Fix crash in invoking microtask on ios arm64

[nihui]

Summary

invoking microtask cause segment fault on ios platform
So I tried to modify __kmp_invoke_microtask() a little to use AArch64 assembly implementation, and the crash solved

cross-compiling for ios
host: fedora linux, clang 3.9.1

[AndreyChurbanov]

Couple notes.

1. Your patch breaks non-Linux ARM implementations I think, like FreeBSD ARM or NetBSD ARM. It would be safer if you changed KMP_OS_LINUX to (KMP_OS_LINUX || KMP_OS_DARWIN) instead of removing KMP_OS_LINUX, so that other platforms were not affected.

2. Please upload the full diff as opposed to stripped, so that people could see more context around your changes.

[nihui]

change KMP_OS_LINUX to (KMP_OS_LINUX || KMP_OS_DARWIN)
upload full diff

Thanks

[AndreyChurbanov]

LGTM

[nihui]

hello
I do not have commit access
please commit the change for me
thanks a lot ;)

openmp/trunk/runtime/src/z_Linux_asm.s

.macro KMP_CFI_REGISTER reg
   .cfi_def_cfa_register \reg
 .endm
 .macro KMP_CFI_DEF reg, sz
   .cfi_def_cfa  \reg,\sz
 .endm
 # endif // KMP_OS_DARWIN
 #endif // KMP_ARCH_X86 || KMP_ARCH_x86_64
 
-#if KMP_OS_LINUX && KMP_ARCH_AARCH64
+#if (KMP_OS_LINUX || KMP_OS_DARWIN) && KMP_ARCH_AARCH64
 
+# if KMP_OS_DARWIN
+#  define KMP_PREFIX_UNDERSCORE(x) _##x  // extra underscore for OS X* symbols
+#  define KMP_LABEL(x) L_##x             // form the name of label
+
+.macro ALIGN
+  .align $0
+.endmacro
+
+.macro DEBUG_INFO
+/* Not sure what .size does in icc, not sure if we need to do something
+   similar for OS X*.
+*/
+.endmacro
+
+.macro PROC
+  ALIGN  4
+  .globl KMP_PREFIX_UNDERSCORE($0)
+KMP_PREFIX_UNDERSCORE($0):
+.endmacro
+# else // KMP_OS_DARWIN
 #  define KMP_PREFIX_UNDERSCORE(x) x  // no extra underscore for Linux* OS symbols
 // Format labels so that they don't override function names in gdb's backtraces
 #  define KMP_LABEL(x) .L_##x         // local label hidden from backtraces
 
 .macro ALIGN size
   .align 1<<(\size)
 .endm
 
 .macro DEBUG_INFO proc
   .cfi_endproc
 // Not sure why we need .type and .size for the functions
   ALIGN 2
   .type  \proc,@function
   .size  \proc,.-\proc
 .endm
 
 .macro PROC proc
   ALIGN 2
   .globl KMP_PREFIX_UNDERSCORE(\proc)
 KMP_PREFIX_UNDERSCORE(\proc):
   .cfi_startproc
 .endm
+# endif // KMP_OS_DARWIN
 
-#endif // KMP_OS_LINUX && KMP_ARCH_AARCH64
+#endif // (KMP_OS_LINUX || KMP_OS_DARWIN) && KMP_ARCH_AARCH64
 
 // -----------------------------------------------------------------------
 // data
 // -----------------------------------------------------------------------
 
 #ifdef KMP_GOMP_COMPAT
 
 //
//
 
         DEBUG_INFO __kmp_bsr32
 
 
 // -----------------------------------------------------------------------
 #endif /* KMP_ARCH_X86_64 */
 
 // '
-#if KMP_OS_LINUX && KMP_ARCH_AARCH64
+#if (KMP_OS_LINUX || KMP_OS_DARWIN) && KMP_ARCH_AARCH64
 
 //------------------------------------------------------------------------
 //
 // typedef void (*microtask_t)( int *gtid, int *tid, ... );
 //
 // int
 // __kmp_invoke_microtask( void (*pkfn) (int gtid, int tid, ...),
 //               int gtid, int tid,
# if OMPT_SUPPORT
   ldp x19, x20, [sp], #16
 # endif
   ldp x29, x30, [sp], #16
   ret
 
   DEBUG_INFO __kmp_invoke_microtask
 // -- End  __kmp_invoke_microtask
 
-#endif /* KMP_OS_LINUX && KMP_ARCH_AARCH64 */
+#endif /* (KMP_OS_LINUX || KMP_OS_DARWIN) && KMP_ARCH_AARCH64 */
 
 #if KMP_ARCH_PPC64
 
 //------------------------------------------------------------------------
 //
 // typedef void (*microtask_t)( int *gtid, int *tid, ... );
 //
 // int

openmp/trunk/runtime/src/z_Linux_util.cpp

__kmp_get_load_balance( int max )
     return running_threads;
 
 } // __kmp_get_load_balance
 
 # endif // KMP_OS_DARWIN
 
 #endif // USE_LOAD_BALANCE
 
-#if !(KMP_ARCH_X86 || KMP_ARCH_X86_64 || KMP_MIC || (KMP_OS_LINUX && KMP_ARCH_AARCH64) || KMP_ARCH_PPC64)
+#if !(KMP_ARCH_X86 || KMP_ARCH_X86_64 || KMP_MIC || ((KMP_OS_LINUX || KMP_OS_DARWIN) && KMP_ARCH_AARCH64) || KMP_ARCH_PPC64)
 
 // we really only need the case with 1 argument, because CLANG always build
 // a struct of pointers to shared variables referenced in the outlined function
 int
 __kmp_invoke_microtask( microtask_t pkfn,
                         int gtid, int tid,
                         int argc, void *p_argv[]
 #if OMPT_SUPPORT