cmake : update CMakeLists.txt to add correct flags (#26)

* modify src/CMakeLists.txt from whisper.cpp

* cmake : remove OpenBLAS stuff

---------

Co-authored-by: Georgi Gerganov <redacted>

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index da46be38f59b800fac351933d38bf40585f72d97..466e2117db4982c9922574b926cf1cf14eea38f2 100644 (file)
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -25,12 +25,94 @@ endif()
 
 message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
 
+if (NOT UNAME_S)
+    execute_process(COMMAND uname -s OUTPUT_VARIABLE UNAME_S)
+endif()
+if (NOT UNAME_P)
+    execute_process(COMMAND uname -p OUTPUT_VARIABLE UNAME_P)
+endif()
+if (NOT UNAME_M)
+    execute_process(COMMAND uname -m OUTPUT_VARIABLE UNAME_M)
+endif()
+message(STATUS "UNAME_S: ${UNAME_S}  UNAME_P: ${UNAME_P}  UNAME_M: ${UNAME_M}")
+
+# Mac OS + Arm can report x86_64
+# ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
+if (UNAME_S MATCHES "Darwin")
+    if (NOT UNAME_P MATCHES "arm")
+        execute_process(COMMAND sysctl -n hw.optional.arm64 OUTPUT_VARIABLE SYSCTL_M)
+       if (SYSCTL_M MATCHES "1")
+            #set(UNAME_P "arm")
+            #set(UNAME_M "arm64")
+           message(WARNING "Your arch is announced as x86_64, but it seems to actually be ARM64. Not fixing that can lead to bad performance. For more info see: https://github.com/ggerganov/whisper.cpp/issues/66\#issuecomment-#1282546789")
+       endif()
+    endif()
+endif()
+
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
     message(STATUS "ARM detected")
     #set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=apple-m1")
 else()
     message(STATUS "x86 detected")
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx -mavx2 -mfma -mf16c")
+    #set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx -mavx2 -mfma -mf16c")
+    if (UNAME_S MATCHES "Darwin")
+        execute_process(COMMAND sysctl machdep.cpu.features OUTPUT_VARIABLE AVX1_M)
+        if (AVX1_M MATCHES "AVX1.0")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx")
+        endif()
+       execute_process(COMMAND sysctl machdep.cpu.leaf7_features OUTPUT_VARIABLE AVX2_M)
+        if (AVX2_M MATCHES "AVX2")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx2")
+        endif()
+       if (AVX1_M MATCHES "FMA")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfma")
+        endif()
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mf16c")
+elseif (UNAME_S MATCHES "Linux")
+        message(STATUS "Linux detected")
+       execute_process(COMMAND grep "avx " /proc/cpuinfo OUTPUT_VARIABLE AVX1_M)
+        if (AVX1_M MATCHES "avx")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx")
+        endif()
+       execute_process(COMMAND grep "avx2 " /proc/cpuinfo OUTPUT_VARIABLE AVX2_M)
+        if (AVX2_M MATCHES "avx2")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx2")
+        endif()
+       execute_process(COMMAND grep "fma " /proc/cpuinfo OUTPUT_VARIABLE FMA_M)
+       if (FMA_M MATCHES "fma")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfma")
+        endif()
+       execute_process(COMMAND grep "f16c " /proc/cpuinfo OUTPUT_VARIABLE F16C_M)
+       if (F16C_M MATCHES "f16c")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mf16c")
+        endif()
+       execute_process(COMMAND grep "sse3 " /proc/cpuinfo OUTPUT_VARIABLE SSE3_M)
+       if (SSE3_M MATCHES "sse3")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse3")
+        endif()
+       message(STATUS "CMAKE_C_FLAGS: ${CMAKE_C_FLAGS}")
+elseif (UNAME_S MATCHES "Haiku")
+       message(STATUS "Haiku detected")
+       execute_process(COMMAND sysinfo -cpu | grep "AVX " OUTPUT_VARIABLE AVX1_M)
+        if (AVX1_M MATCHES "avx")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx")
+        endif()
+       execute_process(COMMAND sysinfo -cpu | grep "AVX2 " OUTPUT_VARIABLE AVX2_M)
+        if (AVX2_M MATCHES "avx2")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx2")
+        endif()
+       execute_process(COMMAND sysinfo -cpu | grep "FMA " OUTPUT_VARIABLE FMA_M)
+       if (FMA_M MATCHES "fma")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfma")
+        endif()
+       execute_process(COMMAND sysinfo -cpu | grep "F16C " OUTPUT_VARIABLE F16C_M)
+       if (F16C_M MATCHES "f16c")
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mf16c")
+        endif()
+       message(STATUS "CMAKE_C_FLAGS: ${CMAKE_C_FLAGS}")
+    else()
+        set(CMAKE_C_FLAGS  "${CMAKE_C_FLAGS} -mfma -mf16c -mavx -mavx2")
+    endif()
 endif()