core: restructure Kleidi-Llama library

examples/llama.android/app/src/main/java/com/example/llama/engine/StubTierDetection.kt

@@ -10,8 +10,7 @@ import android.util.Log
 object StubTierDetection : TierDetection {
     private val tag = StubTierDetection::class.java.simpleName
 
-    override val detectedTier: LLamaTier?
-        get() = LLamaTier.T2
+    override fun getDetectedTier(): LLamaTier? = LLamaTier.T2
 
     override fun clearCache() {
         Log.d(tag, "Cache cleared")

examples/llama.android/app/src/main/java/com/example/llama/viewmodel/SettingsViewModel.kt

@@ -67,7 +67,7 @@ class SettingsViewModel @Inject constructor(
     val darkThemeMode: StateFlow<DarkThemeMode> = _darkThemeMode.asStateFlow()
 
     val detectedTier: LLamaTier?
-        get() = tierDetection.detectedTier
+        get() = tierDetection.getDetectedTier()
 
     init {
         viewModelScope.launch {

examples/llama.android/llama/src/main/cpp/CMakeLists.txt

@@ -17,8 +17,8 @@ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}" CACHE STRING "" FORCE)
 add_subdirectory(
         ${CMAKE_CURRENT_LIST_DIR}/../../../../../../include/cpu_features
         ${CMAKE_BINARY_DIR}/cpu_features_build)
-add_library(llama_cpu_detector SHARED cpu_detector.cpp)
-target_link_libraries(llama_cpu_detector
+add_library(kleidi-llama-cpu-detector SHARED cpu_detector.cpp)
+target_link_libraries(kleidi-llama-cpu-detector
         PRIVATE CpuFeatures::cpu_features
         android
         log)

examples/llama.android/llama/src/main/cpp/cpu_detector.cpp

@@ -12,7 +12,7 @@ static const Aarch64Info info = GetAarch64Info();
 static const Aarch64Features features = info.features;
 
 extern "C" JNIEXPORT jint JNICALL
-Java_android_llama_cpp_internal_InferenceEngineLoader_getOptimalTier(
+Java_android_llama_cpp_internal_TierDetectionImpl_getOptimalTier(
         JNIEnv* env,
         jclass clazz) {
     int tier = 0;  // Default to T0 (baseline)
@@ -46,7 +46,7 @@ Java_android_llama_cpp_internal_InferenceEngineLoader_getOptimalTier(
 
 // Optional: Keep a feature string function for debugging
 extern "C" JNIEXPORT jstring JNICALL
-Java_android_llama_cpp_internal_InferenceEngineLoader_getCpuFeaturesString(
+Java_android_llama_cpp_internal_TierDetectionImpl_getCpuFeaturesString(
         JNIEnv* env,
         jclass clazz) {
     std::string text;

examples/llama.android/llama/src/main/java/android/llama/cpp/ArmFeatures.kt

@@ -84,7 +84,7 @@ object ArmFeaturesMapper {
             LLamaTier.T0 -> listOf(true, false, false, false, false)  // ASIMD only
             LLamaTier.T1 -> listOf(true, true, false, false, false)   // ASIMD + DOTPROD
             LLamaTier.T2 -> listOf(true, true, true, false, false)    // ASIMD + DOTPROD + I8MM
-            LLamaTier.T3 -> listOf(true, true, true, true, false)     // ASIMD + DOTPROD + I8MM + SVE
-            // TODO-han.yin: implement T4 once obtaining an Android device with SME!
+            LLamaTier.T3 -> listOf(true, true, true, true, false)     // ASIMD + DOTPROD + I8MM + SVE/2
+            LLamaTier.T4 -> listOf(true, true, true, true, true)     // ASIMD + DOTPROD + I8MM + SVE/2 + SME/2
         }
 }

examples/llama.android/llama/src/main/java/android/llama/cpp/KleidiLlama.kt

@@ -2,6 +2,7 @@ package android.llama.cpp
 
 import android.content.Context
 import android.llama.cpp.internal.InferenceEngineFactory
+import android.llama.cpp.internal.TierDetectionImpl
 
 /**
  * Main entry point for the Llama Android library.
@@ -11,12 +12,10 @@ object KleidiLlama {
     /**
      * Create an inference engine instance with automatic tier detection.
      */
-    fun createInferenceEngine(context: Context) =
-        InferenceEngineFactory.getInstance(context)
+    fun createInferenceEngine(context: Context) = InferenceEngineFactory.getInstance(context)
 
     /**
      * Get tier detection information for debugging/settings.
      */
-    fun getTierDetection(context: Context) =
-        InferenceEngineFactory.getTierDetection(context)
+    fun getTierDetection(context: Context): TierDetection = TierDetectionImpl(context)
 }

examples/llama.android/llama/src/main/java/android/llama/cpp/TierDetection.kt

@@ -4,7 +4,7 @@ package android.llama.cpp
  * Public interface for [LLamaTier] detection information.
  */
 interface TierDetection {
-    val detectedTier: LLamaTier?
+    fun getDetectedTier(): LLamaTier?
     fun clearCache()
 }
 
@@ -17,8 +17,8 @@ enum class LLamaTier(val rawValue: Int, val libraryName: String, val description
     T0(0, "llama_android_t0", "ARMv8-a baseline with ASIMD"),
     T1(1, "llama_android_t1", "ARMv8.2-a with DotProd"),
     T2(2, "llama_android_t2", "ARMv8.6-a with DotProd + I8MM"),
-    T3(3, "llama_android_t3", "ARMv9-a with DotProd + I8MM + SVE/SVE2");
-    // TODO-han.yin: implement T4 once obtaining an Android device with SME!
+    T3(3, "llama_android_t3", "ARMv9-a with DotProd + I8MM + SVE/SVE2"),
+    T4(4, "llama_android_t4", "ARMv9.2-a with DotProd + I8MM + SVE/SVE2 + SME/SME2");
 
     companion object {
         fun fromRawValue(value: Int): LLamaTier? = entries.find { it.rawValue == value }

examples/llama.android/llama/src/main/java/android/llama/cpp/internal/InferenceEngineFactory.kt

@@ -1,13 +1,45 @@
 package android.llama.cpp.internal
 
 import android.content.Context
+import android.llama.cpp.InferenceEngine
 import android.llama.cpp.TierDetection
+import android.util.Log
+import kotlinx.coroutines.runBlocking
 
 /**
  * Internal factory to create [InferenceEngine] and [TierDetection]
  */
 internal object InferenceEngineFactory {
-    fun getInstance(context: Context) = InferenceEngineLoader.getInstance(context)
+    private val TAG = InferenceEngineFactory::class.simpleName
 
-    fun getTierDetection(context: Context): TierDetection = TierDetectionImpl(context)
+    private var _cachedInstance: InferenceEngineImpl? = null
+
+    /**
+     * Factory method to get a configured [InferenceEngineImpl] instance.
+     * Handles tier detection, caching, and library loading automatically.
+     */
+    @Synchronized
+    fun getInstance(context: Context): InferenceEngine? {
+        // Return cached instance if available
+        _cachedInstance?.let { return it }
+
+        return runBlocking {
+            try {
+                // Create and cache the inference engine instance
+                InferenceEngineImpl.create(context).also {
+                    _cachedInstance = it
+                    Log.i(TAG, "Successfully instantiated Inference Engine")
+                }
+
+            } catch (e: Exception) {
+                Log.e(TAG, "Error instantiating Inference Engine", e)
+                null
+            }
+        }
+    }
+
+    fun clearCache() {
+        _cachedInstance = null
+        Log.i(TAG, "Cleared cached instance of InferenceEngine")
+    }
 }

examples/llama.android/llama/src/main/java/android/llama/cpp/internal/InferenceEngineLoader.kt

@@ -1,159 +0,0 @@
-package android.llama.cpp.internal
-
-import android.content.Context
-import android.llama.cpp.InferenceEngine
-import android.llama.cpp.LLamaTier
-import android.util.Log
-import androidx.datastore.core.DataStore
-import androidx.datastore.preferences.core.Preferences
-import androidx.datastore.preferences.core.edit
-import androidx.datastore.preferences.core.intPreferencesKey
-import androidx.datastore.preferences.preferencesDataStore
-import kotlinx.coroutines.flow.first
-import kotlinx.coroutines.runBlocking
-
-/**
- * Internal [android.llama.cpp.InferenceEngine] loader implementation
- */
-internal object InferenceEngineLoader {
-    private val TAG = InferenceEngineLoader::class.simpleName
-
-    // CPU feature detection preferences
-    private const val DATASTORE_CPU_DETECTION = "llama_cpu_detection"
-    private val Context.llamaTierDataStore: DataStore<Preferences>
-        by preferencesDataStore(name = DATASTORE_CPU_DETECTION)
-
-    private val DETECTION_VERSION = intPreferencesKey("detection_version")
-    private val DETECTED_TIER = intPreferencesKey("detected_tier")
-
-    // Constants
-    private const val DATASTORE_VERSION = 1
-
-    @JvmStatic
-    private external fun getOptimalTier(): Int
-
-    @JvmStatic
-    private external fun getCpuFeaturesString(): String
-
-    private var _cachedInstance: InferenceEngineImpl? = null
-    private var _detectedTier: LLamaTier? = null
-
-    /**
-     * Get the detected tier, loading from cache if needed
-     */
-    fun getDetectedTier(context: Context): LLamaTier? =
-        _detectedTier ?: runBlocking { obtainTier(context) }
-
-    /**
-     * Factory method to get a configured [InferenceEngineImpl] instance.
-     * Handles tier detection, caching, and library loading automatically.
-     */
-    @Synchronized
-    fun getInstance(context: Context): InferenceEngine? {
-        // Return cached instance if available
-        _cachedInstance?.let { return it }
-
-        return runBlocking {
-            try {
-                // Create and cache the inference engine instance
-                InferenceEngineImpl.create(context).also {
-                    _cachedInstance = it
-                    Log.i(TAG, "Successfully instantiated Inference Engine")
-                }
-
-            } catch (e: Exception) {
-                Log.e(TAG, "Error instantiating Inference Engine", e)
-                null
-            }
-        }
-    }
-
-    /**
-     * First attempt to load detected tier from storage, if available;
-     * Otherwise, perform a fresh detection, then save to storage and cache.
-     */
-    private suspend fun obtainTier(context: Context) =
-        loadDetectedTierFromDataStore(context) ?: run {
-            Log.i(TAG, "Performing fresh tier detection")
-            performOptimalTierDetection().also { tier ->
-                tier?.saveToDataStore(context)
-                _detectedTier = tier
-            }
-        }
-
-    /**
-     * Load cached tier from datastore without performing detection
-     */
-    private suspend fun loadDetectedTierFromDataStore(context: Context): LLamaTier? {
-        val preferences = context.llamaTierDataStore.data.first()
-        val cachedVersion = preferences[DETECTION_VERSION] ?: -1
-        val cachedTierValue = preferences[DETECTED_TIER] ?: -1
-
-        return if (cachedVersion == DATASTORE_VERSION && cachedTierValue >= 0) {
-            LLamaTier.fromRawValue(cachedTierValue)?.also {
-                Log.i(TAG, "Loaded cached tier: ${it.name}")
-                _detectedTier = it
-            }
-        } else {
-            Log.i(TAG, "No valid cached tier found")
-            null
-        }
-    }
-
-    /**
-     * Actual implementation of optimal tier detection via native methods
-     */
-    private fun performOptimalTierDetection(): LLamaTier? {
-        try {
-            // Load CPU detection library
-            System.loadLibrary("llama_cpu_detector")
-            Log.i(TAG, "CPU feature detector loaded successfully")
-
-            // Detect optimal tier
-            val tierValue = getOptimalTier()
-            val features = getCpuFeaturesString()
-            Log.i(TAG, "Raw tier $tierValue w/ CPU features: $features")
-
-            // Convert to enum and validate
-            val tier = LLamaTier.fromRawValue(tierValue) ?: run {
-                Log.e(TAG, "Invalid tier value $tierValue")
-                return LLamaTier.NONE
-            }
-
-            // Ensure we don't exceed maximum supported tier
-            val maxTier = LLamaTier.maxSupportedTier
-            return if (tier.rawValue > maxTier.rawValue) {
-                Log.w(TAG, "Detected tier ${tier.name} exceeds max supported, using ${maxTier.name}")
-                maxTier
-            } else {
-                tier
-            }
-
-        } catch (e: UnsatisfiedLinkError) {
-            Log.e(TAG, "Failed to load CPU detection library", e)
-            return null
-
-        } catch (e: Exception) {
-            Log.e(TAG, "Unexpected error during tier detection", e)
-            return null
-        }
-    }
-
-    /**
-     * Clear cached detection results (for testing/debugging)
-     */
-    fun clearCache(context: Context) {
-        runBlocking { context.llamaTierDataStore.edit { it.clear() } }
-        _cachedInstance = null
-        _detectedTier = null
-        Log.i(TAG, "Cleared detection results and cached instance")
-    }
-
-    private suspend fun LLamaTier.saveToDataStore(context: Context) {
-        context.llamaTierDataStore.edit { prefs ->
-            prefs[DETECTED_TIER] = this.rawValue
-            prefs[DETECTION_VERSION] = DATASTORE_VERSION
-        }
-        Log.i(TAG, "Saved ${this.name} to data store")
-    }
-}

examples/llama.android/llama/src/main/java/android/llama/cpp/internal/TierDetectionImpl.kt

@@ -3,13 +3,132 @@ package android.llama.cpp.internal
 import android.content.Context
 import android.llama.cpp.LLamaTier
 import android.llama.cpp.TierDetection
+import android.util.Log
+import androidx.datastore.core.DataStore
+import androidx.datastore.preferences.core.Preferences
+import androidx.datastore.preferences.core.edit
+import androidx.datastore.preferences.core.intPreferencesKey
+import androidx.datastore.preferences.preferencesDataStore
+import kotlinx.coroutines.flow.first
+import kotlinx.coroutines.runBlocking
 
 /**
- * Internal tier detection implementation
+ * Internal [LLamaTier] detection implementation
  */
-internal class TierDetectionImpl(private val context: Context) : TierDetection {
-    override val detectedTier: LLamaTier?
-        get() = InferenceEngineLoader.getDetectedTier(context)
+internal class TierDetectionImpl(
+    private val context: Context
+): TierDetection {
 
-    override fun clearCache() = InferenceEngineLoader.clearCache(context)
+    companion object {
+        private val TAG = TierDetectionImpl::class.simpleName
+
+        // CPU feature detection preferences
+        private const val DATASTORE_CPU_DETECTION = "llama_cpu_detection"
+        private const val DATASTORE_VERSION = 1
+        private val Context.llamaTierDataStore: DataStore<Preferences>
+            by preferencesDataStore(name = DATASTORE_CPU_DETECTION)
+
+        private val DETECTION_VERSION = intPreferencesKey("detection_version")
+        private val DETECTED_TIER = intPreferencesKey("detected_tier")
+    }
+
+    private external fun getOptimalTier(): Int
+
+    private external fun getCpuFeaturesString(): String
+
+    private var _detectedTier: LLamaTier? = null
+
+    /**
+     * Get the detected tier, loading from cache if needed
+     */
+    override fun getDetectedTier(): LLamaTier? =
+        _detectedTier ?: runBlocking { obtainTier() }
+
+    /**
+     * First attempt to load detected tier from storage, if available;
+     * Otherwise, perform a fresh detection, then save to storage and cache.
+     */
+    private suspend fun obtainTier() =
+        loadDetectedTierFromDataStore() ?: run {
+            Log.i(TAG, "Performing fresh tier detection")
+            performOptimalTierDetection().also { tier ->
+                tier?.saveToDataStore()
+                _detectedTier = tier
+            }
+        }
+
+    /**
+     * Load cached tier from datastore without performing detection
+     */
+    private suspend fun loadDetectedTierFromDataStore(): LLamaTier? {
+        val preferences = context.llamaTierDataStore.data.first()
+        val cachedVersion = preferences[DETECTION_VERSION] ?: -1
+        val cachedTierValue = preferences[DETECTED_TIER] ?: -1
+
+        return if (cachedVersion == DATASTORE_VERSION && cachedTierValue >= 0) {
+            LLamaTier.fromRawValue(cachedTierValue)?.also {
+                Log.i(TAG, "Loaded cached tier: ${it.name}")
+                _detectedTier = it
+            }
+        } else {
+            Log.i(TAG, "No valid cached tier found")
+            null
+        }
+    }
+
+    /**
+     * Actual implementation of optimal tier detection via native methods
+     */
+    private fun performOptimalTierDetection(): LLamaTier? {
+        try {
+            // Load CPU detection library
+            System.loadLibrary("kleidi-llama-cpu-detector")
+            Log.i(TAG, "CPU feature detector loaded successfully")
+
+            // Detect optimal tier
+            val tierValue = getOptimalTier()
+            val features = getCpuFeaturesString()
+            Log.i(TAG, "Raw tier $tierValue w/ CPU features: $features")
+
+            // Convert to enum and validate
+            val tier = LLamaTier.fromRawValue(tierValue) ?: run {
+                Log.e(TAG, "Invalid tier value $tierValue")
+                return LLamaTier.NONE
+            }
+
+            // Ensure we don't exceed maximum supported tier
+            val maxTier = LLamaTier.maxSupportedTier
+            return if (tier.rawValue > maxTier.rawValue) {
+                Log.w(TAG, "Detected tier ${tier.name} exceeds max supported, using ${maxTier.name}")
+                maxTier
+            } else {
+                tier
+            }
+
+        } catch (e: UnsatisfiedLinkError) {
+            Log.e(TAG, "Failed to load CPU detection library", e)
+            return null
+
+        } catch (e: Exception) {
+            Log.e(TAG, "Unexpected error during tier detection", e)
+            return null
+        }
+    }
+
+    /**
+     * Clear cached detection results (for testing/debugging)
+     */
+    override fun clearCache() {
+        runBlocking { context.llamaTierDataStore.edit { it.clear() } }
+        _detectedTier = null
+        Log.i(TAG, "Cleared CPU detection results")
+    }
+
+    private suspend fun LLamaTier.saveToDataStore() {
+        context.llamaTierDataStore.edit { prefs ->
+            prefs[DETECTED_TIER] = this.rawValue
+            prefs[DETECTION_VERSION] = DATASTORE_VERSION
+        }
+        Log.i(TAG, "Saved ${this.name} to data store")
+    }
 }