diff --git a/Makefile b/Makefile
index 4e3e67c2b..4bf5a2150 100644
--- a/Makefile
+++ b/Makefile
@@ -254,8 +254,12 @@ TEST_PACKAGES_WASI = \
 .PHONY: tinygo-test
 tinygo-test:
 	$(TINYGO) test $(TEST_PACKAGES)
+tinygo-bench:
+	$(TINYGO) test -bench . $(TEST_PACKAGES)
 tinygo-test-wasi:
 	$(TINYGO) test -target wasi $(TEST_PACKAGES_WASI)
+tinygo-bench-wasi:
+	$(TINYGO) test -target wasi -bench . $(TEST_PACKAGES_WASI)
 
 .PHONY: smoketest
 smoketest:
diff --git a/src/testing/benchmark.go b/src/testing/benchmark.go
index 88aec366f..c293e37da 100644
--- a/src/testing/benchmark.go
+++ b/src/testing/benchmark.go
@@ -8,6 +8,9 @@ package testing
 
 import (
 	"fmt"
+	"io"
+	"math"
+	"strings"
 	"time"
 )
 
@@ -41,14 +44,17 @@ type InternalBenchmark struct {
 // affecting benchmark results.
 type B struct {
 	common
-	hasSub    bool          // TODO: should be in common, and atomic
-	start     time.Time     // TODO: should be in common
-	duration  time.Duration // TODO: should be in common
-	N         int
-	benchFunc func(b *B)
-	benchTime benchTimeFlag
-	timerOn   bool
-	result    BenchmarkResult
+	hasSub       bool          // TODO: should be in common, and atomic
+	start        time.Time     // TODO: should be in common
+	duration     time.Duration // TODO: should be in common
+	context      *benchContext
+	N            int
+	benchFunc    func(b *B)
+	bytes        int64
+	missingBytes bool // one of the subbenchmarks does not have bytes set.
+	benchTime    benchTimeFlag
+	timerOn      bool
+	result       BenchmarkResult
 }
 
 // StartTimer starts timing a test. This function is called automatically
@@ -82,15 +88,13 @@ func (b *B) ResetTimer() {
 
 // SetBytes records the number of bytes processed in a single operation.
 // If this is called, the benchmark will report ns/op and MB/s.
-func (b *B) SetBytes(n int64) {
-	panic("testing: unimplemented: B.SetBytes")
-}
+func (b *B) SetBytes(n int64) { b.bytes = n }
 
 // ReportAllocs enables malloc statistics for this benchmark.
 // It is equivalent to setting -test.benchmem, but it only affects the
 // benchmark function that calls ReportAllocs.
 func (b *B) ReportAllocs() {
-	panic("testing: unimplemented: B.ReportAllocs")
+	return // TODO: implement
 }
 
 // runN runs a single benchmark for the specified number of iterations.
@@ -119,13 +123,31 @@ func max(x, y int64) int64 {
 // run1 runs the first iteration of benchFunc. It reports whether more
 // iterations of this benchmarks should be run.
 func (b *B) run1() bool {
+	if ctx := b.context; ctx != nil {
+		// Extend maxLen, if needed.
+		if n := len(b.name); n > ctx.maxLen {
+			ctx.maxLen = n + 8 // Add additional slack to avoid too many jumps in size.
+		}
+	}
 	b.runN(1)
 	return !b.hasSub
 }
 
 // run executes the benchmark.
 func (b *B) run() {
+	if b.context != nil {
+		// Running go test --test.bench
+		b.processBench(b.context) // calls doBench and prints results
+	} else {
+		// Running func Benchmark.
+		b.doBench()
+	}
+}
+
+func (b *B) doBench() BenchmarkResult {
+	// in upstream, this uses a goroutine
 	b.launch()
+	return b.result
 }
 
 // launch launches the benchmark function. It gradually increases the number
@@ -159,13 +181,14 @@ func (b *B) launch() {
 		n = min(n, 1e9)
 		b.runN(int(n))
 	}
-	b.result = BenchmarkResult{b.N, b.duration}
+	b.result = BenchmarkResult{b.N, b.duration, b.bytes}
 }
 
 // BenchmarkResult contains the results of a benchmark run.
 type BenchmarkResult struct {
-	N int           // The number of iterations.
-	T time.Duration // The total time taken.
+	N     int           // The number of iterations.
+	T     time.Duration // The total time taken.
+	Bytes int64         // Bytes processed in one iteration.
 }
 
 // NsPerOp returns the "ns/op" metric.
@@ -176,6 +199,14 @@ func (r BenchmarkResult) NsPerOp() int64 {
 	return r.T.Nanoseconds() / int64(r.N)
 }
 
+// mbPerSec returns the "MB/s" metric.
+func (r BenchmarkResult) mbPerSec() float64 {
+	if r.Bytes <= 0 || r.T <= 0 || r.N <= 0 {
+		return 0
+	}
+	return (float64(r.Bytes) * float64(r.N) / 1e6) / r.T.Seconds()
+}
+
 // AllocsPerOp returns the "allocs/op" metric,
 // which is calculated as r.MemAllocs / r.N.
 func (r BenchmarkResult) AllocsPerOp() int64 {
@@ -188,10 +219,71 @@ func (r BenchmarkResult) AllocedBytesPerOp() int64 {
 	return 0 // Dummy version to allow running e.g. golang.org/test/fibo.go
 }
 
+// String returns a summary of the benchmark results.
+// It follows the benchmark result line format from
+// https://golang.org/design/14313-benchmark-format, not including the
+// benchmark name.
+// Extra metrics override built-in metrics of the same name.
+// String does not include allocs/op or B/op, since those are reported
+// by MemString.
+func (r BenchmarkResult) String() string {
+	buf := new(strings.Builder)
+	fmt.Fprintf(buf, "%8d", r.N)
+
+	// Get ns/op as a float.
+	ns := float64(r.T.Nanoseconds()) / float64(r.N)
+	if ns != 0 {
+		buf.WriteByte('\t')
+		prettyPrint(buf, ns, "ns/op")
+	}
+
+	if mbs := r.mbPerSec(); mbs != 0 {
+		fmt.Fprintf(buf, "\t%7.2f MB/s", mbs)
+	}
+	return buf.String()
+}
+
+func prettyPrint(w io.Writer, x float64, unit string) {
+	// Print all numbers with 10 places before the decimal point
+	// and small numbers with four sig figs. Field widths are
+	// chosen to fit the whole part in 10 places while aligning
+	// the decimal point of all fractional formats.
+	var format string
+	switch y := math.Abs(x); {
+	case y == 0 || y >= 999.95:
+		format = "%10.0f %s"
+	case y >= 99.995:
+		format = "%12.1f %s"
+	case y >= 9.9995:
+		format = "%13.2f %s"
+	case y >= 0.99995:
+		format = "%14.3f %s"
+	case y >= 0.099995:
+		format = "%15.4f %s"
+	case y >= 0.0099995:
+		format = "%16.5f %s"
+	case y >= 0.00099995:
+		format = "%17.6f %s"
+	default:
+		format = "%18.7f %s"
+	}
+	fmt.Fprintf(w, format, x, unit)
+}
+
+type benchContext struct {
+	maxLen int // The largest recorded benchmark name.
+}
+
 func runBenchmarks(benchmarks []InternalBenchmark) bool {
 	if len(benchmarks) == 0 {
 		return true
 	}
+	ctx := &benchContext{}
+	for _, Benchmark := range benchmarks {
+		if l := len(Benchmark.Name); l > ctx.maxLen {
+			ctx.maxLen = l
+		}
+	}
 	main := &B{
 		common: common{
 			name: "Main",
@@ -199,30 +291,55 @@ func runBenchmarks(benchmarks []InternalBenchmark) bool {
 		benchTime: benchTime,
 		benchFunc: func(b *B) {
 			for _, Benchmark := range benchmarks {
-				if flagVerbose {
-					fmt.Printf("=== RUN   %s\n", Benchmark.Name)
-				}
 				b.Run(Benchmark.Name, Benchmark.F)
-				fmt.Printf("--- Result: %d ns/op\n", b.result.NsPerOp())
 			}
 		},
+		context: ctx,
 	}
 
 	main.runN(1)
 	return true
 }
 
+// processBench runs bench b and prints the results.
+func (b *B) processBench(ctx *benchContext) {
+	benchName := b.name
+
+	if ctx != nil {
+		fmt.Printf("%-*s\t", ctx.maxLen, benchName)
+	}
+	r := b.doBench()
+	if b.failed {
+		// The output could be very long here, but probably isn't.
+		// We print it all, regardless, because we don't want to trim the reason
+		// the benchmark failed.
+		fmt.Printf("--- FAIL: %s\n%s", benchName, "") // b.output)
+		return
+	}
+	if ctx != nil {
+		results := r.String()
+		fmt.Println(results)
+	}
+}
+
 // Run benchmarks f as a subbenchmark with the given name. It reports
 // true if the subbenchmark succeeded.
 //
 // A subbenchmark is like any other benchmark. A benchmark that calls Run at
 // least once will not be measured itself and will be called once with N=1.
 func (b *B) Run(name string, f func(b *B)) bool {
+	if b.level > 0 {
+		name = b.name + "/" + name
+	}
 	b.hasSub = true
 	sub := &B{
-		common:    common{name: name},
+		common: common{
+			name:  name,
+			level: b.level + 1,
+		},
 		benchFunc: f,
 		benchTime: b.benchTime,
+		context:   b.context,
 	}
 	if sub.run1() {
 		sub.run()
@@ -240,6 +357,15 @@ func (b *B) add(other BenchmarkResult) {
 	// in sequence in a single benchmark.
 	r.N = 1
 	r.T += time.Duration(other.NsPerOp())
+	if other.Bytes == 0 {
+		// Summing Bytes is meaningless in aggregate if not all subbenchmarks
+		// set it.
+		b.missingBytes = true
+		r.Bytes = 0
+	}
+	if !b.missingBytes {
+		r.Bytes += other.Bytes
+	}
 }
 
 // A PB is used by RunParallel for running parallel benchmarks.
diff --git a/src/testing/benchmark_test.go b/src/testing/benchmark_test.go
index 49612d3d4..7fbd2cf12 100644
--- a/src/testing/benchmark_test.go
+++ b/src/testing/benchmark_test.go
@@ -48,3 +48,8 @@ func TestBenchmark(t *testing.T) {
 		t.Errorf("Expected speedup >= 0.3, got %f", speedup)
 	}
 }
+
+func BenchmarkSub(b *testing.B) {
+	b.Run("Fast", func(b *testing.B) { BenchmarkFastNonASCII(b) })
+	b.Run("Slow", func(b *testing.B) { BenchmarkSlowNonASCII(b) })
+}
diff --git a/src/testing/testing.go b/src/testing/testing.go
index ee4056ebd..3eac03ddf 100644
--- a/src/testing/testing.go
+++ b/src/testing/testing.go
@@ -48,6 +48,7 @@ type common struct {
 	failed   bool   // Test or benchmark has failed.
 	skipped  bool   // Test of benchmark has been skipped.
 	finished bool   // Test function has completed.
+	level    int    // Nesting depth of test or benchmark.
 	name     string // Name of test or benchmark.
 }