Refactor MNN KleidiAI interface

Refactor MNN_KleidiAI interface to support more model types,
and facilitate subsequent KleidiAI ukernels' integration.

Re-abstract information stored in class KleidiAI:
1) static info: not related to loaded model, initialized when
interface is constructed and never changed.
2) status: will change while pipeline is running.

Let interface and loaded model decouple for more complex mix
of multiple types of models. Add mAccelType in MNN data structure,
kleidiAI interface will rely on this type to decide which branch
to go.

Move some pack functions to mnn_kleidiai_util.cpp.

Add CPU feature detection in source/backend/cpu/CPURuntime.hpp.
Subsequent ukernels need SME information.

source/backend/cpu/CPURuntime.cpp

@@ -30,7 +30,9 @@
 // ref: https://cs.android.com/android/platform/superproject/+/master:bionic/libc/kernel/uapi/asm-arm64/asm/hwcap.h;drc=04da58f5b3bc40dbbafb4f8422aa2991479d9e1e;l=70
 #define CPUINFO_ARM_LINUX_FEATURE_I8MM UINT32_C(0x00002000)
 #define CPUINFO_ARM_LINUX_FEATURE_SVE UINT32_C(0x00400000)
-#define CPUINFO_ARM_LINUX_FEATURE_SVE2 UINT32_C(0x00000002)
+
+#define CPUINFO_ARM_LINUX_FEATURE2_SVE2 UINT32_C(0x00000002)
+#define CPUINFO_ARM_LINUX_FEATURE2_SME2 UINT64_C(0x0000002000000000)
 #endif
 
 #include <algorithm>
@@ -1281,13 +1283,18 @@ static void _getInfoApple(MNNCPUInfo* cpuinfo_isa) {
     if (have_feature("hw.optional.arm.FEAT_I8MM")) {
         cpuinfo_isa->i8mm = true;
     }
+    if (have_feature("hw.optional.arm.FEAT_SME2")) {
+        cpuinfo_isa->sme2 = true;
+    }
 }
 #endif
 
 #if defined(__linux__) && defined(__aarch64__)
 static void _getInfoAux(MNNCPUInfo* cpuinfo_isa) {
     // Use AUX to get info for linux-aarch64
     uint32_t isa_features = 0;
+    uint64_t isa_features2 = 0;
+
     isa_features = (uint32_t)getauxval(AT_HWCAP);
     if (isa_features & CPUINFO_ARM_LINUX_FEATURE_ASIMDDP) {
         cpuinfo_isa->dot = true;
@@ -1299,10 +1306,14 @@ static void _getInfoAux(MNNCPUInfo* cpuinfo_isa) {
     if (isa_features & CPUINFO_ARM_LINUX_FEATURE_I8MM) {
         cpuinfo_isa->i8mm = true;
     }
-    isa_features = (uint32_t)getauxval(AT_HWCAP2);
-    if (isa_features & CPUINFO_ARM_LINUX_FEATURE_SVE2) {
+
+    isa_features2 = (uint64_t)getauxval(AT_HWCAP2);
+    if (isa_features & CPUINFO_ARM_LINUX_FEATURE2_SVE2) {
         cpuinfo_isa->sve2 = true;
     }
+    if (isa_features & CPUINFO_ARM_LINUX_FEATURE2_SME2) {
+        cpuinfo_isa->sme2 = true;
+    }
 }
 #endif
 
@@ -1353,6 +1364,7 @@ static void _fillInfo(MNNCPUInfo* cpuinfo_isa) {
     cpuinfo_isa->fp16arith = false;
     cpuinfo_isa->i8mm = false;
     cpuinfo_isa->sve2 = false;
+    cpuinfo_isa->sme2 = false;
     // android
     /**Get CPU Info*/
 #ifdef __linux__
@@ -1449,6 +1461,7 @@ static void _fillInfo(MNNCPUInfo* cpuinfo_isa) {
     cpuinfo_isa->dot = true;
 #endif
 
-    MNN_PRINT("The device supports: i8sdot:%d, fp16:%d, i8mm: %d, sve2: %d\n", cpuinfo_isa->dot, cpuinfo_isa->fp16arith, cpuinfo_isa->i8mm, cpuinfo_isa->sve2);
+    MNN_PRINT("The device supports: i8sdot:%d, fp16:%d, i8mm: %d, sve2: %d, sme2: %d\n",
+            cpuinfo_isa->dot, cpuinfo_isa->fp16arith, cpuinfo_isa->i8mm, cpuinfo_isa->sve2, cpuinfo_isa->sme2);
     return;
 }

source/backend/cpu/CPURuntime.hpp

@@ -21,6 +21,7 @@ struct MNNCPUInfo {
     bool dot;
     bool i8mm;
     bool sve2;
+    bool sme2;
     std::vector<CPUGroup> groups;
     int cpuNumber = 0;
 };

source/backend/cpu/arm/CMakeLists.txt

@@ -47,7 +47,9 @@ if (MNN_KLEIDIAI)
     endif()
 
     list(APPEND MNN_SOURCES_KLEIDIAI ${CMAKE_CURRENT_LIST_DIR}/mnn_kleidiai.cpp)
+    list(APPEND MNN_SOURCES_KLEIDIAI ${CMAKE_CURRENT_LIST_DIR}/mnn_kleidiai_util.cpp)
     list(APPEND MNN_HEADERS_KLEIDIAI ${CMAKE_CURRENT_LIST_DIR}/mnn_kleidiai.h)
+    list(APPEND MNN_HEADERS_KLEIDIAI ${CMAKE_CURRENT_LIST_DIR}/mnn_kleidiai_util.h)
 
     # KleidiAI
     include_directories(