diff options
Diffstat (limited to '')
| -rw-r--r-- | src/citra_qt/debugger/graphics.cpp | 34 | ||||
| -rw-r--r-- | src/common/register_set.h | 14 | ||||
| -rw-r--r-- | src/core/hle/config_mem.h | 6 | ||||
| -rw-r--r-- | src/core/hle/service/gsp.cpp | 143 | ||||
| -rw-r--r-- | src/core/hle/service/gsp.h | 60 | ||||
| -rw-r--r-- | src/core/hw/gpu.cpp | 253 | ||||
| -rw-r--r-- | src/core/hw/gpu.h | 219 | ||||
| -rw-r--r-- | src/core/hw/hw.h | 4 | ||||
| -rw-r--r-- | src/video_core/gpu_debugger.h | 11 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/renderer_opengl.cpp | 62 | ||||
| -rw-r--r-- | src/video_core/renderer_opengl/renderer_opengl.h | 7 |
11 files changed, 585 insertions, 228 deletions
diff --git a/src/citra_qt/debugger/graphics.cpp b/src/citra_qt/debugger/graphics.cpp index 9aaade8f9..0f911a015 100644 --- a/src/citra_qt/debugger/graphics.cpp +++ b/src/citra_qt/debugger/graphics.cpp | |||
| @@ -28,22 +28,24 @@ QVariant GPUCommandStreamItemModel::data(const QModelIndex& index, int role) con | |||
| 28 | const GSP_GPU::GXCommand& command = GetDebugger()->ReadGXCommandHistory(command_index); | 28 | const GSP_GPU::GXCommand& command = GetDebugger()->ReadGXCommandHistory(command_index); |
| 29 | if (role == Qt::DisplayRole) | 29 | if (role == Qt::DisplayRole) |
| 30 | { | 30 | { |
| 31 | std::map<GSP_GPU::GXCommandId, const char*> command_names; | 31 | std::map<GSP_GPU::GXCommandId, const char*> command_names = { |
| 32 | command_names[GSP_GPU::GXCommandId::REQUEST_DMA] = "REQUEST_DMA"; | 32 | { GSP_GPU::GXCommandId::REQUEST_DMA, "REQUEST_DMA" }, |
| 33 | command_names[GSP_GPU::GXCommandId::SET_COMMAND_LIST_FIRST] = "SET_COMMAND_LIST_FIRST"; | 33 | { GSP_GPU::GXCommandId::SET_COMMAND_LIST_FIRST, "SET_COMMAND_LIST_FIRST" }, |
| 34 | command_names[GSP_GPU::GXCommandId::SET_MEMORY_FILL] = "SET_MEMORY_FILL"; | 34 | { GSP_GPU::GXCommandId::SET_MEMORY_FILL, "SET_MEMORY_FILL" }, |
| 35 | command_names[GSP_GPU::GXCommandId::SET_DISPLAY_TRANSFER] = "SET_DISPLAY_TRANSFER"; | 35 | { GSP_GPU::GXCommandId::SET_DISPLAY_TRANSFER, "SET_DISPLAY_TRANSFER" }, |
| 36 | command_names[GSP_GPU::GXCommandId::SET_TEXTURE_COPY] = "SET_TEXTURE_COPY"; | 36 | { GSP_GPU::GXCommandId::SET_TEXTURE_COPY, "SET_TEXTURE_COPY" }, |
| 37 | command_names[GSP_GPU::GXCommandId::SET_COMMAND_LIST_LAST] = "SET_COMMAND_LIST_LAST"; | 37 | { GSP_GPU::GXCommandId::SET_COMMAND_LIST_LAST, "SET_COMMAND_LIST_LAST" } |
| 38 | QString str = QString("%1 %2 %3 %4 %5 %6 %7 %8 %9").arg(command_names[static_cast<GSP_GPU::GXCommandId>(command.id)]) | 38 | }; |
| 39 | .arg(command.data[0], 8, 16, QLatin1Char('0')) | 39 | const u32* command_data = reinterpret_cast<const u32*>(&command); |
| 40 | .arg(command.data[1], 8, 16, QLatin1Char('0')) | 40 | QString str = QString("%1 %2 %3 %4 %5 %6 %7 %8 %9").arg(command_names[command.id]) |
| 41 | .arg(command.data[2], 8, 16, QLatin1Char('0')) | 41 | .arg(command_data[0], 8, 16, QLatin1Char('0')) |
| 42 | .arg(command.data[3], 8, 16, QLatin1Char('0')) | 42 | .arg(command_data[1], 8, 16, QLatin1Char('0')) |
| 43 | .arg(command.data[4], 8, 16, QLatin1Char('0')) | 43 | .arg(command_data[2], 8, 16, QLatin1Char('0')) |
| 44 | .arg(command.data[5], 8, 16, QLatin1Char('0')) | 44 | .arg(command_data[3], 8, 16, QLatin1Char('0')) |
| 45 | .arg(command.data[6], 8, 16, QLatin1Char('0')) | 45 | .arg(command_data[4], 8, 16, QLatin1Char('0')) |
| 46 | .arg(command.data[7], 8, 16, QLatin1Char('0')); | 46 | .arg(command_data[5], 8, 16, QLatin1Char('0')) |
| 47 | .arg(command_data[6], 8, 16, QLatin1Char('0')) | ||
| 48 | .arg(command_data[7], 8, 16, QLatin1Char('0')); | ||
| 47 | return QVariant(str); | 49 | return QVariant(str); |
| 48 | } | 50 | } |
| 49 | else | 51 | else |
diff --git a/src/common/register_set.h b/src/common/register_set.h index 0418551b3..ba19a2614 100644 --- a/src/common/register_set.h +++ b/src/common/register_set.h | |||
| @@ -34,7 +34,7 @@ | |||
| 34 | /* | 34 | /* |
| 35 | * Standardized way to define a group of registers and corresponding data structures. To define | 35 | * Standardized way to define a group of registers and corresponding data structures. To define |
| 36 | * a new register set, first define struct containing an enumeration called "Id" containing | 36 | * a new register set, first define struct containing an enumeration called "Id" containing |
| 37 | * all register IDs and a template union called "Struct". Specialize the Struct union for any | 37 | * all register IDs and a template struct called "Struct". Specialize the Struct struct for any |
| 38 | * register ID which needs to be accessed in a specialized way. You can then declare the object | 38 | * register ID which needs to be accessed in a specialized way. You can then declare the object |
| 39 | * containing all register values using the RegisterSet<BaseType, DefiningStruct> type, where | 39 | * containing all register values using the RegisterSet<BaseType, DefiningStruct> type, where |
| 40 | * BaseType is the underlying type of each register (e.g. u32). | 40 | * BaseType is the underlying type of each register (e.g. u32). |
| @@ -54,7 +54,7 @@ | |||
| 54 | * | 54 | * |
| 55 | * // declare register definition structures | 55 | * // declare register definition structures |
| 56 | * template<Id id> | 56 | * template<Id id> |
| 57 | * union Struct; | 57 | * struct Struct; |
| 58 | * }; | 58 | * }; |
| 59 | * | 59 | * |
| 60 | * // Define register set object | 60 | * // Define register set object |
| @@ -62,9 +62,11 @@ | |||
| 62 | * | 62 | * |
| 63 | * // define register definition structures | 63 | * // define register definition structures |
| 64 | * template<> | 64 | * template<> |
| 65 | * union Regs::Struct<Regs::Value1> { | 65 | * struct Regs::Struct<Regs::Value1> { |
| 66 | * BitField<0, 4, u32> some_field; | 66 | * union { |
| 67 | * BitField<4, 3, u32> some_other_field; | 67 | * BitField<0, 4, u32> some_field; |
| 68 | * BitField<4, 3, u32> some_other_field; | ||
| 69 | * }; | ||
| 68 | * }; | 70 | * }; |
| 69 | * | 71 | * |
| 70 | * Usage in external code (within SomeNamespace scope): | 72 | * Usage in external code (within SomeNamespace scope): |
| @@ -77,7 +79,7 @@ | |||
| 77 | * | 79 | * |
| 78 | * | 80 | * |
| 79 | * @tparam BaseType Base type used for storing individual registers, e.g. u32 | 81 | * @tparam BaseType Base type used for storing individual registers, e.g. u32 |
| 80 | * @tparam RegDefinition Class defining an enumeration called "Id" and a template<Id id> union, as described above. | 82 | * @tparam RegDefinition Class defining an enumeration called "Id" and a template<Id id> struct, as described above. |
| 81 | * @note RegDefinition::Id needs to have an enum value called NumIds defining the number of registers to be allocated. | 83 | * @note RegDefinition::Id needs to have an enum value called NumIds defining the number of registers to be allocated. |
| 82 | */ | 84 | */ |
| 83 | template<typename BaseType, typename RegDefinition> | 85 | template<typename BaseType, typename RegDefinition> |
diff --git a/src/core/hle/config_mem.h b/src/core/hle/config_mem.h index da396a3e6..fa01b5cdb 100644 --- a/src/core/hle/config_mem.h +++ b/src/core/hle/config_mem.h | |||
| @@ -1,10 +1,10 @@ | |||
| 1 | // Copyright 2014 Citra Emulator Project | 1 | // Copyright 2014 Citra Emulator Project |
| 2 | // Licensed under GPLv2 | 2 | // Licensed under GPLv2 |
| 3 | // Refer to the license.txt file included. | 3 | // Refer to the license.txt file included. |
| 4 | 4 | ||
| 5 | #pragma once | 5 | #pragma once |
| 6 | 6 | ||
| 7 | // Configuration memory stores various hardware/kernel configuration settings. This memory page is | 7 | // Configuration memory stores various hardware/kernel configuration settings. This memory page is |
| 8 | // read-only for ARM11 processes. I'm guessing this would normally be written to by the firmware/ | 8 | // read-only for ARM11 processes. I'm guessing this would normally be written to by the firmware/ |
| 9 | // bootrom. Because we're not emulating this, and essentially just "stubbing" the functionality, I'm | 9 | // bootrom. Because we're not emulating this, and essentially just "stubbing" the functionality, I'm |
| 10 | // putting this as a subset of HLE for now. | 10 | // putting this as a subset of HLE for now. |
| @@ -16,6 +16,6 @@ | |||
| 16 | namespace ConfigMem { | 16 | namespace ConfigMem { |
| 17 | 17 | ||
| 18 | template <typename T> | 18 | template <typename T> |
| 19 | inline void Read(T &var, const u32 addr); | 19 | void Read(T &var, const u32 addr); |
| 20 | 20 | ||
| 21 | } // namespace | 21 | } // namespace |
diff --git a/src/core/hle/service/gsp.cpp b/src/core/hle/service/gsp.cpp index 1fdbdf342..b20203e27 100644 --- a/src/core/hle/service/gsp.cpp +++ b/src/core/hle/service/gsp.cpp | |||
| @@ -47,11 +47,6 @@ Handle g_shared_memory = 0; | |||
| 47 | 47 | ||
| 48 | u32 g_thread_id = 0; | 48 | u32 g_thread_id = 0; |
| 49 | 49 | ||
| 50 | enum { | ||
| 51 | REG_FRAMEBUFFER_1 = 0x00400468, | ||
| 52 | REG_FRAMEBUFFER_2 = 0x00400494, | ||
| 53 | }; | ||
| 54 | |||
| 55 | /// Gets a pointer to the start (header) of a command buffer in GSP shared memory | 50 | /// Gets a pointer to the start (header) of a command buffer in GSP shared memory |
| 56 | static inline u8* GX_GetCmdBufferPointer(u32 thread_id, u32 offset=0) { | 51 | static inline u8* GX_GetCmdBufferPointer(u32 thread_id, u32 offset=0) { |
| 57 | return Kernel::GetSharedMemoryPointer(g_shared_memory, 0x800 + (thread_id * 0x200) + offset); | 52 | return Kernel::GetSharedMemoryPointer(g_shared_memory, 0x800 + (thread_id * 0x200) + offset); |
| @@ -67,38 +62,62 @@ void GX_FinishCommand(u32 thread_id) { | |||
| 67 | // TODO: Increment header->index? | 62 | // TODO: Increment header->index? |
| 68 | } | 63 | } |
| 69 | 64 | ||
| 65 | /// Write a GSP GPU hardware register | ||
| 66 | void WriteHWRegs(Service::Interface* self) { | ||
| 67 | u32* cmd_buff = Service::GetCommandBuffer(); | ||
| 68 | u32 reg_addr = cmd_buff[1]; | ||
| 69 | u32 size = cmd_buff[2]; | ||
| 70 | |||
| 71 | // TODO: Return proper error codes | ||
| 72 | if (reg_addr + size >= 0x420000) { | ||
| 73 | ERROR_LOG(GPU, "Write address out of range! (address=0x%08x, size=0x%08x)", reg_addr, size); | ||
| 74 | return; | ||
| 75 | } | ||
| 76 | |||
| 77 | // size should be word-aligned | ||
| 78 | if ((size % 4) != 0) { | ||
| 79 | ERROR_LOG(GPU, "Invalid size 0x%08x", size); | ||
| 80 | return; | ||
| 81 | } | ||
| 82 | |||
| 83 | u32* src = (u32*)Memory::GetPointer(cmd_buff[0x4]); | ||
| 84 | |||
| 85 | while (size > 0) { | ||
| 86 | GPU::Write<u32>(reg_addr + 0x1EB00000, *src); | ||
| 87 | |||
| 88 | size -= 4; | ||
| 89 | ++src; | ||
| 90 | reg_addr += 4; | ||
| 91 | } | ||
| 92 | } | ||
| 93 | |||
| 70 | /// Read a GSP GPU hardware register | 94 | /// Read a GSP GPU hardware register |
| 71 | void ReadHWRegs(Service::Interface* self) { | 95 | void ReadHWRegs(Service::Interface* self) { |
| 72 | static const u32 framebuffer_1[] = {GPU::PADDR_VRAM_TOP_LEFT_FRAME1, GPU::PADDR_VRAM_TOP_RIGHT_FRAME1}; | ||
| 73 | static const u32 framebuffer_2[] = {GPU::PADDR_VRAM_TOP_LEFT_FRAME2, GPU::PADDR_VRAM_TOP_RIGHT_FRAME2}; | ||
| 74 | |||
| 75 | u32* cmd_buff = Service::GetCommandBuffer(); | 96 | u32* cmd_buff = Service::GetCommandBuffer(); |
| 76 | u32 reg_addr = cmd_buff[1]; | 97 | u32 reg_addr = cmd_buff[1]; |
| 77 | u32 size = cmd_buff[2]; | 98 | u32 size = cmd_buff[2]; |
| 78 | u32* dst = (u32*)Memory::GetPointer(cmd_buff[0x41]); | ||
| 79 | 99 | ||
| 80 | switch (reg_addr) { | 100 | // TODO: Return proper error codes |
| 101 | if (reg_addr + size >= 0x420000) { | ||
| 102 | ERROR_LOG(GPU, "Read address out of range! (address=0x%08x, size=0x%08x)", reg_addr, size); | ||
| 103 | return; | ||
| 104 | } | ||
| 81 | 105 | ||
| 82 | // NOTE: Calling SetFramebufferLocation here is a hack... Not sure the correct way yet to set | 106 | // size should be word-aligned |
| 83 | // whether the framebuffers should be in VRAM or GSP heap, but from what I understand, if the | 107 | if ((size % 4) != 0) { |
| 84 | // user application is reading from either of these registers, then its going to be in VRAM. | 108 | ERROR_LOG(GPU, "Invalid size 0x%08x", size); |
| 109 | return; | ||
| 110 | } | ||
| 85 | 111 | ||
| 86 | // Top framebuffer 1 addresses | 112 | u32* dst = (u32*)Memory::GetPointer(cmd_buff[0x41]); |
| 87 | case REG_FRAMEBUFFER_1: | ||
| 88 | GPU::SetFramebufferLocation(GPU::FRAMEBUFFER_LOCATION_VRAM); | ||
| 89 | memcpy(dst, framebuffer_1, size); | ||
| 90 | break; | ||
| 91 | 113 | ||
| 92 | // Top framebuffer 2 addresses | 114 | while (size > 0) { |
| 93 | case REG_FRAMEBUFFER_2: | 115 | GPU::Read<u32>(*dst, reg_addr + 0x1EB00000); |
| 94 | GPU::SetFramebufferLocation(GPU::FRAMEBUFFER_LOCATION_VRAM); | ||
| 95 | memcpy(dst, framebuffer_2, size); | ||
| 96 | break; | ||
| 97 | 116 | ||
| 98 | default: | 117 | size -= 4; |
| 99 | ERROR_LOG(GSP, "unknown register read at address %08X", reg_addr); | 118 | ++dst; |
| 119 | reg_addr += 4; | ||
| 100 | } | 120 | } |
| 101 | |||
| 102 | } | 121 | } |
| 103 | 122 | ||
| 104 | /** | 123 | /** |
| @@ -120,8 +139,8 @@ void RegisterInterruptRelayQueue(Service::Interface* self) { | |||
| 120 | 139 | ||
| 121 | Kernel::SetEventLocked(g_event, false); | 140 | Kernel::SetEventLocked(g_event, false); |
| 122 | 141 | ||
| 123 | // Hack - This function will permanently set the state of the GSP event such that GPU command | 142 | // Hack - This function will permanently set the state of the GSP event such that GPU command |
| 124 | // synchronization barriers always passthrough. Correct solution would be to set this after the | 143 | // synchronization barriers always passthrough. Correct solution would be to set this after the |
| 125 | // GPU as processed all queued up commands, but due to the emulator being single-threaded they | 144 | // GPU as processed all queued up commands, but due to the emulator being single-threaded they |
| 126 | // will always be ready. | 145 | // will always be ready. |
| 127 | Kernel::SetPermanentLock(g_event, true); | 146 | Kernel::SetPermanentLock(g_event, true); |
| @@ -134,52 +153,92 @@ void RegisterInterruptRelayQueue(Service::Interface* self) { | |||
| 134 | 153 | ||
| 135 | /// This triggers handling of the GX command written to the command buffer in shared memory. | 154 | /// This triggers handling of the GX command written to the command buffer in shared memory. |
| 136 | void TriggerCmdReqQueue(Service::Interface* self) { | 155 | void TriggerCmdReqQueue(Service::Interface* self) { |
| 156 | |||
| 157 | // Utility function to convert register ID to address | ||
| 158 | auto WriteGPURegister = [](u32 id, u32 data) { | ||
| 159 | GPU::Write<u32>(0x1EF00000 + 4 * id, data); | ||
| 160 | }; | ||
| 161 | |||
| 137 | GX_CmdBufferHeader* header = (GX_CmdBufferHeader*)GX_GetCmdBufferPointer(g_thread_id); | 162 | GX_CmdBufferHeader* header = (GX_CmdBufferHeader*)GX_GetCmdBufferPointer(g_thread_id); |
| 138 | u32* cmd_buff = (u32*)GX_GetCmdBufferPointer(g_thread_id, 0x20 + (header->index * 0x20)); | 163 | auto& command = *(const GXCommand*)GX_GetCmdBufferPointer(g_thread_id, 0x20 + (header->index * 0x20)); |
| 139 | 164 | ||
| 140 | switch (static_cast<GXCommandId>(cmd_buff[0])) { | 165 | switch (command.id) { |
| 141 | 166 | ||
| 142 | // GX request DMA - typically used for copying memory from GSP heap to VRAM | 167 | // GX request DMA - typically used for copying memory from GSP heap to VRAM |
| 143 | case GXCommandId::REQUEST_DMA: | 168 | case GXCommandId::REQUEST_DMA: |
| 144 | memcpy(Memory::GetPointer(cmd_buff[2]), Memory::GetPointer(cmd_buff[1]), cmd_buff[3]); | 169 | memcpy(Memory::GetPointer(command.dma_request.dest_address), |
| 170 | Memory::GetPointer(command.dma_request.source_address), | ||
| 171 | command.dma_request.size); | ||
| 145 | break; | 172 | break; |
| 146 | 173 | ||
| 174 | // ctrulib homebrew sends all relevant command list data with this command, | ||
| 175 | // hence we do all "interesting" stuff here and do nothing in SET_COMMAND_LIST_FIRST. | ||
| 176 | // TODO: This will need some rework in the future. | ||
| 147 | case GXCommandId::SET_COMMAND_LIST_LAST: | 177 | case GXCommandId::SET_COMMAND_LIST_LAST: |
| 148 | GPU::Write<u32>(GPU::Registers::CommandListAddress, cmd_buff[1] >> 3); | 178 | { |
| 149 | GPU::Write<u32>(GPU::Registers::CommandListSize, cmd_buff[2] >> 3); | 179 | auto& params = command.set_command_list_last; |
| 150 | GPU::Write<u32>(GPU::Registers::ProcessCommandList, 1); // TODO: Not sure if we are supposed to always write this | 180 | WriteGPURegister(GPU::Regs::CommandProcessor + 2, params.address >> 3); |
| 181 | WriteGPURegister(GPU::Regs::CommandProcessor, params.size >> 3); | ||
| 182 | WriteGPURegister(GPU::Regs::CommandProcessor + 4, 1); // TODO: Not sure if we are supposed to always write this .. seems to trigger processing though | ||
| 151 | 183 | ||
| 152 | // TODO: Move this to GPU | 184 | // TODO: Move this to GPU |
| 153 | // TODO: Not sure what units the size is measured in | 185 | // TODO: Not sure what units the size is measured in |
| 154 | g_debugger.CommandListCalled(cmd_buff[1], (u32*)Memory::GetPointer(cmd_buff[1]), cmd_buff[2]); | 186 | g_debugger.CommandListCalled(params.address, |
| 187 | (u32*)Memory::GetPointer(params.address), | ||
| 188 | params.size); | ||
| 155 | break; | 189 | break; |
| 190 | } | ||
| 156 | 191 | ||
| 192 | // It's assumed that the two "blocks" behave equivalently. | ||
| 193 | // Presumably this is done simply to allow two memory fills to run in parallel. | ||
| 157 | case GXCommandId::SET_MEMORY_FILL: | 194 | case GXCommandId::SET_MEMORY_FILL: |
| 195 | { | ||
| 196 | auto& params = command.memory_fill; | ||
| 197 | WriteGPURegister(GPU::Regs::MemoryFill, params.start1 >> 3); | ||
| 198 | WriteGPURegister(GPU::Regs::MemoryFill + 1, params.end1 >> 3); | ||
| 199 | WriteGPURegister(GPU::Regs::MemoryFill + 2, params.end1 - params.start1); | ||
| 200 | WriteGPURegister(GPU::Regs::MemoryFill + 3, params.value1); | ||
| 201 | |||
| 202 | WriteGPURegister(GPU::Regs::MemoryFill + 4, params.start2 >> 3); | ||
| 203 | WriteGPURegister(GPU::Regs::MemoryFill + 5, params.end2 >> 3); | ||
| 204 | WriteGPURegister(GPU::Regs::MemoryFill + 6, params.end2 - params.start2); | ||
| 205 | WriteGPURegister(GPU::Regs::MemoryFill + 7, params.value2); | ||
| 158 | break; | 206 | break; |
| 207 | } | ||
| 159 | 208 | ||
| 209 | // TODO: Check if texture copies are implemented correctly.. | ||
| 160 | case GXCommandId::SET_DISPLAY_TRANSFER: | 210 | case GXCommandId::SET_DISPLAY_TRANSFER: |
| 161 | break; | ||
| 162 | |||
| 163 | case GXCommandId::SET_TEXTURE_COPY: | 211 | case GXCommandId::SET_TEXTURE_COPY: |
| 212 | { | ||
| 213 | auto& params = command.image_copy; | ||
| 214 | WriteGPURegister(GPU::Regs::DisplayTransfer, params.in_buffer_address >> 3); | ||
| 215 | WriteGPURegister(GPU::Regs::DisplayTransfer + 1, params.out_buffer_address >> 3); | ||
| 216 | WriteGPURegister(GPU::Regs::DisplayTransfer + 3, params.in_buffer_size); | ||
| 217 | WriteGPURegister(GPU::Regs::DisplayTransfer + 2, params.out_buffer_size); | ||
| 218 | WriteGPURegister(GPU::Regs::DisplayTransfer + 4, params.flags); | ||
| 219 | |||
| 220 | // TODO: Should this only be ORed with 1 for texture copies? | ||
| 221 | // trigger transfer | ||
| 222 | WriteGPURegister(GPU::Regs::DisplayTransfer + 6, 1); | ||
| 164 | break; | 223 | break; |
| 224 | } | ||
| 165 | 225 | ||
| 226 | // TODO: Figure out what exactly SET_COMMAND_LIST_FIRST and SET_COMMAND_LIST_LAST | ||
| 227 | // are supposed to do. | ||
| 166 | case GXCommandId::SET_COMMAND_LIST_FIRST: | 228 | case GXCommandId::SET_COMMAND_LIST_FIRST: |
| 167 | { | 229 | { |
| 168 | //u32* buf0_data = (u32*)Memory::GetPointer(cmd_buff[1]); | ||
| 169 | //u32* buf1_data = (u32*)Memory::GetPointer(cmd_buff[3]); | ||
| 170 | //u32* buf2_data = (u32*)Memory::GetPointer(cmd_buff[5]); | ||
| 171 | break; | 230 | break; |
| 172 | } | 231 | } |
| 173 | 232 | ||
| 174 | default: | 233 | default: |
| 175 | ERROR_LOG(GSP, "unknown command 0x%08X", cmd_buff[0]); | 234 | ERROR_LOG(GSP, "unknown command 0x%08X", (int)command.id.Value()); |
| 176 | } | 235 | } |
| 177 | 236 | ||
| 178 | GX_FinishCommand(g_thread_id); | 237 | GX_FinishCommand(g_thread_id); |
| 179 | } | 238 | } |
| 180 | 239 | ||
| 181 | const Interface::FunctionInfo FunctionTable[] = { | 240 | const Interface::FunctionInfo FunctionTable[] = { |
| 182 | {0x00010082, nullptr, "WriteHWRegs"}, | 241 | {0x00010082, WriteHWRegs, "WriteHWRegs"}, |
| 183 | {0x00020084, nullptr, "WriteHWRegsWithMask"}, | 242 | {0x00020084, nullptr, "WriteHWRegsWithMask"}, |
| 184 | {0x00030082, nullptr, "WriteHWRegRepeat"}, | 243 | {0x00030082, nullptr, "WriteHWRegRepeat"}, |
| 185 | {0x00040080, ReadHWRegs, "ReadHWRegs"}, | 244 | {0x00040080, ReadHWRegs, "ReadHWRegs"}, |
diff --git a/src/core/hle/service/gsp.h b/src/core/hle/service/gsp.h index 214de140f..a83cb4846 100644 --- a/src/core/hle/service/gsp.h +++ b/src/core/hle/service/gsp.h | |||
| @@ -4,6 +4,7 @@ | |||
| 4 | 4 | ||
| 5 | #pragma once | 5 | #pragma once |
| 6 | 6 | ||
| 7 | #include "common/bit_field.h" | ||
| 7 | #include "core/hle/service/service.h" | 8 | #include "core/hle/service/service.h" |
| 8 | 9 | ||
| 9 | //////////////////////////////////////////////////////////////////////////////////////////////////// | 10 | //////////////////////////////////////////////////////////////////////////////////////////////////// |
| @@ -12,21 +13,58 @@ | |||
| 12 | namespace GSP_GPU { | 13 | namespace GSP_GPU { |
| 13 | 14 | ||
| 14 | enum class GXCommandId : u32 { | 15 | enum class GXCommandId : u32 { |
| 15 | REQUEST_DMA = 0x00000000, | 16 | REQUEST_DMA = 0x00, |
| 16 | SET_COMMAND_LIST_LAST = 0x00000001, | 17 | SET_COMMAND_LIST_LAST = 0x01, |
| 17 | SET_MEMORY_FILL = 0x00000002, // TODO: Confirm? (lictru uses 0x01000102) | 18 | |
| 18 | SET_DISPLAY_TRANSFER = 0x00000003, | 19 | // Fills a given memory range with a particular value |
| 19 | SET_TEXTURE_COPY = 0x00000004, | 20 | SET_MEMORY_FILL = 0x02, |
| 20 | SET_COMMAND_LIST_FIRST = 0x00000005, | 21 | |
| 22 | // Copies an image and optionally performs color-conversion or scaling. | ||
| 23 | // This is highly similar to the GameCube's EFB copy feature | ||
| 24 | SET_DISPLAY_TRANSFER = 0x03, | ||
| 25 | |||
| 26 | // Conceptionally similar to SET_DISPLAY_TRANSFER and presumable uses the same hardware path | ||
| 27 | SET_TEXTURE_COPY = 0x04, | ||
| 28 | |||
| 29 | SET_COMMAND_LIST_FIRST = 0x05, | ||
| 21 | }; | 30 | }; |
| 22 | 31 | ||
| 23 | union GXCommand { | 32 | struct GXCommand { |
| 24 | struct { | 33 | BitField<0, 8, GXCommandId> id; |
| 25 | GXCommandId id; | ||
| 26 | }; | ||
| 27 | 34 | ||
| 28 | u32 data[0x20]; | 35 | union { |
| 36 | struct { | ||
| 37 | u32 source_address; | ||
| 38 | u32 dest_address; | ||
| 39 | u32 size; | ||
| 40 | } dma_request; | ||
| 41 | |||
| 42 | struct { | ||
| 43 | u32 address; | ||
| 44 | u32 size; | ||
| 45 | } set_command_list_last; | ||
| 46 | |||
| 47 | struct { | ||
| 48 | u32 start1; | ||
| 49 | u32 value1; | ||
| 50 | u32 end1; | ||
| 51 | u32 start2; | ||
| 52 | u32 value2; | ||
| 53 | u32 end2; | ||
| 54 | } memory_fill; | ||
| 55 | |||
| 56 | struct { | ||
| 57 | u32 in_buffer_address; | ||
| 58 | u32 out_buffer_address; | ||
| 59 | u32 in_buffer_size; | ||
| 60 | u32 out_buffer_size; | ||
| 61 | u32 flags; | ||
| 62 | } image_copy; | ||
| 63 | |||
| 64 | u8 raw_data[0x1C]; | ||
| 65 | }; | ||
| 29 | }; | 66 | }; |
| 67 | static_assert(sizeof(GXCommand) == 0x20, "GXCommand struct has incorrect size"); | ||
| 30 | 68 | ||
| 31 | /// Interface to "srv:" service | 69 | /// Interface to "srv:" service |
| 32 | class Interface : public Service::Interface { | 70 | class Interface : public Service::Interface { |
diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp index f0ca4eada..c00be2a83 100644 --- a/src/core/hw/gpu.cpp +++ b/src/core/hw/gpu.cpp | |||
| @@ -15,48 +15,58 @@ | |||
| 15 | 15 | ||
| 16 | namespace GPU { | 16 | namespace GPU { |
| 17 | 17 | ||
| 18 | Registers g_regs; | 18 | RegisterSet<u32, Regs> g_regs; |
| 19 | 19 | ||
| 20 | u64 g_last_ticks = 0; ///< Last CPU ticks | 20 | u64 g_last_ticks = 0; ///< Last CPU ticks |
| 21 | 21 | ||
| 22 | /** | 22 | /** |
| 23 | * Sets whether the framebuffers are in the GSP heap (FCRAM) or VRAM | 23 | * Sets whether the framebuffers are in the GSP heap (FCRAM) or VRAM |
| 24 | * @param | 24 | * @param |
| 25 | */ | 25 | */ |
| 26 | void SetFramebufferLocation(const FramebufferLocation mode) { | 26 | void SetFramebufferLocation(const FramebufferLocation mode) { |
| 27 | switch (mode) { | 27 | switch (mode) { |
| 28 | case FRAMEBUFFER_LOCATION_FCRAM: | 28 | case FRAMEBUFFER_LOCATION_FCRAM: |
| 29 | g_regs.framebuffer_top_left_1 = PADDR_TOP_LEFT_FRAME1; | 29 | { |
| 30 | g_regs.framebuffer_top_left_2 = PADDR_TOP_LEFT_FRAME2; | 30 | auto& framebuffer_top = g_regs.Get<Regs::FramebufferTop>(); |
| 31 | g_regs.framebuffer_top_right_1 = PADDR_TOP_RIGHT_FRAME1; | 31 | auto& framebuffer_sub = g_regs.Get<Regs::FramebufferBottom>(); |
| 32 | g_regs.framebuffer_top_right_2 = PADDR_TOP_RIGHT_FRAME2; | 32 | |
| 33 | g_regs.framebuffer_sub_left_1 = PADDR_SUB_FRAME1; | 33 | framebuffer_top.address_left1 = PADDR_TOP_LEFT_FRAME1; |
| 34 | //g_regs.framebuffer_sub_left_2 = unknown; | 34 | framebuffer_top.address_left2 = PADDR_TOP_LEFT_FRAME2; |
| 35 | g_regs.framebuffer_sub_right_1 = PADDR_SUB_FRAME2; | 35 | framebuffer_top.address_right1 = PADDR_TOP_RIGHT_FRAME1; |
| 36 | //g_regs.framebufferr_sub_right_2 = unknown; | 36 | framebuffer_top.address_right2 = PADDR_TOP_RIGHT_FRAME2; |
| 37 | framebuffer_sub.address_left1 = PADDR_SUB_FRAME1; | ||
| 38 | //framebuffer_sub.address_left2 = unknown; | ||
| 39 | framebuffer_sub.address_right1 = PADDR_SUB_FRAME2; | ||
| 40 | //framebuffer_sub.address_right2 = unknown; | ||
| 37 | break; | 41 | break; |
| 42 | } | ||
| 38 | 43 | ||
| 39 | case FRAMEBUFFER_LOCATION_VRAM: | 44 | case FRAMEBUFFER_LOCATION_VRAM: |
| 40 | g_regs.framebuffer_top_left_1 = PADDR_VRAM_TOP_LEFT_FRAME1; | 45 | { |
| 41 | g_regs.framebuffer_top_left_2 = PADDR_VRAM_TOP_LEFT_FRAME2; | 46 | auto& framebuffer_top = g_regs.Get<Regs::FramebufferTop>(); |
| 42 | g_regs.framebuffer_top_right_1 = PADDR_VRAM_TOP_RIGHT_FRAME1; | 47 | auto& framebuffer_sub = g_regs.Get<Regs::FramebufferBottom>(); |
| 43 | g_regs.framebuffer_top_right_2 = PADDR_VRAM_TOP_RIGHT_FRAME2; | 48 | |
| 44 | g_regs.framebuffer_sub_left_1 = PADDR_VRAM_SUB_FRAME1; | 49 | framebuffer_top.address_left1 = PADDR_VRAM_TOP_LEFT_FRAME1; |
| 45 | //g_regs.framebuffer_sub_left_2 = unknown; | 50 | framebuffer_top.address_left2 = PADDR_VRAM_TOP_LEFT_FRAME2; |
| 46 | g_regs.framebuffer_sub_right_1 = PADDR_VRAM_SUB_FRAME2; | 51 | framebuffer_top.address_right1 = PADDR_VRAM_TOP_RIGHT_FRAME1; |
| 47 | //g_regs.framebufferr_sub_right_2 = unknown; | 52 | framebuffer_top.address_right2 = PADDR_VRAM_TOP_RIGHT_FRAME2; |
| 53 | framebuffer_sub.address_left1 = PADDR_VRAM_SUB_FRAME1; | ||
| 54 | //framebuffer_sub.address_left2 = unknown; | ||
| 55 | framebuffer_sub.address_right1 = PADDR_VRAM_SUB_FRAME2; | ||
| 56 | //framebuffer_sub.address_right2 = unknown; | ||
| 48 | break; | 57 | break; |
| 49 | } | 58 | } |
| 59 | } | ||
| 50 | } | 60 | } |
| 51 | 61 | ||
| 52 | /** | 62 | /** |
| 53 | * Gets the location of the framebuffers | 63 | * Gets the location of the framebuffers |
| 54 | * @return Location of framebuffers as FramebufferLocation enum | 64 | * @return Location of framebuffers as FramebufferLocation enum |
| 55 | */ | 65 | */ |
| 56 | const FramebufferLocation GetFramebufferLocation() { | 66 | FramebufferLocation GetFramebufferLocation(u32 address) { |
| 57 | if ((g_regs.framebuffer_top_right_1 & ~Memory::VRAM_MASK) == Memory::VRAM_PADDR) { | 67 | if ((address & ~Memory::VRAM_MASK) == Memory::VRAM_PADDR) { |
| 58 | return FRAMEBUFFER_LOCATION_VRAM; | 68 | return FRAMEBUFFER_LOCATION_VRAM; |
| 59 | } else if ((g_regs.framebuffer_top_right_1 & ~Memory::FCRAM_MASK) == Memory::FCRAM_PADDR) { | 69 | } else if ((address & ~Memory::FCRAM_MASK) == Memory::FCRAM_PADDR) { |
| 60 | return FRAMEBUFFER_LOCATION_FCRAM; | 70 | return FRAMEBUFFER_LOCATION_FCRAM; |
| 61 | } else { | 71 | } else { |
| 62 | ERROR_LOG(GPU, "unknown framebuffer location!"); | 72 | ERROR_LOG(GPU, "unknown framebuffer location!"); |
| @@ -64,91 +74,161 @@ const FramebufferLocation GetFramebufferLocation() { | |||
| 64 | return FRAMEBUFFER_LOCATION_UNKNOWN; | 74 | return FRAMEBUFFER_LOCATION_UNKNOWN; |
| 65 | } | 75 | } |
| 66 | 76 | ||
| 77 | u32 GetFramebufferAddr(const u32 address) { | ||
| 78 | switch (GetFramebufferLocation(address)) { | ||
| 79 | case FRAMEBUFFER_LOCATION_FCRAM: | ||
| 80 | return Memory::VirtualAddressFromPhysical_FCRAM(address); | ||
| 81 | case FRAMEBUFFER_LOCATION_VRAM: | ||
| 82 | return Memory::VirtualAddressFromPhysical_VRAM(address); | ||
| 83 | default: | ||
| 84 | ERROR_LOG(GPU, "unknown framebuffer location"); | ||
| 85 | } | ||
| 86 | return 0; | ||
| 87 | } | ||
| 88 | |||
| 67 | /** | 89 | /** |
| 68 | * Gets a read-only pointer to a framebuffer in memory | 90 | * Gets a read-only pointer to a framebuffer in memory |
| 69 | * @param address Physical address of framebuffer | 91 | * @param address Physical address of framebuffer |
| 70 | * @return Returns const pointer to raw framebuffer | 92 | * @return Returns const pointer to raw framebuffer |
| 71 | */ | 93 | */ |
| 72 | const u8* GetFramebufferPointer(const u32 address) { | 94 | const u8* GetFramebufferPointer(const u32 address) { |
| 73 | switch (GetFramebufferLocation()) { | 95 | u32 addr = GetFramebufferAddr(address); |
| 74 | case FRAMEBUFFER_LOCATION_FCRAM: | 96 | return (addr != 0) ? Memory::GetPointer(addr) : nullptr; |
| 75 | return (const u8*)Memory::GetPointer(Memory::VirtualAddressFromPhysical_FCRAM(address)); | ||
| 76 | case FRAMEBUFFER_LOCATION_VRAM: | ||
| 77 | return (const u8*)Memory::GetPointer(Memory::VirtualAddressFromPhysical_VRAM(address)); | ||
| 78 | default: | ||
| 79 | ERROR_LOG(GPU, "unknown framebuffer location"); | ||
| 80 | } | ||
| 81 | return NULL; | ||
| 82 | } | 97 | } |
| 83 | 98 | ||
| 84 | template <typename T> | 99 | template <typename T> |
| 85 | inline void Read(T &var, const u32 addr) { | 100 | inline void Read(T &var, const u32 raw_addr) { |
| 86 | switch (addr) { | 101 | u32 addr = raw_addr - 0x1EF00000; |
| 87 | case Registers::FramebufferTopLeft1: | 102 | int index = addr / 4; |
| 88 | var = g_regs.framebuffer_top_left_1; | ||
| 89 | break; | ||
| 90 | 103 | ||
| 91 | case Registers::FramebufferTopLeft2: | 104 | // Reads other than u32 are untested, so I'd rather have them abort than silently fail |
| 92 | var = g_regs.framebuffer_top_left_2; | 105 | if (index >= Regs::NumIds || !std::is_same<T,u32>::value) |
| 93 | break; | 106 | { |
| 107 | ERROR_LOG(GPU, "unknown Read%d @ 0x%08X", sizeof(var) * 8, addr); | ||
| 108 | return; | ||
| 109 | } | ||
| 94 | 110 | ||
| 95 | case Registers::FramebufferTopRight1: | 111 | var = g_regs[static_cast<Regs::Id>(addr / 4)]; |
| 96 | var = g_regs.framebuffer_top_right_1; | 112 | } |
| 97 | break; | ||
| 98 | 113 | ||
| 99 | case Registers::FramebufferTopRight2: | 114 | template <typename T> |
| 100 | var = g_regs.framebuffer_top_right_2; | 115 | inline void Write(u32 addr, const T data) { |
| 101 | break; | 116 | addr -= 0x1EF00000; |
| 117 | int index = addr / 4; | ||
| 102 | 118 | ||
| 103 | case Registers::FramebufferSubLeft1: | 119 | // Writes other than u32 are untested, so I'd rather have them abort than silently fail |
| 104 | var = g_regs.framebuffer_sub_left_1; | 120 | if (index >= Regs::NumIds || !std::is_same<T,u32>::value) |
| 105 | break; | 121 | { |
| 122 | ERROR_LOG(GPU, "unknown Write%d 0x%08X @ 0x%08X", sizeof(data) * 8, data, addr); | ||
| 123 | return; | ||
| 124 | } | ||
| 106 | 125 | ||
| 107 | case Registers::FramebufferSubRight1: | 126 | g_regs[static_cast<Regs::Id>(index)] = data; |
| 108 | var = g_regs.framebuffer_sub_right_1; | ||
| 109 | break; | ||
| 110 | 127 | ||
| 111 | case Registers::CommandListSize: | 128 | switch (static_cast<Regs::Id>(index)) { |
| 112 | var = g_regs.command_list_size; | ||
| 113 | break; | ||
| 114 | 129 | ||
| 115 | case Registers::CommandListAddress: | 130 | // Memory fills are triggered once the fill value is written. |
| 116 | var = g_regs.command_list_address; | 131 | // NOTE: This is not verified. |
| 117 | break; | 132 | case Regs::MemoryFill + 3: |
| 133 | case Regs::MemoryFill + 7: | ||
| 134 | { | ||
| 135 | const auto& config = g_regs.Get<Regs::MemoryFill>(static_cast<Regs::Id>(index - 3)); | ||
| 118 | 136 | ||
| 119 | case Registers::ProcessCommandList: | 137 | // TODO: Not sure if this check should be done at GSP level instead |
| 120 | var = g_regs.command_processing_enabled; | 138 | if (config.address_start) { |
| 121 | break; | 139 | // TODO: Not sure if this algorithm is correct, particularly because it doesn't use the size member at all |
| 140 | u32* start = (u32*)Memory::GetPointer(config.GetStartAddress()); | ||
| 141 | u32* end = (u32*)Memory::GetPointer(config.GetEndAddress()); | ||
| 142 | for (u32* ptr = start; ptr < end; ++ptr) | ||
| 143 | *ptr = bswap32(config.value); // TODO: This is just a workaround to missing framebuffer format emulation | ||
| 122 | 144 | ||
| 123 | default: | 145 | DEBUG_LOG(GPU, "MemoryFill from 0x%08x to 0x%08x", config.GetStartAddress(), config.GetEndAddress()); |
| 124 | ERROR_LOG(GPU, "unknown Read%d @ 0x%08X", sizeof(var) * 8, addr); | 146 | } |
| 125 | break; | 147 | break; |
| 126 | } | 148 | } |
| 127 | } | ||
| 128 | |||
| 129 | template <typename T> | ||
| 130 | inline void Write(u32 addr, const T data) { | ||
| 131 | switch (static_cast<Registers::Id>(addr)) { | ||
| 132 | case Registers::CommandListSize: | ||
| 133 | g_regs.command_list_size = data; | ||
| 134 | break; | ||
| 135 | 149 | ||
| 136 | case Registers::CommandListAddress: | 150 | case Regs::DisplayTransfer + 6: |
| 137 | g_regs.command_list_address = data; | 151 | { |
| 152 | const auto& config = g_regs.Get<Regs::DisplayTransfer>(); | ||
| 153 | if (config.trigger & 1) { | ||
| 154 | u8* source_pointer = Memory::GetPointer(config.GetPhysicalInputAddress()); | ||
| 155 | u8* dest_pointer = Memory::GetPointer(config.GetPhysicalOutputAddress()); | ||
| 156 | |||
| 157 | for (int y = 0; y < config.output_height; ++y) { | ||
| 158 | // TODO: Why does the register seem to hold twice the framebuffer width? | ||
| 159 | for (int x = 0; x < config.output_width / 2; ++x) { | ||
| 160 | struct { | ||
| 161 | int r, g, b, a; | ||
| 162 | } source_color = { 0, 0, 0, 0 }; | ||
| 163 | |||
| 164 | switch (config.input_format) { | ||
| 165 | case Regs::FramebufferFormat::RGBA8: | ||
| 166 | { | ||
| 167 | // TODO: Most likely got the component order messed up. | ||
| 168 | u8* srcptr = source_pointer + x * 4 + y * config.input_width * 4 / 2; | ||
| 169 | source_color.r = srcptr[0]; // blue | ||
| 170 | source_color.g = srcptr[1]; // green | ||
| 171 | source_color.b = srcptr[2]; // red | ||
| 172 | source_color.a = srcptr[3]; // alpha | ||
| 173 | break; | ||
| 174 | } | ||
| 175 | |||
| 176 | default: | ||
| 177 | ERROR_LOG(GPU, "Unknown source framebuffer format %x", config.input_format.Value()); | ||
| 178 | break; | ||
| 179 | } | ||
| 180 | |||
| 181 | switch (config.output_format) { | ||
| 182 | /*case Regs::FramebufferFormat::RGBA8: | ||
| 183 | { | ||
| 184 | // TODO: Untested | ||
| 185 | u8* dstptr = (u32*)(dest_pointer + x * 4 + y * config.output_width * 4); | ||
| 186 | dstptr[0] = source_color.r; | ||
| 187 | dstptr[1] = source_color.g; | ||
| 188 | dstptr[2] = source_color.b; | ||
| 189 | dstptr[3] = source_color.a; | ||
| 190 | break; | ||
| 191 | }*/ | ||
| 192 | |||
| 193 | case Regs::FramebufferFormat::RGB8: | ||
| 194 | { | ||
| 195 | // TODO: Most likely got the component order messed up. | ||
| 196 | u8* dstptr = dest_pointer + x * 3 + y * config.output_width * 3 / 2; | ||
| 197 | dstptr[0] = source_color.r; // blue | ||
| 198 | dstptr[1] = source_color.g; // green | ||
| 199 | dstptr[2] = source_color.b; // red | ||
| 200 | break; | ||
| 201 | } | ||
| 202 | |||
| 203 | default: | ||
| 204 | ERROR_LOG(GPU, "Unknown destination framebuffer format %x", config.output_format.Value()); | ||
| 205 | break; | ||
| 206 | } | ||
| 207 | } | ||
| 208 | } | ||
| 209 | |||
| 210 | DEBUG_LOG(GPU, "DisplayTriggerTransfer: 0x%08x bytes from 0x%08x(%dx%d)-> 0x%08x(%dx%d), dst format %x", | ||
| 211 | config.output_height * config.output_width * 4, | ||
| 212 | config.GetPhysicalInputAddress(), (int)config.input_width, (int)config.input_height, | ||
| 213 | config.GetPhysicalOutputAddress(), (int)config.output_width, (int)config.output_height, | ||
| 214 | config.output_format.Value()); | ||
| 215 | } | ||
| 138 | break; | 216 | break; |
| 217 | } | ||
| 139 | 218 | ||
| 140 | case Registers::ProcessCommandList: | 219 | case Regs::CommandProcessor + 4: |
| 141 | g_regs.command_processing_enabled = data; | 220 | { |
| 142 | if (g_regs.command_processing_enabled & 1) | 221 | const auto& config = g_regs.Get<Regs::CommandProcessor>(); |
| 222 | if (config.trigger & 1) | ||
| 143 | { | 223 | { |
| 144 | // u32* buffer = (u32*)Memory::GetPointer(g_regs.command_list_address << 3); | 224 | // u32* buffer = (u32*)Memory::GetPointer(config.address << 3); |
| 145 | ERROR_LOG(GPU, "Beginning %x bytes of commands from address %x", g_regs.command_list_size, g_regs.command_list_address << 3); | 225 | ERROR_LOG(GPU, "Beginning 0x%08x bytes of commands from address 0x%08x", config.size, config.address << 3); |
| 146 | // TODO: Process command list! | 226 | // TODO: Process command list! |
| 147 | } | 227 | } |
| 148 | break; | 228 | break; |
| 229 | } | ||
| 149 | 230 | ||
| 150 | default: | 231 | default: |
| 151 | ERROR_LOG(GPU, "unknown Write%d 0x%08X @ 0x%08X", sizeof(data) * 8, data, addr); | ||
| 152 | break; | 232 | break; |
| 153 | } | 233 | } |
| 154 | } | 234 | } |
| @@ -180,7 +260,24 @@ void Update() { | |||
| 180 | /// Initialize hardware | 260 | /// Initialize hardware |
| 181 | void Init() { | 261 | void Init() { |
| 182 | g_last_ticks = Core::g_app_core->GetTicks(); | 262 | g_last_ticks = Core::g_app_core->GetTicks(); |
| 183 | SetFramebufferLocation(FRAMEBUFFER_LOCATION_FCRAM); | 263 | // SetFramebufferLocation(FRAMEBUFFER_LOCATION_FCRAM); |
| 264 | SetFramebufferLocation(FRAMEBUFFER_LOCATION_VRAM); | ||
| 265 | |||
| 266 | auto& framebuffer_top = g_regs.Get<Regs::FramebufferTop>(); | ||
| 267 | auto& framebuffer_sub = g_regs.Get<Regs::FramebufferBottom>(); | ||
| 268 | // TODO: Width should be 240 instead? | ||
| 269 | framebuffer_top.width = 480; | ||
| 270 | framebuffer_top.height = 400; | ||
| 271 | framebuffer_top.stride = 480*3; | ||
| 272 | framebuffer_top.color_format = Regs::FramebufferFormat::RGB8; | ||
| 273 | framebuffer_top.active_fb = 0; | ||
| 274 | |||
| 275 | framebuffer_sub.width = 480; | ||
| 276 | framebuffer_sub.height = 400; | ||
| 277 | framebuffer_sub.stride = 480*3; | ||
| 278 | framebuffer_sub.color_format = Regs::FramebufferFormat::RGB8; | ||
| 279 | framebuffer_sub.active_fb = 0; | ||
| 280 | |||
| 184 | NOTICE_LOG(GPU, "initialized OK"); | 281 | NOTICE_LOG(GPU, "initialized OK"); |
| 185 | } | 282 | } |
| 186 | 283 | ||
diff --git a/src/core/hw/gpu.h b/src/core/hw/gpu.h index 3314ba989..42f18a0e7 100644 --- a/src/core/hw/gpu.h +++ b/src/core/hw/gpu.h | |||
| @@ -5,43 +5,168 @@ | |||
| 5 | #pragma once | 5 | #pragma once |
| 6 | 6 | ||
| 7 | #include "common/common_types.h" | 7 | #include "common/common_types.h" |
| 8 | #include "common/bit_field.h" | ||
| 9 | #include "common/register_set.h" | ||
| 8 | 10 | ||
| 9 | namespace GPU { | 11 | namespace GPU { |
| 10 | 12 | ||
| 11 | static const u32 kFrameCycles = 268123480 / 60; ///< 268MHz / 60 frames per second | 13 | static const u32 kFrameCycles = 268123480 / 60; ///< 268MHz / 60 frames per second |
| 12 | static const u32 kFrameTicks = kFrameCycles / 3; ///< Approximate number of instructions/frame | 14 | static const u32 kFrameTicks = kFrameCycles / 3; ///< Approximate number of instructions/frame |
| 13 | 15 | ||
| 14 | struct Registers { | 16 | // MMIO region 0x1EFxxxxx |
| 17 | struct Regs { | ||
| 15 | enum Id : u32 { | 18 | enum Id : u32 { |
| 16 | FramebufferTopLeft1 = 0x1EF00468, // Main LCD, first framebuffer for 3D left | 19 | MemoryFill = 0x00004, // + 5,6,7; second block at 8-11 |
| 17 | FramebufferTopLeft2 = 0x1EF0046C, // Main LCD, second framebuffer for 3D left | 20 | |
| 18 | FramebufferTopRight1 = 0x1EF00494, // Main LCD, first framebuffer for 3D right | 21 | FramebufferTop = 0x00117, // + 11a,11b,11c,11d(?),11e...126 |
| 19 | FramebufferTopRight2 = 0x1EF00498, // Main LCD, second framebuffer for 3D right | 22 | FramebufferBottom = 0x00157, // + 15a,15b,15c,15d(?),15e...166 |
| 20 | FramebufferSubLeft1 = 0x1EF00568, // Sub LCD, first framebuffer | 23 | |
| 21 | FramebufferSubLeft2 = 0x1EF0056C, // Sub LCD, second framebuffer | 24 | DisplayTransfer = 0x00300, // + 301,302,303,304,305,306 |
| 22 | FramebufferSubRight1 = 0x1EF00594, // Sub LCD, unused first framebuffer | 25 | |
| 23 | FramebufferSubRight2 = 0x1EF00598, // Sub LCD, unused second framebuffer | 26 | CommandProcessor = 0x00638, // + 63a,63c |
| 24 | 27 | ||
| 25 | CommandListSize = 0x1EF018E0, | 28 | NumIds = 0x01000 |
| 26 | CommandListAddress = 0x1EF018E8, | 29 | }; |
| 27 | ProcessCommandList = 0x1EF018F0, | 30 | |
| 31 | template<Id id> | ||
| 32 | struct Struct; | ||
| 33 | |||
| 34 | enum class FramebufferFormat : u32 { | ||
| 35 | RGBA8 = 0, | ||
| 36 | RGB8 = 1, | ||
| 37 | RGB565 = 2, | ||
| 38 | RGB5A1 = 3, | ||
| 39 | RGBA4 = 4, | ||
| 40 | }; | ||
| 41 | }; | ||
| 42 | |||
| 43 | template<> | ||
| 44 | struct Regs::Struct<Regs::MemoryFill> { | ||
| 45 | u32 address_start; | ||
| 46 | u32 address_end; // ? | ||
| 47 | u32 size; | ||
| 48 | u32 value; // ? | ||
| 49 | |||
| 50 | inline u32 GetStartAddress() const { | ||
| 51 | return address_start * 8; | ||
| 52 | } | ||
| 53 | |||
| 54 | inline u32 GetEndAddress() const { | ||
| 55 | return address_end * 8; | ||
| 56 | } | ||
| 57 | }; | ||
| 58 | static_assert(sizeof(Regs::Struct<Regs::MemoryFill>) == 0x10, "Structure size and register block length don't match"); | ||
| 59 | |||
| 60 | template<> | ||
| 61 | struct Regs::Struct<Regs::FramebufferTop> { | ||
| 62 | using Format = Regs::FramebufferFormat; | ||
| 63 | |||
| 64 | union { | ||
| 65 | u32 size; | ||
| 66 | |||
| 67 | BitField< 0, 16, u32> width; | ||
| 68 | BitField<16, 16, u32> height; | ||
| 69 | }; | ||
| 70 | |||
| 71 | u32 pad0[2]; | ||
| 72 | |||
| 73 | u32 address_left1; | ||
| 74 | u32 address_left2; | ||
| 75 | |||
| 76 | union { | ||
| 77 | u32 format; | ||
| 78 | |||
| 79 | BitField< 0, 3, Format> color_format; | ||
| 80 | }; | ||
| 81 | |||
| 82 | u32 pad1; | ||
| 83 | |||
| 84 | union { | ||
| 85 | u32 active_fb; | ||
| 86 | |||
| 87 | // 0: Use parameters ending with "1" | ||
| 88 | // 1: Use parameters ending with "2" | ||
| 89 | BitField<0, 1, u32> second_fb_active; | ||
| 90 | }; | ||
| 91 | |||
| 92 | u32 pad2[5]; | ||
| 93 | |||
| 94 | // Distance between two pixel rows, in bytes | ||
| 95 | u32 stride; | ||
| 96 | |||
| 97 | u32 address_right1; | ||
| 98 | u32 address_right2; | ||
| 99 | }; | ||
| 100 | |||
| 101 | template<> | ||
| 102 | struct Regs::Struct<Regs::FramebufferBottom> : public Regs::Struct<Regs::FramebufferTop> { | ||
| 103 | }; | ||
| 104 | static_assert(sizeof(Regs::Struct<Regs::FramebufferTop>) == 0x40, "Structure size and register block length don't match"); | ||
| 105 | |||
| 106 | template<> | ||
| 107 | struct Regs::Struct<Regs::DisplayTransfer> { | ||
| 108 | using Format = Regs::FramebufferFormat; | ||
| 109 | |||
| 110 | u32 input_address; | ||
| 111 | u32 output_address; | ||
| 112 | |||
| 113 | inline u32 GetPhysicalInputAddress() const { | ||
| 114 | return input_address * 8; | ||
| 115 | } | ||
| 116 | |||
| 117 | inline u32 GetPhysicalOutputAddress() const { | ||
| 118 | return output_address * 8; | ||
| 119 | } | ||
| 120 | |||
| 121 | union { | ||
| 122 | u32 output_size; | ||
| 123 | |||
| 124 | BitField< 0, 16, u32> output_width; | ||
| 125 | BitField<16, 16, u32> output_height; | ||
| 126 | }; | ||
| 127 | |||
| 128 | union { | ||
| 129 | u32 input_size; | ||
| 130 | |||
| 131 | BitField< 0, 16, u32> input_width; | ||
| 132 | BitField<16, 16, u32> input_height; | ||
| 28 | }; | 133 | }; |
| 29 | 134 | ||
| 30 | u32 framebuffer_top_left_1; | 135 | union { |
| 31 | u32 framebuffer_top_left_2; | 136 | u32 flags; |
| 32 | u32 framebuffer_top_right_1; | 137 | |
| 33 | u32 framebuffer_top_right_2; | 138 | BitField< 0, 1, u32> flip_data; // flips input data horizontally (TODO) if true |
| 34 | u32 framebuffer_sub_left_1; | 139 | BitField< 8, 3, Format> input_format; |
| 35 | u32 framebuffer_sub_left_2; | 140 | BitField<12, 3, Format> output_format; |
| 36 | u32 framebuffer_sub_right_1; | 141 | BitField<16, 1, u32> output_tiled; // stores output in a tiled format |
| 37 | u32 framebuffer_sub_right_2; | 142 | }; |
| 38 | 143 | ||
| 39 | u32 command_list_size; | 144 | u32 unknown; |
| 40 | u32 command_list_address; | 145 | |
| 41 | u32 command_processing_enabled; | 146 | // it seems that writing to this field triggers the display transfer |
| 147 | u32 trigger; | ||
| 42 | }; | 148 | }; |
| 149 | static_assert(sizeof(Regs::Struct<Regs::DisplayTransfer>) == 0x1C, "Structure size and register block length don't match"); | ||
| 150 | |||
| 151 | template<> | ||
| 152 | struct Regs::Struct<Regs::CommandProcessor> { | ||
| 153 | // command list size | ||
| 154 | u32 size; | ||
| 155 | |||
| 156 | u32 pad0; | ||
| 157 | |||
| 158 | // command list address | ||
| 159 | u32 address; | ||
| 43 | 160 | ||
| 44 | extern Registers g_regs; | 161 | u32 pad1; |
| 162 | |||
| 163 | // it seems that writing to this field triggers command list processing | ||
| 164 | u32 trigger; | ||
| 165 | }; | ||
| 166 | static_assert(sizeof(Regs::Struct<Regs::CommandProcessor>) == 0x14, "Structure size and register block length don't match"); | ||
| 167 | |||
| 168 | |||
| 169 | extern RegisterSet<u32, Regs> g_regs; | ||
| 45 | 170 | ||
| 46 | enum { | 171 | enum { |
| 47 | TOP_ASPECT_X = 0x5, | 172 | TOP_ASPECT_X = 0x5, |
| @@ -51,23 +176,35 @@ enum { | |||
| 51 | TOP_WIDTH = 400, | 176 | TOP_WIDTH = 400, |
| 52 | BOTTOM_WIDTH = 320, | 177 | BOTTOM_WIDTH = 320, |
| 53 | 178 | ||
| 54 | // Physical addresses in FCRAM used by ARM9 applications - these are correct for real hardware | 179 | // Physical addresses in FCRAM (chosen arbitrarily) |
| 55 | PADDR_FRAMEBUFFER_SEL = 0x20184E59, | 180 | PADDR_TOP_LEFT_FRAME1 = 0x201D4C00, |
| 56 | PADDR_TOP_LEFT_FRAME1 = 0x20184E60, | 181 | PADDR_TOP_LEFT_FRAME2 = 0x202D4C00, |
| 182 | PADDR_TOP_RIGHT_FRAME1 = 0x203D4C00, | ||
| 183 | PADDR_TOP_RIGHT_FRAME2 = 0x204D4C00, | ||
| 184 | PADDR_SUB_FRAME1 = 0x205D4C00, | ||
| 185 | PADDR_SUB_FRAME2 = 0x206D4C00, | ||
| 186 | // Physical addresses in FCRAM used by ARM9 applications | ||
| 187 | /* PADDR_TOP_LEFT_FRAME1 = 0x20184E60, | ||
| 57 | PADDR_TOP_LEFT_FRAME2 = 0x201CB370, | 188 | PADDR_TOP_LEFT_FRAME2 = 0x201CB370, |
| 58 | PADDR_TOP_RIGHT_FRAME1 = 0x20282160, | 189 | PADDR_TOP_RIGHT_FRAME1 = 0x20282160, |
| 59 | PADDR_TOP_RIGHT_FRAME2 = 0x202C8670, | 190 | PADDR_TOP_RIGHT_FRAME2 = 0x202C8670, |
| 60 | PADDR_SUB_FRAME1 = 0x202118E0, | 191 | PADDR_SUB_FRAME1 = 0x202118E0, |
| 61 | PADDR_SUB_FRAME2 = 0x20249CF0, | 192 | PADDR_SUB_FRAME2 = 0x20249CF0,*/ |
| 62 | 193 | ||
| 63 | // Physical addresses in VRAM - I'm not sure how these are actually allocated (so not real) | 194 | // Physical addresses in VRAM |
| 64 | PADDR_VRAM_FRAMEBUFFER_SEL = 0x18184E59, | 195 | // TODO: These should just be deduced from the ones above |
| 65 | PADDR_VRAM_TOP_LEFT_FRAME1 = 0x18184E60, | 196 | PADDR_VRAM_TOP_LEFT_FRAME1 = 0x181D4C00, |
| 66 | PADDR_VRAM_TOP_LEFT_FRAME2 = 0x181CB370, | 197 | PADDR_VRAM_TOP_LEFT_FRAME2 = 0x182D4C00, |
| 198 | PADDR_VRAM_TOP_RIGHT_FRAME1 = 0x183D4C00, | ||
| 199 | PADDR_VRAM_TOP_RIGHT_FRAME2 = 0x184D4C00, | ||
| 200 | PADDR_VRAM_SUB_FRAME1 = 0x185D4C00, | ||
| 201 | PADDR_VRAM_SUB_FRAME2 = 0x186D4C00, | ||
| 202 | // Physical addresses in VRAM used by ARM9 applications | ||
| 203 | /* PADDR_VRAM_TOP_LEFT_FRAME2 = 0x181CB370, | ||
| 67 | PADDR_VRAM_TOP_RIGHT_FRAME1 = 0x18282160, | 204 | PADDR_VRAM_TOP_RIGHT_FRAME1 = 0x18282160, |
| 68 | PADDR_VRAM_TOP_RIGHT_FRAME2 = 0x182C8670, | 205 | PADDR_VRAM_TOP_RIGHT_FRAME2 = 0x182C8670, |
| 69 | PADDR_VRAM_SUB_FRAME1 = 0x182118E0, | 206 | PADDR_VRAM_SUB_FRAME1 = 0x182118E0, |
| 70 | PADDR_VRAM_SUB_FRAME2 = 0x18249CF0, | 207 | PADDR_VRAM_SUB_FRAME2 = 0x18249CF0,*/ |
| 71 | }; | 208 | }; |
| 72 | 209 | ||
| 73 | /// Framebuffer location | 210 | /// Framebuffer location |
| @@ -79,7 +216,7 @@ enum FramebufferLocation { | |||
| 79 | 216 | ||
| 80 | /** | 217 | /** |
| 81 | * Sets whether the framebuffers are in the GSP heap (FCRAM) or VRAM | 218 | * Sets whether the framebuffers are in the GSP heap (FCRAM) or VRAM |
| 82 | * @param | 219 | * @param |
| 83 | */ | 220 | */ |
| 84 | void SetFramebufferLocation(const FramebufferLocation mode); | 221 | void SetFramebufferLocation(const FramebufferLocation mode); |
| 85 | 222 | ||
| @@ -90,16 +227,18 @@ void SetFramebufferLocation(const FramebufferLocation mode); | |||
| 90 | */ | 227 | */ |
| 91 | const u8* GetFramebufferPointer(const u32 address); | 228 | const u8* GetFramebufferPointer(const u32 address); |
| 92 | 229 | ||
| 230 | u32 GetFramebufferAddr(const u32 address); | ||
| 231 | |||
| 93 | /** | 232 | /** |
| 94 | * Gets the location of the framebuffers | 233 | * Gets the location of the framebuffers |
| 95 | */ | 234 | */ |
| 96 | const FramebufferLocation GetFramebufferLocation(); | 235 | FramebufferLocation GetFramebufferLocation(u32 address); |
| 97 | 236 | ||
| 98 | template <typename T> | 237 | template <typename T> |
| 99 | inline void Read(T &var, const u32 addr); | 238 | void Read(T &var, const u32 addr); |
| 100 | 239 | ||
| 101 | template <typename T> | 240 | template <typename T> |
| 102 | inline void Write(u32 addr, const T data); | 241 | void Write(u32 addr, const T data); |
| 103 | 242 | ||
| 104 | /// Update hardware | 243 | /// Update hardware |
| 105 | void Update(); | 244 | void Update(); |
diff --git a/src/core/hw/hw.h b/src/core/hw/hw.h index 92e9304ca..1055ed94f 100644 --- a/src/core/hw/hw.h +++ b/src/core/hw/hw.h | |||
| @@ -9,10 +9,10 @@ | |||
| 9 | namespace HW { | 9 | namespace HW { |
| 10 | 10 | ||
| 11 | template <typename T> | 11 | template <typename T> |
| 12 | inline void Read(T &var, const u32 addr); | 12 | void Read(T &var, const u32 addr); |
| 13 | 13 | ||
| 14 | template <typename T> | 14 | template <typename T> |
| 15 | inline void Write(u32 addr, const T data); | 15 | void Write(u32 addr, const T data); |
| 16 | 16 | ||
| 17 | /// Update hardware | 17 | /// Update hardware |
| 18 | void Update(); | 18 | void Update(); |
diff --git a/src/video_core/gpu_debugger.h b/src/video_core/gpu_debugger.h index 5d909beba..d92ceaa72 100644 --- a/src/video_core/gpu_debugger.h +++ b/src/video_core/gpu_debugger.h | |||
| @@ -50,7 +50,7 @@ public: | |||
| 50 | virtual void GXCommandProcessed(int total_command_count) | 50 | virtual void GXCommandProcessed(int total_command_count) |
| 51 | { | 51 | { |
| 52 | const GSP_GPU::GXCommand& cmd = observed->ReadGXCommandHistory(total_command_count-1); | 52 | const GSP_GPU::GXCommand& cmd = observed->ReadGXCommandHistory(total_command_count-1); |
| 53 | ERROR_LOG(GSP, "Received command: id=%x", cmd.id); | 53 | ERROR_LOG(GSP, "Received command: id=%x", (int)cmd.id.Value()); |
| 54 | } | 54 | } |
| 55 | 55 | ||
| 56 | /** | 56 | /** |
| @@ -78,11 +78,13 @@ public: | |||
| 78 | 78 | ||
| 79 | void GXCommandProcessed(u8* command_data) | 79 | void GXCommandProcessed(u8* command_data) |
| 80 | { | 80 | { |
| 81 | if (observers.empty()) | ||
| 82 | return; | ||
| 83 | |||
| 81 | gx_command_history.push_back(GSP_GPU::GXCommand()); | 84 | gx_command_history.push_back(GSP_GPU::GXCommand()); |
| 82 | GSP_GPU::GXCommand& cmd = gx_command_history[gx_command_history.size()-1]; | 85 | GSP_GPU::GXCommand& cmd = gx_command_history[gx_command_history.size()-1]; |
| 83 | 86 | ||
| 84 | const int cmd_length = sizeof(GSP_GPU::GXCommand); | 87 | memcpy(&cmd, command_data, sizeof(GSP_GPU::GXCommand)); |
| 85 | memcpy(cmd.data, command_data, cmd_length); | ||
| 86 | 88 | ||
| 87 | ForEachObserver([this](DebuggerObserver* observer) { | 89 | ForEachObserver([this](DebuggerObserver* observer) { |
| 88 | observer->GXCommandProcessed(this->gx_command_history.size()); | 90 | observer->GXCommandProcessed(this->gx_command_history.size()); |
| @@ -91,6 +93,9 @@ public: | |||
| 91 | 93 | ||
| 92 | void CommandListCalled(u32 address, u32* command_list, u32 size_in_words) | 94 | void CommandListCalled(u32 address, u32* command_list, u32 size_in_words) |
| 93 | { | 95 | { |
| 96 | if (observers.empty()) | ||
| 97 | return; | ||
| 98 | |||
| 94 | PicaCommandList cmdlist; | 99 | PicaCommandList cmdlist; |
| 95 | for (u32* parse_pointer = command_list; parse_pointer < command_list + size_in_words;) | 100 | for (u32* parse_pointer = command_list; parse_pointer < command_list + size_in_words;) |
| 96 | { | 101 | { |
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 70af47c59..d0a8ec1da 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp | |||
| @@ -12,8 +12,8 @@ | |||
| 12 | 12 | ||
| 13 | /// RendererOpenGL constructor | 13 | /// RendererOpenGL constructor |
| 14 | RendererOpenGL::RendererOpenGL() { | 14 | RendererOpenGL::RendererOpenGL() { |
| 15 | memset(m_fbo, 0, sizeof(m_fbo)); | 15 | memset(m_fbo, 0, sizeof(m_fbo)); |
| 16 | memset(m_fbo_rbo, 0, sizeof(m_fbo_rbo)); | 16 | memset(m_fbo_rbo, 0, sizeof(m_fbo_rbo)); |
| 17 | memset(m_fbo_depth_buffers, 0, sizeof(m_fbo_depth_buffers)); | 17 | memset(m_fbo_depth_buffers, 0, sizeof(m_fbo_depth_buffers)); |
| 18 | 18 | ||
| 19 | m_resolution_width = max(VideoCore::kScreenTopWidth, VideoCore::kScreenBottomWidth); | 19 | m_resolution_width = max(VideoCore::kScreenTopWidth, VideoCore::kScreenBottomWidth); |
| @@ -35,7 +35,7 @@ void RendererOpenGL::SwapBuffers() { | |||
| 35 | m_render_window->MakeCurrent(); | 35 | m_render_window->MakeCurrent(); |
| 36 | 36 | ||
| 37 | // EFB->XFB copy | 37 | // EFB->XFB copy |
| 38 | // TODO(bunnei): This is a hack and does not belong here. The copy should be triggered by some | 38 | // TODO(bunnei): This is a hack and does not belong here. The copy should be triggered by some |
| 39 | // register write We're also treating both framebuffers as a single one in OpenGL. | 39 | // register write We're also treating both framebuffers as a single one in OpenGL. |
| 40 | common::Rect framebuffer_size(0, 0, m_resolution_width, m_resolution_height); | 40 | common::Rect framebuffer_size(0, 0, m_resolution_width, m_resolution_height); |
| 41 | RenderXFB(framebuffer_size, framebuffer_size); | 41 | RenderXFB(framebuffer_size, framebuffer_size); |
| @@ -61,24 +61,40 @@ void RendererOpenGL::FlipFramebuffer(const u8* in, u8* out) { | |||
| 61 | int in_coord = 0; | 61 | int in_coord = 0; |
| 62 | for (int x = 0; x < VideoCore::kScreenTopWidth; x++) { | 62 | for (int x = 0; x < VideoCore::kScreenTopWidth; x++) { |
| 63 | for (int y = VideoCore::kScreenTopHeight-1; y >= 0; y--) { | 63 | for (int y = VideoCore::kScreenTopHeight-1; y >= 0; y--) { |
| 64 | // TODO: Properly support other framebuffer formats | ||
| 64 | int out_coord = (x + y * VideoCore::kScreenTopWidth) * 3; | 65 | int out_coord = (x + y * VideoCore::kScreenTopWidth) * 3; |
| 65 | out[out_coord] = in[in_coord]; | 66 | out[out_coord] = in[in_coord]; // blue? |
| 66 | out[out_coord + 1] = in[in_coord + 1]; | 67 | out[out_coord + 1] = in[in_coord + 1]; // green? |
| 67 | out[out_coord + 2] = in[in_coord + 2]; | 68 | out[out_coord + 2] = in[in_coord + 2]; // red? |
| 68 | in_coord+=3; | 69 | in_coord+=3; |
| 69 | } | 70 | } |
| 70 | } | 71 | } |
| 71 | } | 72 | } |
| 72 | 73 | ||
| 73 | /** | 74 | /** |
| 74 | * Renders external framebuffer (XFB) | 75 | * Renders external framebuffer (XFB) |
| 75 | * @param src_rect Source rectangle in XFB to copy | 76 | * @param src_rect Source rectangle in XFB to copy |
| 76 | * @param dst_rect Destination rectangle in output framebuffer to copy to | 77 | * @param dst_rect Destination rectangle in output framebuffer to copy to |
| 77 | */ | 78 | */ |
| 78 | void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& dst_rect) { | 79 | void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& dst_rect) { |
| 79 | 80 | ||
| 80 | FlipFramebuffer(GPU::GetFramebufferPointer(GPU::g_regs.framebuffer_top_left_1), m_xfb_top_flipped); | 81 | const auto& framebuffer_top = GPU::g_regs.Get<GPU::Regs::FramebufferTop>(); |
| 81 | FlipFramebuffer(GPU::GetFramebufferPointer(GPU::g_regs.framebuffer_sub_left_1), m_xfb_bottom_flipped); | 82 | const auto& framebuffer_sub = GPU::g_regs.Get<GPU::Regs::FramebufferBottom>(); |
| 83 | const u32 active_fb_top = (framebuffer_top.active_fb == 1) | ||
| 84 | ? framebuffer_top.address_left2 | ||
| 85 | : framebuffer_top.address_left1; | ||
| 86 | const u32 active_fb_sub = (framebuffer_sub.active_fb == 1) | ||
| 87 | ? framebuffer_sub.address_left2 | ||
| 88 | : framebuffer_sub.address_left1; | ||
| 89 | |||
| 90 | DEBUG_LOG(GPU, "RenderXFB: 0x%08x bytes from 0x%08x(%dx%d), fmt %x", | ||
| 91 | framebuffer_top.stride * framebuffer_top.height, | ||
| 92 | GPU::GetFramebufferAddr(active_fb_top), (int)framebuffer_top.width, | ||
| 93 | (int)framebuffer_top.height, (int)framebuffer_top.format); | ||
| 94 | |||
| 95 | // TODO: This should consider the GPU registers for framebuffer width, height and stride. | ||
| 96 | FlipFramebuffer(GPU::GetFramebufferPointer(active_fb_top), m_xfb_top_flipped); | ||
| 97 | FlipFramebuffer(GPU::GetFramebufferPointer(active_fb_sub), m_xfb_bottom_flipped); | ||
| 82 | 98 | ||
| 83 | // Blit the top framebuffer | 99 | // Blit the top framebuffer |
| 84 | // ------------------------ | 100 | // ------------------------ |
| @@ -98,7 +114,7 @@ void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& | |||
| 98 | glReadBuffer(GL_COLOR_ATTACHMENT0); | 114 | glReadBuffer(GL_COLOR_ATTACHMENT0); |
| 99 | 115 | ||
| 100 | // Blit | 116 | // Blit |
| 101 | glBlitFramebuffer(src_rect.x0_, src_rect.y0_, src_rect.x1_, src_rect.y1_, | 117 | glBlitFramebuffer(src_rect.x0_, src_rect.y0_, src_rect.x1_, src_rect.y1_, |
| 102 | dst_rect.x0_, dst_rect.y1_, dst_rect.x1_, dst_rect.y0_, | 118 | dst_rect.x0_, dst_rect.y1_, dst_rect.x1_, dst_rect.y0_, |
| 103 | GL_COLOR_BUFFER_BIT, GL_LINEAR); | 119 | GL_COLOR_BUFFER_BIT, GL_LINEAR); |
| 104 | 120 | ||
| @@ -110,7 +126,7 @@ void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& | |||
| 110 | // Update textures with contents of XFB in RAM - bottom | 126 | // Update textures with contents of XFB in RAM - bottom |
| 111 | glBindTexture(GL_TEXTURE_2D, m_xfb_texture_bottom); | 127 | glBindTexture(GL_TEXTURE_2D, m_xfb_texture_bottom); |
| 112 | glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, VideoCore::kScreenTopWidth, VideoCore::kScreenTopHeight, | 128 | glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, VideoCore::kScreenTopWidth, VideoCore::kScreenTopHeight, |
| 113 | GL_RGB, GL_UNSIGNED_BYTE, m_xfb_bottom_flipped); | 129 | GL_BGR, GL_UNSIGNED_BYTE, m_xfb_bottom_flipped); |
| 114 | glBindTexture(GL_TEXTURE_2D, 0); | 130 | glBindTexture(GL_TEXTURE_2D, 0); |
| 115 | 131 | ||
| 116 | // Render target is destination framebuffer | 132 | // Render target is destination framebuffer |
| @@ -124,7 +140,7 @@ void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& | |||
| 124 | 140 | ||
| 125 | // Blit | 141 | // Blit |
| 126 | int offset = (VideoCore::kScreenTopWidth - VideoCore::kScreenBottomWidth) / 2; | 142 | int offset = (VideoCore::kScreenTopWidth - VideoCore::kScreenBottomWidth) / 2; |
| 127 | glBlitFramebuffer(0,0, VideoCore::kScreenBottomWidth, VideoCore::kScreenBottomHeight, | 143 | glBlitFramebuffer(0,0, VideoCore::kScreenBottomWidth, VideoCore::kScreenBottomHeight, |
| 128 | offset, VideoCore::kScreenBottomHeight, VideoCore::kScreenBottomWidth + offset, 0, | 144 | offset, VideoCore::kScreenBottomHeight, VideoCore::kScreenBottomWidth + offset, 0, |
| 129 | GL_COLOR_BUFFER_BIT, GL_LINEAR); | 145 | GL_COLOR_BUFFER_BIT, GL_LINEAR); |
| 130 | 146 | ||
| @@ -133,7 +149,7 @@ void RendererOpenGL::RenderXFB(const common::Rect& src_rect, const common::Rect& | |||
| 133 | 149 | ||
| 134 | /// Initialize the FBO | 150 | /// Initialize the FBO |
| 135 | void RendererOpenGL::InitFramebuffer() { | 151 | void RendererOpenGL::InitFramebuffer() { |
| 136 | // TODO(bunnei): This should probably be implemented with the top screen and bottom screen as | 152 | // TODO(bunnei): This should probably be implemented with the top screen and bottom screen as |
| 137 | // separate framebuffers | 153 | // separate framebuffers |
| 138 | 154 | ||
| 139 | // Init the FBOs | 155 | // Init the FBOs |
| @@ -146,12 +162,12 @@ void RendererOpenGL::InitFramebuffer() { | |||
| 146 | for (int i = 0; i < kMaxFramebuffers; i++) { | 162 | for (int i = 0; i < kMaxFramebuffers; i++) { |
| 147 | // Generate color buffer storage | 163 | // Generate color buffer storage |
| 148 | glBindRenderbuffer(GL_RENDERBUFFER, m_fbo_rbo[i]); | 164 | glBindRenderbuffer(GL_RENDERBUFFER, m_fbo_rbo[i]); |
| 149 | glRenderbufferStorage(GL_RENDERBUFFER, GL_RGBA8, VideoCore::kScreenTopWidth, | 165 | glRenderbufferStorage(GL_RENDERBUFFER, GL_RGBA8, VideoCore::kScreenTopWidth, |
| 150 | VideoCore::kScreenTopHeight + VideoCore::kScreenBottomHeight); | 166 | VideoCore::kScreenTopHeight + VideoCore::kScreenBottomHeight); |
| 151 | 167 | ||
| 152 | // Generate depth buffer storage | 168 | // Generate depth buffer storage |
| 153 | glBindRenderbuffer(GL_RENDERBUFFER, m_fbo_depth_buffers[i]); | 169 | glBindRenderbuffer(GL_RENDERBUFFER, m_fbo_depth_buffers[i]); |
| 154 | glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT32, VideoCore::kScreenTopWidth, | 170 | glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT32, VideoCore::kScreenTopWidth, |
| 155 | VideoCore::kScreenTopHeight + VideoCore::kScreenBottomHeight); | 171 | VideoCore::kScreenTopHeight + VideoCore::kScreenBottomHeight); |
| 156 | 172 | ||
| 157 | // Attach the buffers | 173 | // Attach the buffers |
| @@ -167,7 +183,7 @@ void RendererOpenGL::InitFramebuffer() { | |||
| 167 | } else { | 183 | } else { |
| 168 | ERROR_LOG(RENDER, "couldn't create OpenGL frame buffer"); | 184 | ERROR_LOG(RENDER, "couldn't create OpenGL frame buffer"); |
| 169 | exit(1); | 185 | exit(1); |
| 170 | } | 186 | } |
| 171 | } | 187 | } |
| 172 | glBindFramebuffer(GL_FRAMEBUFFER, 0); // Unbind our frame buffer(s) | 188 | glBindFramebuffer(GL_FRAMEBUFFER, 0); // Unbind our frame buffer(s) |
| 173 | 189 | ||
| @@ -175,8 +191,8 @@ void RendererOpenGL::InitFramebuffer() { | |||
| 175 | // ------------------------------- | 191 | // ------------------------------- |
| 176 | 192 | ||
| 177 | // Create XFB textures | 193 | // Create XFB textures |
| 178 | glGenTextures(1, &m_xfb_texture_top); | 194 | glGenTextures(1, &m_xfb_texture_top); |
| 179 | glGenTextures(1, &m_xfb_texture_bottom); | 195 | glGenTextures(1, &m_xfb_texture_bottom); |
| 180 | 196 | ||
| 181 | // Alocate video memorry for XFB textures | 197 | // Alocate video memorry for XFB textures |
| 182 | glBindTexture(GL_TEXTURE_2D, m_xfb_texture_top); | 198 | glBindTexture(GL_TEXTURE_2D, m_xfb_texture_top); |
| @@ -192,13 +208,13 @@ void RendererOpenGL::InitFramebuffer() { | |||
| 192 | // Create the FBO and attach color/depth textures | 208 | // Create the FBO and attach color/depth textures |
| 193 | glGenFramebuffers(1, &m_xfb_top); // Generate framebuffer | 209 | glGenFramebuffers(1, &m_xfb_top); // Generate framebuffer |
| 194 | glBindFramebuffer(GL_DRAW_FRAMEBUFFER, m_xfb_top); | 210 | glBindFramebuffer(GL_DRAW_FRAMEBUFFER, m_xfb_top); |
| 195 | glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, | 211 | glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, |
| 196 | m_xfb_texture_top, 0); | 212 | m_xfb_texture_top, 0); |
| 197 | glBindFramebuffer(GL_FRAMEBUFFER, 0); | 213 | glBindFramebuffer(GL_FRAMEBUFFER, 0); |
| 198 | 214 | ||
| 199 | glGenFramebuffers(1, &m_xfb_bottom); // Generate framebuffer | 215 | glGenFramebuffers(1, &m_xfb_bottom); // Generate framebuffer |
| 200 | glBindFramebuffer(GL_DRAW_FRAMEBUFFER, m_xfb_bottom); | 216 | glBindFramebuffer(GL_DRAW_FRAMEBUFFER, m_xfb_bottom); |
| 201 | glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, | 217 | glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, |
| 202 | m_xfb_texture_bottom, 0); | 218 | m_xfb_texture_bottom, 0); |
| 203 | glBindFramebuffer(GL_FRAMEBUFFER, 0); | 219 | glBindFramebuffer(GL_FRAMEBUFFER, 0); |
| 204 | } | 220 | } |
| @@ -214,7 +230,7 @@ void RendererOpenGL::RenderFramebuffer() { | |||
| 214 | glReadBuffer(GL_COLOR_ATTACHMENT0); | 230 | glReadBuffer(GL_COLOR_ATTACHMENT0); |
| 215 | 231 | ||
| 216 | // Blit | 232 | // Blit |
| 217 | glBlitFramebuffer(0, 0, m_resolution_width, m_resolution_height, 0, 0, m_resolution_width, | 233 | glBlitFramebuffer(0, 0, m_resolution_width, m_resolution_height, 0, 0, m_resolution_width, |
| 218 | m_resolution_height, GL_COLOR_BUFFER_BIT, GL_LINEAR); | 234 | m_resolution_height, GL_COLOR_BUFFER_BIT, GL_LINEAR); |
| 219 | 235 | ||
| 220 | // Update the FPS count | 236 | // Update the FPS count |
| @@ -230,7 +246,7 @@ void RendererOpenGL::RenderFramebuffer() { | |||
| 230 | void RendererOpenGL::UpdateFramerate() { | 246 | void RendererOpenGL::UpdateFramerate() { |
| 231 | } | 247 | } |
| 232 | 248 | ||
| 233 | /** | 249 | /** |
| 234 | * Set the emulator window to use for renderer | 250 | * Set the emulator window to use for renderer |
| 235 | * @param window EmuWindow handle to emulator window to use for rendering | 251 | * @param window EmuWindow handle to emulator window to use for rendering |
| 236 | */ | 252 | */ |
| @@ -264,7 +280,7 @@ void RendererOpenGL::Init() { | |||
| 264 | 280 | ||
| 265 | GLenum err = glewInit(); | 281 | GLenum err = glewInit(); |
| 266 | if (GLEW_OK != err) { | 282 | if (GLEW_OK != err) { |
| 267 | ERROR_LOG(RENDER, "Failed to initialize GLEW! Error message: \"%s\". Exiting...", | 283 | ERROR_LOG(RENDER, "Failed to initialize GLEW! Error message: \"%s\". Exiting...", |
| 268 | glewGetErrorString(err)); | 284 | glewGetErrorString(err)); |
| 269 | exit(-1); | 285 | exit(-1); |
| 270 | } | 286 | } |
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index dd811cad6..30f4febe0 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h | |||
| @@ -84,7 +84,6 @@ private: | |||
| 84 | // "Flipped" framebuffers translate scanlines from native 3DS left-to-right to top-to-bottom | 84 | // "Flipped" framebuffers translate scanlines from native 3DS left-to-right to top-to-bottom |
| 85 | // as OpenGL expects them in a texture. There probably is a more efficient way of doing this: | 85 | // as OpenGL expects them in a texture. There probably is a more efficient way of doing this: |
| 86 | 86 | ||
| 87 | u8 m_xfb_top_flipped[VideoCore::kScreenTopWidth * VideoCore::kScreenTopWidth * 4]; | 87 | u8 m_xfb_top_flipped[VideoCore::kScreenTopWidth * VideoCore::kScreenTopHeight * 4]; |
| 88 | u8 m_xfb_bottom_flipped[VideoCore::kScreenTopWidth * VideoCore::kScreenTopWidth * 4]; | 88 | u8 m_xfb_bottom_flipped[VideoCore::kScreenBottomWidth * VideoCore::kScreenBottomHeight * 4]; |
| 89 | 89 | }; | |
| 90 | }; \ No newline at end of file | ||