diff options
Diffstat (limited to 'src/video_core/shader/decode/memory.cpp')
-rw-r--r-- | src/video_core/shader/decode/memory.cpp | 83 |
1 files changed, 44 insertions, 39 deletions
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp index 80fc0ccfc..7923d4d69 100644 --- a/src/video_core/shader/decode/memory.cpp +++ b/src/video_core/shader/decode/memory.cpp @@ -35,7 +35,7 @@ u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) { return 1; } } -} // namespace +} // Anonymous namespace u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; @@ -95,10 +95,10 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { const Node op_b = GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.GetOffset() + 4, index); - SetTemporal(bb, 0, op_a); - SetTemporal(bb, 1, op_b); - SetRegister(bb, instr.gpr0, GetTemporal(0)); - SetRegister(bb, instr.gpr0.Value() + 1, GetTemporal(1)); + SetTemporary(bb, 0, op_a); + SetTemporary(bb, 1, op_b); + SetRegister(bb, instr.gpr0, GetTemporary(0)); + SetRegister(bb, instr.gpr0.Value() + 1, GetTemporary(1)); break; } default: @@ -106,16 +106,17 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { } break; } - case OpCode::Id::LD_L: { - LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}", - static_cast<u64>(instr.ld_l.unknown.Value())); - - const auto GetLmem = [&](s32 offset) { + case OpCode::Id::LD_L: + LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}", static_cast<u64>(instr.ld_l.unknown)); + [[fallthrough]]; + case OpCode::Id::LD_S: { + const auto GetMemory = [&](s32 offset) { ASSERT(offset % 4 == 0); const Node immediate_offset = Immediate(static_cast<s32>(instr.smem_imm) + offset); const Node address = Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8), immediate_offset); - return GetLocalMemory(address); + return opcode->get().GetId() == OpCode::Id::LD_S ? GetSharedMemory(address) + : GetLocalMemory(address); }; switch (instr.ldst_sl.type.Value()) { @@ -135,14 +136,16 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { return 0; } }(); - for (u32 i = 0; i < count; ++i) - SetTemporal(bb, i, GetLmem(i * 4)); - for (u32 i = 0; i < count; ++i) - SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i)); + for (u32 i = 0; i < count; ++i) { + SetTemporary(bb, i, GetMemory(i * 4)); + } + for (u32 i = 0; i < count; ++i) { + SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); + } break; } default: - UNIMPLEMENTED_MSG("LD_L Unhandled type: {}", + UNIMPLEMENTED_MSG("{} Unhandled type: {}", opcode->get().GetName(), static_cast<u32>(instr.ldst_sl.type.Value())); } break; @@ -172,10 +175,10 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset); const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); - SetTemporal(bb, i, gmem); + SetTemporary(bb, i, gmem); } for (u32 i = 0; i < count; ++i) { - SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i)); + SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); } break; } @@ -209,27 +212,34 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { break; } - case OpCode::Id::ST_L: { + case OpCode::Id::ST_L: LOG_DEBUG(HW_GPU, "ST_L cache management mode: {}", static_cast<u64>(instr.st_l.cache_management.Value())); - - const auto GetLmemAddr = [&](s32 offset) { + [[fallthrough]]; + case OpCode::Id::ST_S: { + const auto GetAddress = [&](s32 offset) { ASSERT(offset % 4 == 0); const Node immediate = Immediate(static_cast<s32>(instr.smem_imm) + offset); return Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8), immediate); }; + const auto set_memory = opcode->get().GetId() == OpCode::Id::ST_L + ? &ShaderIR::SetLocalMemory + : &ShaderIR::SetSharedMemory; + switch (instr.ldst_sl.type.Value()) { case Tegra::Shader::StoreType::Bits128: - SetLocalMemory(bb, GetLmemAddr(12), GetRegister(instr.gpr0.Value() + 3)); - SetLocalMemory(bb, GetLmemAddr(8), GetRegister(instr.gpr0.Value() + 2)); + (this->*set_memory)(bb, GetAddress(12), GetRegister(instr.gpr0.Value() + 3)); + (this->*set_memory)(bb, GetAddress(8), GetRegister(instr.gpr0.Value() + 2)); + [[fallthrough]]; case Tegra::Shader::StoreType::Bits64: - SetLocalMemory(bb, GetLmemAddr(4), GetRegister(instr.gpr0.Value() + 1)); + (this->*set_memory)(bb, GetAddress(4), GetRegister(instr.gpr0.Value() + 1)); + [[fallthrough]]; case Tegra::Shader::StoreType::Bits32: - SetLocalMemory(bb, GetLmemAddr(0), GetRegister(instr.gpr0)); + (this->*set_memory)(bb, GetAddress(0), GetRegister(instr.gpr0)); break; default: - UNIMPLEMENTED_MSG("ST_L Unhandled type: {}", + UNIMPLEMENTED_MSG("{} unhandled type: {}", opcode->get().GetName(), static_cast<u32>(instr.ldst_sl.type.Value())); } break; @@ -253,11 +263,11 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { TrackAndGetGlobalMemory(bb, instr, true); // Encode in temporary registers like this: real_base_address, {registers_to_be_written...} - SetTemporal(bb, 0, real_address_base); + SetTemporary(bb, 0, real_address_base); const u32 count = GetUniformTypeElementsCount(type); for (u32 i = 0; i < count; ++i) { - SetTemporal(bb, i + 1, GetRegister(instr.gpr0.Value() + i)); + SetTemporary(bb, i + 1, GetRegister(instr.gpr0.Value() + i)); } for (u32 i = 0; i < count; ++i) { const Node it_offset = Immediate(i * 4); @@ -265,7 +275,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset); const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); - bb.push_back(Operation(OperationCode::Assign, gmem, GetTemporal(i + 1))); + bb.push_back(Operation(OperationCode::Assign, gmem, GetTemporary(i + 1))); } break; } @@ -297,18 +307,13 @@ std::tuple<Node, Node, GlobalMemoryBase> ShaderIR::TrackAndGetGlobalMemory(NodeB const auto addr_register{GetRegister(instr.gmem.gpr)}; const auto immediate_offset{static_cast<u32>(instr.gmem.offset)}; - const Node base_address{ - TrackCbuf(addr_register, global_code, static_cast<s64>(global_code.size()))}; - const auto cbuf = std::get_if<CbufNode>(&*base_address); - ASSERT(cbuf != nullptr); - const auto cbuf_offset_imm = std::get_if<ImmediateNode>(&*cbuf->GetOffset()); - ASSERT(cbuf_offset_imm != nullptr); - const auto cbuf_offset = cbuf_offset_imm->GetValue(); + const auto [base_address, index, offset] = + TrackCbuf(addr_register, global_code, static_cast<s64>(global_code.size())); + ASSERT(base_address != nullptr); - bb.push_back( - Comment(fmt::format("Base address is c[0x{:x}][0x{:x}]", cbuf->GetIndex(), cbuf_offset))); + bb.push_back(Comment(fmt::format("Base address is c[0x{:x}][0x{:x}]", index, offset))); - const GlobalMemoryBase descriptor{cbuf->GetIndex(), cbuf_offset}; + const GlobalMemoryBase descriptor{index, offset}; const auto& [entry, is_new] = used_global_memory.try_emplace(descriptor); auto& usage = entry->second; if (is_write) { |