61 - # define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active ); 62 62 - # define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active ); 61 61 + # define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); 62 62 + # define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); 63 63 # define TracyGpuZone( name ) TracyGpuNamedZone( ___tracy_gpu_zone, name, true ) 64 64 # define TracyGpuZoneC( name, color ) TracyGpuNamedZoneC( ___tracy_gpu_zone, name, color, true ) 65 65 - # define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), active ); 65 65 + # define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ); 66 66 #endif 67 67 #define TracyGpuCollect tracy::GetGpuCtx().ptr->Collect(); 68 68 69 69 #ifdef TRACY_HAS_CALLSTACK 70 70 - # define TracyGpuNamedZoneS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active ); 71 71 - # define TracyGpuNamedZoneCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active ); 70 70 + # define TracyGpuNamedZoneS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); 71 71 + # define TracyGpuNamedZoneCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); 72 72 # define TracyGpuZoneS( name, depth ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, depth, true ) 73 73 # define TracyGpuZoneCS( name, color, depth ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, depth, true ) 74 74 - # define TracyGpuZoneTransientS( varname, name, depth, active ) tracy::GpuCtxScope varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), depth, active ); 74 74 + # define TracyGpuZoneTransientS( varname, name, depth, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ); 75 75 #else 76 76 # define TracyGpuNamedZoneS( varname, name, depth, active ) TracyGpuNamedZone( varname, name, active ) 77 77 # define TracyGpuNamedZoneCS( varname, name, color, depth, active ) TracyGpuNamedZoneC( varname, name, color, active )

+8 -8

src/external/tracy/tracy/TracyVulkan.hpp

··· 479 479 #define TracyVkDestroy( ctx ) tracy::DestroyVkContext( ctx ); 480 480 #define TracyVkContextName( ctx, name, size ) ctx->Name( name, size ); 481 481 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK 482 482 - # define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active ); 483 483 - # define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active ); 482 482 + # define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, TRACY_CALLSTACK, active ); 483 483 + # define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, TRACY_CALLSTACK, active ); 484 484 # define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, TRACY_CALLSTACK, true ) 485 485 # define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, TRACY_CALLSTACK, true ) 486 486 # define TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) TracyVkZoneTransientS( ctx, varname, cmdbuf, name, TRACY_CALLSTACK, active ) 487 487 #else 488 488 - # define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active ); 489 489 - # define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active ); 488 488 + # define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, active ); 489 489 + # define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, active ); 490 490 # define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZone( ctx, ___tracy_gpu_zone, cmdbuf, name, true ) 491 491 # define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneC( ctx, ___tracy_gpu_zone, cmdbuf, name, color, true ) 492 492 - # define TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) tracy::VkCtxScope varname( ctx, __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), cmdbuf, active ); 492 492 + # define TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) tracy::VkCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), cmdbuf, active ); 493 493 #endif 494 494 #define TracyVkCollect( ctx, cmdbuf ) ctx->Collect( cmdbuf ); 495 495 496 496 #ifdef TRACY_HAS_CALLSTACK 497 497 - # define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active ); 498 498 - # define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__, __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active ); 497 497 + # define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, depth, active ); 498 498 + # define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, depth, active ); 499 499 # define TracyVkZoneS( ctx, cmdbuf, name, depth ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, depth, true ) 500 500 # define TracyVkZoneCS( ctx, cmdbuf, name, color, depth ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, depth, true ) 501 501 - # define TracyVkZoneTransientS( ctx, varname, cmdbuf, name, depth, active ) tracy::VkCtxScope varname( ctx, __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), cmdbuf, depth, active ); 501 501 + # define TracyVkZoneTransientS( ctx, varname, cmdbuf, name, depth, active ) tracy::VkCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), cmdbuf, depth, active ); 502 502 #else 503 503 # define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) 504 504 # define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active )