drop binding benchmarks to a single core

This was added for to mainain clock and cpu-cache coherence when moving between cores. However on linux this means that spawning threads from a benchmark will inherit the afinity from the spawner and will themselves be pinned on this core.

On the other hand modern hardware has gotten really good with clocks and the clock coherence is mostly a non-issue for cpus made after 2019.

So, instead of being this aggressive, we drop this completely. Users must decide wether they want this and do in in their own code.

Author: iboB

include/picobench/picobench.hpp

@@ -1,4 +1,4 @@
-// picobench v2.8.0
+// picobench v2.9.0
 // https://github.com/iboB/picobench
 //
 // A micro microbenchmarking library in a single header file
@@ -30,6 +30,7 @@
 //
 //                  VERSION HISTORY
 //
+//  2.9.0 (2026-04-30) * Completely drop binding benchmarks to a single core
 //  2.8.0 (2025-12-15) Switch to SemVer to appease certain package managers
 //
 //  2.08 (2025-04-04) Internal. This file was not affected
@@ -131,8 +132,8 @@
 #   include <functional>
 #endif
 
-#define PICOBENCH_VERSION 2.8.0
-#define PICOBENCH_VERSION_STR "2.8.0"
+#define PICOBENCH_VERSION 2.9.0
+#define PICOBENCH_VERSION_STR "2.9.0"
 
 #if defined(PICOBENCH_DEBUG)
 #   include <cassert>
@@ -378,14 +379,6 @@ class global_registry
 #if defined(_WIN32)
 #   define WIN32_LEAN_AND_MEAN
 #   include <Windows.h>
-#else
-#   if !defined(PICOBENCH_DONT_BIND_TO_ONE_CORE)
-#       if defined(__APPLE__)
-#           include <mach/mach.h>
-#       else
-#           include <sched.h>
-#       endif
-#   endif
 #endif
 
 namespace PICOBENCH_NAMESPACE