sve: Try to add ARM SVE compile support

This might be simple, this might be tough. Let's go step-by-step.

Signed-off-by: Johannes Demel <[email protected]>

Author: jdemel

gen/archs.xml

@@ -105,6 +105,20 @@ at the top, as a last resort.
   <check name="neon"></check>
 </arch>
 
+<arch name="sve">
+  <flag compiler="gnu">-march=armv8-a+sve</flag>
+  <flag compiler="clang">-march=armv8-a+sve</flag>
+  <alignment>16</alignment>
+  <check name="sve"></check>
+</arch>
+
+<arch name="sve2">
+  <flag compiler="gnu">-march=armv8-a+sve2</flag>
+  <flag compiler="clang">-march=armv8-a+sve2</flag>
+  <alignment>16</alignment>
+  <check name="sve2"></check>
+</arch>
+
 <arch name="sse3">
   <check name="sse3"></check>
   <flag compiler="gnu">-msse3</flag>

gen/machines.xml

@@ -16,6 +16,14 @@
 <archs>generic neon neonv8 orc|</archs>
 </machine>
 
+<machine name="sve">
+<archs>generic neon neonv8 sve orc|</archs>
+</machine>
+
+<machine name="sve2">
+<archs>generic neon neonv8 sve sve2 orc|</archs>
+</machine>
+
 <!-- trailing | bar means generate without either for MSVC -->
 <machine name="sse2">
 <archs>generic 32|64| mmx| sse sse2 orc|</archs>

lib/CMakeLists.txt

@@ -203,6 +203,8 @@ if(NOT CPU_IS_x86)
     overrule_arch(sse4_1 "Architecture is not x86 or x86_64")
     overrule_arch(sse4_2 "Architecture is not x86 or x86_64")
     overrule_arch(avx "Architecture is not x86 or x86_64")
+    overrule_arch(avx2 "Architecture is not x86 or x86_64")
+    overrule_arch(fma "Architecture is not x86 or x86_64")
     overrule_arch(avx512f "Architecture is not x86 or x86_64")
     overrule_arch(avx512cd "Architecture is not x86 or x86_64")
 endif(NOT CPU_IS_x86)
@@ -239,6 +241,8 @@ else(neon_compile_result)
     overrule_arch(neon "Compiler doesn't support NEON")
     overrule_arch(neonv7 "Compiler doesn't support NEON")
     overrule_arch(neonv8 "Compiler doesn't support NEON")
+    overrule_arch(sve "Compiler doesn't support SVE")
+    overrule_arch(sve2 "Compiler doesn't support SVE2")
 endif(neon_compile_result)
 
 ########################################################################

tmpl/volk_cpu.tmpl.c

@@ -44,6 +44,10 @@ static int i_can_has_${arch.name} (void) {
         %if "neon" in arch.name:
 #if defined(CPU_FEATURES_ARCH_ARM)
     if (GetArmInfo().features.${check} == 0){ return 0; }
+#endif
+        %elif "neon" in arch.name or "sve" in arch.name:
+#if defined(CPU_FEATURES_ARCH_AARCH64)
+    if (GetAarch64Info().features.${check} == 0){ return 0; }
 #endif
         %elif "mips" in arch.name:
 #if defined(CPU_FEATURES_ARCH_MIPS)