dev: improve memory tracking

Author: jirfag

.gitignore

@@ -1,4 +1,5 @@
 /*.txt
+/*.pdf
 /*.pprof
 /dist/
 /.idea/

README.md

@@ -341,6 +341,22 @@ On average golangci-lint consumes 26% less memory.
 Golangci-lint directly calls linters (no forking) and reuses 80% of work by parsing program only once.
 Read [this section](#internals) for details.
 
+### Memory Usage of Golangci-lint
+
+A trade-off between memory usage and execution time can be controlled by [`GOCC`](https://golang.org/pkg/runtime/#hdr-Environment_Variables) environment variable.
+Less `GOGC` values trigger garbage collection more frequently and golangci-lint consumes less memory and more CPU. Below is the trade-off table for running on this repo:
+
+|`GOGC`|Peak Memory, GB|Executon Time, s|
+|------|---------------|----------------|
+|`5`   |1.1            |60              |
+|`10`  |1.1            |34              |
+|`20`  |1.3            |25              |
+|`30`  |1.6            |20.2            |
+|`50`  |2.0            |17.1            |
+|`80`  |2.2            |14.1            |
+|`100` (default)|2.2   |13.8            |
+|`off` |3.2            |9.3             |
+
 ## Internals
 
 1. Work sharing

README.tmpl.md

@@ -310,6 +310,22 @@ On average golangci-lint consumes 26% less memory.
 Golangci-lint directly calls linters (no forking) and reuses 80% of work by parsing program only once.
 Read [this section](#internals) for details.
 
+### Memory Usage of Golangci-lint
+
+A trade-off between memory usage and execution time can be controlled by [`GOCC`](https://golang.org/pkg/runtime/#hdr-Environment_Variables) environment variable.
+Less `GOGC` values trigger garbage collection more frequently and golangci-lint consumes less memory and more CPU. Below is the trade-off table for running on this repo:
+
+|`GOGC`|Peak Memory, GB|Executon Time, s|
+|------|---------------|----------------|
+|`5`   |1.1            |60              |
+|`10`  |1.1            |34              |
+|`20`  |1.3            |25              |
+|`30`  |1.6            |20.2            |
+|`50`  |2.0            |17.1            |
+|`80`  |2.2            |14.1            |
+|`100` (default)|2.2   |13.8            |
+|`off` |3.2            |9.3             |
+
 ## Internals
 
 1. Work sharing

pkg/commands/root.go

@@ -5,6 +5,7 @@ import (
 	"os"
 	"runtime"
 	"runtime/pprof"
+	"strconv"
 
 	"github.com/spf13/cobra"
 	"github.com/spf13/pflag"
@@ -30,6 +31,12 @@ func (e *Executor) persistentPreRun(_ *cobra.Command, _ []string) {
 			e.log.Fatalf("Can't start CPU profiling: %s", err)
 		}
 	}
+
+	if e.cfg.Run.MemProfilePath != "" {
+		if rate := os.Getenv("GL_MEMPROFILE_RATE"); rate != "" {
+			runtime.MemProfileRate, _ = strconv.Atoi(rate)
+		}
+	}
 }
 
 func (e *Executor) persistentPostRun(_ *cobra.Command, _ []string) {
@@ -41,15 +48,45 @@ func (e *Executor) persistentPostRun(_ *cobra.Command, _ []string) {
 		if err != nil {
 			e.log.Fatalf("Can't create file %s: %s", e.cfg.Run.MemProfilePath, err)
 		}
-		runtime.GC() // get up-to-date statistics
+
+		var ms runtime.MemStats
+		runtime.ReadMemStats(&ms)
+		printMemStats(&ms, e.log)
+
 		if err := pprof.WriteHeapProfile(f); err != nil {
 			e.log.Fatalf("Can't write heap profile: %s", err)
 		}
+		f.Close()
 	}
 
 	os.Exit(e.exitCode)
 }
 
+func printMemStats(ms *runtime.MemStats, logger logutils.Log) {
+	logger.Infof("Mem stats: alloc=%s total_alloc=%s sys=%s "+
+		"heap_alloc=%s heap_sys=%s heap_idle=%s heap_released=%s heap_in_use=%s "+
+		"stack_in_use=%s stack_sys=%s "+
+		"mspan_sys=%s mcache_sys=%s buck_hash_sys=%s gc_sys=%s other_sys=%s "+
+		"mallocs_n=%d frees_n=%d heap_objects_n=%d gc_cpu_fraction=%.2f",
+		formatMemory(ms.Alloc), formatMemory(ms.TotalAlloc), formatMemory(ms.Sys),
+		formatMemory(ms.HeapAlloc), formatMemory(ms.HeapSys),
+		formatMemory(ms.HeapIdle), formatMemory(ms.HeapReleased), formatMemory(ms.HeapInuse),
+		formatMemory(ms.StackInuse), formatMemory(ms.StackSys),
+		formatMemory(ms.MSpanSys), formatMemory(ms.MCacheSys), formatMemory(ms.BuckHashSys),
+		formatMemory(ms.GCSys), formatMemory(ms.OtherSys),
+		ms.Mallocs, ms.Frees, ms.HeapObjects, ms.GCCPUFraction)
+}
+
+func formatMemory(memBytes uint64) string {
+	if memBytes < 1024 {
+		return fmt.Sprintf("%db", memBytes)
+	}
+	if memBytes < 1024*1024 {
+		return fmt.Sprintf("%dkb", memBytes/1024)
+	}
+	return fmt.Sprintf("%dmb", memBytes/1024/1024)
+}
+
 func getDefaultConcurrency() int {
 	if os.Getenv("HELP_RUN") == "1" {
 		return 8 // to make stable concurrency for README help generating builds

pkg/commands/run.go

@@ -430,11 +430,21 @@ func watchResources(ctx context.Context, done chan struct{}, logger logutils.Log
 	ticker := time.NewTicker(100 * time.Millisecond)
 	defer ticker.Stop()
 
-	for {
+	logEveryRecord := os.Getenv("GL_MEM_LOG_EVERY") == "1"
+
+	track := func() {
 		var m runtime.MemStats
 		runtime.ReadMemStats(&m)
 
+		if logEveryRecord {
+			printMemStats(&m, logger)
+		}
+
 		rssValues = append(rssValues, m.Sys)
+	}
+
+	for {
+		track()
 
 		stop := false
 		select {
@@ -447,6 +457,7 @@ func watchResources(ctx context.Context, done chan struct{}, logger logutils.Log
 			break
 		}
 	}
+	track()
 
 	var avg, max uint64
 	for _, v := range rssValues {

pkg/result/processors/skip_dirs.go

@@ -14,7 +14,7 @@ import (
 type SkipDirs struct {
 	patterns    []*regexp.Regexp
 	log         logutils.Log
-	skippedDirs map[string][]string // regexp to dir mapping
+	skippedDirs map[string]string // dir to the last regexp mapping
 	absArgsDirs []string
 }
 
@@ -52,7 +52,7 @@ func NewSkipDirs(patterns []string, log logutils.Log, runArgs []string) (*SkipDi
 	return &SkipDirs{
 		patterns:    patternsRe,
 		log:         log,
-		skippedDirs: map[string][]string{},
+		skippedDirs: map[string]string{},
 		absArgsDirs: absArgsDirs,
 	}, nil
 }
@@ -99,7 +99,7 @@ func (p *SkipDirs) shouldPassIssue(i *result.Issue) bool {
 	for _, pattern := range p.patterns {
 		if pattern.MatchString(issueRelDir) {
 			ps := pattern.String()
-			p.skippedDirs[ps] = append(p.skippedDirs[ps], issueRelDir)
+			p.skippedDirs[issueRelDir] = ps
 			return false
 		}
 	}
@@ -108,7 +108,7 @@ func (p *SkipDirs) shouldPassIssue(i *result.Issue) bool {
 }
 
 func (p SkipDirs) Finish() {
-	for pattern, dirs := range p.skippedDirs {
-		p.log.Infof("Skipped by pattern %s dirs: %s", pattern, dirs)
+	for dir, pattern := range p.skippedDirs {
+		p.log.Infof("Skipped dir %s by pattern %s", dir, pattern)
 	}
 }