From 25f9af666194f81741aed8ea422661bd58149d30 Mon Sep 17 00:00:00 2001
From: =?utf8?q?Felix=20Geisend=C3=B6rfer?=
 <felix.geisendoerfer@datadoghq.com>
Date: Mon, 13 Mar 2023 09:56:45 +0100
Subject: [PATCH] runtime: add a benchmark of FPCallers
MIME-Version: 1.0
Content-Type: text/plain; charset=utf8
Content-Transfer-Encoding: 8bit

This allows comparing frame pointer unwinding against the default
unwinder as shown below.

goos: linux
goarch: amd64
pkg: runtime
cpu: Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz
                    │ callers.txt │
                    │   sec/op    │
Callers/cached-32     1.254µ ± 0%
FPCallers/cached-32   24.99n ± 0%

For #16638

Change-Id: I4dd05f82254726152ef4a5d5beceab33641e9d2b
Reviewed-on: https://go-review.googlesource.com/c/go/+/475795
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Felix Geisendörfer <felix.geisendoerfer@datadoghq.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
---
 src/runtime/callers_test.go | 20 ++++++++++++++++++++
 src/runtime/export_test.go  |  6 ++++++
 2 files changed, 26 insertions(+)

diff --git a/src/runtime/callers_test.go b/src/runtime/callers_test.go
index d738076a2f..e52357f175 100644
--- a/src/runtime/callers_test.go
+++ b/src/runtime/callers_test.go
@@ -430,3 +430,23 @@ func callersNoCache(b *testing.B, n int) int {
 		return 1 + callersNoCache(b, n-1)
 	}
 }
+
+func BenchmarkFPCallers(b *testing.B) {
+	b.Run("cached", func(b *testing.B) {
+		// Very pcvalueCache-friendly, no inlining.
+		fpCallersCached(b, 100)
+	})
+}
+
+func fpCallersCached(b *testing.B, n int) int {
+	if n <= 0 {
+		pcs := make([]uintptr, 32)
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			runtime.FPCallers(0, pcs)
+		}
+		b.StopTimer()
+		return 0
+	}
+	return 1 + fpCallersCached(b, n-1)
+}
diff --git a/src/runtime/export_test.go b/src/runtime/export_test.go
index 367362b672..498c63f5b6 100644
--- a/src/runtime/export_test.go
+++ b/src/runtime/export_test.go
@@ -1727,3 +1727,9 @@ func FrameStartLine(f *Frame) int {
 func PersistentAlloc(n uintptr) unsafe.Pointer {
 	return persistentalloc(n, 0, &memstats.other_sys)
 }
+
+// FPCallers works like Callers and uses frame pointer unwinding to populate
+// pcBuf with the return addresses of the physical frames on the stack.
+func FPCallers(skip int, pcBuf []uintptr) int {
+	return fpTracebackPCs(unsafe.Pointer(getcallerfp()), skip, pcBuf)
+}
-- 
2.50.0