SVE enabled on user input. updated README

tune and arch flags will be applied from autodetect only if they have been created by the process, otherwise the old logical flow remains wrt the flags

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 011bfec56bee2dc1b1f3b5fc779aeb745433f50d..b26fcc40db12042d6b98fb9d30c2d8f0725184f9 100644 (file)
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -171,7 +171,7 @@ if (CMAKE_COMPILER_IS_GNUCC AND NOT CROSS_COMPILE)
 
     # arg1 might exist if using ccache
     string (STRIP "${CMAKE_C_COMPILER_ARG1}" CC_ARG1)
-    set (EXEC_ARGS ${CC_ARG1} -c -Q --help=target -${ARCH_FLAG}=native -mtune=native)
+    set (EXEC_ARGS ${CC_ARG1} -c -Q --help=target -${ARCH_FLAG}=native -${TUNE_FLAG}=native)
     execute_process(COMMAND ${CMAKE_C_COMPILER} ${EXEC_ARGS}
         OUTPUT_VARIABLE _GCC_OUTPUT)
     set(_GCC_OUTPUT_TUNE ${_GCC_OUTPUT})
@@ -187,11 +187,11 @@ if (CMAKE_COMPILER_IS_GNUCC AND NOT CROSS_COMPILE)
     string(FIND "${GNUCC_ARCH}" "sve2" POS_SVE2)
     string(FIND "${GNUCC_ARCH}" "sve2-bitperm" POS_SVE2_BITPERM)
     if (NOT POS_SVE EQUAL 0)
-        set(BUILD_SVE 1)
+        set(SVE_FOUND 1)
     elseif(NOT POS_SVE2 EQUAL 0)
-        set(BUILD_SVE2 1)
+        set(SVE2_FOUND 1)
     elseif(NOT POS_SVE2_BITPERM EQUAL 0)
-        set(BUILD_SVE2_BITPERM 1)
+        set(SVE2_BITPERM_FOUND 1)
     endif()
 
     message(STATUS "ARCH_FLAG '${ARCH_FLAG}' '${GNUCC_ARCH}', TUNE_FLAG '${TUNE_FLAG}' '${GNUCC_TUNE}' ")
@@ -203,11 +203,11 @@ if (CMAKE_COMPILER_IS_GNUCC AND NOT CROSS_COMPILE)
         INPUT_FILE /dev/null
         RESULT_VARIABLE GNUCC_TUNE_TEST)
     if (NOT GNUCC_TUNE_TEST EQUAL 0)
-        message(WARNING "Something went wrong determining gcc tune: -mtune=${GNUCC_ARCH} not valid, falling back to -mtune=native")
+        message(WARNING "Something went wrong determining gcc tune: -mtune=${GNUCC_TUNE} not valid, falling back to -mtune=native")
         set(GNUCC_TUNE native)
     else()
         set(GNUCC_TUNE ${GNUCC_TUNE})
-        message(STATUS "gcc will tune for ${GNUCC_ARCH}, ${TUNE_FLAG}")
+        message(STATUS "gcc will tune for ${GNUCC_ARCH}, ${GNUCC_TUNE}")
     endif()
 elseif (CMAKE_COMPILER_IS_CLANG AND NOT CROSS_COMPILE)
     if (ARCH_IA32 OR ARCH_X86_64)
@@ -247,12 +247,27 @@ if (ARCH_IA32 OR ARCH_X86_64)
     endif()
 endif()
 
+if (ARCH_AARCH64)
+    if (BUILD_SVE2_BITPERM AND NOT SVE2_BITPERM_FOUND)
+        set(GNUCC_ARCH "${GNUCC_ARCH}+sve2-bitperm")
+    elseif (BUILD_SVE2 AND NOT SVE2_FOUND)
+        set(GNUCC_ARCH "${GNUCC_ARCH}+sve2")
+    elseif (BUILD_SVE AND NOT SVE_FOUND)
+        set(GNUCC_ARCH "${GNUCC_ARCH}+sve")
+    endif ()
+endif(ARCH_AARCH64)
+
 message(STATUS "ARCH_C_FLAGS   : ${ARCH_C_FLAGS}")
 message(STATUS "ARCH_CXX_FLAGS : ${ARCH_CXX_FLAGS}")
 
 if (NOT FAT_RUNTIME)
-    set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
-    set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
+    if (GNUCC_TUNE)
+        set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
+        set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -${TUNE_FLAG}=${GNUCC_TUNE}")
+    else()
+        set(ARCH_C_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_C_FLAGS}")
+        set(ARCH_CXX_FLAGS "-${ARCH_FLAG}=${GNUCC_ARCH} -mtune=${TUNE_FLAG} ${ARCH_CXX_FLAGS}")
+    endif()
 endif()
 
 #if (ARCH_IA32 OR ARCH_X86_64 OR ARCH_ARM32 OR ARCH_AARCH64)

diff --git a/README.md b/README.md
index f9d2708f4ad1572395d82244834623c80a8adb14..8bc7aff64540df34d73c19cccd72debcff40988b 100644 (file)
--- a/README.md
+++ b/README.md
@@ -47,8 +47,6 @@ Vectorscan is typically used in a DPI library stack, just like Hyperscan.
 
 # Compiling for SVE
 
-When compiling on AARCH64 machine with support for either of the SVE flags, it will be detected and applied automatically.
-
 The following cmake variables can be set in order to target Arm's Scalable
 Vector Extension. They are listed in ascending order of strength, with cmake
 detecting whether the feature is available in the compiler and falling back to