You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
koreader/frontend/ui/timeval.lua

283 lines
8.9 KiB
Lua

--[[--
A simple module to module to compare and do arithmetic with time values.
@usage
local TimeVal = require("ui/timeval")
local tv_start = TimeVal:now()
-- Do some stuff.
-- You can add and subtract `TimeVal` objects.
local tv_duration = TimeVal:now() - tv_start
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..
3 years ago
-- And convert that object to various more human-readable formats, e.g.,
print(string.format("Stuff took %.3fms", tv_duration:tomsecs()))
]]
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..
3 years ago
local ffi = require("ffi")
require("ffi/posix_h")
local logger = require("logger")
local util = require("ffi/util")
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..
3 years ago
local C = ffi.C
-- We prefer CLOCK_MONOTONIC_COARSE if it's available and has a decent resolution,
-- as we generally don't need nano/micro second precision,
-- and it can be more than twice as fast as CLOCK_MONOTONIC/CLOCK_REALTIME/gettimeofday...
local PREFERRED_MONOTONIC_CLOCKID = C.CLOCK_MONOTONIC
-- Ditto for REALTIME (for :realtime_coarse only, :realtime uses gettimeofday ;)).
local PREFERRED_REALTIME_CLOCKID = C.CLOCK_REALTIME
if ffi.os == "Linux" then
-- Unfortunately, it was only implemented in Linux 2.6.32, and we may run on older kernels than that...
-- So, just probe it to see if we can rely on it.
local probe_ts = ffi.new("struct timespec")
if C.clock_getres(C.CLOCK_MONOTONIC_COARSE, probe_ts) == 0 then
-- Now, it usually has a 1ms resolution on modern x86_64 systems,
-- but it only provides a 10ms resolution on all my armv7 devices :/.
if probe_ts.tv_sec == 0 and probe_ts.tv_nsec <= 1000000 then
PREFERRED_MONOTONIC_CLOCKID = C.CLOCK_MONOTONIC_COARSE
end
end
logger.dbg("TimeVal: Preferred MONOTONIC clock source is", PREFERRED_MONOTONIC_CLOCKID == C.CLOCK_MONOTONIC_COARSE and "CLOCK_MONOTONIC_COARSE" or "CLOCK_MONOTONIC")
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..
3 years ago
if C.clock_getres(C.CLOCK_REALTIME_COARSE, probe_ts) == 0 then
if probe_ts.tv_sec == 0 and probe_ts.tv_nsec <= 1000000 then
PREFERRED_REALTIME_CLOCKID = C.CLOCK_REALTIME_COARSE
end
end
logger.dbg("TimeVal: Preferred REALTIME clock source is", PREFERRED_REALTIME_CLOCKID == C.CLOCK_REALTIME_COARSE and "CLOCK_REALTIME_COARSE" or "CLOCK_REALTIME")
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..
3 years ago
probe_ts = nil --luacheck: ignore
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..
3 years ago
TimeVal object. Maps to a POSIX struct timeval (<sys/time.h>).
@table TimeVal
@int sec floored number of seconds
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..
3 years ago
@int usec number of microseconds past that second.
]]
local TimeVal = {
sec = 0,
usec = 0,
12 years ago
}
--[[--
Creates a new TimeVal object.
@usage
local timev = TimeVal:new{
sec = 10,
usec = 10000,
}
@treturn TimeVal
]]
function TimeVal:new(from_o)
local o = from_o or {}
if o.sec == nil then
o.sec = 0
end
if o.usec == nil then
o.usec = 0
elseif o.usec > 1000000 then
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..
3 years ago
o.sec = o.sec + math.floor(o.usec / 1000000)
o.usec = o.usec % 1000000
end
setmetatable(o, self)
self.__index = self
return o
12 years ago
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..
3 years ago
-- Based on <bsd/sys/time.h>
12 years ago
function TimeVal:__lt(time_b)
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..
3 years ago
if self.sec == time_b.sec then
return self.usec < time_b.usec
else
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..
3 years ago
return self.sec < time_b.sec
end
12 years ago
end
function TimeVal:__le(time_b)
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..
3 years ago
if self.sec == time_b.sec then
return self.usec <= time_b.usec
else
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..
3 years ago
return self.sec <= time_b.sec
end
12 years ago
end
function TimeVal:__eq(time_b)
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..
3 years ago
if self.sec == time_b.sec then
return self.usec == time_b.usec
else
return false
end
12 years ago
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..
3 years ago
-- If sec is negative, time went backwards!
12 years ago
function TimeVal:__sub(time_b)
local diff = TimeVal:new{ sec = 0, usec = 0 }
12 years ago
diff.sec = self.sec - time_b.sec
diff.usec = self.usec - time_b.usec
12 years ago
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..
3 years ago
if diff.usec < 0 then
diff.sec = diff.sec - 1
diff.usec = diff.usec + 1000000
end
12 years ago
return diff
12 years ago
end
function TimeVal:__add(time_b)
local sum = TimeVal:new{ sec = 0, usec = 0 }
12 years ago
sum.sec = self.sec + time_b.sec
sum.usec = self.usec + time_b.usec
12 years ago
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..
3 years ago
if sum.usec >= 1000000 then
sum.sec = sum.sec + 1
sum.usec = sum.usec - 1000000
end
12 years ago
return sum
12 years ago
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..
3 years ago
Creates a new TimeVal object based on the current wall clock time.
(e.g., gettimeofday / clock_gettime(CLOCK_REALTIME).
This is a simple wrapper around util.gettime() to get all the niceties of a TimeVal object.
If you don't need sub-second precision, prefer os.time().
Which means that, yes, this is a fancier POSIX Epoch ;).
@usage
local TimeVal = require("ui/timeval")
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..
3 years ago
local tv_start = TimeVal:realtime()
-- Do some stuff.
-- You can add and substract `TimeVal` objects.
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..
3 years ago
local tv_duration = TimeVal:realtime() - tv_start
@treturn TimeVal
]]
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..
3 years ago
function TimeVal:realtime()
local sec, usec = util.gettime()
return TimeVal:new{ sec = sec, usec = usec }
12 years ago
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..
3 years ago
--[[--
Creates a new TimeVal object based on the current value from the system's MONOTONIC clock source.
(e.g., clock_gettime(CLOCK_MONOTONIC).)
POSIX guarantees that this clock source will *never* go backwards (but it *may* return the same value multiple times).
On Linux, this will not account for time spent with the device in suspend (unlike CLOCK_BOOTTIME).
@treturn TimeVal
]]
function TimeVal:monotonic()
local timespec = ffi.new("struct timespec")
C.clock_gettime(C.CLOCK_MONOTONIC, timespec)
-- TIMESPEC_TO_TIMEVAL
return TimeVal:new{ sec = tonumber(timespec.tv_sec), usec = math.floor(tonumber(timespec.tv_nsec / 1000)) }
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..
3 years ago
end
--- Ditto, but w/ CLOCK_MONOTONIC_COARSE if it's available and has a 1ms resolution or better (uses CLOCK_MONOTONIC otherwise).
function TimeVal:monotonic_coarse()
local timespec = ffi.new("struct timespec")
C.clock_gettime(PREFERRED_MONOTONIC_CLOCKID, timespec)
-- TIMESPEC_TO_TIMEVAL
return TimeVal:new{ sec = tonumber(timespec.tv_sec), usec = math.floor(tonumber(timespec.tv_nsec / 1000)) }
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..
3 years ago
end
--- Ditto, but w/ CLOCK_REALTIME_COARSE if it's available and has a 1ms resolution or better (uses CLOCK_REALTIME otherwise).
function TimeVal:realtime_coarse()
local timespec = ffi.new("struct timespec")
C.clock_gettime(PREFERRED_REALTIME_CLOCKID, timespec)
-- TIMESPEC_TO_TIMEVAL
return TimeVal:new{ sec = tonumber(timespec.tv_sec), usec = math.floor(tonumber(timespec.tv_nsec / 1000)) }
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..
3 years ago
end
--- Ditto, but w/ CLOCK_BOOTTIME (will return a TimeVal set to 0, 0 if the clock source is unsupported, as it's 2.6.39+)
function TimeVal:boottime()
local timespec = ffi.new("struct timespec")
C.clock_gettime(C.CLOCK_BOOTTIME, timespec)
-- TIMESPEC_TO_TIMEVAL
return TimeVal:new{ sec = tonumber(timespec.tv_sec), usec = math.floor(tonumber(timespec.tv_nsec / 1000)) }
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..
3 years ago
end
--[[-- Alias for `monotonic_coarse`.
The assumption being anything that requires accurate timestamps expects a monotonic clock source.
This is certainly true for KOReader's UI scheduling.
]]
TimeVal.now = TimeVal.monotonic_coarse
--- Converts a TimeVal object to a Lua (decimal) number (sec.usecs) (accurate to the ms, rounded to 4 decimal places)
function TimeVal:tonumber()
-- Round to 4 decimal places
return math.floor((self.sec + self.usec / 1000000) * 10000) / 10000
end
--- Converts a TimeVal object to a Lua (int) number (resolution: 1µs)
function TimeVal:tousecs()
return math.floor(self.sec * 1000000 + self.usec + 0.5)
end
--[[-- Converts a TimeVal object to a Lua (int) number (resolution: 1ms).
(Mainly useful when computing a time lapse for benchmarking purposes).
]]
function TimeVal:tomsecs()
return self:tousecs() / 1000
end
--- Converts a Lua (decimal) number (sec.usecs) to a TimeVal object
function TimeVal:fromnumber(seconds)
local sec = math.floor(seconds)
local usec = math.floor((seconds - sec) * 1000000 + 0.5)
return TimeVal:new{ sec = sec, usec = usec }
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..
3 years ago
end
--[[-- Compare a past *MONOTONIC* TimeVal object to *now*, returning the elapsed time between the two. (sec.usecs variant)
Returns a Lua (decimal) number (sec.usecs) (accurate to the ms, rounded to 4 decimal places) (i.e., :tonumber())
]]
function TimeVal:getDuration(start_tv)
return (TimeVal:now() - start_tv):tonumber()
end
--[[-- Compare a past *MONOTONIC* TimeVal object to *now*, returning the elapsed time between the two. (µs variant)
Returns a Lua (int) number (resolution: 1µs) (i.e., :tousecs())
]]
function TimeVal:getDurationUs(start_tv)
return (TimeVal:now() - start_tv):tousecs()
end
--[[-- Compare a past *MONOTONIC* TimeVal object to *now*, returning the elapsed time between the two. (ms variant)
Returns a Lua (int) number (resolution: 1ms) (i.e., :tomsecs())
]]
function TimeVal:getDurationMs(start_tv)
return (TimeVal:now() - start_tv):tomsecs()
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..
3 years ago
--- Checks if a TimeVal object is positive
function TimeVal:isPositive()
return self.sec >= 0
end
--- Checks if a TimeVal object is zero
function TimeVal:isZero()
return self.sec == 0 and self.usec == 0
end
--- We often need a const TimeVal set to zero...
--- LuaJIT doesn't actually support const values (Lua 5.4+): Do *NOT* modify it.
TimeVal.zero = TimeVal:new{ sec = 0, usec = 0 }
--- Ditto for one set to math.huge
TimeVal.huge = TimeVal:new{ sec = math.huge, usec = 0 }
return TimeVal