From d1139b700db0ed640e216f76564278f14fd83dca Mon Sep 17 00:00:00 2001
From: Thorsten Hater <24411438+thorstenhater@users.noreply.github.com>
Date: Wed, 3 May 2023 13:15:31 +0200
Subject: [PATCH] Add LTO and M1 arch to CMake (#2106)

Tweak CMake to set up Link Time Optimizations (helps pybind11) and set
arch/tune for Arm Macs (M1 for now).
---
 CMakeLists.txt                                 | 9 +++++++++
 arbor/backends/multicore/threshold_watcher.hpp | 1 +
 cmake/CompilerOptions.cmake                    | 9 ++++-----
 3 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index c4aa73c9b0..d2a7d20acc 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,5 +1,6 @@
 cmake_minimum_required(VERSION 3.19)
 include(CMakeDependentOption)
+include(CheckIPOSupported)
 
 # Make CUDA support throw errors if architectures remain unclear
 cmake_policy(SET CMP0104 NEW)
@@ -16,6 +17,14 @@ include(GNUInstallDirs)
 # Effectively adds '-fpic' flag to CXX_FLAGS. Needed for dynamic catalogues.
 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 
+# Have LTO where possible, ie add -flto
+check_ipo_supported(RESULT HAVE_LTO OUTPUT ERR_LTO)
+if(HAVE_LTO AND NOT DEFINED CMAKE_INTERPROCEDURAL_OPTIMIZATION)
+  message (STATUS "LTO support found, enabling")
+  set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE)
+else()
+  message(STATUS "No LTO: ${ERR_LTO}")
+endif()
 
 # Turn on this option to force the compilers to produce color output when output is
 # redirected from the terminal (e.g. when using ninja or a pager).
diff --git a/arbor/backends/multicore/threshold_watcher.hpp b/arbor/backends/multicore/threshold_watcher.hpp
index b3419cdbb8..ed075ba8b6 100644
--- a/arbor/backends/multicore/threshold_watcher.hpp
+++ b/arbor/backends/multicore/threshold_watcher.hpp
@@ -58,6 +58,7 @@ class threshold_watcher {
     /// Crossing events are recorded for each threshold that
     /// is crossed since the last call to test
     void test(array& time_since_spike, const arb_value_type& t_before, const arb_value_type& t_after) {
+        if (cv_index_.empty() || n_detectors_ == 0) return;
         arb_assert(values_!=nullptr);
 
         // Reset all spike times to -1.0 indicating no spike has been recorded on the detector
diff --git a/cmake/CompilerOptions.cmake b/cmake/CompilerOptions.cmake
index 5a8b2bcab0..824bb4e45e 100644
--- a/cmake/CompilerOptions.cmake
+++ b/cmake/CompilerOptions.cmake
@@ -110,12 +110,11 @@ function(set_arch_target optvar arch)
         endforeach(line)
         string(REGEX REPLACE "-.*" "" target_model "${target}")
 
-        # Use -mcpu for all supported targets _except_ for x86, where it should be -march.
-
-        if("${target}" MATCHES "aarch64-apple-darwin")
-            set(arch_opt "-mcpu=${arch}")
+        # Use -mcpu for all supported targets _except_ for x86 and Apple arm64, where it should be -march.
+        if("${target}" MATCHES "aarch64-apple-darwin" OR "${target}" MATCHES "arm64-apple-darwin")
+            set(arch_opt "-march=${arch} -mtune=${arch}")
         elseif(target_model MATCHES "x86|i[3456]86" OR target_model MATCHES "amd64" OR target_model MATCHES "aarch64")
-            set(arch_opt "-march=${arch}")
+            set(arch_opt "-march=${arch} -mtune=${arch}")
         else()
             set(arch_opt "-mcpu=${arch}")
         endif()