From 8e56a84566036cfff0aa5c3d80ae1b051d2bd0bf Mon Sep 17 00:00:00 2001 From: Morph <39850852+Morph1984@users.noreply.github.com> Date: Sun, 23 Apr 2023 00:01:08 -0400 Subject: [PATCH] core_timing: Use CNTPCT as the guest CPU tick Previously, we were mixing the raw CPU frequency and CNTFRQ. The raw CPU frequency (1020 MHz) should've never been used as CNTPCT (whose frequency is CNTFRQ) is the only counter available. --- src/audio_core/renderer/adsp/adsp.cpp | 1 - .../renderer/adsp/audio_renderer.cpp | 5 +- .../renderer/adsp/command_list_processor.cpp | 1 - .../command/performance/performance.cpp | 15 ++--- src/audio_core/sink/sink_stream.cpp | 1 - src/common/wall_clock.h | 17 ++++++ src/core/CMakeLists.txt | 1 - src/core/core_timing.cpp | 35 +++-------- src/core/core_timing.h | 11 +--- src/core/core_timing_util.h | 58 ------------------- src/core/hle/kernel/k_scheduler.cpp | 5 +- src/core/hle/kernel/svc/svc_info.cpp | 4 +- src/core/hle/service/hid/hidbus.cpp | 1 - src/video_core/gpu.cpp | 14 ++--- 14 files changed, 47 insertions(+), 122 deletions(-) delete mode 100644 src/core/core_timing_util.h diff --git a/src/audio_core/renderer/adsp/adsp.cpp b/src/audio_core/renderer/adsp/adsp.cpp index 74772fc50c..b1db31e931 100644 --- a/src/audio_core/renderer/adsp/adsp.cpp +++ b/src/audio_core/renderer/adsp/adsp.cpp @@ -7,7 +7,6 @@ #include "common/logging/log.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" #include "core/memory.h" namespace AudioCore::AudioRenderer::ADSP { diff --git a/src/audio_core/renderer/adsp/audio_renderer.cpp b/src/audio_core/renderer/adsp/audio_renderer.cpp index 8bc39f9f9e..9ca716b604 100644 --- a/src/audio_core/renderer/adsp/audio_renderer.cpp +++ b/src/audio_core/renderer/adsp/audio_renderer.cpp @@ -13,7 +13,6 @@ #include "common/thread.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" MICROPROFILE_DEFINE(Audio_Renderer, "Audio", "DSP", MP_RGB(60, 19, 97)); @@ -144,6 +143,7 @@ void AudioRenderer::ThreadFunc(std::stop_token stop_token) { mailbox->ADSPSendMessage(RenderMessage::AudioRenderer_InitializeOK); + // 0.12 seconds (2304000 / 19200000) constexpr u64 max_process_time{2'304'000ULL}; while (!stop_token.stop_requested()) { @@ -184,8 +184,7 @@ void AudioRenderer::ThreadFunc(std::stop_token stop_token) { u64 max_time{max_process_time}; if (index == 1 && command_buffer.applet_resource_user_id == mailbox->GetCommandBuffer(0).applet_resource_user_id) { - max_time = max_process_time - - Core::Timing::CyclesToNs(render_times_taken[0]).count(); + max_time = max_process_time - render_times_taken[0]; if (render_times_taken[0] > max_process_time) { max_time = 0; } diff --git a/src/audio_core/renderer/adsp/command_list_processor.cpp b/src/audio_core/renderer/adsp/command_list_processor.cpp index 7a300d2162..3a0f1ae389 100644 --- a/src/audio_core/renderer/adsp/command_list_processor.cpp +++ b/src/audio_core/renderer/adsp/command_list_processor.cpp @@ -9,7 +9,6 @@ #include "common/settings.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" #include "core/memory.h" namespace AudioCore::AudioRenderer::ADSP { diff --git a/src/audio_core/renderer/command/performance/performance.cpp b/src/audio_core/renderer/command/performance/performance.cpp index 985958b039..4a881547f2 100644 --- a/src/audio_core/renderer/command/performance/performance.cpp +++ b/src/audio_core/renderer/command/performance/performance.cpp @@ -5,7 +5,6 @@ #include "audio_core/renderer/command/performance/performance.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" namespace AudioCore::AudioRenderer { @@ -18,20 +17,18 @@ void PerformanceCommand::Process(const ADSP::CommandListProcessor& processor) { auto base{entry_address.translated_address}; if (state == PerformanceState::Start) { auto start_time_ptr{reinterpret_cast(base + entry_address.entry_start_time_offset)}; - *start_time_ptr = static_cast( - Core::Timing::CyclesToUs(processor.system->CoreTiming().GetClockTicks() - - processor.start_time - processor.current_processing_time) - .count()); + *start_time_ptr = + static_cast(processor.system->CoreTiming().GetClockTicks() - processor.start_time - + processor.current_processing_time); } else if (state == PerformanceState::Stop) { auto processed_time_ptr{ reinterpret_cast(base + entry_address.entry_processed_time_offset)}; auto entry_count_ptr{ reinterpret_cast(base + entry_address.header_entry_count_offset)}; - *processed_time_ptr = static_cast( - Core::Timing::CyclesToUs(processor.system->CoreTiming().GetClockTicks() - - processor.start_time - processor.current_processing_time) - .count()); + *processed_time_ptr = + static_cast(processor.system->CoreTiming().GetClockTicks() - processor.start_time - + processor.current_processing_time); (*entry_count_ptr)++; } } diff --git a/src/audio_core/sink/sink_stream.cpp b/src/audio_core/sink/sink_stream.cpp index f44fedfd5e..9a718a9cc5 100644 --- a/src/audio_core/sink/sink_stream.cpp +++ b/src/audio_core/sink/sink_stream.cpp @@ -15,7 +15,6 @@ #include "common/settings.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" namespace AudioCore::Sink { diff --git a/src/common/wall_clock.h b/src/common/wall_clock.h index a73e6e6443..56c18ca250 100644 --- a/src/common/wall_clock.h +++ b/src/common/wall_clock.h @@ -38,6 +38,22 @@ public: /// @returns Whether the clock directly uses the host's hardware clock. virtual bool IsNative() const = 0; + static inline u64 NSToCNTPCT(u64 ns) { + return ns * NsToCNTPCTRatio::num / NsToCNTPCTRatio::den; + } + + static inline u64 USToCNTPCT(u64 us) { + return us * UsToCNTPCTRatio::num / UsToCNTPCTRatio::den; + } + + static inline u64 CNTPCTToNS(u64 cntpct) { + return cntpct * NsToCNTPCTRatio::den / NsToCNTPCTRatio::num; + } + + static inline u64 CNTPCTToUS(u64 cntpct) { + return cntpct * UsToCNTPCTRatio::den / UsToCNTPCTRatio::num; + } + protected: using NsRatio = std::nano; using UsRatio = std::micro; @@ -46,6 +62,7 @@ protected: using NsToUsRatio = std::ratio_divide; using NsToMsRatio = std::ratio_divide; using NsToCNTPCTRatio = std::ratio; + using UsToCNTPCTRatio = std::ratio; }; std::unique_ptr CreateOptimalClock(); diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 99602699a2..3df4094a71 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -16,7 +16,6 @@ add_library(core STATIC core.h core_timing.cpp core_timing.h - core_timing_util.h cpu_manager.cpp cpu_manager.h crypto/aes_util.cpp diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp index 4f2692b05c..9a1d5a69aa 100644 --- a/src/core/core_timing.cpp +++ b/src/core/core_timing.cpp @@ -16,7 +16,6 @@ #include "common/microprofile.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" #include "core/hardware_properties.h" namespace Core::Timing { @@ -45,9 +44,7 @@ struct CoreTiming::Event { } }; -CoreTiming::CoreTiming() - : cpu_clock{Common::CreateBestMatchingClock(Hardware::BASE_CLOCK_RATE, Hardware::CNTFREQ)}, - event_clock{Common::CreateStandardWallClock(Hardware::BASE_CLOCK_RATE, Hardware::CNTFREQ)} {} +CoreTiming::CoreTiming() : clock{Common::CreateOptimalClock()} {} CoreTiming::~CoreTiming() { Reset(); @@ -180,7 +177,7 @@ void CoreTiming::AddTicks(u64 ticks_to_add) { void CoreTiming::Idle() { if (!event_queue.empty()) { const u64 next_event_time = event_queue.front().time; - const u64 next_ticks = nsToCycles(std::chrono::nanoseconds(next_event_time)) + 10U; + const u64 next_ticks = Common::WallClock::NSToCNTPCT(next_event_time) + 10U; if (next_ticks > ticks) { ticks = next_ticks; } @@ -193,18 +190,11 @@ void CoreTiming::ResetTicks() { downcount = MAX_SLICE_LENGTH; } -u64 CoreTiming::GetCPUTicks() const { - if (is_multicore) [[likely]] { - return cpu_clock->GetCPUCycles(); - } - return ticks; -} - u64 CoreTiming::GetClockTicks() const { if (is_multicore) [[likely]] { - return cpu_clock->GetClockCycles(); + return clock->GetCNTPCT(); } - return CpuCyclesToClockCycles(ticks); + return ticks; } std::optional CoreTiming::Advance() { @@ -297,9 +287,7 @@ void CoreTiming::ThreadLoop() { } paused_set = true; - event_clock->Pause(true); pause_event.Wait(); - event_clock->Pause(false); } } @@ -315,25 +303,18 @@ void CoreTiming::Reset() { has_started = false; } -std::chrono::nanoseconds CoreTiming::GetCPUTimeNs() const { - if (is_multicore) [[likely]] { - return cpu_clock->GetTimeNS(); - } - return CyclesToNs(ticks); -} - std::chrono::nanoseconds CoreTiming::GetGlobalTimeNs() const { if (is_multicore) [[likely]] { - return event_clock->GetTimeNS(); + return clock->GetTimeNS(); } - return CyclesToNs(ticks); + return std::chrono::nanoseconds{Common::WallClock::CNTPCTToNS(ticks)}; } std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const { if (is_multicore) [[likely]] { - return event_clock->GetTimeUS(); + return clock->GetTimeUS(); } - return CyclesToUs(ticks); + return std::chrono::microseconds{Common::WallClock::CNTPCTToUS(ticks)}; } } // namespace Core::Timing diff --git a/src/core/core_timing.h b/src/core/core_timing.h index e7c4a949f6..fdacdd94a6 100644 --- a/src/core/core_timing.h +++ b/src/core/core_timing.h @@ -116,15 +116,9 @@ public: return downcount; } - /// Returns current time in emulated CPU cycles - u64 GetCPUTicks() const; - - /// Returns current time in emulated in Clock cycles + /// Returns the current CNTPCT tick value. u64 GetClockTicks() const; - /// Returns current time in nanoseconds. - std::chrono::nanoseconds GetCPUTimeNs() const; - /// Returns current time in microseconds. std::chrono::microseconds GetGlobalTimeUs() const; @@ -142,8 +136,7 @@ private: void Reset(); - std::unique_ptr cpu_clock; - std::unique_ptr event_clock; + std::unique_ptr clock; s64 global_timer = 0; diff --git a/src/core/core_timing_util.h b/src/core/core_timing_util.h deleted file mode 100644 index fe5aaefc70..0000000000 --- a/src/core/core_timing_util.h +++ /dev/null @@ -1,58 +0,0 @@ -// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project -// SPDX-License-Identifier: GPL-2.0-or-later - -#pragma once - -#include - -#include "common/common_types.h" -#include "core/hardware_properties.h" - -namespace Core::Timing { - -namespace detail { -constexpr u64 CNTFREQ_ADJUSTED = Hardware::CNTFREQ / 1000; -constexpr u64 BASE_CLOCK_RATE_ADJUSTED = Hardware::BASE_CLOCK_RATE / 1000; -} // namespace detail - -[[nodiscard]] constexpr s64 msToCycles(std::chrono::milliseconds ms) { - return ms.count() * detail::BASE_CLOCK_RATE_ADJUSTED; -} - -[[nodiscard]] constexpr s64 usToCycles(std::chrono::microseconds us) { - return us.count() * detail::BASE_CLOCK_RATE_ADJUSTED / 1000; -} - -[[nodiscard]] constexpr s64 nsToCycles(std::chrono::nanoseconds ns) { - return ns.count() * detail::BASE_CLOCK_RATE_ADJUSTED / 1000000; -} - -[[nodiscard]] constexpr u64 msToClockCycles(std::chrono::milliseconds ms) { - return static_cast(ms.count()) * detail::CNTFREQ_ADJUSTED; -} - -[[nodiscard]] constexpr u64 usToClockCycles(std::chrono::microseconds us) { - return us.count() * detail::CNTFREQ_ADJUSTED / 1000; -} - -[[nodiscard]] constexpr u64 nsToClockCycles(std::chrono::nanoseconds ns) { - return ns.count() * detail::CNTFREQ_ADJUSTED / 1000000; -} - -[[nodiscard]] constexpr u64 CpuCyclesToClockCycles(u64 ticks) { - return ticks * detail::CNTFREQ_ADJUSTED / detail::BASE_CLOCK_RATE_ADJUSTED; -} - -[[nodiscard]] constexpr std::chrono::milliseconds CyclesToMs(s64 cycles) { - return std::chrono::milliseconds(cycles / detail::BASE_CLOCK_RATE_ADJUSTED); -} - -[[nodiscard]] constexpr std::chrono::nanoseconds CyclesToNs(s64 cycles) { - return std::chrono::nanoseconds(cycles * 1000000 / detail::BASE_CLOCK_RATE_ADJUSTED); -} - -[[nodiscard]] constexpr std::chrono::microseconds CyclesToUs(s64 cycles) { - return std::chrono::microseconds(cycles * 1000 / detail::BASE_CLOCK_RATE_ADJUSTED); -} - -} // namespace Core::Timing diff --git a/src/core/hle/kernel/k_scheduler.cpp b/src/core/hle/kernel/k_scheduler.cpp index faa12b4f05..75ce5a23c0 100644 --- a/src/core/hle/kernel/k_scheduler.cpp +++ b/src/core/hle/kernel/k_scheduler.cpp @@ -184,7 +184,8 @@ u64 KScheduler::UpdateHighestPriorityThread(KThread* highest_thread) { prev_highest_thread != highest_thread) [[likely]] { if (prev_highest_thread != nullptr) [[likely]] { IncrementScheduledCount(prev_highest_thread); - prev_highest_thread->SetLastScheduledTick(m_kernel.System().CoreTiming().GetCPUTicks()); + prev_highest_thread->SetLastScheduledTick( + m_kernel.System().CoreTiming().GetClockTicks()); } if (m_state.should_count_idle) { if (highest_thread != nullptr) [[likely]] { @@ -351,7 +352,7 @@ void KScheduler::SwitchThread(KThread* next_thread) { // Update the CPU time tracking variables. const s64 prev_tick = m_last_context_switch_time; - const s64 cur_tick = m_kernel.System().CoreTiming().GetCPUTicks(); + const s64 cur_tick = m_kernel.System().CoreTiming().GetClockTicks(); const s64 tick_diff = cur_tick - prev_tick; cur_thread->AddCpuTime(m_core_id, tick_diff); if (cur_process != nullptr) { diff --git a/src/core/hle/kernel/svc/svc_info.cpp b/src/core/hle/kernel/svc/svc_info.cpp index 2b2c878b5a..445cdd87b3 100644 --- a/src/core/hle/kernel/svc/svc_info.cpp +++ b/src/core/hle/kernel/svc/svc_info.cpp @@ -199,9 +199,9 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle if (same_thread && info_sub_id == 0xFFFFFFFFFFFFFFFF) { const u64 thread_ticks = current_thread->GetCpuTime(); - out_ticks = thread_ticks + (core_timing.GetCPUTicks() - prev_ctx_ticks); + out_ticks = thread_ticks + (core_timing.GetClockTicks() - prev_ctx_ticks); } else if (same_thread && info_sub_id == system.Kernel().CurrentPhysicalCoreIndex()) { - out_ticks = core_timing.GetCPUTicks() - prev_ctx_ticks; + out_ticks = core_timing.GetClockTicks() - prev_ctx_ticks; } *result = out_ticks; diff --git a/src/core/hle/service/hid/hidbus.cpp b/src/core/hle/service/hid/hidbus.cpp index 5604a6fdad..80aac221b8 100644 --- a/src/core/hle/service/hid/hidbus.cpp +++ b/src/core/hle/service/hid/hidbus.cpp @@ -5,7 +5,6 @@ #include "common/settings.h" #include "core/core.h" #include "core/core_timing.h" -#include "core/core_timing_util.h" #include "core/hid/hid_types.h" #include "core/hle/kernel/k_event.h" #include "core/hle/kernel/k_readable_event.h" diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 456f733cf3..70762c51a4 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -194,17 +194,17 @@ struct GPU::Impl { [[nodiscard]] u64 GetTicks() const { // This values were reversed engineered by fincs from NVN - // The gpu clock is reported in units of 385/625 nanoseconds - constexpr u64 gpu_ticks_num = 384; - constexpr u64 gpu_ticks_den = 625; + // The GPU clock is 614.4 MHz + using NsToGPUTickRatio = std::ratio<614'400'000, std::nano::den>; + static_assert(NsToGPUTickRatio::num == 384 && NsToGPUTickRatio::den == 625); + + u64 nanoseconds = system.CoreTiming().GetGlobalTimeNs().count(); - u64 nanoseconds = system.CoreTiming().GetCPUTimeNs().count(); if (Settings::values.use_fast_gpu_time.GetValue()) { nanoseconds /= 256; } - const u64 nanoseconds_num = nanoseconds / gpu_ticks_den; - const u64 nanoseconds_rem = nanoseconds % gpu_ticks_den; - return nanoseconds_num * gpu_ticks_num + (nanoseconds_rem * gpu_ticks_num) / gpu_ticks_den; + + return nanoseconds * NsToGPUTickRatio::num / NsToGPUTickRatio::den; } [[nodiscard]] bool IsAsync() const {