userland/arch/aarch64/freestanding/linux/gem5_exit.S to benchmark immediate gem5 exit

ldxr_stxr does an ldxr on thread 0 to fall in the ISA case where event is certain.

Also add ldxr_str which tests yet another case where event is certain: non conditional
write.

README.adoc

@@ -9951,7 +9951,9 @@ The `--verbose` is optional, but shows clearly each GCC build command so that yo
 
 The build outputs are automatically stored in a different directories for optimized and debug builds, which prevents `debug` files from overwriting `opt` ones. Therefore, `--gem5-build-id` is not required.
 
-The price to pay for debuggability is high however: a Linux kernel boot was about 3x slower in QEMU and 14 times slower in gem5 debug compared to opt, see benchmarks at: xref:benchmark-linux-kernel-boot[xrefstyle=full]
+The price to pay for debuggability is high however: a Linux kernel boot was about 3x slower in QEMU and 14 times slower in gem5 debug compared to opt, see benchmarks at: xref:benchmark-linux-kernel-boot[xrefstyle=full].
+
+Similar slowdowns can be observed at: xref:benchmark-emulators-on-userland-executables[xrefstyle=full].
 
 When in <<qemu-text-mode>>, using `--debug-vm` makes Ctrl-C not get passed to the QEMU guest anymore: it is instead captured by GDB itself, so allow breaking. So e.g. you won't be able to easily quit from a guest program like:
 
@@ -14069,7 +14071,9 @@ Programs under link:userland/cpp/[] are examples of https://en.wikipedia.org/wik
 * random
 ** link:userland/cpp/random.cpp[]
 * containers
-** link:userland/cpp/set.cpp[]: `std::set` contains unique keys
+** associative
+*** <<algorithms>> contains a benchmark comparison of different c++ containers
+*** link:userland/cpp/set.cpp[]: `std::set` contains unique keys
 
 [[cpp-multithreading]]
 ==== C++ multithreading
@@ -18715,7 +18719,10 @@ For how userland spinlocks and mutexes are implemented see <<userland-mutex-impl
 
 ====== ARM WFE global monitor events
 
-link:userland/arch/aarch64/inline_asm/wfe_ldxr_stxr.cpp[]
+Examples:
+
+* link:userland/arch/aarch64/inline_asm/wfe_ldxr_stxr.cpp[]
+* link:userland/arch/aarch64/inline_asm/wfe_ldxr_str.cpp[]
 
 SEV is not the only thing that can wake up a WFE, it is only an explicit software way to do it.
 
@@ -19544,9 +19551,9 @@ cd -
 ./bench-all -A
 ....
 
-=== Continuous integraion
+=== Continuous integration
 
-We have exploreed a few Continuous integration solutions.
+We have explored a few Continuous integration solutions.
 
 We haven't setup any of them yet.
 
@@ -19556,7 +19563,7 @@ We tried to automate it on Travis with link:.travis.yml[] but it hits the curren
 
 ==== CircleCI
 
-This setup sucessfully built gem5 on every commit: link:.circleci/config.yml[]
+This setup successfully built gem5 on every commit: link:.circleci/config.yml[]
 
 Enabling it is however blocked on: https://github.com/cirosantilli/linux-kernel-module-cheat/issues/79 so we disabled the builds on the web UI.
 
@@ -19714,6 +19721,14 @@ Summary of manually collected results on <<p51>> at LKMC a18f28e263c91362519ef55
 |1.10018162 * 10^8
 |1
 
+|a18f28e263c91362519ef550150b5c9d75fa3679 + 1
+|link:userland/gcc/busy_loop.c[] `-O0`
+|`gem5 --arch aarch64 --gem5-build-id debug`
+|10^5
+|32
+|2.528728 * 10^6
+|0.08
+
 |a18f28e263c91362519ef550150b5c9d75fa3679 + 1
 |link:userland/gcc/busy_loop.c[] `-O0`
 |`+gem5 --arch aarch64 -- --cpu-type MinorCPU --caches+`
@@ -19755,7 +19770,7 @@ Summary of manually collected results on <<p51>> at LKMC a18f28e263c91362519ef55
 |1.6
 
 |ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
-|link:userland/c/m5ops.c[] `-O0`
+|glibc C pre-main link:userland/c/m5ops.c[] `-O0`
 |`gem5 --arch aarch64 --userland-args e`
 |1
 |2
@@ -19763,13 +19778,45 @@ Summary of manually collected results on <<p51>> at LKMC a18f28e263c91362519ef55
 |0.05
 
 |ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
-|link:userland/cpp/m5ops.cpp[] `-O0`
+|glibc C pre-main link:userland/c/m5ops.c[] `-O0`
+|`gem5 --arch aarch64 --userland-args e --gem5-build-type debug`
+|1
+|2
+|1.26479 * 10^5
+|0.05
+
+|ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
+|glibc C++ pre-main link:userland/cpp/m5ops.cpp[] `-O0`
 |`gem5 --arch aarch64 --userland-args e`
 |1
 |2
 |2.385012 * 10^6
 |1
 
+|ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
+|glibc C++ pre-main link:userland/cpp/m5ops.cpp[] `-O0`
+|`gem5 --arch aarch64 --userland-args e --gem5-build-type debug`
+|1
+|25
+|2.385012 * 10^6
+|0.1
+
+|ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
+|immediate exit link:userland/arch/aarch64/freestanding/linux/gem5_exit.S[] `-O0`
+|`gem5 --arch aarch64`
+|1
+|1
+|
+|
+
+|ab6f7331406b22f8ab6e2df5f8b8e464fb35b611
+|immediate exit link:userland/arch/aarch64/freestanding/linux/gem5_exit.S[] `-O0`
+|`gem5 --arch aarch64 --gem5-build-type debug`
+|1
+|1
+|
+|
+
 |===
 
 The first step is to determine a number of loops that will run long enough to have meaningful results, but not too long that we will get bored, so about 1 minute.

path_properties.py

@@ -542,6 +542,7 @@ path_properties_tuples = (
                                         'linux': (
                                             {},
                                             {
+                                                'gem5_exit.S': {'allowed_emulators': {'gem5'}},
                                                 'wfe.S': {'more_than_1s': True},
                                                 'wfe_wfe.S': {'more_than_1s': True},
                                             }

userland/arch/aarch64/freestanding/linux/gem5_exit.S

@@ -0,0 +1,10 @@
+/* https://cirosantilli.com/linux-kernel-module-cheat#benchmark-emulators-on-userland-executables */
+
+#define LKMC_M5OPS_ENABLE 1
+#include "lkmc/m5ops.h"
+
+.text
+.global _start
+_start:
+asm_main_after_prologue:
+    LKMC_M5OPS_EXIT_ASM

userland/arch/aarch64/inline_asm/wfe_ldxr_str.cpp

@@ -0,0 +1,35 @@
+// https://cirosantilli.com/linux-kernel-module-cheat#arm-wfe-global-monitor-events
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+#include <atomic>
+#include <iostream>
+#include <mutex>
+#include <thread>
+
+#include <lkmc/futex.h>
+
+std::atomic_ulong done;
+int futex = 1;
+
+void myfunc() {
+    __asm__ __volatile__ ("ldxr x0, [%0];wfe;wfe" : : "r" (&futex) : "x0");
+    done.store(futex);
+}
+
+int main(int argc, char **argv) {
+    bool do_sev = true;
+    if (argc > 1) {
+        do_sev = (argv[1][0] != '0');
+    }
+    done.store(0);
+    std::thread thread;
+    thread = std::thread(myfunc);
+    while (!done.load()) {
+        if (do_sev) {
+            __asm__ __volatile__ ("mov x0, 1;str x0, [%0]" : : "r" (&futex) : "x0", "x1");
+        }
+    }
+    thread.join();
+}

userland/arch/aarch64/inline_asm/wfe_ldxr_stxr.cpp

@@ -28,7 +28,7 @@ int main(int argc, char **argv) {
     thread = std::thread(myfunc);
     while (!done.load()) {
         if (do_sev) {
-            __asm__ __volatile__ ("mov x0, 1;stxr w1, x0, [%0]" : : "r" (&futex) : "x0", "x1");
+            __asm__ __volatile__ ("mov x0, 1;ldxr x0, [%0];stxr w1, x0, [%0]" : : "r" (&futex) : "x0", "x1");
         }
     }
     thread.join();