From 2f3631914f1a88657e1485bbc9bfb12cc4ec5fc4 Mon Sep 17 00:00:00 2001 From: pineappleEA Date: Thu, 27 Jan 2022 22:36:15 +0100 Subject: [PATCH] early-access version 2439 --- README.md | 2 +- src/core/hle/kernel/k_priority_queue.h | 2 +- src/core/hle/kernel/k_scheduler.cpp | 45 ++++++++++--------- .../translate/impl/video_minimum_maximum.cpp | 18 +++----- src/video_core/buffer_cache/buffer_cache.h | 2 + 5 files changed, 33 insertions(+), 36 deletions(-) diff --git a/README.md b/README.md index f7ff7e9a3..a29f6fbe2 100755 --- a/README.md +++ b/README.md @@ -1,7 +1,7 @@ yuzu emulator early access ============= -This is the source code for early-access 2438. +This is the source code for early-access 2439. ## Legal Notice diff --git a/src/core/hle/kernel/k_priority_queue.h b/src/core/hle/kernel/k_priority_queue.h index 0b894c8cf..bd779739d 100755 --- a/src/core/hle/kernel/k_priority_queue.h +++ b/src/core/hle/kernel/k_priority_queue.h @@ -258,7 +258,7 @@ private: private: constexpr void ClearAffinityBit(u64& affinity, s32 core) { - affinity &= ~(u64(1) << core); + affinity &= ~(UINT64_C(1) << core); } constexpr s32 GetNextCore(u64& affinity) { diff --git a/src/core/hle/kernel/k_scheduler.cpp b/src/core/hle/kernel/k_scheduler.cpp index b32d4f285..c96520828 100755 --- a/src/core/hle/kernel/k_scheduler.cpp +++ b/src/core/hle/kernel/k_scheduler.cpp @@ -710,23 +710,19 @@ void KScheduler::Unload(KThread* thread) { } void KScheduler::Reload(KThread* thread) { - LOG_TRACE(Kernel, "core {}, reload thread {}", core_id, thread ? thread->GetName() : "nullptr"); + LOG_TRACE(Kernel, "core {}, reload thread {}", core_id, thread->GetName()); - if (thread) { - ASSERT_MSG(thread->GetState() == ThreadState::Runnable, "Thread must be runnable."); - - Core::ARM_Interface& cpu_core = system.ArmInterface(core_id); - cpu_core.LoadContext(thread->GetContext32()); - cpu_core.LoadContext(thread->GetContext64()); - cpu_core.SetTlsAddress(thread->GetTLSAddress()); - cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0()); - cpu_core.ClearExclusiveState(); - } + Core::ARM_Interface& cpu_core = system.ArmInterface(core_id); + cpu_core.LoadContext(thread->GetContext32()); + cpu_core.LoadContext(thread->GetContext64()); + cpu_core.SetTlsAddress(thread->GetTLSAddress()); + cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0()); + cpu_core.ClearExclusiveState(); } void KScheduler::SwitchContextStep2() { // Load context of new thread - Reload(current_thread.load()); + Reload(GetCurrentThread()); RescheduleCurrentCore(); } @@ -735,13 +731,17 @@ void KScheduler::ScheduleImpl() { KThread* previous_thread = GetCurrentThread(); KThread* next_thread = state.highest_priority_thread; - state.needs_scheduling = false; + state.needs_scheduling.store(false); // We never want to schedule a null thread, so use the idle thread if we don't have a next. if (next_thread == nullptr) { next_thread = idle_thread; } + if (next_thread->GetCurrentCore() != core_id) { + next_thread->SetCurrentCore(core_id); + } + // We never want to schedule a dummy thread, as these are only used by host threads for locking. if (next_thread->GetThreadType() == ThreadType::Dummy) { ASSERT_MSG(false, "Dummy threads should never be scheduled!"); @@ -755,14 +755,8 @@ void KScheduler::ScheduleImpl() { return; } - if (next_thread->GetCurrentCore() != core_id) { - next_thread->SetCurrentCore(core_id); - } - - current_thread.store(next_thread); - + // Update the CPU time tracking variables. KProcess* const previous_process = system.Kernel().CurrentProcess(); - UpdateLastContextSwitchTime(previous_thread, previous_process); // Save context for previous thread @@ -770,6 +764,10 @@ void KScheduler::ScheduleImpl() { std::shared_ptr* old_context; old_context = &previous_thread->GetHostContext(); + + // Set the new thread. + current_thread.store(next_thread); + guard.Unlock(); Common::Fiber::YieldTo(*old_context, *switch_fiber); @@ -797,8 +795,8 @@ void KScheduler::SwitchToCurrent() { do { auto next_thread = current_thread.load(); if (next_thread != nullptr) { - next_thread->context_guard.Lock(); - if (next_thread->GetRawState() != ThreadState::Runnable) { + const auto locked = next_thread->context_guard.TryLock(); + if (state.needs_scheduling.load()) { next_thread->context_guard.Unlock(); break; } @@ -806,6 +804,9 @@ void KScheduler::SwitchToCurrent() { next_thread->context_guard.Unlock(); break; } + if (!locked) { + continue; + } } auto thread = next_thread ? next_thread : idle_thread; Common::Fiber::YieldTo(switch_fiber, *thread->GetHostContext()); diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/video_minimum_maximum.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/video_minimum_maximum.cpp index 78869601f..4851b0b8d 100755 --- a/src/shader_recompiler/frontend/maxwell/translate/impl/video_minimum_maximum.cpp +++ b/src/shader_recompiler/frontend/maxwell/translate/impl/video_minimum_maximum.cpp @@ -57,16 +57,6 @@ void TranslatorVisitor::VMNMX(u64 insn) { if (vmnmx.sat != 0) { throw NotImplementedException("VMNMX SAT"); } - // Selectors were shown to default to 2 in unit tests - if (vmnmx.src_a_selector != 2) { - throw NotImplementedException("VMNMX Selector {}", vmnmx.src_a_selector.Value()); - } - if (vmnmx.src_b_selector != 2) { - throw NotImplementedException("VMNMX Selector {}", vmnmx.src_b_selector.Value()); - } - if (vmnmx.src_a_width != VideoWidth::Word) { - throw NotImplementedException("VMNMX Source Width {}", vmnmx.src_a_width.Value()); - } const bool is_b_imm{vmnmx.is_src_b_reg == 0}; const IR::U32 src_a{GetReg8(insn)}; @@ -76,10 +66,14 @@ void TranslatorVisitor::VMNMX(u64 insn) { const VideoWidth a_width{vmnmx.src_a_width}; const VideoWidth b_width{GetVideoSourceWidth(vmnmx.src_b_width, is_b_imm)}; + const u32 a_selector{static_cast(vmnmx.src_a_selector)}; + // Immediate values can't have a selector + const u32 b_selector{is_b_imm ? 0U : static_cast(vmnmx.src_b_selector)}; + const bool src_a_signed{vmnmx.src_a_sign != 0}; const bool src_b_signed{vmnmx.src_b_sign != 0}; - const IR::U32 op_a{ExtractVideoOperandValue(ir, src_a, a_width, 0, src_a_signed)}; - const IR::U32 op_b{ExtractVideoOperandValue(ir, src_b, b_width, 0, src_b_signed)}; + const IR::U32 op_a{ExtractVideoOperandValue(ir, src_a, a_width, a_selector, src_a_signed)}; + const IR::U32 op_b{ExtractVideoOperandValue(ir, src_b, b_width, b_selector, src_b_signed)}; // First operation's sign is only dependent on operand b's sign const bool op_1_signed{src_b_signed}; diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index e6cd73d46..d34a706df 100755 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -1500,6 +1500,8 @@ typename BufferCache

::OverlapResult BufferCache

::ResolveOverlaps(VAddr cpu // When this memory region has been joined a bunch of times, we assume it's being used // as a stream buffer. Increase the size to skip constantly recreating buffers. has_stream_leap = true; + begin -= PAGE_SIZE * 256; + cpu_addr = begin; end += PAGE_SIZE * 256; } }