Streamly.Internal.Data.Channel

Latency collection

minThreadDelay :: NanoSecond64

This is a magic number and it is overloaded, and used at several places to achieve batching:

1. If we have to sleep to slowdown this is the minimum period that we accumulate before we sleep. Also, workers do not stop until this much sleep time is accumulated.
2. Collected latencies are computed and transferred to measured latency after a minimum of this period.

collectLatency :: Bool -> SVarStats -> YieldRateInfo -> Bool -> IO (Count, AbsTime, NanoSecond64)

Thread accounting

addThread :: MonadIO m => IORef (Set ThreadId) -> ThreadId -> m ()

delThread :: MonadIO m => IORef (Set ThreadId) -> ThreadId -> m ()

modifyThread :: MonadIO m => IORef (Set ThreadId) -> MVar () -> ThreadId -> m ()

allThreadsDone :: MonadIO m => IORef (Set ThreadId) -> m Bool

This is safe even if we are adding more threads concurrently because if a child thread is adding another thread then anyway workerThreads will not be empty.

recordMaxWorkers :: MonadIO m => IORef Int -> SVarStats -> m ()

Diagnostics

dumpSVarStats :: Bool -> Maybe YieldRateInfo -> SVarStats -> IO String

Types

newtype Count

Constructors

Count Int64

Instances

Bounded Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

minBound :: Count

maxBound :: Count

Enum Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

succ :: Count -> Count

pred :: Count -> Count

toEnum :: Int -> Count

fromEnum :: Count -> Int

enumFrom :: Count -> [Count]

enumFromThen :: Count -> Count -> [Count]

enumFromTo :: Count -> Count -> [Count]

enumFromThenTo :: Count -> Count -> Count -> [Count]

Num Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

(+) :: Count -> Count -> Count

(-) :: Count -> Count -> Count

(*) :: Count -> Count -> Count

negate :: Count -> Count

abs :: Count -> Count

signum :: Count -> Count

fromInteger :: Integer -> Count

Read Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

readsPrec :: Int -> ReadS Count

readList :: ReadS [Count]

readPrec :: ReadPrec Count

readListPrec :: ReadPrec [Count]

Integral Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

quot :: Count -> Count -> Count

rem :: Count -> Count -> Count

div :: Count -> Count -> Count

mod :: Count -> Count -> Count

quotRem :: Count -> Count -> (Count, Count)

divMod :: Count -> Count -> (Count, Count)

toInteger :: Count -> Integer

Real Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

toRational :: Count -> Rational

Show Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

showsPrec :: Int -> Count -> ShowS

show :: Count -> String

showList :: [Count] -> ShowS

Eq Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

(==) :: Count -> Count -> Bool

(/=) :: Count -> Count -> Bool

Ord Count

Defined in Streamly.Internal.Data.Channel.Types

Methods

compare :: Count -> Count -> Ordering

(<) :: Count -> Count -> Bool

(<=) :: Count -> Count -> Bool

(>) :: Count -> Count -> Bool

(>=) :: Count -> Count -> Bool

max :: Count -> Count -> Count

min :: Count -> Count -> Count

data Limit

Constructors

Unlimited

Limited Word

Instances

Show Limit

Defined in Streamly.Internal.Data.Channel.Types

Methods

showsPrec :: Int -> Limit -> ShowS

show :: Limit -> String

showList :: [Limit] -> ShowS

Eq Limit

Defined in Streamly.Internal.Data.Channel.Types

Methods

(==) :: Limit -> Limit -> Bool

(/=) :: Limit -> Limit -> Bool

Ord Limit

Defined in Streamly.Internal.Data.Channel.Types

Methods

compare :: Limit -> Limit -> Ordering

(<) :: Limit -> Limit -> Bool

(<=) :: Limit -> Limit -> Bool

(>) :: Limit -> Limit -> Bool

(>=) :: Limit -> Limit -> Bool

max :: Limit -> Limit -> Limit

min :: Limit -> Limit -> Limit

data ThreadAbort

Channel driver throws this exception to all active workers to clean up the channel.

Constructors

ThreadAbort

Instances

Exception ThreadAbort

Defined in Streamly.Internal.Data.Channel.Types

Methods

toException :: ThreadAbort -> SomeException

fromException :: SomeException -> Maybe ThreadAbort

displayException :: ThreadAbort -> String

Show ThreadAbort

Defined in Streamly.Internal.Data.Channel.Types

Methods

showsPrec :: Int -> ThreadAbort -> ShowS

show :: ThreadAbort -> String

showList :: [ThreadAbort] -> ShowS

data ChildEvent a

Events that a child thread may send to a parent thread.

Constructors

ChildYield a

ChildStopChannel

ChildStop ThreadId (Maybe SomeException)

Stats

data SVarStats

Constructors

SVarStats

Fields

• totalDispatches :: IORef Int
• maxWorkers :: IORef Int
• maxOutQSize :: IORef Int
• maxHeapSize :: IORef Int
• maxWorkQSize :: IORef Int
• avgWorkerLatency :: IORef (Count, NanoSecond64)
• minWorkerLatency :: IORef NanoSecond64
• maxWorkerLatency :: IORef NanoSecond64
• svarStopTime :: IORef (Maybe AbsTime)

newSVarStats :: IO SVarStats

Rate Control

data WorkerInfo

We measure the individual worker latencies to estimate the number of workers needed or the amount of time we have to sleep between dispatches to achieve a particular rate when controlled pace mode it used.

Constructors

WorkerInfo

Fields

• workerYieldMax :: Count

  0 means unlimited

• workerYieldCount :: IORef Count

  total number of yields by the worker till now

• workerLatencyStart :: IORef (Count, AbsTime)

  yieldCount at start, timestamp

data LatencyRange

Constructors

LatencyRange

Fields

• minLatency :: NanoSecond64
• maxLatency :: NanoSecond64

Instances

Show LatencyRange

Defined in Streamly.Internal.Data.Channel.Types

Methods

showsPrec :: Int -> LatencyRange -> ShowS

show :: LatencyRange -> String

showList :: [LatencyRange] -> ShowS

data YieldRateInfo

Rate control.

Constructors

YieldRateInfo

Fields

• svarLatencyTarget :: NanoSecond64
• svarLatencyRange :: LatencyRange
• svarRateBuffer :: Int
• svarGainedLostYields :: IORef Count

  LOCKING

  Unlocked access. Modified by the consumer thread and unsafely read by the worker threads

• svarAllTimeLatency :: IORef (Count, AbsTime)

  Actual latency/througput as seen from the consumer side, we count the yields and the time it took to generates those yields. This is used to increase or decrease the number of workers needed to achieve the desired rate. The idle time of workers is adjusted in this, so that we only account for the rate when the consumer actually demands data. XXX interval latency is enough, we can move this under diagnostics build [LOCKING] Unlocked access. Modified by the consumer thread and unsafely read by the worker threads

• workerBootstrapLatency :: Maybe NanoSecond64
• workerPollingInterval :: IORef Count

  After how many yields the worker should update the latency information. If the latency is high, this count is kept lower and vice-versa. XXX If the latency suddenly becomes too high this count may remain too high for long time, in such cases the consumer can change it. 0 means no latency computation XXX this is derivable from workerMeasuredLatency, can be removed. [LOCKING] Unlocked access. Modified by the consumer thread and unsafely read by the worker threads

• workerPendingLatency :: IORef (Count, Count, NanoSecond64)

  This is in progress latency stats maintained by the workers which we empty into workerCollectedLatency stats at certain intervals - whenever we process the stream elements yielded in this period. The first count is all yields, the second count is only those yields for which the latency was measured to be non-zero (note that if the timer resolution is low the measured latency may be zero e.g. on JS platform). [LOCKING] Locked access. Modified by the consumer thread as well as worker threads. Workers modify it periodically based on workerPollingInterval and not on every yield to reduce the locking overhead. (allYieldCount, yieldCount, timeTaken)

• workerCollectedLatency :: IORef (Count, Count, NanoSecond64)

  This is the second level stat which is an accmulation from workerPendingLatency stats. We keep accumulating latencies in this bucket until we have stats for a sufficient period and then we reset it to start collecting for the next period and retain the computed average latency for the last period in workerMeasuredLatency. [LOCKING] Unlocked access. Modified by the consumer thread and unsafely read by the worker threads (allYieldCount, yieldCount, timeTaken)

• workerMeasuredLatency :: IORef NanoSecond64

  Latency as measured by workers, aggregated for the last period. [LOCKING] Unlocked access. Modified by the consumer thread and unsafely read by the worker threads

newRateInfo :: Config -> IO (Maybe YieldRateInfo)

Output queue

readOutputQRaw :: IORef ([ChildEvent a], Int) -> Maybe SVarStats -> IO ([ChildEvent a], Int)

readOutputQBasic :: IORef ([ChildEvent a], Int) -> IO ([ChildEvent a], Int)

ringDoorBell :: IORef Bool -> MVar () -> IO ()

Yield Limit

decrementYieldLimit :: Maybe (IORef Count) -> IO Bool

A worker decrements the yield limit before it executes an action. However, the action may not result in an element being yielded, in that case we have to increment the yield limit.

Note that we need it to be an Int type so that we have the ability to undo a decrement that takes it below zero.

incrementYieldLimit :: Maybe (IORef Count) -> IO ()

Configuration

data Rate

Specifies the stream yield rate in yields per second (Hertz). We keep accumulating yield credits at rateGoal. At any point of time we allow only as many yields as we have accumulated as per rateGoal since the start of time. If the consumer or the producer is slower or faster, the actual rate may fall behind or exceed rateGoal. We try to recover the gap between the two by increasing or decreasing the pull rate from the producer. However, if the gap becomes more than rateBuffer we try to recover only as much as rateBuffer.

rateLow puts a bound on how low the instantaneous rate can go when recovering the rate gap. In other words, it determines the maximum yield latency. Similarly, rateHigh puts a bound on how high the instantaneous rate can go when recovering the rate gap. In other words, it determines the minimum yield latency. We reduce the latency by increasing concurrency, therefore we can say that it puts an upper bound on concurrency.

If the rateGoal is 0 or negative the stream never yields a value. If the rateBuffer is 0 or negative we do not attempt to recover.

Constructors

Rate

Fields

• rateLow :: Double

  The lower rate limit

• rateGoal :: Double

  The target rate we want to achieve

• rateHigh :: Double

  The upper rate limit

• rateBuffer :: Int

  Maximum slack from the goal

data StopWhen

Specify when the Channel should stop.

Constructors

FirstStops

Stop when the first stream ends.

AllStop

Stop when all the streams end.

AnyStops

Stop when any one stream ends.

data Config

An abstract type for specifying the configuration parameters of a Channel. Use Config -> Config modifier functions to modify the default configuration. See the individual modifier documentation for default values.

Default config

magicMaxBuffer :: Word

A magical value for the buffer size arrived at by running the smallest possible task and measuring the optimal value of the buffer for that. This is obviously dependent on hardware, this figure is based on a 2.2GHz intel core-i7 processor.

defaultConfig :: Config

The fields prefixed by an _ are not to be accessed or updated directly but via smart accessor APIs. Use get/set routines instead of directly accessing the Config fields

Set config

maxThreads :: Int -> Config -> Config

Specify the maximum number of threads that can be spawned by the channel. A value of 0 resets the thread limit to default, a negative value means there is no limit. The default value is 1500.

When the actions in a stream are IO bound, having blocking IO calls, this option can be used to control the maximum number of in-flight IO requests. When the actions are CPU bound this option can be used to control the amount of CPU used by the stream.

maxBuffer :: Int -> Config -> Config

Specify the maximum size of the buffer for storing the results from concurrent computations. If the buffer becomes full we stop spawning more concurrent tasks until there is space in the buffer. A value of 0 resets the buffer size to default, a negative value means there is no limit. The default value is 1500.

CAUTION! using an unbounded maxBuffer value (i.e. a negative value) coupled with an unbounded maxThreads value is a recipe for disaster in presence of infinite streams, or very large streams. Especially, it must not be used when pure is used in ZipAsyncM streams as pure in applicative zip streams generates an infinite stream causing unbounded concurrent generation with no limit on the buffer or threads.

maxYields :: Maybe Int64 -> Config -> Config

inspect :: Bool -> Config -> Config

Print debug information about the Channel when the stream ends.

eager :: Bool -> Config -> Config

By default, processing of output from the worker threads is given priority over dispatching new workers. More workers are dispatched only when there is no output to process. When eager is set to True, workers are dispatched aggresively as long as there is more work to do irrespective of whether there is output pending to be processed by the stream consumer. However, dispatching may stop if maxThreads or maxBuffer is reached.

Note: This option has no effect when rate has been specified.

Note: Not supported with interleaved.

stopWhen :: StopWhen -> Config -> Config

Specify when the Channel should stop.

ordered :: Bool -> Config -> Config

When enabled the streams may be evaluated cocnurrently but the results are produced in the same sequence as a serial evaluation would produce.

Note: Not supported with interleaved.

interleaved :: Bool -> Config -> Config

Interleave the streams fairly instead of prioritizing the left stream. This schedules all streams in a round robin fashion over limited number of threads.

Note: Can only be used on finite number of streams.

Note: Not supported with ordered.

boundThreads :: Bool -> Config -> Config

Spawn bound threads (i.e., spawn threads using forkOS instead of forkIO). The default value is False.

Currently, this only takes effect only for concurrent folds.

rate :: Maybe Rate -> Config -> Config

Specify the stream evaluation rate of a channel.

A Nothing value means there is no smart rate control, concurrent execution blocks only if maxThreads or maxBuffer is reached, or there are no more concurrent tasks to execute. This is the default.

When rate (throughput) is specified, concurrent production may be ramped up or down automatically to achieve the specified stream throughput. The specific behavior for different styles of Rate specifications is documented under Rate. The effective maximum production rate achieved by a channel is governed by:

• The maxThreads limit
• The maxBuffer limit
• The maximum rate that the stream producer can achieve
• The maximum rate that the stream consumer can achieve

Maximum production rate is given by:

\(rate = \frac{maxThreads}{latency}\)

If we know the average latency of the tasks we can set maxThreads accordingly.

avgRate :: Double -> Config -> Config

Same as rate (Just $ Rate (r/2) r (2*r) maxBound)

Specifies the average production rate of a stream in number of yields per second (i.e. Hertz). Concurrent production is ramped up or down automatically to achieve the specified average yield rate. The rate can go down to half of the specified rate on the lower side and double of the specified rate on the higher side.

minRate :: Double -> Config -> Config

Same as rate (Just $ Rate r r (2*r) maxBound)

Specifies the minimum rate at which the stream should yield values. As far as possible the yield rate would never be allowed to go below the specified rate, even though it may possibly go above it at times, the upper limit is double of the specified rate.

maxRate :: Double -> Config -> Config

Same as rate (Just $ Rate (r/2) r r maxBound)

Specifies the maximum rate at which the stream should yield values. As far as possible the yield rate would never be allowed to go above the specified rate, even though it may possibly go below it at times, the lower limit is half of the specified rate. This can be useful in applications where certain resource usage must not be allowed to go beyond certain limits.

constRate :: Double -> Config -> Config

Same as rate (Just $ Rate r r r 0)

Specifies a constant yield rate. If for some reason the actual rate goes above or below the specified rate we do not try to recover it by increasing or decreasing the rate in future. This can be useful in applications like graphics frame refresh where we need to maintain a constant refresh rate.

Get config

getMaxThreads :: Config -> Limit

getMaxBuffer :: Config -> Limit

getStreamRate :: Config -> Maybe Rate

getStreamLatency :: Config -> Maybe NanoSecond64

setStreamLatency :: Int -> Config -> Config

getYieldLimit :: Config -> Maybe Count

getInspectMode :: Config -> Bool

getEagerDispatch :: Config -> Bool

getStopWhen :: Config -> StopWhen

getOrdered :: Config -> Bool

getInterleaved :: Config -> Bool

getBound :: Config -> Bool

Cleanup

cleanupSVar :: IORef (Set ThreadId) -> IO ()

Never called from a worker thread.

Diagnostics

dumpCreator :: Show a => a -> String

dumpOutputQ :: (Foldable t, Show a1) => IORef (t a2, a1) -> IO String

dumpDoorBell :: Show a => MVar a -> IO String

dumpNeedDoorBell :: Show a => IORef a -> IO String

dumpRunningThreads :: Show a => IORef a -> IO String

dumpWorkerCount :: Show a => IORef a -> IO String

withDiagMVar :: Bool -> IO String -> String -> IO () -> IO ()

MVar diagnostics has some overhead - around 5% on AsyncT null benchmark, we can keep it on in production to debug problems quickly if and when they happen, but it may result in unexpected output when threads are left hanging until they are GCed because the consumer went away.

printSVar :: IO String -> String -> IO ()

data Work

Constructors

BlockWait NanoSecond64

PartialWorker Count

ManyWorkers Int Count

Instances

Show Work

Defined in Streamly.Internal.Data.Channel.Worker

Methods

showsPrec :: Int -> Work -> ShowS

show :: Work -> String

showList :: [Work] -> ShowS

estimateWorkers :: Limit -> Count -> Count -> NanoSecond64 -> NanoSecond64 -> NanoSecond64 -> LatencyRange -> Work

isBeyondMaxRate :: Limit -> IORef Int -> YieldRateInfo -> IO Bool

workerRateControl :: Limit -> IORef Int -> YieldRateInfo -> WorkerInfo -> IO Bool

CAUTION! this also updates the yield count and therefore should be called only when we are actually yielding an element.

Send Events

sendWithDoorBell :: IORef ([ChildEvent a], Int) -> MVar () -> ChildEvent a -> IO Int

sendYield :: Limit -> Limit -> IORef Int -> Maybe WorkerInfo -> Maybe YieldRateInfo -> IORef ([ChildEvent a], Int) -> MVar () -> ChildEvent a -> IO Bool

Returns whether the worker should continue (True) or stop (False).

sendStop :: IORef Int -> Maybe WorkerInfo -> Maybe YieldRateInfo -> IORef ([ChildEvent a], Int) -> MVar () -> IO ()

handleChildException :: IORef ([ChildEvent a], Int) -> MVar () -> SomeException -> IO ()