this library eating CPU time even when idle (not accessed) about ttlcache HOT 6 CLOSED

for that 14.83s part, probably replacing with https://github.com/kpango/fastime works
or probably increase the time resolution will result in better performance (eg. only check time every 20ms)

for the selectgo part probably there's select channel but with default (it would eat CPU time)

from ttlcache.

OK, will check this out. Definitely hasn't received attention so far

from ttlcache.

Are you running with a very low TTL ? I find 'meaningful' cpu usage for example if the global TTL is set to 1 milisecond. It would be nice to know your configuration and usage pattern.

from ttlcache.

I'm using 30s-90s TTL
only calling SetExpirationReasonCallback, the rest are default

from ttlcache.

When checking one of my production services that does 1-3 cache lookups per REST request, and then some queries, i don't see significant usage. When i create a synthetic case with SetExpirationReasonCallback i do see some usage. However the time spent is in a different part than you show.

I'm using Go 1.16 and latest version of the library on Darwin. Is your deployment also on latest versions?

Bench:

package main

import (
	"fmt"
	"log"
	"math/rand"
	"os"
	"runtime"
	"runtime/pprof"
	"time"

	"github.com/ReneKroon/ttlcache/v2"
)

var notFound = ttlcache.ErrNotFound

func main() {
	cache := ttlcache.NewCache()
	cache.SetExpirationReasonCallback(func(key string, reason ttlcache.EvictionReason, value interface{}) {
		if key == value {
			reason = 0
		}
	})
	f, err := os.Create("barf.pprof")
	if err != nil {
		log.Fatal(err)
	}
	pprof.StartCPUProfile(f)
	defer pprof.StopCPUProfile()

	cache.SetTTL(time.Second * 3)

	for j := 1; j < 30; j++ {
		for i := 1; i < 10000; i++ {

			cache.SetWithTTL(fmt.Sprintf("%d", i), "MyValue", time.Millisecond*time.Duration(rand.Int31n(10)))
			cache.Set("MyNumber", 1000)

			if val, err := cache.Get(fmt.Sprintf("%d", i)); err != notFound {
				a := val
				if a == "ADF" {

				}
			}

		}
		time.Sleep(time.Millisecond * 10)
	}

	cache.Purge()

	cache.Close()
}

With a usage of 0.47s on 2.94s of total usage in the program.

from ttlcache.

I think you were looking at the 'blocking' profile instead of cpu profile?

from ttlcache.