2
0
mirror of https://github.com/koreader/koreader synced 2024-11-04 12:00:25 +00:00
koreader/spec/unit/mock_time.lua

316 lines
11 KiB
Lua
Raw Normal View History

2017-10-10 10:17:58 +00:00
require("commonrequire")
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local TimeVal = require("ui/timeval")
local time = require("ui/time")
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local ffi = require("ffi")
local dummy = require("ffi/posix_h")
local logger = require("logger")
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local util = require("ffi/util")
local C = ffi.C
local MockTime = {
original_os_time = os.time,
original_util_time = nil,
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
original_tv_realtime = nil,
original_tv_realtime_coarse = nil,
original_tv_monotonic = nil,
original_tv_monotonic_coarse = nil,
original_tv_boottime = nil,
original_tv_boottime_or_realtime_coarse = nil,
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
original_tv_now = nil,
monotonic = 0,
realtime = 0,
boottime = 0,
boottime_or_realtime_coarse = 0,
monotonic_time = 0,
realtime_time = 0,
boottime_time = 0,
boottime_or_realtime_coarse_time = 0,
}
function MockTime:install()
if self.original_util_time == nil then
self.original_util_time = util.gettime
assert(self.original_util_time ~= nil)
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if self.original_tv_realtime == nil then
self.original_tv_realtime = TimeVal.realtime
assert(self.original_tv_realtime ~= nil)
end
if self.original_tv_realtime_coarse == nil then
self.original_tv_realtime_coarse = TimeVal.realtime_coarse
assert(self.original_tv_realtime_coarse ~= nil)
end
if self.original_tv_monotonic == nil then
self.original_tv_monotonic = TimeVal.monotonic
assert(self.original_tv_monotonic ~= nil)
end
if self.original_tv_monotonic_coarse == nil then
self.original_tv_monotonic_coarse = TimeVal.monotonic_coarse
assert(self.original_tv_monotonic_coarse ~= nil)
end
if self.original_tv_boottime == nil then
self.original_tv_boottime = TimeVal.boottime
assert(self.original_tv_boottime ~= nil)
end
if self.original_tv_boottime_or_realtime_coarse == nil then
self.original_tv_boottime_or_realtime_coarse = TimeVal.boottime_or_realtime_coarse
assert(self.original_tv_boottime_or_realtime_coarse ~= nil)
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if self.original_tv_now == nil then
self.original_tv_now = TimeVal.now
assert(self.original_tv_now ~= nil)
end
-- Store both REALTIME & MONOTONIC clocks
self.realtime = os.time()
local timespec = ffi.new("struct timespec")
C.clock_gettime(C.CLOCK_MONOTONIC_COARSE, timespec)
self.monotonic = tonumber(timespec.tv_sec)
os.time = function() --luacheck: ignore
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
logger.dbg("MockTime:os.time: ", self.realtime)
return self.realtime
end
util.gettime = function()
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
logger.dbg("MockTime:util.gettime: ", self.realtime)
return self.realtime, 0
end
TimeVal.realtime = function()
logger.dbg("MockTime:TimeVal.realtime: ", self.realtime)
return TimeVal:new{ sec = self.realtime }
end
TimeVal.realtime_coarse = function()
logger.dbg("MockTime:TimeVal.realtime_coarse: ", self.realtime)
return TimeVal:new{ sec = self.realtime }
end
TimeVal.monotonic = function()
logger.dbg("MockTime:TimeVal.monotonic: ", self.monotonic)
return TimeVal:new{ sec = self.monotonic }
end
TimeVal.monotonic_coarse = function()
logger.dbg("MockTime:TimeVal.monotonic_coarse: ", self.monotonic)
return TimeVal:new{ sec = self.monotonic }
end
TimeVal.boottime = function()
logger.dbg("MockTime:TimeVal.boottime: ", self.boottime)
return TimeVal:new{ sec = self.boottime }
end
TimeVal.boottime_or_realtime_coarse = function()
logger.dbg("MockTime:TimeVal.boottime: ", self.boottime_or_realtime_coarse)
return TimeVal:new{ sec = self.boottime_or_realtime_coarse }
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
TimeVal.now = function()
logger.dbg("MockTime:TimeVal.now: ", self.monotonic)
return TimeVal:new{ sec = self.monotonic }
end
if self.original_tv_realtime_time == nil then
self.original_tv_realtime_time = time.realtime
assert(self.original_tv_realtime_time ~= nil)
end
if self.original_tv_realtime_coarse_time == nil then
self.original_tv_realtime_coarse_time = time.realtime_coarse
assert(self.original_tv_realtime_coarse_time ~= nil)
end
if self.original_tv_monotonic_time == nil then
self.original_tv_monotonic_time = time.monotonic
assert(self.original_tv_monotonic_time ~= nil)
end
if self.original_tv_monotonic_coarse_time == nil then
self.original_tv_monotonic_coarse_time = time.monotonic_coarse
assert(self.original_tv_monotonic_coarse_time ~= nil)
end
if self.original_tv_boottime_time == nil then
self.original_tv_boottime_time = time.boottime
assert(self.original_tv_boottime_time ~= nil)
end
if self.original_tv_boottime_or_realtime_coarse_time == nil then
self.original_tv_boottime_or_realtime_coarse_time = time.boottime_or_realtime_coarse
assert(self.original_tv_boottime_or_realtime_coarse_time ~= nil)
end
if self.original_tv_now == nil then
self.original_tv_now = time.now
assert(self.original_tv_now ~= nil)
end
-- Store both REALTIME & MONOTONIC clocks for fts
self.realtime_time = os.time() * 1e6
local timespec_time = ffi.new("struct timespec")
C.clock_gettime(C.CLOCK_MONOTONIC_COARSE, timespec_time)
self.monotonic = tonumber(timespec.tv_sec) * 1e6
time.realtime = function()
logger.dbg("MockTime:TimeVal.realtime: ", self.realtime_time)
return self.realtime_time
end
time.realtime_coarse = function()
logger.dbg("MockTime:TimeVal.realtime_coarse: ", self.realtime_coarse_time)
return self.realtime_coarse_time
end
time.monotonic = function()
logger.dbg("MockTime:TimeVal.monotonic: ", self.monotonic)
return self.monotonic_time
end
time.monotonic_coarse = function()
logger.dbg("MockTime:TimeVal.monotonic_coarse: ", self.monotonic)
return self.monotonic_time
end
time.boottime = function()
logger.dbg("MockTime:TimeVal.boottime: ", self.boottime_time)
return self.boottime_time
end
time.boottime_or_realtime_coarse = function()
logger.dbg("MockTime:TimeVal.boottime: ", self.boottime_or_realtime_coarse_time)
return self.boottime_or_realtime_coarse_time
end
time.now = function()
logger.dbg("MockTime:TimeVal.now: ", self.monotonic)
return self.monotonic_time
end
end
function MockTime:uninstall()
os.time = self.original_os_time --luacheck: ignore
if self.original_util_time ~= nil then
util.gettime = self.original_util_time
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if self.original_tv_realtime ~= nil then
TimeVal.realtime = self.original_tv_realtime
end
if self.original_tv_realtime_coarse ~= nil then
TimeVal.realtime_coarse = self.original_tv_realtime_coarse
end
if self.original_tv_monotonic ~= nil then
TimeVal.monotonic = self.original_tv_monotonic
end
if self.original_tv_monotonic_coarse ~= nil then
TimeVal.monotonic_coarse = self.original_tv_monotonic_coarse
end
if self.original_tv_boottime ~= nil then
TimeVal.boottime = self.original_tv_boottime
end
if self.original_tv_boottime_or_realtime_coarse ~= nil then
TimeVal.boottime_or_realtime_coarse = self.original_tv_boottime_or_realtime_coarse
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if self.original_tv_now ~= nil then
TimeVal.now = self.original_tv_now
end
end
function MockTime:set_realtime(value)
if type(value) ~= "number" then
return false
end
self.realtime = math.floor(value)
logger.dbg("MockTime:set_realtime ", self.realtime)
return true
end
function MockTime:increase_realtime(value)
if type(value) ~= "number" then
return false
end
self.realtime = math.floor(self.realtime + value)
logger.dbg("MockTime:increase_realtime ", self.realtime)
return true
end
function MockTime:set_monotonic(value)
if type(value) ~= "number" then
return false
end
self.monotonic = math.floor(value)
logger.dbg("MockTime:set_monotonic ", self.monotonic)
return true
end
function MockTime:increase_monotonic(value)
if type(value) ~= "number" then
return false
end
self.monotonic = math.floor(self.monotonic + value)
logger.dbg("MockTime:increase_monotonic ", self.monotonic)
return true
end
function MockTime:set_boottime(value)
if type(value) ~= "number" then
return false
end
self.boottime = math.floor(value)
logger.dbg("MockTime:set_boottime ", self.boottime)
return true
end
function MockTime:increase_boottime(value)
if type(value) ~= "number" then
return false
end
self.boottime = math.floor(self.boottime + value)
logger.dbg("MockTime:increase_boottime ", self.boottime)
return true
end
function MockTime:set_boottime_or_realtime_coarse(value)
if type(value) ~= "number" then
return false
end
self.boottime_or_realtime_coarse = math.floor(value)
logger.dbg("MockTime:set_boottime ", self.boottime_or_realtime_coarse)
return true
end
function MockTime:increase_boottime_or_realtime_coarse(value)
if type(value) ~= "number" then
return false
end
self.boottime_or_realtime_coarse = math.floor(self.boottime_or_realtime_coarse + value)
logger.dbg("MockTime:increase_boottime ", self.boottime_or_realtime_coarse)
return true
end
function MockTime:set(value)
if type(value) ~= "number" then
return false
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
self.realtime = math.floor(value)
logger.dbg("MockTime:set (realtime) ", self.realtime)
self.monotonic = math.floor(value)
logger.dbg("MockTime:set (monotonic) ", self.monotonic)
self.boottime = math.floor(value)
logger.dbg("MockTime:set (boottime) ", self.boottime)
self.boottime_or_realtime_coarse = math.floor(value)
logger.dbg("MockTime:set (boottime) ", self.boottime_or_realtime_coarse)
return true
end
function MockTime:increase(value)
if type(value) ~= "number" then
return false
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415) * ReaderDictionary: Port delay computations to TimeVal * ReaderHighlight: Port delay computations to TimeVal * ReaderView: Port delay computations to TimeVal * Android: Reset gesture detection state on APP_CMD_TERM_WINDOW. This prevents potentially being stuck in bogus gesture states when switching apps. * GestureDetector: * Port delay computations to TimeVal * Fixed delay computations to handle time warps (large and negative deltas). * Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture. * Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1. * Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy. * The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events. The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use a clock source that is *NOT* MONOTONIC. AFAICT, that's pretty much... PocketBook, and that's it? * Input: * Use the <linux/input.h> FFI module instead of re-declaring every constant * Fixed (verbose) debug logging of input events to actually translate said constants properly. * Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume. * Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient. Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts, as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector. * reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary. * TimeVal: * Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>). * Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC. * Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence. * New methods: * Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime * Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero * UIManager: * Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base. This ensures reliable and consistent scheduling, as time is ensured never to go backwards. * Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame. It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value, because very few time has passed. The only code left that does live syscalls does it because it's actually necessary for accuracy, and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics). * DictQuickLookup: Port delay computations to TimeVal * FootNoteWidget: Port delay computations to TimeVal * HTMLBoxWidget: Port delay computations to TimeVal * Notification: Port delay computations to TimeVal * TextBoxWidget: Port delay computations to TimeVal * AutoSuspend: Port to TimeVal * AutoTurn: * Fix it so that settings are actually honored. * Port to TimeVal * BackgroundRunner: Port to TimeVal * Calibre: Port benchmarking code to TimeVal * BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly. * All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
self.realtime = math.floor(self.realtime + value)
logger.dbg("MockTime:increase (realtime) ", self.realtime)
self.monotonic = math.floor(self.monotonic + value)
logger.dbg("MockTime:increase (monotonic) ", self.monotonic)
self.boottime = math.floor(self.boottime + value)
logger.dbg("MockTime:increase (boottime) ", self.boottime)
self.boottime_or_realtime_coarse = math.floor(self.boottime_or_realtime_coarse + value)
logger.dbg("MockTime:increase (boottime) ", self.boottime_or_realtime_coarse)
local value_time = value * 1e6
self.realtime_time = math.floor(self.realtime_time + value_time)
logger.dbg("MockTime:increase (realtime) ", self.realtime_time)
self.monotonic_time = math.floor(self.monotonic_time + value_time)
logger.dbg("MockTime:increase (monotonic) ", self.monotonic)
self.boottime_time = math.floor(self.boottime_time + value_time)
logger.dbg("MockTime:increase (boottime) ", self.boottime_time)
self.boottime_or_realtime_coarse_time = math.floor(self.boottime_or_realtime_coarse_time + value_time)
logger.dbg("MockTime:increase (boottime) ", self.boottime_or_realtime_coarse_time)
return true
end
return MockTime