diff options
Diffstat (limited to 'src')
69 files changed, 908 insertions, 252 deletions
diff --git a/src/citra/resource.h b/src/citra/resource.h Binary files differindex 0d42c8a8..12789642 100644 --- a/src/citra/resource.h +++ b/src/citra/resource.h diff --git a/src/citra_qt/CMakeLists.txt b/src/citra_qt/CMakeLists.txt index efccdbec..c0577938 100644 --- a/src/citra_qt/CMakeLists.txt +++ b/src/citra_qt/CMakeLists.txt @@ -89,15 +89,15 @@ if (Qt5_FOUND AND MSVC) ) set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/") set(PLATFORMS ${DLL_DEST}platforms/) - + # windows commandline expects the / to be \ so switch them string(REPLACE "/" "\\" Qt5_DLL_DIR ${Qt5_DLL_DIR}) string(REPLACE "/" "\\" Qt5_PLATFORMS_DIR ${Qt5_PLATFORMS_DIR}) string(REPLACE "/" "\\" DLL_DEST ${DLL_DEST}) string(REPLACE "/" "\\" PLATFORMS ${PLATFORMS}) - + # /NJH /NJS /NDL /NFL /NC /NS /NP - Silence any output - # cmake adds an extra check for command success which doesn't work too well with robocopy + # cmake adds an extra check for command success which doesn't work too well with robocopy # so trick it into thinking the command was successful with the || cmd /c "exit /b 0" add_custom_command(TARGET citra-qt POST_BUILD COMMAND robocopy ${Qt5_DLL_DIR} ${DLL_DEST} ${Qt5_DLLS} /NJH /NJS /NDL /NFL /NC /NS /NP || cmd /c "exit /b 0" diff --git a/src/citra_qt/bootmanager.cpp b/src/citra_qt/bootmanager.cpp index 72b55e94..3db09c65 100644 --- a/src/citra_qt/bootmanager.cpp +++ b/src/citra_qt/bootmanager.cpp @@ -57,7 +57,7 @@ void EmuThread::run() { Core::SingleStep(); emit DebugModeEntered(); yieldCurrentThread(); - + was_active = false; } else { std::unique_lock<std::mutex> lock(running_mutex); diff --git a/src/citra_qt/bootmanager.h b/src/citra_qt/bootmanager.h index 16809eaa..47512431 100644 --- a/src/citra_qt/bootmanager.h +++ b/src/citra_qt/bootmanager.h @@ -80,7 +80,7 @@ signals: * @warning When connecting to this signal from other threads, make sure to specify either Qt::QueuedConnection (invoke slot within the destination object's message thread) or even Qt::BlockingQueuedConnection (additionally block source thread until slot returns) */ void DebugModeEntered(); - + /** * Emitted right before the CPU continues execution * diff --git a/src/citra_qt/debugger/callstack.cpp b/src/citra_qt/debugger/callstack.cpp index 94e20471..6799ce84 100644 --- a/src/citra_qt/debugger/callstack.cpp +++ b/src/citra_qt/debugger/callstack.cpp @@ -39,7 +39,7 @@ void CallstackWidget::OnDebugModeEntered() { ret_addr = Memory::Read32(addr); call_addr = ret_addr - 4; //get call address??? - + if (Memory::GetPointer(call_addr) == nullptr) break; diff --git a/src/citra_qt/main.cpp b/src/citra_qt/main.cpp index f6010459..8041816a 100644 --- a/src/citra_qt/main.cpp +++ b/src/citra_qt/main.cpp @@ -127,7 +127,7 @@ GMainWindow::GMainWindow() : emu_thread(nullptr) ui.action_Use_Hardware_Renderer->setChecked(Settings::values.use_hw_renderer); SetHardwareRendererEnabled(ui.action_Use_Hardware_Renderer->isChecked()); - + ui.action_Single_Window_Mode->setChecked(settings.value("singleWindowMode", true).toBool()); ToggleWindowMode(); diff --git a/src/citra_qt/main.ui b/src/citra_qt/main.ui index 0942c28c..9a809ee6 100644 --- a/src/citra_qt/main.ui +++ b/src/citra_qt/main.ui @@ -24,7 +24,20 @@ <bool>true</bool> </property> <widget class="QWidget" name="centralwidget"> - <layout class="QHBoxLayout" name="horizontalLayout"/> + <layout class="QHBoxLayout" name="horizontalLayout"> + <property name="leftMargin"> + <number>0</number> + </property> + <property name="topMargin"> + <number>0</number> + </property> + <property name="rightMargin"> + <number>0</number> + </property> + <property name="bottomMargin"> + <number>0</number> + </property> + </layout> </widget> <widget class="QMenuBar" name="menubar"> <property name="geometry"> @@ -92,7 +105,7 @@ </action> <action name="action_Start"> <property name="enabled"> - <bool>false</bool> + <bool>false</bool> </property> <property name="text"> <string>&Start</string> diff --git a/src/common/emu_window.cpp b/src/common/emu_window.cpp index f5b6c730..43facb85 100644 --- a/src/common/emu_window.cpp +++ b/src/common/emu_window.cpp @@ -32,7 +32,7 @@ std::tuple<unsigned,unsigned> EmuWindow::ClipToTouchScreen(unsigned new_x, unsig new_x = std::max(new_x, framebuffer_layout.bottom_screen.left); new_x = std::min(new_x, framebuffer_layout.bottom_screen.right-1); - + new_y = std::max(new_y, framebuffer_layout.bottom_screen.top); new_y = std::min(new_y, framebuffer_layout.bottom_screen.bottom-1); diff --git a/src/common/file_util.cpp b/src/common/file_util.cpp index 7cdd1484..24648ea3 100644 --- a/src/common/file_util.cpp +++ b/src/common/file_util.cpp @@ -16,7 +16,7 @@ #include <io.h> #include <direct.h> // getcwd #include <tchar.h> - + // 64 bit offsets for windows #define fseeko _fseeki64 #define ftello _ftelli64 diff --git a/src/common/swap.h b/src/common/swap.h index 7e37655b..588cebc7 100644 --- a/src/common/swap.h +++ b/src/common/swap.h @@ -135,7 +135,7 @@ template <> inline void swap<8>(u8* data) { *reinterpret_cast<u64*>(data) = swap64(data); } - + } // Namespace Common diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 42b4be93..bbc28516 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -36,6 +36,7 @@ set(SRCS hle/kernel/shared_memory.cpp hle/kernel/thread.cpp hle/kernel/timer.cpp + hle/kernel/vm_manager.cpp hle/service/ac_u.cpp hle/service/act_u.cpp hle/service/am_app.cpp @@ -147,6 +148,7 @@ set(HEADERS hle/kernel/shared_memory.h hle/kernel/thread.h hle/kernel/timer.h + hle/kernel/vm_manager.h hle/result.h hle/service/ac_u.h hle/service/act_u.h diff --git a/src/core/arm/disassembler/arm_disasm.cpp b/src/core/arm/disassembler/arm_disasm.cpp index 913dc145..f6d44d85 100644 --- a/src/core/arm/disassembler/arm_disasm.cpp +++ b/src/core/arm/disassembler/arm_disasm.cpp @@ -813,7 +813,7 @@ Opcode ARM_Disasm::Decode11(uint32_t insn) { // SWI return OP_SWI; } - + uint8_t bit4 = (insn >> 4) & 0x1; uint8_t cpnum = (insn >> 8) & 0xf; diff --git a/src/core/arm/dyncom/arm_dyncom_interpreter.cpp b/src/core/arm/dyncom/arm_dyncom_interpreter.cpp index b5d1b43c..b00eb49a 100644 --- a/src/core/arm/dyncom/arm_dyncom_interpreter.cpp +++ b/src/core/arm/dyncom/arm_dyncom_interpreter.cpp @@ -134,7 +134,7 @@ static unsigned int DPO(Immediate)(ARMul_State* cpu, unsigned int sht_oper) { unsigned int immed_8 = BITS(sht_oper, 0, 7); unsigned int rotate_imm = BITS(sht_oper, 8, 11); unsigned int shifter_operand = ROTATE_RIGHT_32(immed_8, rotate_imm * 2); - if (rotate_imm == 0) + if (rotate_imm == 0) cpu->shifter_carry_out = cpu->CFlag; else cpu->shifter_carry_out = BIT(shifter_operand, 31); @@ -521,7 +521,7 @@ static void MLnS(ImmediateOffset)(ARMul_State* cpu, unsigned int inst, unsigned addr = CHECK_READ_REG15_WA(cpu, Rn) + offset_8; else addr = CHECK_READ_REG15_WA(cpu, Rn) - offset_8; - + virt_addr = addr; } @@ -550,7 +550,7 @@ static void MLnS(ImmediatePreIndexed)(ARMul_State* cpu, unsigned int inst, unsig if (U_BIT) addr = rn + offset_8; - else + else addr = rn - offset_8; virt_addr = addr; @@ -1306,8 +1306,8 @@ static ARM_INST_PTR INTERPRETER_TRANSLATE(and)(unsigned int inst, int index) inst_cream->Rd = BITS(inst, 12, 15); inst_cream->shifter_operand = BITS(inst, 0, 11); inst_cream->shtop_func = get_shtop(inst); - - if (inst_cream->Rd == 15) + + if (inst_cream->Rd == 15) inst_base->br = INDIRECT_BRANCH; return inst_base; @@ -1350,7 +1350,7 @@ static ARM_INST_PTR INTERPRETER_TRANSLATE(bic)(unsigned int inst, int index) inst_cream->shifter_operand = BITS(inst, 0, 11); inst_cream->shtop_func = get_shtop(inst); - if (inst_cream->Rd == 15) + if (inst_cream->Rd == 15) inst_base->br = INDIRECT_BRANCH; return inst_base; } @@ -3269,7 +3269,7 @@ static ARM_INST_PTR INTERPRETER_TRANSLATE(yield)(unsigned int inst, int index) #define VFP_INTERPRETER_STRUCT #include "core/arm/skyeye_common/vfp/vfpinstr.cpp" #undef VFP_INTERPRETER_STRUCT - + #define VFP_INTERPRETER_TRANS #include "core/arm/skyeye_common/vfp/vfpinstr.cpp" #undef VFP_INTERPRETER_TRANS @@ -3478,9 +3478,9 @@ const transop_fp_t arm_instruction_trans[] = { INTERPRETER_TRANSLATE(bbl), // All the thumb instructions should be placed the end of table - INTERPRETER_TRANSLATE(b_2_thumb), - INTERPRETER_TRANSLATE(b_cond_thumb), - INTERPRETER_TRANSLATE(bl_1_thumb), + INTERPRETER_TRANSLATE(b_2_thumb), + INTERPRETER_TRANSLATE(b_cond_thumb), + INTERPRETER_TRANSLATE(bl_1_thumb), INTERPRETER_TRANSLATE(bl_2_thumb), INTERPRETER_TRANSLATE(blx_1_thumb) }; @@ -4338,7 +4338,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { } } if (BIT(inst, 13)) { - if (cpu->Mode == USER32MODE) + if (cpu->Mode == USER32MODE) cpu->Reg[13] = ReadMemory32(cpu, addr); else cpu->Reg_usr[0] = ReadMemory32(cpu, addr); @@ -4346,7 +4346,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { addr += 4; } if (BIT(inst, 14)) { - if (cpu->Mode == USER32MODE) + if (cpu->Mode == USER32MODE) cpu->Reg[14] = ReadMemory32(cpu, addr); else cpu->Reg_usr[1] = ReadMemory32(cpu, addr); @@ -5148,7 +5148,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { REV16_INST: REVSH_INST: { - + if (inst_base->cond == 0xE || CondPassed(cpu, inst_base->cond)) { rev_inst* const inst_cream = (rev_inst*)inst_base->component; @@ -5721,7 +5721,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { if (do_swap) rm_val = (((rm_val & 0xFFFF) << 16) | (rm_val >> 16)); - + const s32 product1 = (s16)(rn_val & 0xFFFF) * (s16)(rm_val & 0xFFFF); const s32 product2 = (s16)((rn_val >> 16) & 0xFFFF) * (s16)((rm_val >> 16) & 0xFFFF); s64 result; @@ -6583,7 +6583,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { { u32 lo_val = 0; u32 hi_val = 0; - + // UHADD16 if (op2 == 0x00) { lo_val = (rn_val & 0xFFFF) + (rm_val & 0xFFFF); @@ -6772,7 +6772,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) { u16 lo_val = 0; u16 hi_val = 0; - + // UQADD16 if (op2 == 0x00) { lo_val = ARMul_UnsignedSaturatedAdd16(rn_val & 0xFFFF, rm_val & 0xFFFF); diff --git a/src/core/arm/dyncom/arm_dyncom_thumb.cpp b/src/core/arm/dyncom/arm_dyncom_thumb.cpp index cdaf2145..2fc8170b 100644 --- a/src/core/arm/dyncom/arm_dyncom_thumb.cpp +++ b/src/core/arm/dyncom/arm_dyncom_thumb.cpp @@ -184,7 +184,7 @@ tdstate thumb_translate(u32 addr, u32 instr, u32* ainstr, u32* inst_size) { case 9: // LDR Rd,[PC,#imm8] *ainstr = 0xE59F0000 // base | ((tinstr & 0x0700) << (12 - 8)) // Rd - |((tinstr & 0x00FF) << (2 - 0)); // off8 + |((tinstr & 0x00FF) << (2 - 0)); // off8 break; case 10: diff --git a/src/core/arm/interpreter/armsupp.cpp b/src/core/arm/interpreter/armsupp.cpp index 1b078dc7..83f7f3e2 100644 --- a/src/core/arm/interpreter/armsupp.cpp +++ b/src/core/arm/interpreter/armsupp.cpp @@ -628,7 +628,7 @@ void WriteCP15Register(ARMul_State* cpu, u32 value, u32 crn, u32 opcode_1, u32 c cpu->CP15[CP15_DATA_SYNC_BARRIER] = value; else if (opcode_2 == 5) cpu->CP15[CP15_DATA_MEMORY_BARRIER] = value; - + } else if (crn == 13 && opcode_1 == 0 && crm == 0 && opcode_2 == 2) { diff --git a/src/core/arm/skyeye_common/armdefs.h b/src/core/arm/skyeye_common/armdefs.h index 08ece69b..d2c90110 100644 --- a/src/core/arm/skyeye_common/armdefs.h +++ b/src/core/arm/skyeye_common/armdefs.h @@ -1,16 +1,16 @@ /* armdefs.h -- ARMulator common definitions: ARM6 Instruction Emulator. Copyright (C) 1994 Advanced RISC Machines Ltd. - + This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. - + This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. - + You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ diff --git a/src/core/arm/skyeye_common/vfp/vfp_helper.h b/src/core/arm/skyeye_common/vfp/vfp_helper.h index ccc0212a..2007d6dc 100644 --- a/src/core/arm/skyeye_common/vfp/vfp_helper.h +++ b/src/core/arm/skyeye_common/vfp/vfp_helper.h @@ -18,10 +18,10 @@ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ -/* +/* * The following code is derivative from Linux Android kernel vfp * floating point support. - * + * * Copyright (C) 2004 ARM Limited. * Written by Deep Blue Solutions Limited. * diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp index f70c84c3..e53c2e60 100644 --- a/src/core/core_timing.cpp +++ b/src/core/core_timing.cpp @@ -549,7 +549,7 @@ std::string GetScheduledEventsSummary() { const char* name = event_types[event->type].name; if (!name) name = "[unknown]"; - text += Common::StringFromFormat("%s : %i %08x%08x\n", name, (int)event->time, + text += Common::StringFromFormat("%s : %i %08x%08x\n", name, (int)event->time, (u32)(event->userdata >> 32), (u32)(event->userdata)); event = event->next; } diff --git a/src/core/core_timing.h b/src/core/core_timing.h index 01519608..64f5b06d 100644 --- a/src/core/core_timing.h +++ b/src/core/core_timing.h @@ -87,7 +87,7 @@ void UnregisterAllEvents(); /// userdata MAY NOT CONTAIN POINTERS. userdata might get written and reloaded from disk, /// when we implement state saves. /** - * Schedules an event to run after the specified number of cycles, + * Schedules an event to run after the specified number of cycles, * with an optional parameter to be passed to the callback handler. * This must be run ONLY from within the cpu thread. * @param cycles_into_future The number of cycles after which this event will be fired diff --git a/src/core/file_sys/archive_extsavedata.cpp b/src/core/file_sys/archive_extsavedata.cpp index 38d498d0..975aafb3 100644 --- a/src/core/file_sys/archive_extsavedata.cpp +++ b/src/core/file_sys/archive_extsavedata.cpp @@ -30,8 +30,8 @@ std::string GetExtSaveDataPath(const std::string& mount_point, const Path& path) std::string GetExtDataContainerPath(const std::string& mount_point, bool shared) { if (shared) return Common::StringFromFormat("%sdata/%s/extdata/", mount_point.c_str(), SYSTEM_ID.c_str()); - - return Common::StringFromFormat("%sNintendo 3DS/%s/%s/extdata/", mount_point.c_str(), + + return Common::StringFromFormat("%sNintendo 3DS/%s/%s/extdata/", mount_point.c_str(), SYSTEM_ID.c_str(), SDCARD_ID.c_str()); } diff --git a/src/core/file_sys/archive_extsavedata.h b/src/core/file_sys/archive_extsavedata.h index c77c04e4..ef0b27bd 100644 --- a/src/core/file_sys/archive_extsavedata.h +++ b/src/core/file_sys/archive_extsavedata.h @@ -35,14 +35,14 @@ public: private: /** * This holds the full directory path for this archive, it is only set after a successful call - * to Open, this is formed as <base extsavedatapath>/<type>/<high>/<low>. + * to Open, this is formed as <base extsavedatapath>/<type>/<high>/<low>. * See GetExtSaveDataPath for the code that extracts this data from an archive path. */ std::string mount_point; }; /** - * Constructs a path to the concrete ExtData archive in the host filesystem based on the + * Constructs a path to the concrete ExtData archive in the host filesystem based on the * input Path and base mount point. * @param mount_point The base mount point of the ExtSaveData archives. * @param path The path that identifies the requested concrete ExtSaveData archive. diff --git a/src/core/file_sys/archive_savedata.cpp b/src/core/file_sys/archive_savedata.cpp index 8dff5196..a9230937 100644 --- a/src/core/file_sys/archive_savedata.cpp +++ b/src/core/file_sys/archive_savedata.cpp @@ -21,7 +21,7 @@ namespace FileSys { static std::string GetSaveDataContainerPath(const std::string& sdmc_directory) { - return Common::StringFromFormat("%sNintendo 3DS/%s/%s/title/", sdmc_directory.c_str(), + return Common::StringFromFormat("%sNintendo 3DS/%s/%s/title/", sdmc_directory.c_str(), SYSTEM_ID.c_str(), SDCARD_ID.c_str()); } diff --git a/src/core/hle/kernel/resource_limit.h b/src/core/hle/kernel/resource_limit.h index 201ec0db..1b8249c7 100644 --- a/src/core/hle/kernel/resource_limit.h +++ b/src/core/hle/kernel/resource_limit.h @@ -81,13 +81,13 @@ public: s32 max_timers = 0; s32 max_shared_mems = 0; s32 max_address_arbiters = 0; - + /// Max CPU time that the processes in this category can utilize s32 max_cpu_time = 0; - // TODO(Subv): Increment these in their respective Kernel::T::Create functions, keeping in mind that - // APPLICATION resource limits should not be affected by the objects created by service modules. - // Currently we have no way of distinguishing if a Create was called by the running application, + // TODO(Subv): Increment these in their respective Kernel::T::Create functions, keeping in mind that + // APPLICATION resource limits should not be affected by the objects created by service modules. + // Currently we have no way of distinguishing if a Create was called by the running application, // or by a service module. Approach this once we have separated the service modules into their own processes /// Current memory that the processes in this category are using diff --git a/src/core/hle/kernel/semaphore.cpp b/src/core/hle/kernel/semaphore.cpp index dbb4c9b7..96d61ed3 100644 --- a/src/core/hle/kernel/semaphore.cpp +++ b/src/core/hle/kernel/semaphore.cpp @@ -42,7 +42,7 @@ void Semaphore::Acquire() { ResultVal<s32> Semaphore::Release(s32 release_count) { if (max_count - available_count < release_count) - return ResultCode(ErrorDescription::OutOfRange, ErrorModule::Kernel, + return ResultCode(ErrorDescription::OutOfRange, ErrorModule::Kernel, ErrorSummary::InvalidArgument, ErrorLevel::Permanent); s32 previous_count = available_count; diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp index 690d33b5..22c795ad 100644 --- a/src/core/hle/kernel/thread.cpp +++ b/src/core/hle/kernel/thread.cpp @@ -100,7 +100,7 @@ void Thread::Stop() { } status = THREADSTATUS_DEAD; - + WakeupAllWaitingThreads(); // Clean up any dangling references in objects that this thread was waiting for @@ -169,7 +169,7 @@ static void PriorityBoostStarvedThreads() { } } -/** +/** * Switches the CPU's active thread context to that of the specified thread * @param new_thread The thread to switch to */ @@ -353,7 +353,7 @@ void Thread::ResumeFromWait() { GetObjectId()); return; } - + ready_queue.push_back(current_priority, this); status = THREADSTATUS_READY; } @@ -504,7 +504,7 @@ void Reschedule() { } else if (next) { LOG_TRACE(Kernel, "context switch idle -> %u", next->GetObjectId()); } - + SwitchContext(next); } diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h index 38992817..2c65419c 100644 --- a/src/core/hle/kernel/thread.h +++ b/src/core/hle/kernel/thread.h @@ -94,7 +94,7 @@ public: * @return The thread's ID */ u32 GetThreadId() const { return thread_id; } - + /** * Release an acquired wait object * @param wait_object WaitObject to release diff --git a/src/core/hle/kernel/timer.cpp b/src/core/hle/kernel/timer.cpp index 25d066bf..8aa4110a 100644 --- a/src/core/hle/kernel/timer.cpp +++ b/src/core/hle/kernel/timer.cpp @@ -88,7 +88,7 @@ static void TimerCallback(u64 timer_handle, int cycles_late) { if (timer->interval_delay != 0) { // Reschedule the timer with the interval delay u64 interval_microseconds = timer->interval_delay / 1000; - CoreTiming::ScheduleEvent(usToCycles(interval_microseconds) - cycles_late, + CoreTiming::ScheduleEvent(usToCycles(interval_microseconds) - cycles_late, timer_callback_event_type, timer_handle); } } diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp new file mode 100644 index 00000000..b2dd2154 --- /dev/null +++ b/src/core/hle/kernel/vm_manager.cpp @@ -0,0 +1,245 @@ +// Copyright 2015 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "common/assert.h" + +#include "core/hle/kernel/vm_manager.h" +#include "core/memory_setup.h" + +namespace Kernel { + +bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const { + ASSERT(base + size == next.base); + if (permissions != next.permissions || + meminfo_state != next.meminfo_state || + type != next.type) { + return false; + } + if (type == VMAType::AllocatedMemoryBlock && + (backing_block != next.backing_block || offset + size != next.offset)) { + return false; + } + if (type == VMAType::BackingMemory && backing_memory + size != next.backing_memory) { + return false; + } + if (type == VMAType::MMIO && paddr + size != next.paddr) { + return false; + } + return true; +} + +VMManager::VMManager() { + Reset(); +} + +void VMManager::Reset() { + vma_map.clear(); + + // Initialize the map with a single free region covering the entire managed space. + VirtualMemoryArea initial_vma; + initial_vma.size = MAX_ADDRESS; + vma_map.emplace(initial_vma.base, initial_vma); + + UpdatePageTableForVMA(initial_vma); +} + +VMManager::VMAHandle VMManager::FindVMA(VAddr target) const { + return std::prev(vma_map.upper_bound(target)); +} + +ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target, + std::shared_ptr<std::vector<u8>> block, u32 offset, u32 size, MemoryState state) { + ASSERT(block != nullptr); + ASSERT(offset + size <= block->size()); + + // This is the appropriately sized VMA that will turn into our allocation. + CASCADE_RESULT(VMAIter vma_handle, CarveVMA(target, size)); + VirtualMemoryArea& final_vma = vma_handle->second; + ASSERT(final_vma.size == size); + + final_vma.type = VMAType::AllocatedMemoryBlock; + final_vma.permissions = VMAPermission::ReadWrite; + final_vma.meminfo_state = state; + final_vma.backing_block = block; + final_vma.offset = offset; + UpdatePageTableForVMA(final_vma); + + return MakeResult<VMAHandle>(MergeAdjacent(vma_handle)); +} + +ResultVal<VMManager::VMAHandle> VMManager::MapBackingMemory(VAddr target, u8 * memory, u32 size, MemoryState state) { + ASSERT(memory != nullptr); + + // This is the appropriately sized VMA that will turn into our allocation. + CASCADE_RESULT(VMAIter vma_handle, CarveVMA(target, size)); + VirtualMemoryArea& final_vma = vma_handle->second; + ASSERT(final_vma.size == size); + + final_vma.type = VMAType::BackingMemory; + final_vma.permissions = VMAPermission::ReadWrite; + final_vma.meminfo_state = state; + final_vma.backing_memory = memory; + UpdatePageTableForVMA(final_vma); + + return MakeResult<VMAHandle>(MergeAdjacent(vma_handle)); +} + +ResultVal<VMManager::VMAHandle> VMManager::MapMMIO(VAddr target, PAddr paddr, u32 size, MemoryState state) { + // This is the appropriately sized VMA that will turn into our allocation. + CASCADE_RESULT(VMAIter vma_handle, CarveVMA(target, size)); + VirtualMemoryArea& final_vma = vma_handle->second; + ASSERT(final_vma.size == size); + + final_vma.type = VMAType::MMIO; + final_vma.permissions = VMAPermission::ReadWrite; + final_vma.meminfo_state = state; + final_vma.paddr = paddr; + UpdatePageTableForVMA(final_vma); + + return MakeResult<VMAHandle>(MergeAdjacent(vma_handle)); +} + +void VMManager::Unmap(VMAHandle vma_handle) { + VMAIter iter = StripIterConstness(vma_handle); + + VirtualMemoryArea& vma = iter->second; + vma.type = VMAType::Free; + vma.permissions = VMAPermission::None; + vma.meminfo_state = MemoryState::Free; + + vma.backing_block = nullptr; + vma.offset = 0; + vma.backing_memory = nullptr; + vma.paddr = 0; + + UpdatePageTableForVMA(vma); + + MergeAdjacent(iter); +} + +void VMManager::Reprotect(VMAHandle vma_handle, VMAPermission new_perms) { + VMAIter iter = StripIterConstness(vma_handle); + + VirtualMemoryArea& vma = iter->second; + vma.permissions = new_perms; + UpdatePageTableForVMA(vma); + + MergeAdjacent(iter); +} + +VMManager::VMAIter VMManager::StripIterConstness(const VMAHandle & iter) { + // This uses a neat C++ trick to convert a const_iterator to a regular iterator, given + // non-const access to its container. + return vma_map.erase(iter, iter); // Erases an empty range of elements +} + +ResultVal<VMManager::VMAIter> VMManager::CarveVMA(VAddr base, u32 size) { + ASSERT_MSG((size & Memory::PAGE_MASK) == 0, "non-page aligned size: %8X", size); + ASSERT_MSG((base & Memory::PAGE_MASK) == 0, "non-page aligned base: %08X", base); + + VMAIter vma_handle = StripIterConstness(FindVMA(base)); + if (vma_handle == vma_map.end()) { + // Target address is outside the range managed by the kernel + return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS, + ErrorSummary::InvalidArgument, ErrorLevel::Usage); // 0xE0E01BF5 + } + + VirtualMemoryArea& vma = vma_handle->second; + if (vma.type != VMAType::Free) { + // Region is already allocated + return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS, + ErrorSummary::InvalidState, ErrorLevel::Usage); // 0xE0A01BF5 + } + + u32 start_in_vma = base - vma.base; + u32 end_in_vma = start_in_vma + size; + + if (end_in_vma > vma.size) { + // Requested allocation doesn't fit inside VMA + return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS, + ErrorSummary::InvalidState, ErrorLevel::Usage); // 0xE0A01BF5 + } + + if (end_in_vma != vma.size) { + // Split VMA at the end of the allocated region + SplitVMA(vma_handle, end_in_vma); + } + if (start_in_vma != 0) { + // Split VMA at the start of the allocated region + vma_handle = SplitVMA(vma_handle, start_in_vma); + } + + return MakeResult<VMAIter>(vma_handle); +} + +VMManager::VMAIter VMManager::SplitVMA(VMAIter vma_handle, u32 offset_in_vma) { + VirtualMemoryArea& old_vma = vma_handle->second; + VirtualMemoryArea new_vma = old_vma; // Make a copy of the VMA + + // For now, don't allow no-op VMA splits (trying to split at a boundary) because it's probably + // a bug. This restriction might be removed later. + ASSERT(offset_in_vma < old_vma.size); + ASSERT(offset_in_vma > 0); + + old_vma.size = offset_in_vma; + new_vma.base += offset_in_vma; + new_vma.size -= offset_in_vma; + + switch (new_vma.type) { + case VMAType::Free: + break; + case VMAType::AllocatedMemoryBlock: + new_vma.offset += offset_in_vma; + break; + case VMAType::BackingMemory: + new_vma.backing_memory += offset_in_vma; + break; + case VMAType::MMIO: + new_vma.paddr += offset_in_vma; + break; + } + + ASSERT(old_vma.CanBeMergedWith(new_vma)); + + return vma_map.emplace_hint(std::next(vma_handle), new_vma.base, new_vma); +} + +VMManager::VMAIter VMManager::MergeAdjacent(VMAIter iter) { + VMAIter next_vma = std::next(iter); + if (next_vma != vma_map.end() && iter->second.CanBeMergedWith(next_vma->second)) { + iter->second.size += next_vma->second.size; + vma_map.erase(next_vma); + } + + if (iter != vma_map.begin()) { + VMAIter prev_vma = std::prev(iter); + if (prev_vma->second.CanBeMergedWith(iter->second)) { + prev_vma->second.size += iter->second.size; + vma_map.erase(iter); + iter = prev_vma; + } + } + + return iter; +} + +void VMManager::UpdatePageTableForVMA(const VirtualMemoryArea& vma) { + switch (vma.type) { + case VMAType::Free: + Memory::UnmapRegion(vma.base, vma.size); + break; + case VMAType::AllocatedMemoryBlock: + Memory::MapMemoryRegion(vma.base, vma.size, vma.backing_block->data() + vma.offset); + break; + case VMAType::BackingMemory: + Memory::MapMemoryRegion(vma.base, vma.size, vma.backing_memory); + break; + case VMAType::MMIO: + // TODO(yuriks): Add support for MMIO handlers. + Memory::MapIoRegion(vma.base, vma.size); + break; + } +} + +} diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h new file mode 100644 index 00000000..22b72460 --- /dev/null +++ b/src/core/hle/kernel/vm_manager.h @@ -0,0 +1,200 @@ +// Copyright 2015 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <map> +#include <memory> +#include <string> +#include <vector> + +#include "common/common_types.h" + +#include "core/hle/result.h" + +namespace Kernel { + +enum class VMAType : u8 { + /// VMA represents an unmapped region of the address space. + Free, + /// VMA is backed by a ref-counted allocate memory block. + AllocatedMemoryBlock, + /// VMA is backed by a raw, unmanaged pointer. + BackingMemory, + /// VMA is mapped to MMIO registers at a fixed PAddr. + MMIO, + // TODO(yuriks): Implement MemoryAlias to support MAP/UNMAP +}; + +/// Permissions for mapped memory blocks +enum class VMAPermission : u8 { + None = 0, + Read = 1, + Write = 2, + Execute = 4, + + ReadWrite = Read | Write, + ReadExecute = Read | Execute, + WriteExecute = Write | Execute, + ReadWriteExecute = Read | Write | Execute, +}; + +/// Set of values returned in MemoryInfo.state by svcQueryMemory. +enum class MemoryState : u8 { + Free = 0, + Reserved = 1, + IO = 2, + Static = 3, + Code = 4, + Private = 5, + Shared = 6, + Continuous = 7, + Aliased = 8, + Alias = 9, + AliasCode = 10, + Locked = 11, +}; + +/** + * Represents a VMA in an address space. A VMA is a contiguous region of virtual addressing space + * with homogeneous attributes across its extents. In this particular implementation each VMA is + * also backed by a single host memory allocation. + */ +struct VirtualMemoryArea { + /// Virtual base address of the region. + VAddr base = 0; + /// Size of the region. + u32 size = 0; + + VMAType type = VMAType::Free; + VMAPermission permissions = VMAPermission::None; + /// Tag returned by svcQueryMemory. Not otherwise used. + MemoryState meminfo_state = MemoryState::Free; + + // Settings for type = AllocatedMemoryBlock + /// Memory block backing this VMA. + std::shared_ptr<std::vector<u8>> backing_block = nullptr; + /// Offset into the backing_memory the mapping starts from. + u32 offset = 0; + + // Settings for type = BackingMemory + /// Pointer backing this VMA. It will not be destroyed or freed when the VMA is removed. + u8* backing_memory = nullptr; + + // Settings for type = MMIO + /// Physical address of the register area this VMA maps to. + PAddr paddr = 0; + + /// Tests if this area can be merged to the right with `next`. + bool CanBeMergedWith(const VirtualMemoryArea& next) const; +}; + +/** + * Manages a process' virtual addressing space. This class maintains a list of allocated and free + * regions in the address space, along with their attributes, and allows kernel clients to + * manipulate it, adjusting the page table to match. + * + * This is similar in idea and purpose to the VM manager present in operating system kernels, with + * the main difference being that it doesn't have to support swapping or memory mapping of files. + * The implementation is also simplified by not having to allocate page frames. See these articles + * about the Linux kernel for an explantion of the concept and implementation: + * - http://duartes.org/gustavo/blog/post/how-the-kernel-manages-your-memory/ + * - http://duartes.org/gustavo/blog/post/page-cache-the-affair-between-memory-and-files/ + */ +class VMManager { + // TODO(yuriks): Make page tables switchable to support multiple VMManagers +public: + /** + * The maximum amount of address space managed by the kernel. Addresses above this are never used. + * @note This is the limit used by the New 3DS kernel. Old 3DS used 0x20000000. + */ + static const u32 MAX_ADDRESS = 0x40000000; + + /** + * A map covering the entirety of the managed address space, keyed by the `base` field of each + * VMA. It must always be modified by splitting or merging VMAs, so that the invariant + * `elem.base + elem.size == next.base` is preserved, and mergeable regions must always be + * merged when possible so that no two similar and adjacent regions exist that have not been + * merged. + */ + std::map<VAddr, VirtualMemoryArea> vma_map; + using VMAHandle = decltype(vma_map)::const_iterator; + + VMManager(); + + /// Clears the address space map, re-initializing with a single free area. + void Reset(); + + /// Finds the VMA in which the given address is included in, or `vma_map.end()`. + VMAHandle FindVMA(VAddr target) const; + + // TODO(yuriks): Should these functions actually return the handle? + + /** + * Maps part of a ref-counted block of memory at a given address. + * + * @param target The guest address to start the mapping at. + * @param block The block to be mapped. + * @param offset Offset into `block` to map from. + * @param size Size of the mapping. + * @param state MemoryState tag to attach to the VMA. + */ + ResultVal<VMAHandle> MapMemoryBlock(VAddr target, std::shared_ptr<std::vector<u8>> block, + u32 offset, u32 size, MemoryState state); + + /** + * Maps an unmanaged host memory pointer at a given address. + * + * @param target The guest address to start the mapping at. + * @param memory The memory to be mapped. + * @param size Size of the mapping. + * @param state MemoryState tag to attach to the VMA. + */ + ResultVal<VMAHandle> MapBackingMemory(VAddr target, u8* memory, u32 size, MemoryState state); + + /** + * Maps a memory-mapped IO region at a given address. + * + * @param target The guest address to start the mapping at. + * @param paddr The physical address where the registers are present. + * @param size Size of the mapping. + * @param state MemoryState tag to attach to the VMA. + */ + ResultVal<VMAHandle> MapMMIO(VAddr target, PAddr paddr, u32 size, MemoryState state); + + /// Unmaps the given VMA. + void Unmap(VMAHandle vma); + + /// Changes the permissions of the given VMA. + void Reprotect(VMAHandle vma, VMAPermission new_perms); + +private: + using VMAIter = decltype(vma_map)::iterator; + + /// Converts a VMAHandle to a mutable VMAIter. + VMAIter StripIterConstness(const VMAHandle& iter); + + /** + * Carves a VMA of a specific size at the specified address by splitting Free VMAs while doing + * the appropriate error checking. + */ + ResultVal<VMAIter> CarveVMA(VAddr base, u32 size); + + /** + * Splits a VMA in two, at the specified offset. + * @returns the right side of the split, with the original iterator becoming the left side. + */ + VMAIter SplitVMA(VMAIter vma, u32 offset_in_vma); + + /** + * Checks for and merges the specified VMA with adjacent ones if possible. + * @returns the merged VMA or the original if no merging was possible. + */ + VMAIter MergeAdjacent(VMAIter vma); + + /// Updates the pages corresponding to this VMA so they match the VMA's attributes. + void UpdatePageTableForVMA(const VirtualMemoryArea& vma); +}; + +} diff --git a/src/core/hle/service/apt/apt.cpp b/src/core/hle/service/apt/apt.cpp index 3fd4cfb0..5d14f393 100644 --- a/src/core/hle/service/apt/apt.cpp +++ b/src/core/hle/service/apt/apt.cpp @@ -151,7 +151,7 @@ void SendParameter(Service::Interface* self) { u32 handle = cmd_buff[6]; u32 size = cmd_buff[7]; u32 in_param_buffer_ptr = cmd_buff[8]; - + cmd_buff[1] = RESULT_SUCCESS.raw; // No error LOG_WARNING(Service_APT, "(STUBBED) called src_app_id=0x%08X, dst_app_id=0x%08X, signal_type=0x%08X," @@ -283,7 +283,7 @@ void Init() { AddService(new APT_A_Interface); AddService(new APT_S_Interface); AddService(new APT_U_Interface); - + // Load the shared system font (if available). // The expected format is a decrypted, uncompressed BCFNT file with the 0x80 byte header // generated by the APT:U service. The best way to get is by dumping it from RAM. We've provided diff --git a/src/core/hle/service/apt/apt.h b/src/core/hle/service/apt/apt.h index e7fa3932..a03e1712 100644 --- a/src/core/hle/service/apt/apt.h +++ b/src/core/hle/service/apt/apt.h @@ -63,7 +63,7 @@ void Initialize(Service::Interface* self); * 4 : Handle to shared font memory */ void GetSharedFont(Service::Interface* self); - + /** * APT::NotifyToWait service function * Inputs: @@ -88,7 +88,7 @@ void Enable(Service::Interface* self); * 4 : Home Menu AppId * 5 : AppID of currently active app */ -void GetAppletManInfo(Service::Interface* self); +void GetAppletManInfo(Service::Interface* self); /** * APT::IsRegistered service function. This returns whether the specified AppID is registered with NS yet. @@ -100,14 +100,14 @@ void GetAppletManInfo(Service::Interface* self); * Outputs: * 0 : Return header * 1 : Result of function, 0 on success, otherwise error code - * 2 : Output, 0 = not registered, 1 = registered. + * 2 : Output, 0 = not registered, 1 = registered. */ void IsRegistered(Service::Interface* self); void InquireNotification(Service::Interface* self); /** - * APT::SendParameter service function. This sets the parameter data state. + * APT::SendParameter service function. This sets the parameter data state. * Inputs: * 1 : Source AppID * 2 : Destination AppID diff --git a/src/core/hle/service/boss_p.h b/src/core/hle/service/boss_p.h index 71f1e746..6fb51d57 100644 --- a/src/core/hle/service/boss_p.h +++ b/src/core/hle/service/boss_p.h @@ -14,7 +14,7 @@ namespace BOSS_P { class Interface : public Service::Interface { public: Interface(); - + std::string GetPortName() const override { return "boss:P"; } diff --git a/src/core/hle/service/boss_u.h b/src/core/hle/service/boss_u.h index 2668f2df..89e77fe4 100644 --- a/src/core/hle/service/boss_u.h +++ b/src/core/hle/service/boss_u.h @@ -14,7 +14,7 @@ namespace BOSS_U { class Interface : public Service::Interface { public: Interface(); - + std::string GetPortName() const override { return "boss:U"; } diff --git a/src/core/hle/service/cam_u.cpp b/src/core/hle/service/cam_u.cpp index fcfd8771..ecda0dbd 100644 --- a/src/core/hle/service/cam_u.cpp +++ b/src/core/hle/service/cam_u.cpp @@ -19,5 +19,5 @@ namespace CAM_U { Interface::Interface() { //Register(FunctionTable); } - + } // namespace diff --git a/src/core/hle/service/cfg/cfg.cpp b/src/core/hle/service/cfg/cfg.cpp index d4268288..62ad90fd 100644 --- a/src/core/hle/service/cfg/cfg.cpp +++ b/src/core/hle/service/cfg/cfg.cpp @@ -315,11 +315,11 @@ void Init() { AddService(new CFG_I_Interface); AddService(new CFG_S_Interface); AddService(new CFG_U_Interface); - + // Open the SystemSaveData archive 0x00010017 FileSys::Path archive_path(cfg_system_savedata_id); auto archive_result = Service::FS::OpenArchive(Service::FS::ArchiveIdCode::SystemSaveData, archive_path); - + // If the archive didn't exist, create the files inside if (archive_result.Code().description == ErrorDescription::FS_NotFormatted) { // Format the archive to create the directories diff --git a/src/core/hle/service/fs/archive.cpp b/src/core/hle/service/fs/archive.cpp index 6d4a9c7c..7cab6802 100644 --- a/src/core/hle/service/fs/archive.cpp +++ b/src/core/hle/service/fs/archive.cpp @@ -254,7 +254,7 @@ ResultVal<ArchiveHandle> OpenArchive(ArchiveIdCode id_code, FileSys::Path& archi CASCADE_RESULT(std::unique_ptr<ArchiveBackend> res, itr->second->Open(archive_path)); - // This should never even happen in the first place with 64-bit handles, + // This should never even happen in the first place with 64-bit handles, while (handle_map.count(next_handle) != 0) { ++next_handle; } @@ -488,7 +488,7 @@ void ArchiveInit() { RegisterArchiveType(std::move(sdmc_factory), ArchiveIdCode::SDMC); else LOG_ERROR(Service_FS, "Can't instantiate SDMC archive with path %s", sdmc_directory.c_str()); - + // Create the SaveData archive auto savedata_factory = Common::make_unique<FileSys::ArchiveFactory_SaveData>(sdmc_directory); RegisterArchiveType(std::move(savedata_factory), ArchiveIdCode::SaveData); @@ -503,7 +503,7 @@ void ArchiveInit() { if (sharedextsavedata_factory->Initialize()) RegisterArchiveType(std::move(sharedextsavedata_factory), ArchiveIdCode::SharedExtSaveData); else - LOG_ERROR(Service_FS, "Can't instantiate SharedExtSaveData archive with path %s", + LOG_ERROR(Service_FS, "Can't instantiate SharedExtSaveData archive with path %s", sharedextsavedata_factory->GetMountPoint().c_str()); // Create the SaveDataCheck archive, basically a small variation of the RomFS archive diff --git a/src/core/hle/service/fs/fs_user.cpp b/src/core/hle/service/fs/fs_user.cpp index b25c8941..32db773b 100644 --- a/src/core/hle/service/fs/fs_user.cpp +++ b/src/core/hle/service/fs/fs_user.cpp @@ -434,7 +434,7 @@ static void IsSdmcWriteable(Service::Interface* self) { } /** - * FS_User::FormatSaveData service function, + * FS_User::FormatSaveData service function, * formats the SaveData specified by the input path. * Inputs: * 0 : 0x084C0242 @@ -520,7 +520,7 @@ static void CreateExtSaveData(Service::Interface* self) { LOG_WARNING(Service_FS, "(STUBBED) savedata_high=%08X savedata_low=%08X cmd_buff[3]=%08X " "cmd_buff[4]=%08X cmd_buff[5]=%08X cmd_buff[6]=%08X cmd_buff[7]=%08X cmd_buff[8]=%08X " "cmd_buff[9]=%08X cmd_buff[10]=%08X cmd_buff[11]=%08X", save_high, save_low, - cmd_buff[3], cmd_buff[4], cmd_buff[5], cmd_buff[6], cmd_buff[7], cmd_buff[8], cmd_buff[9], + cmd_buff[3], cmd_buff[4], cmd_buff[5], cmd_buff[6], cmd_buff[7], cmd_buff[8], cmd_buff[9], cmd_buff[10], cmd_buff[11]); cmd_buff[1] = CreateExtSaveData(media_type, save_high, save_low).raw; @@ -544,7 +544,7 @@ static void DeleteExtSaveData(Service::Interface* self) { u32 save_high = cmd_buff[3]; u32 unknown = cmd_buff[4]; // TODO(Subv): Figure out what this is - LOG_WARNING(Service_FS, "(STUBBED) save_low=%08X save_high=%08X media_type=%08X unknown=%08X", + LOG_WARNING(Service_FS, "(STUBBED) save_low=%08X save_high=%08X media_type=%08X unknown=%08X", save_low, save_high, cmd_buff[1] & 0xFF, unknown); cmd_buff[1] = DeleteExtSaveData(media_type, save_high, save_low).raw; diff --git a/src/core/hle/service/gsp_gpu.cpp b/src/core/hle/service/gsp_gpu.cpp index 4af168bf..4b0b4229 100644 --- a/src/core/hle/service/gsp_gpu.cpp +++ b/src/core/hle/service/gsp_gpu.cpp @@ -167,7 +167,7 @@ static void WriteHWRegsWithMask(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); u32 reg_addr = cmd_buff[1]; u32 size = cmd_buff[2]; - + u32* src_data = (u32*)Memory::GetPointer(cmd_buff[4]); u32* mask_data = (u32*)Memory::GetPointer(cmd_buff[6]); @@ -208,21 +208,21 @@ static void SetBufferSwap(u32 screen_id, const FrameBufferInfo& info) { PAddr phys_address_left = Memory::VirtualToPhysicalAddress(info.address_left); PAddr phys_address_right = Memory::VirtualToPhysicalAddress(info.address_right); if (info.active_fb == 0) { - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_left1)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_left1)), 4, &phys_address_left); - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_right1)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_right1)), 4, &phys_address_right); } else { - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_left2)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_left2)), 4, &phys_address_left); - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_right2)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].address_right2)), 4, &phys_address_right); } - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].stride)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].stride)), 4, &info.stride); - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].color_format)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].color_format)), 4, &info.format); - WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].active_fb)), 4, + WriteHWRegs(base_address + 4 * static_cast<u32>(GPU_REG_INDEX(framebuffer_config[screen_id].active_fb)), 4, &info.shown_fb); } @@ -374,7 +374,7 @@ static void ExecuteCommand(const Command& command, u32 thread_id) { { auto& params = command.set_command_list_last; - WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(command_processor_config.address)), + WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(command_processor_config.address)), Memory::VirtualToPhysicalAddress(params.address) >> 3); WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(command_processor_config.size)), params.size); @@ -470,7 +470,7 @@ static void SetLcdForceBlack(Service::Interface* self) { LCD::Write(HW::VADDR_LCD + 4 * LCD_REG_INDEX(color_fill_top), data.raw); // Top LCD LCD::Write(HW::VADDR_LCD + 4 * LCD_REG_INDEX(color_fill_bottom), data.raw); // Bottom LCD - + cmd_buff[1] = RESULT_SUCCESS.raw; } @@ -516,8 +516,8 @@ static void TriggerCmdReqQueue(Service::Interface* self) { */ static void ImportDisplayCaptureInfo(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); - - // TODO(Subv): We're always returning the framebuffer structures for thread_id = 0, + + // TODO(Subv): We're always returning the framebuffer structures for thread_id = 0, // because we only support a single running application at a time. // This should always return the framebuffer data that is currently displayed on the screen. diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp index feac5381..c7c1bb5a 100644 --- a/src/core/hle/service/hid/hid.cpp +++ b/src/core/hle/service/hid/hid.cpp @@ -106,7 +106,7 @@ void Update() { mem->touch.index_reset_ticks_previous = mem->touch.index_reset_ticks; mem->touch.index_reset_ticks = (s64)CoreTiming::GetTicks(); } - + // Signal both handles when there's an update to Pad or touch event_pad_or_touch_1->Signal(); event_pad_or_touch_2->Signal(); diff --git a/src/core/hle/service/hid/hid_spvr.cpp b/src/core/hle/service/hid/hid_spvr.cpp index 02db12ef..532931ae 100644 --- a/src/core/hle/service/hid/hid_spvr.cpp +++ b/src/core/hle/service/hid/hid_spvr.cpp @@ -25,6 +25,6 @@ const Interface::FunctionInfo FunctionTable[] = { HID_SPVR_Interface::HID_SPVR_Interface() { Register(FunctionTable); } - + } // namespace HID } // namespace Service diff --git a/src/core/hle/service/hid/hid_user.h b/src/core/hle/service/hid/hid_user.h index 0eeec2c2..baf7fed7 100644 --- a/src/core/hle/service/hid/hid_user.h +++ b/src/core/hle/service/hid/hid_user.h @@ -11,7 +11,7 @@ namespace Service { namespace HID { - + /** * HID service interface. */ diff --git a/src/core/hle/service/ptm/ptm.h b/src/core/hle/service/ptm/ptm.h index 493e6a11..b690003c 100644 --- a/src/core/hle/service/ptm/ptm.h +++ b/src/core/hle/service/ptm/ptm.h @@ -20,15 +20,15 @@ enum class ChargeLevels : u32 { CompletelyFull = 5, }; -/** +/** * Represents the gamecoin file structure in the SharedExtData archive * More information in 3dbrew (http://www.3dbrew.org/wiki/Extdata#Shared_Extdata_0xf000000b_gamecoin.dat) */ struct GameCoin { u32 magic; ///< Magic number: 0x4F00 - u16 total_coins; ///< Total Play Coins + u16 total_coins; ///< Total Play Coins u16 total_coins_on_date; ///< Total Play Coins obtained on the date stored below. - u32 step_count; ///< Total step count at the time a new Play Coin was obtained. + u32 step_count; ///< Total step count at the time a new Play Coin was obtained. u32 last_step_count; ///< Step count for the day the last Play Coin was obtained u16 year; u8 month; diff --git a/src/core/hle/service/ptm/ptm_play.cpp b/src/core/hle/service/ptm/ptm_play.cpp index 48e68a3d..7bb99019 100644 --- a/src/core/hle/service/ptm/ptm_play.cpp +++ b/src/core/hle/service/ptm/ptm_play.cpp @@ -18,6 +18,6 @@ const Interface::FunctionInfo FunctionTable[] = { PTM_Play_Interface::PTM_Play_Interface() { Register(FunctionTable); } - + } // namespace PTM } // namespace Service
\ No newline at end of file diff --git a/src/core/hle/service/service.cpp b/src/core/hle/service/service.cpp index 64185c62..dc667500 100644 --- a/src/core/hle/service/service.cpp +++ b/src/core/hle/service/service.cpp @@ -52,7 +52,7 @@ std::unordered_map<std::string, Kernel::SharedPtr<Interface>> g_kernel_named_por std::unordered_map<std::string, Kernel::SharedPtr<Interface>> g_srv_services; /** - * Creates a function string for logging, complete with the name (or header code, depending + * Creates a function string for logging, complete with the name (or header code, depending * on what's passed in) the port name, and all the cmd_buff arguments. */ static std::string MakeFunctionString(const char* name, const char* port_name, const u32* cmd_buff) { diff --git a/src/core/hle/service/soc_u.cpp b/src/core/hle/service/soc_u.cpp index 39b8d65f..1e0f5df9 100644 --- a/src/core/hle/service/soc_u.cpp +++ b/src/core/hle/service/soc_u.cpp @@ -139,7 +139,7 @@ static int TranslateError(int error) { auto found = error_map.find(error); if (found != error_map.end()) return -found->second; - + return error; } @@ -346,7 +346,7 @@ static void Bind(Service::Interface* self) { sockaddr sock_addr = CTRSockAddr::ToPlatform(*ctr_sock_addr); int res = ::bind(socket_handle, &sock_addr, std::max<u32>(sizeof(sock_addr), len)); - + int result = 0; if (res != 0) result = TranslateError(GET_ERRNO); @@ -360,14 +360,14 @@ static void Fcntl(Service::Interface* self) { u32 socket_handle = cmd_buffer[1]; u32 ctr_cmd = cmd_buffer[2]; u32 ctr_arg = cmd_buffer[3]; - + int result = 0; u32 posix_ret = 0; // TODO: Check what hardware returns for F_SETFL (unspecified by POSIX) SCOPE_EXIT({ cmd_buffer[1] = result; cmd_buffer[2] = posix_ret; }); - + if (ctr_cmd == 3) { // F_GETFL #if EMU_PLATFORM == PLATFORM_WINDOWS posix_ret = 0; @@ -404,11 +404,11 @@ static void Fcntl(Service::Interface* self) { posix_ret = -1; return; } - + flags &= ~O_NONBLOCK; if (ctr_arg & 4) // O_NONBLOCK flags |= O_NONBLOCK; - + int ret = ::fcntl(socket_handle, F_SETFL, flags); if (ret == SOCKET_ERROR_VALUE) { result = TranslateError(GET_ERRNO); @@ -439,8 +439,8 @@ static void Listen(Service::Interface* self) { } static void Accept(Service::Interface* self) { - // TODO(Subv): Calling this function on a blocking socket will block the emu thread, - // preventing graceful shutdown when closing the emulator, this can be fixed by always + // TODO(Subv): Calling this function on a blocking socket will block the emu thread, + // preventing graceful shutdown when closing the emulator, this can be fixed by always // performing nonblocking operations and spinlock until the data is available u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; @@ -448,7 +448,7 @@ static void Accept(Service::Interface* self) { sockaddr addr; socklen_t addr_len = sizeof(addr); u32 ret = static_cast<u32>(::accept(socket_handle, &addr, &addr_len)); - + if ((s32)ret != SOCKET_ERROR_VALUE) open_sockets[ret] = { ret, true }; @@ -525,8 +525,8 @@ static void SendTo(Service::Interface* self) { } static void RecvFrom(Service::Interface* self) { - // TODO(Subv): Calling this function on a blocking socket will block the emu thread, - // preventing graceful shutdown when closing the emulator, this can be fixed by always + // TODO(Subv): Calling this function on a blocking socket will block the emu thread, + // preventing graceful shutdown when closing the emulator, this can be fixed by always // performing nonblocking operations and spinlock until the data is available u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; @@ -568,7 +568,7 @@ static void Poll(Service::Interface* self) { pollfd* platform_pollfd = new pollfd[nfds]; for (unsigned current_fds = 0; current_fds < nfds; ++current_fds) platform_pollfd[current_fds] = CTRPollFD::ToPlatform(input_fds[current_fds]); - + int ret = ::poll(platform_pollfd, nfds, timeout); // Now update the output pollfd structure @@ -630,7 +630,7 @@ static void GetPeerName(Service::Interface* self) { socklen_t len = cmd_buffer[2]; CTRSockAddr* ctr_dest_addr = reinterpret_cast<CTRSockAddr*>(Memory::GetPointer(cmd_buffer[0x104 >> 2])); - + sockaddr dest_addr; socklen_t dest_addr_len = sizeof(dest_addr); int ret = ::getpeername(socket_handle, &dest_addr, &dest_addr_len); @@ -651,8 +651,8 @@ static void GetPeerName(Service::Interface* self) { } static void Connect(Service::Interface* self) { - // TODO(Subv): Calling this function on a blocking socket will block the emu thread, - // preventing graceful shutdown when closing the emulator, this can be fixed by always + // TODO(Subv): Calling this function on a blocking socket will block the emu thread, + // preventing graceful shutdown when closing the emulator, this can be fixed by always // performing nonblocking operations and spinlock until the data is available u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; diff --git a/src/core/hle/svc.cpp b/src/core/hle/svc.cpp index ca3ff332..d1555c75 100644 --- a/src/core/hle/svc.cpp +++ b/src/core/hle/svc.cpp @@ -228,7 +228,7 @@ static ResultCode WaitSynchronizationN(s32* out, Handle* handles, s32 handle_cou // Actually wait the current thread on each object if we decided to wait... std::vector<SharedPtr<Kernel::WaitObject>> wait_objects; wait_objects.reserve(handle_count); - + for (int i = 0; i < handle_count; ++i) { auto object = Kernel::g_handle_table.GetWaitObject(handles[i]); object->AddWaitingThread(Kernel::GetCurrentThread()); @@ -475,7 +475,7 @@ static ResultCode GetProcessIdOfThread(u32* process_id, Handle thread_handle) { return ERR_INVALID_HANDLE; const SharedPtr<Kernel::Process> process = thread->owner_process; - + ASSERT_MSG(process != nullptr, "Invalid parent process for thread=0x%08X", thread_handle); *process_id = process->process_id; diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp index 99cfdaaf..789d3dda 100644 --- a/src/core/hw/gpu.cpp +++ b/src/core/hw/gpu.cpp @@ -140,7 +140,7 @@ inline void Write(u32 addr, const T data) { // Raw copies do not perform color conversion nor tiled->linear / linear->tiled conversions // TODO(Subv): Verify if raw copies perform scaling memcpy(dst_pointer, src_pointer, output_size); - + LOG_TRACE(HW_GPU, "DisplayTriggerTransfer: 0x%08x bytes from 0x%08x(%ux%u)-> 0x%08x(%ux%u), output format: %x, flags 0x%08X, Raw copy", output_size, config.GetPhysicalInputAddress(), config.input_width.Value(), config.input_height.Value(), @@ -159,14 +159,14 @@ inline void Write(u32 addr, const T data) { for (u32 x = 0; x < output_width; ++x) { Math::Vec4<u8> src_color = { 0, 0, 0, 0 }; - // Calculate the [x,y] position of the input image + // Calculate the [x,y] position of the input image // based on the current output position and the scale u32 input_x = x * horizontal_scale; u32 input_y = y * vertical_scale; if (config.flip_vertically) { - // Flip the y value of the output data, - // we do this after calculating the [x,y] position of the input image + // Flip the y value of the output data, + // we do this after calculating the [x,y] position of the input image // to account for the scaling options. y = output_height - y - 1; } @@ -302,7 +302,7 @@ static void VBlankCallback(u64 userdata, int cycles_late) { // - If frameskip == 0 (disabled), always swap buffers // - If frameskip == 1, swap buffers every other frame (starting from the first frame) // - If frameskip > 1, swap buffers every frameskip^n frames (starting from the second frame) - if ((((Settings::values.frame_skip != 1) ^ last_skip_frame) && last_skip_frame != g_skip_frame) || + if ((((Settings::values.frame_skip != 1) ^ last_skip_frame) && last_skip_frame != g_skip_frame) || Settings::values.frame_skip == 0) { VideoCore::g_renderer->SwapBuffers(); } diff --git a/src/core/hw/lcd.cpp b/src/core/hw/lcd.cpp index 09134c95..963c8d98 100644 --- a/src/core/hw/lcd.cpp +++ b/src/core/hw/lcd.cpp @@ -66,5 +66,5 @@ void Init() { void Shutdown() { LOG_DEBUG(HW_LCD, "shutdown OK"); } - + } // namespace diff --git a/src/core/hw/lcd.h b/src/core/hw/lcd.h index fb14c3b2..8631eb20 100644 --- a/src/core/hw/lcd.h +++ b/src/core/hw/lcd.h @@ -85,5 +85,5 @@ void Init(); /// Shutdown hardware void Shutdown(); - + } // namespace diff --git a/src/core/loader/3dsx.cpp b/src/core/loader/3dsx.cpp index ad5e929c..14aeebeb 100644 --- a/src/core/loader/3dsx.cpp +++ b/src/core/loader/3dsx.cpp @@ -234,7 +234,7 @@ ResultStatus AppLoader_THREEDSX::Load() { Kernel::g_current_process = Kernel::Process::Create(filename, 0); Kernel::g_current_process->svc_access_mask.set(); Kernel::g_current_process->address_mappings = default_address_mappings; - + // Attach the default resource limit (APPLICATION) to the process Kernel::g_current_process->resource_limit = Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION); diff --git a/src/core/mem_map.cpp b/src/core/mem_map.cpp index 5ecec956..bf814b94 100644 --- a/src/core/mem_map.cpp +++ b/src/core/mem_map.cpp @@ -8,6 +8,10 @@ #include "common/logging/log.h" #include "core/hle/config_mem.h" +#include "core/hle/kernel/kernel.h" +#include "core/hle/kernel/shared_memory.h" +#include "core/hle/kernel/vm_manager.h" +#include "core/hle/result.h" #include "core/hle/shared_page.h" #include "core/mem_map.h" #include "core/memory.h" @@ -17,31 +21,23 @@ namespace Memory { -u8* g_exefs_code; ///< ExeFS:/.code is loaded here -u8* g_heap; ///< Application heap (main memory) -u8* g_shared_mem; ///< Shared memory -u8* g_heap_linear; ///< Linear heap -u8* g_vram; ///< Video memory (VRAM) pointer -u8* g_dsp_mem; ///< DSP memory -u8* g_tls_mem; ///< TLS memory - namespace { struct MemoryArea { - u8** ptr; u32 base; u32 size; + const char* name; }; // We don't declare the IO regions in here since its handled by other means. static MemoryArea memory_areas[] = { - {&g_exefs_code, PROCESS_IMAGE_VADDR, PROCESS_IMAGE_MAX_SIZE}, - {&g_heap, HEAP_VADDR, HEAP_SIZE }, - {&g_shared_mem, SHARED_MEMORY_VADDR, SHARED_MEMORY_SIZE }, - {&g_heap_linear, LINEAR_HEAP_VADDR, LINEAR_HEAP_SIZE }, - {&g_vram, VRAM_VADDR, VRAM_SIZE }, - {&g_dsp_mem, DSP_RAM_VADDR, DSP_RAM_SIZE }, - {&g_tls_mem, TLS_AREA_VADDR, TLS_AREA_SIZE }, + {PROCESS_IMAGE_VADDR, PROCESS_IMAGE_MAX_SIZE, "Process Image"}, // ExeFS:/.code is loaded here + {HEAP_VADDR, HEAP_SIZE, "Heap"}, // Application heap (main memory) + {SHARED_MEMORY_VADDR, SHARED_MEMORY_SIZE, "Shared Memory"}, // Shared memory + {LINEAR_HEAP_VADDR, LINEAR_HEAP_SIZE, "Linear Heap"}, // Linear heap (main memory) + {VRAM_VADDR, VRAM_SIZE, "VRAM"}, // Video memory (VRAM) + {DSP_RAM_VADDR, DSP_RAM_SIZE, "DSP RAM"}, // DSP memory + {TLS_AREA_VADDR, TLS_AREA_SIZE, "TLS Area"}, // TLS memory }; /// Represents a block of memory mapped by ControlMemory/MapMemoryBlock @@ -135,27 +131,34 @@ VAddr PhysicalToVirtualAddress(const PAddr addr) { return addr | 0x80000000; } +// TODO(yuriks): Move this into Process +static Kernel::VMManager address_space; + void Init() { + using namespace Kernel; + InitMemoryMap(); for (MemoryArea& area : memory_areas) { - *area.ptr = new u8[area.size]; - MapMemoryRegion(area.base, area.size, *area.ptr); + auto block = std::make_shared<std::vector<u8>>(area.size); + address_space.MapMemoryBlock(area.base, std::move(block), 0, area.size, MemoryState::Private).Unwrap(); } - MapMemoryRegion(CONFIG_MEMORY_VADDR, CONFIG_MEMORY_SIZE, (u8*)&ConfigMem::config_mem); - MapMemoryRegion(SHARED_PAGE_VADDR, SHARED_PAGE_SIZE, (u8*)&SharedPage::shared_page); - LOG_DEBUG(HW_Memory, "initialized OK, RAM at %p", g_heap); + auto cfg_mem_vma = address_space.MapBackingMemory(CONFIG_MEMORY_VADDR, + (u8*)&ConfigMem::config_mem, CONFIG_MEMORY_SIZE, MemoryState::Shared).MoveFrom(); + address_space.Reprotect(cfg_mem_vma, VMAPermission::Read); + + auto shared_page_vma = address_space.MapBackingMemory(SHARED_PAGE_VADDR, + (u8*)&SharedPage::shared_page, SHARED_PAGE_SIZE, MemoryState::Shared).MoveFrom(); + address_space.Reprotect(shared_page_vma, VMAPermission::Read); + + LOG_DEBUG(HW_Memory, "initialized OK"); } void Shutdown() { heap_map.clear(); heap_linear_map.clear(); - - for (MemoryArea& area : memory_areas) { - delete[] *area.ptr; - *area.ptr = nullptr; - } + address_space.Reset(); LOG_DEBUG(HW_Memory, "shutdown OK"); } diff --git a/src/core/mem_map.h b/src/core/mem_map.h index 945815cd..ba50914a 100644 --- a/src/core/mem_map.h +++ b/src/core/mem_map.h @@ -8,14 +8,6 @@ namespace Memory { -extern u8* g_exefs_code; ///< ExeFS:/.code is loaded here -extern u8* g_heap; ///< Application heap (main memory) -extern u8* g_shared_mem; ///< Shared memory -extern u8* g_heap_linear; ///< Linear heap (main memory) -extern u8* g_vram; ///< Video memory (VRAM) -extern u8* g_dsp_mem; ///< DSP memory -extern u8* g_tls_mem; ///< TLS memory - void Init(); void Shutdown(); diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 5d8069ac..28844a91 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -14,12 +14,10 @@ #include "core/hw/hw.h" #include "core/mem_map.h" #include "core/memory.h" +#include "core/memory_setup.h" namespace Memory { -const u32 PAGE_MASK = PAGE_SIZE - 1; -const int PAGE_BITS = 12; - enum class PageType { /// Page is unmapped and should cause an access error. Unmapped, @@ -64,7 +62,7 @@ static void MapPages(u32 base, u32 size, u8* memory, PageType type) { while (base != end) { ASSERT_MSG(base < PageTable::NUM_ENTRIES, "out of range mapping at %08X", base); - if (current_page_table->attributes[base] != PageType::Unmapped) { + if (current_page_table->attributes[base] != PageType::Unmapped && type != PageType::Unmapped) { LOG_ERROR(HW_Memory, "overlapping memory ranges at %08X", base * PAGE_SIZE); } current_page_table->attributes[base] = type; @@ -92,6 +90,12 @@ void MapIoRegion(VAddr base, u32 size) { MapPages(base / PAGE_SIZE, size / PAGE_SIZE, nullptr, PageType::Special); } +void UnmapRegion(VAddr base, u32 size) { + ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: %08X", size); + ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: %08X", base); + MapPages(base / PAGE_SIZE, size / PAGE_SIZE, nullptr, PageType::Unmapped); +} + template <typename T> T Read(const VAddr vaddr) { const u8* page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS]; diff --git a/src/core/memory.h b/src/core/memory.h index 2d225801..0b8ff9ec 100644 --- a/src/core/memory.h +++ b/src/core/memory.h @@ -97,7 +97,7 @@ enum : VAddr { SHARED_PAGE_VADDR_END = SHARED_PAGE_VADDR + SHARED_PAGE_SIZE, // TODO(yuriks): The size of this area is dynamic, the kernel grows - // it as more and more threads are created. For now we'll just use a + // it as more and more threads are created. For now we'll just use a // hardcoded value. /// Area where TLS (Thread-Local Storage) buffers are allocated. TLS_AREA_VADDR = 0x1FF82000, diff --git a/src/core/memory_setup.h b/src/core/memory_setup.h index 46263495..361bfc81 100644 --- a/src/core/memory_setup.h +++ b/src/core/memory_setup.h @@ -6,8 +6,13 @@ #include "common/common_types.h" +#include "core/memory.h" + namespace Memory { +const u32 PAGE_MASK = PAGE_SIZE - 1; +const int PAGE_BITS = 12; + void InitMemoryMap(); /** @@ -26,4 +31,6 @@ void MapMemoryRegion(VAddr base, u32 size, u8* target); */ void MapIoRegion(VAddr base, u32 size); +void UnmapRegion(VAddr base, u32 size); + } diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index 100d8c7c..b46fadd9 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -56,7 +56,17 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { // Trigger IRQ case PICA_REG_INDEX(trigger_irq): GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::P3D); - return; + break; + + case PICA_REG_INDEX_WORKAROUND(command_buffer.trigger[0], 0x23c): + case PICA_REG_INDEX_WORKAROUND(command_buffer.trigger[1], 0x23d): + { + unsigned index = id - PICA_REG_INDEX(command_buffer.trigger[0]); + u32* head_ptr = (u32*)Memory::GetPhysicalPointer(regs.command_buffer.GetPhysicalAddress(index)); + g_state.cmd_list.head_ptr = g_state.cmd_list.current_ptr = head_ptr; + g_state.cmd_list.length = regs.command_buffer.GetSize(index) / sizeof(u32); + break; + } // It seems like these trigger vertex rendering case PICA_REG_INDEX(trigger_draw): @@ -136,7 +146,7 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { input.attr[i][0].ToFloat32(), input.attr[i][1].ToFloat32(), input.attr[i][2].ToFloat32(), input.attr[i][3].ToFloat32()); } - + // Load per-vertex data from the loader arrays for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) { const u8* srcdata = Memory::GetPhysicalPointer(vertex_attribute_sources[i] + vertex_attribute_strides[i] * vertex + comp * vertex_attribute_element_size[i]); @@ -193,7 +203,7 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { const Pica::VertexShader::OutputVertex& v2) { VideoCore::g_renderer->hw_rasterizer->AddTriangle(v0, v1, v2); }; - + primitive_assembler.SubmitVertex(output, AddHWTriangle); } else { // Send to triangle clipper @@ -282,7 +292,7 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { } break; } - + // Load default vertex input attributes case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.set_value[0], 0x233): case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.set_value[1], 0x234): @@ -306,7 +316,7 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { } Math::Vec4<float24>& attribute = g_state.vs.default_attributes[setup.index]; - + // NOTE: The destination component order indeed is "backwards" attribute.w = float24::FromRawFloat24(default_attr_write_buffer[0] >> 8); attribute.z = float24::FromRawFloat24(((default_attr_write_buffer[0] & 0xFF) << 16) | ((default_attr_write_buffer[1] >> 16) & 0xFFFF)); @@ -363,38 +373,34 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::CommandProcessed, reinterpret_cast<void*>(&id)); } -static std::ptrdiff_t ExecuteCommandBlock(const u32* first_command_word) { - const CommandHeader& header = *(const CommandHeader*)(&first_command_word[1]); - - u32* read_pointer = (u32*)first_command_word; - - const u32 write_mask = ((header.parameter_mask & 0x1) ? (0xFFu << 0) : 0u) | - ((header.parameter_mask & 0x2) ? (0xFFu << 8) : 0u) | - ((header.parameter_mask & 0x4) ? (0xFFu << 16) : 0u) | - ((header.parameter_mask & 0x8) ? (0xFFu << 24) : 0u); - - WritePicaReg(header.cmd_id, *read_pointer, write_mask); - read_pointer += 2; - - for (unsigned int i = 1; i < 1+header.extra_data_length; ++i) { - u32 cmd = header.cmd_id + ((header.group_commands) ? i : 0); - WritePicaReg(cmd, *read_pointer, write_mask); - ++read_pointer; - } - - // align read pointer to 8 bytes - if ((first_command_word - read_pointer) % 2) - ++read_pointer; - - return read_pointer - first_command_word; -} - void ProcessCommandList(const u32* list, u32 size) { - u32* read_pointer = (u32*)list; - u32 list_length = size / sizeof(u32); - - while (read_pointer < list + list_length) { - read_pointer += ExecuteCommandBlock(read_pointer); + g_state.cmd_list.head_ptr = g_state.cmd_list.current_ptr = list; + g_state.cmd_list.length = size / sizeof(u32); + + while (g_state.cmd_list.current_ptr < g_state.cmd_list.head_ptr + g_state.cmd_list.length) { + // Expand a 4-bit mask to 4-byte mask, e.g. 0b0101 -> 0x00FF00FF + static const u32 expand_bits_to_bytes[] = { + 0x00000000, 0x000000ff, 0x0000ff00, 0x0000ffff, + 0x00ff0000, 0x00ff00ff, 0x00ffff00, 0x00ffffff, + 0xff000000, 0xff0000ff, 0xff00ff00, 0xff00ffff, + 0xffff0000, 0xffff00ff, 0xffffff00, 0xffffffff + }; + + // Align read pointer to 8 bytes + if ((g_state.cmd_list.head_ptr - g_state.cmd_list.current_ptr) % 2 != 0) + ++g_state.cmd_list.current_ptr; + + u32 value = *g_state.cmd_list.current_ptr++; + const CommandHeader header = { *g_state.cmd_list.current_ptr++ }; + const u32 write_mask = expand_bits_to_bytes[header.parameter_mask]; + u32 cmd = header.cmd_id; + + WritePicaReg(cmd, value, write_mask); + + for (unsigned i = 0; i < header.extra_data_length; ++i) { + u32 cmd = header.cmd_id + (header.group_commands ? i + 1 : 0); + WritePicaReg(cmd, *g_state.cmd_list.current_ptr++, write_mask); + } } } diff --git a/src/video_core/debug_utils/debug_utils.cpp b/src/video_core/debug_utils/debug_utils.cpp index 7987b922..b92cd1a7 100644 --- a/src/video_core/debug_utils/debug_utils.cpp +++ b/src/video_core/debug_utils/debug_utils.cpp @@ -319,7 +319,7 @@ const Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const Texture // TODO(neobrain): Fix code design to unify vertical block offsets! source += coarse_y * info.stride; } - + // TODO: Assert that width/height are multiples of block dimensions switch (info.format) { diff --git a/src/video_core/math.h b/src/video_core/math.h index f9a82265..4928c9bf 100644 --- a/src/video_core/math.h +++ b/src/video_core/math.h @@ -461,7 +461,7 @@ public: // e.g. Vec2 uv() { return Vec2(x,y); } // _DEFINE_SWIZZLER2 defines a single such function - // DEFINE_SWIZZLER2_COMP1 defines one-component functions for all component names (x<->r) + // DEFINE_SWIZZLER2_COMP1 defines one-component functions for all component names (x<->r) // DEFINE_SWIZZLER2_COMP2 defines two component functions for all component names (x<->r) and permutations (xy<->yx) #define _DEFINE_SWIZZLER2(a, b, name) const Vec2<T> name() const { return Vec2<T>(a, b); } #define DEFINE_SWIZZLER2_COMP1(a, a2) \ diff --git a/src/video_core/pica.h b/src/video_core/pica.h index b67dce1a..8ad47a92 100644 --- a/src/video_core/pica.h +++ b/src/video_core/pica.h @@ -16,7 +16,7 @@ #include "common/common_types.h" #include "common/logging/log.h" -#include "math.h" +#include "math.h" namespace Pica { @@ -162,6 +162,25 @@ struct Regs { ETC1A4 = 13, // compressed }; + enum class LogicOp : u32 { + Clear = 0, + And = 1, + AndReverse = 2, + Copy = 3, + Set = 4, + CopyInverted = 5, + NoOp = 6, + Invert = 7, + Nand = 8, + Or = 9, + Nor = 10, + Xor = 11, + Equiv = 12, + AndInverted = 13, + OrReverse = 14, + OrInverted = 15, + }; + static unsigned NibblesPerPixel(TextureFormat format) { switch (format) { case TextureFormat::RGBA8: @@ -221,6 +240,7 @@ struct Regs { enum class Source : u32 { PrimaryColor = 0x0, PrimaryFragmentColor = 0x1, + SecondaryFragmentColor = 0x2, Texture0 = 0x3, Texture1 = 0x4, @@ -337,7 +357,7 @@ struct Regs { return (stage_index < 4) && (update_mask_a & (1 << stage_index)); } } tev_combiner_buffer_input; - + INSERT_PADDING_WORDS(0xf); TevStageConfig tev_stage4; INSERT_PADDING_WORDS(0x3); @@ -413,12 +433,8 @@ struct Regs { } alpha_blending; union { - enum Op { - Set = 4, - }; - - BitField<0, 4, Op> op; - } logic_op; + BitField<0, 4, LogicOp> logic_op; + }; union { BitField< 0, 8, u32> r; @@ -703,12 +719,38 @@ struct Regs { struct { // Index of the current default attribute u32 index; - + // Writing to these registers sets the "current" default attribute. u32 set_value[3]; } vs_default_attributes_setup; - - INSERT_PADDING_WORDS(0x28); + + INSERT_PADDING_WORDS(0x2); + + struct { + // There are two channels that can be used to configure the next command buffer, which + // can be then executed by writing to the "trigger" registers. There are two reasons why a + // game might use this feature: + // 1) With this, an arbitrary number of additional command buffers may be executed in + // sequence without requiring any intervention of the CPU after the initial one is + // kicked off. + // 2) Games can configure these registers to provide a command list subroutine mechanism. + + BitField< 0, 20, u32> size[2]; ///< Size (in bytes / 8) of each channel's command buffer + BitField< 0, 28, u32> addr[2]; ///< Physical address / 8 of each channel's command buffer + u32 trigger[2]; ///< Triggers execution of the channel's command buffer when written to + + unsigned GetSize(unsigned index) const { + ASSERT(index < 2); + return 8 * size[index]; + } + + PAddr GetPhysicalAddress(unsigned index) const { + ASSERT(index < 2); + return (PAddr)(8 * addr[index]); + } + } command_buffer; + + INSERT_PADDING_WORDS(0x20); enum class TriangleTopology : u32 { List = 0, @@ -861,6 +903,7 @@ struct Regs { ADD_FIELD(trigger_draw); ADD_FIELD(trigger_draw_indexed); ADD_FIELD(vs_default_attributes_setup); + ADD_FIELD(command_buffer); ADD_FIELD(triangle_topology); ADD_FIELD(vs_bool_uniforms); ADD_FIELD(vs_int_uniforms); @@ -938,6 +981,7 @@ ASSERT_REG_POSITION(num_vertices, 0x228); ASSERT_REG_POSITION(trigger_draw, 0x22e); ASSERT_REG_POSITION(trigger_draw_indexed, 0x22f); ASSERT_REG_POSITION(vs_default_attributes_setup, 0x232); +ASSERT_REG_POSITION(command_buffer, 0x238); ASSERT_REG_POSITION(triangle_topology, 0x25e); ASSERT_REG_POSITION(vs_bool_uniforms, 0x2b0); ASSERT_REG_POSITION(vs_int_uniforms, 0x2b1); @@ -1053,21 +1097,12 @@ private: float value; }; -union CommandHeader { - CommandHeader(u32 h) : hex(h) {} - - u32 hex; - - BitField< 0, 16, u32> cmd_id; - BitField<16, 4, u32> parameter_mask; - BitField<20, 11, u32> extra_data_length; - BitField<31, 1, u32> group_commands; -}; - /// Struct used to describe current Pica state struct State { + /// Pica registers Regs regs; + /// Vertex shader memory struct { struct { Math::Vec4<float24> f[96]; @@ -1080,6 +1115,13 @@ struct State { std::array<u32, 1024> program_code; std::array<u32, 1024> swizzle_data; } vs; + + /// Current Pica command list + struct { + const u32* head_ptr; + const u32* current_ptr; + u32 length; + } cmd_list; }; /// Initialize Pica state diff --git a/src/video_core/rasterizer.cpp b/src/video_core/rasterizer.cpp index 767ff420..113b573f 100644 --- a/src/video_core/rasterizer.cpp +++ b/src/video_core/rasterizer.cpp @@ -104,7 +104,7 @@ static u32 GetDepth(int x, int y) { u8* depth_buffer = Memory::GetPhysicalPointer(addr); y = framebuffer.height - y; - + const u32 coarse_y = y & ~7; u32 bytes_per_pixel = Regs::BytesPerDepthPixel(framebuffer.depth_format); u32 stride = framebuffer.width * bytes_per_pixel; @@ -402,11 +402,16 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0, auto GetSource = [&](Source source) -> Math::Vec4<u8> { switch (source) { - // TODO: What's the difference between these two? case Source::PrimaryColor: + + // HACK: Until we implement fragment lighting, use primary_color case Source::PrimaryFragmentColor: return primary_color; + // HACK: Until we implement fragment lighting, use zero + case Source::SecondaryFragmentColor: + return {0, 0, 0, 0}; + case Source::Texture0: return texture_color[0]; @@ -570,6 +575,13 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0, case Operation::Add: return std::min(255, input[0] + input[1]); + case Operation::AddSigned: + { + // TODO(bunnei): Verify that the color conversion from (float) 0.5f to (byte) 128 is correct + auto result = static_cast<int>(input[0]) + static_cast<int>(input[1]) - 128; + return static_cast<u8>(MathUtil::Clamp<int>(result, 0, 255)); + } + case Operation::Lerp: return (input[0] * input[2] + input[1] * (255 - input[2])) / 255; @@ -808,10 +820,9 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0, } }; - using BlendEquation = Regs::BlendEquation; static auto EvaluateBlendEquation = [](const Math::Vec4<u8>& src, const Math::Vec4<u8>& srcfactor, const Math::Vec4<u8>& dest, const Math::Vec4<u8>& destfactor, - BlendEquation equation) { + Regs::BlendEquation equation) { Math::Vec4<int> result; auto src_result = (src * srcfactor).Cast<int>(); @@ -866,8 +877,63 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0, blend_output = EvaluateBlendEquation(combiner_output, srcfactor, dest, dstfactor, params.blend_equation_rgb); blend_output.a() = EvaluateBlendEquation(combiner_output, srcfactor, dest, dstfactor, params.blend_equation_a).a(); } else { - LOG_CRITICAL(HW_GPU, "logic op: %x", output_merger.logic_op); - UNIMPLEMENTED(); + static auto LogicOp = [](u8 src, u8 dest, Regs::LogicOp op) -> u8 { + switch (op) { + case Regs::LogicOp::Clear: + return 0; + + case Regs::LogicOp::And: + return src & dest; + + case Regs::LogicOp::AndReverse: + return src & ~dest; + + case Regs::LogicOp::Copy: + return src; + + case Regs::LogicOp::Set: + return 255; + + case Regs::LogicOp::CopyInverted: + return ~src; + + case Regs::LogicOp::NoOp: + return dest; + + case Regs::LogicOp::Invert: + return ~dest; + + case Regs::LogicOp::Nand: + return ~(src & dest); + + case Regs::LogicOp::Or: + return src | dest; + + case Regs::LogicOp::Nor: + return ~(src | dest); + + case Regs::LogicOp::Xor: + return src ^ dest; + + case Regs::LogicOp::Equiv: + return ~(src ^ dest); + + case Regs::LogicOp::AndInverted: + return ~src & dest; + + case Regs::LogicOp::OrReverse: + return src | ~dest; + + case Regs::LogicOp::OrInverted: + return ~src | dest; + } + }; + + blend_output = Math::MakeVec( + LogicOp(combiner_output.r(), dest.r(), output_merger.logic_op), + LogicOp(combiner_output.g(), dest.g(), output_merger.logic_op), + LogicOp(combiner_output.b(), dest.b(), output_merger.logic_op), + LogicOp(combiner_output.a(), dest.a(), output_merger.logic_op)); } const Math::Vec4<u8> result = { diff --git a/src/video_core/renderer_opengl/generated/gl_3_2_core.c b/src/video_core/renderer_opengl/generated/gl_3_2_core.c index ef29972d..95fd29c0 100644 --- a/src/video_core/renderer_opengl/generated/gl_3_2_core.c +++ b/src/video_core/renderer_opengl/generated/gl_3_2_core.c @@ -62,9 +62,9 @@ static int TestPointer(const PROC pTest) ptrdiff_t iTest; if(!pTest) return 0; iTest = (ptrdiff_t)pTest; - + if(iTest == 1 || iTest == 2 || iTest == 3 || iTest == -1) return 0; - + return 1; } @@ -79,7 +79,7 @@ static PROC WinGetProcAddress(const char *name) glMod = GetModuleHandleA("OpenGL32.dll"); return (PROC)GetProcAddress(glMod, (LPCSTR)name); } - + #define IntGetProcAddress(name) WinGetProcAddress(name) #else #if defined(__APPLE__) @@ -1083,7 +1083,7 @@ static ogl_StrToExtMap *FindExtEntry(const char *extensionName) if(strcmp(extensionName, currLoc->extensionName) == 0) return currLoc; } - + return NULL; } @@ -1135,15 +1135,15 @@ int ogl_LoadFunctions() { int numFailed = 0; ClearExtensionVars(); - + _ptrc_glGetIntegerv = (void (CODEGEN_FUNCPTR *)(GLenum, GLint *))IntGetProcAddress("glGetIntegerv"); if(!_ptrc_glGetIntegerv) return ogl_LOAD_FAILED; _ptrc_glGetStringi = (const GLubyte * (CODEGEN_FUNCPTR *)(GLenum, GLuint))IntGetProcAddress("glGetStringi"); if(!_ptrc_glGetStringi) return ogl_LOAD_FAILED; - + ProcExtsFromExtList(); numFailed = Load_Version_3_2(); - + if(numFailed == 0) return ogl_LOAD_SUCCEEDED; else @@ -1177,7 +1177,7 @@ int ogl_IsVersionGEQ(int majorVersion, int minorVersion) { if(g_major_version == 0) GetGLVersion(); - + if(majorVersion > g_major_version) return 1; if(majorVersion < g_major_version) return 0; if(minorVersion >= g_minor_version) return 1; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 4b7d099a..b51f8efd 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -135,6 +135,7 @@ void RasterizerOpenGL::Reset() { SyncBlendFuncs(); SyncBlendColor(); SyncAlphaTest(); + SyncLogicOp(); SyncStencilTest(); SyncDepthTest(); @@ -249,6 +250,11 @@ void RasterizerOpenGL::NotifyPicaRegisterChanged(u32 id) { SyncDepthTest(); break; + // Logic op + case PICA_REG_INDEX(output_merger.logic_op): + SyncLogicOp(); + break; + // TEV stage 0 case PICA_REG_INDEX(tev_stage0.color_source1): SyncTevSources(0, regs.tev_stage0); @@ -350,7 +356,7 @@ void RasterizerOpenGL::NotifyPicaRegisterChanged(u32 id) { case PICA_REG_INDEX(tev_stage5.color_scale): SyncTevMultipliers(5, regs.tev_stage5); break; - + // TEV combiner buffer color case PICA_REG_INDEX(tev_combiner_buffer_color): SyncCombinerColor(); @@ -633,6 +639,10 @@ void RasterizerOpenGL::SyncAlphaTest() { glUniform1f(uniform_alphatest_ref, regs.output_merger.alpha_test.ref / 255.0f); } +void RasterizerOpenGL::SyncLogicOp() { + state.logic_op = PicaToGL::LogicOp(Pica::g_state.regs.output_merger.logic_op); +} + void RasterizerOpenGL::SyncStencilTest() { // TODO: Implement stencil test, mask, and op } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 9896f8d0..d7d422b1 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -125,6 +125,9 @@ private: /// Syncs the alpha test states to match the PICA register void SyncAlphaTest(); + /// Syncs the logic op states to match the PICA register + void SyncLogicOp(); + /// Syncs the stencil test states to match the PICA register void SyncStencilTest(); diff --git a/src/video_core/renderer_opengl/gl_shaders.h b/src/video_core/renderer_opengl/gl_shaders.h index 8f094123..a8cb2f59 100644 --- a/src/video_core/renderer_opengl/gl_shaders.h +++ b/src/video_core/renderer_opengl/gl_shaders.h @@ -69,15 +69,16 @@ const char g_fragment_shader_hw[] = R"( #define NUM_VTX_ATTR 7 #define NUM_TEV_STAGES 6 -#define SOURCE_PRIMARYCOLOR 0x0 -#define SOURCE_PRIMARYFRAGMENTCOLOR 0x1 -#define SOURCE_TEXTURE0 0x3 -#define SOURCE_TEXTURE1 0x4 -#define SOURCE_TEXTURE2 0x5 -#define SOURCE_TEXTURE3 0x6 -#define SOURCE_PREVIOUSBUFFER 0xd -#define SOURCE_CONSTANT 0xe -#define SOURCE_PREVIOUS 0xf +#define SOURCE_PRIMARYCOLOR 0x0 +#define SOURCE_PRIMARYFRAGMENTCOLOR 0x1 +#define SOURCE_SECONDARYFRAGMENTCOLOR 0x2 +#define SOURCE_TEXTURE0 0x3 +#define SOURCE_TEXTURE1 0x4 +#define SOURCE_TEXTURE2 0x5 +#define SOURCE_TEXTURE3 0x6 +#define SOURCE_PREVIOUSBUFFER 0xd +#define SOURCE_CONSTANT 0xe +#define SOURCE_PREVIOUS 0xf #define COLORMODIFIER_SOURCECOLOR 0x0 #define COLORMODIFIER_ONEMINUSSOURCECOLOR 0x1 @@ -151,8 +152,11 @@ vec4 GetSource(int source) { if (source == SOURCE_PRIMARYCOLOR) { return o[2]; } else if (source == SOURCE_PRIMARYFRAGMENTCOLOR) { - // HACK: Uses color value, but should really use fragment lighting output + // HACK: Until we implement fragment lighting, use primary_color return o[2]; + } else if (source == SOURCE_SECONDARYFRAGMENTCOLOR) { + // HACK: Until we implement fragment lighting, use zero + return vec4(0.0, 0.0, 0.0, 0.0); } else if (source == SOURCE_TEXTURE0) { return texture(tex[0], o[3].xy); } else if (source == SOURCE_TEXTURE1) { diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp index 1afa58c9..9c5f38f9 100644 --- a/src/video_core/renderer_opengl/gl_state.cpp +++ b/src/video_core/renderer_opengl/gl_state.cpp @@ -32,6 +32,8 @@ OpenGLState::OpenGLState() { blend.color.blue = 0.0f; blend.color.alpha = 0.0f; + logic_op = GL_COPY; + for (auto& texture_unit : texture_units) { texture_unit.enabled_2d = false; texture_unit.texture_2d = 0; @@ -82,7 +84,7 @@ void OpenGLState::Apply() { } else { glDisable(GL_STENCIL_TEST); } - } + } if (stencil.test_func != cur_state.stencil.test_func || stencil.test_ref != cur_state.stencil.test_ref || @@ -99,8 +101,13 @@ void OpenGLState::Apply() { if (blend.enabled != cur_state.blend.enabled) { if (blend.enabled) { glEnable(GL_BLEND); + + cur_state.logic_op = GL_COPY; + glLogicOp(cur_state.logic_op); + glDisable(GL_COLOR_LOGIC_OP); } else { glDisable(GL_BLEND); + glEnable(GL_COLOR_LOGIC_OP); } } @@ -118,6 +125,10 @@ void OpenGLState::Apply() { glBlendFuncSeparate(blend.src_rgb_func, blend.dst_rgb_func, blend.src_a_func, blend.dst_a_func); } + if (logic_op != cur_state.logic_op) { + glLogicOp(logic_op); + } + // Textures for (unsigned texture_index = 0; texture_index < ARRAY_SIZE(texture_units); ++texture_index) { if (texture_units[texture_index].enabled_2d != cur_state.texture_units[texture_index].enabled_2d) { diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h index 281b7cad..6b97721d 100644 --- a/src/video_core/renderer_opengl/gl_state.h +++ b/src/video_core/renderer_opengl/gl_state.h @@ -42,6 +42,8 @@ public: } color; // GL_BLEND_COLOR } blend; + GLenum logic_op; // GL_LOGIC_OP_MODE + // 3 texture units - one for each that is used in PICA fragment shader emulation struct { bool enabled_2d; // GL_TEXTURE_2D @@ -61,7 +63,7 @@ public: static const OpenGLState& GetCurState() { return cur_state; } - + /// Apply this state as the current OpenGL state void Apply(); diff --git a/src/video_core/renderer_opengl/pica_to_gl.h b/src/video_core/renderer_opengl/pica_to_gl.h index f8763e71..e566f9f7 100644 --- a/src/video_core/renderer_opengl/pica_to_gl.h +++ b/src/video_core/renderer_opengl/pica_to_gl.h @@ -71,6 +71,37 @@ inline GLenum BlendFunc(Pica::Regs::BlendFactor factor) { return blend_func_table[(unsigned)factor]; } +inline GLenum LogicOp(Pica::Regs::LogicOp op) { + static const GLenum logic_op_table[] = { + GL_CLEAR, // Clear + GL_AND, // And + GL_AND_REVERSE, // AndReverse + GL_COPY, // Copy + GL_SET, // Set + GL_COPY_INVERTED, // CopyInverted + GL_NOOP, // NoOp + GL_INVERT, // Invert + GL_NAND, // Nand + GL_OR, // Or + GL_NOR, // Nor + GL_XOR, // Xor + GL_EQUIV, // Equiv + GL_AND_INVERTED, // AndInverted + GL_OR_REVERSE, // OrReverse + GL_OR_INVERTED, // OrInverted + }; + + // Range check table for input + if ((unsigned)op >= ARRAY_SIZE(logic_op_table)) { + LOG_CRITICAL(Render_OpenGL, "Unknown logic op %d", op); + UNREACHABLE(); + + return GL_COPY; + } + + return logic_op_table[(unsigned)op]; +} + inline GLenum CompareFunc(Pica::Regs::CompareFunc func) { static const GLenum compare_func_table[] = { GL_NEVER, // CompareFunc::Never diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 16cf92e2..382aeaa0 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -157,7 +157,7 @@ void RendererOpenGL::LoadFBToActiveGLTexture(const GPU::Regs::FramebufferConfig& state.texture_units[0].enabled_2d = true; state.texture_units[0].texture_2d = texture.handle; state.Apply(); - + glActiveTexture(GL_TEXTURE0); glPixelStorei(GL_UNPACK_ROW_LENGTH, (GLint)pixel_stride); diff --git a/src/video_core/vertex_shader.cpp b/src/video_core/vertex_shader.cpp index 7d68998f..87006a83 100644 --- a/src/video_core/vertex_shader.cpp +++ b/src/video_core/vertex_shader.cpp @@ -119,17 +119,13 @@ static void ProcessShaderCode(VertexShaderState& state) { switch (instr.opcode.Value().GetInfo().type) { case OpCode::Type::Arithmetic: { - bool is_inverted = 0 != (instr.opcode.Value().GetInfo().subtype & OpCode::Info::SrcInversed); - // TODO: We don't really support this properly: For instance, the address register - // offset needs to be applied to SRC2 instead, etc. - // For now, we just abort in this situation. - ASSERT_MSG(!is_inverted, "Bad condition..."); + const bool is_inverted = (0 != (instr.opcode.Value().GetInfo().subtype & OpCode::Info::SrcInversed)); const int address_offset = (instr.common.address_register_index == 0) ? 0 : state.address_registers[instr.common.address_register_index - 1]; - const float24* src1_ = LookupSourceRegister(instr.common.GetSrc1(is_inverted) + address_offset); - const float24* src2_ = LookupSourceRegister(instr.common.GetSrc2(is_inverted)); + const float24* src1_ = LookupSourceRegister(instr.common.GetSrc1(is_inverted) + (!is_inverted * address_offset)); + const float24* src2_ = LookupSourceRegister(instr.common.GetSrc2(is_inverted) + ( is_inverted * address_offset)); const bool negate_src1 = ((bool)swizzle.negate_src1 != false); const bool negate_src2 = ((bool)swizzle.negate_src2 != false); @@ -208,6 +204,15 @@ static void ProcessShaderCode(VertexShaderState& state) { } break; + case OpCode::Id::MIN: + for (int i = 0; i < 4; ++i) { + if (!swizzle.DestComponentEnabled(i)) + continue; + + dest[i] = std::min(src1[i], src2[i]); + } + break; + case OpCode::Id::DP3: case OpCode::Id::DP4: { @@ -279,6 +284,16 @@ static void ProcessShaderCode(VertexShaderState& state) { break; } + case OpCode::Id::SLT: + case OpCode::Id::SLTI: + for (int i = 0; i < 4; ++i) { + if (!swizzle.DestComponentEnabled(i)) + continue; + + dest[i] = (src1[i] < src2[i]) ? float24::FromFloat32(1.0f) : float24::FromFloat32(0.0f); + } + break; + case OpCode::Id::CMP: for (int i = 0; i < 2; ++i) { // TODO: Can you restrict to one compare via dest masking? @@ -330,7 +345,7 @@ static void ProcessShaderCode(VertexShaderState& state) { case OpCode::Type::MultiplyAdd: { - if ((instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD) || + if ((instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD) || (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI)) { const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.mad.operand_desc_id]; @@ -547,7 +562,7 @@ OutputVertex RunShader(const InputVertex& input, int num_attributes) { const auto& attribute_register_map = regs.vs_input_register_map; float24 dummy_register; boost::fill(state.input_register_table, &dummy_register); - + if (num_attributes > 0) state.input_register_table[attribute_register_map.attribute0_register] = &input.attr[0].x; if (num_attributes > 1) state.input_register_table[attribute_register_map.attribute1_register] = &input.attr[1].x; if (num_attributes > 2) state.input_register_table[attribute_register_map.attribute2_register] = &input.attr[2].x; |