ggml-qnn: rebase to upstream

Author: zhouwg

CMakeLists.txt

@@ -7,6 +7,16 @@ set(CMAKE_WARN_UNUSED_CLI YES)
 
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
+if(CMAKE_SYSTEM_NAME STREQUAL "Android")
+    set(TARGET_SNAPDRAGON8GEN3    ON)
+    if(TARGET_SNAPDRAGON8GEN3)
+       #works fine on Snapdragon 8Gen3 with 1.5x(45+ tokens/second)-3x(70+ tokens/second) performance gain through the default ggml backend
+       add_definitions(-march=armv8.7-a)
+       add_definitions(-mcpu=cortex-x1)
+       add_definitions(-mtune=cortex-x1)
+   endif()
+endif()
+
 if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
     set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
     set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
@@ -119,6 +129,7 @@ llama_option_depr(WARNING     LLAMA_RPC                 GGML_RPC)
 llama_option_depr(WARNING     LLAMA_SYCL                GGML_SYCL)
 llama_option_depr(WARNING     LLAMA_SYCL_F16            GGML_SYCL_F16)
 llama_option_depr(WARNING     LLAMA_CANN                GGML_CANN)
+llama_option_depr(WARNING     LLAMA_QNN                 GGML_QNN)
 
 if (NOT MSVC)
     if (LLAMA_SANITIZE_THREAD)