[OpenMP][libomp] Cleanup __kmpc_flush() code

[jlpeyton]

Summary

Have __kmpc_flush() be simple KMP_MFENCE() which incorporates x86-specific logic and reduces to KMP_MB() for other platforms.

[hbae]

LGTM

openmp/runtime/src/kmp_csupport.cpp

 Execute <tt>flush</tt>. This is implemented as a full memory fence. (Though
 depending on the memory ordering convention obeyed by the compiler
 even that may not be necessary).
 */
 void __kmpc_flush(ident_t *loc) {
   KC_TRACE(10, ("__kmpc_flush: called\n"));
 
   /* need explicit __mf() here since use volatile instead in library */
-  KMP_MB(); /* Flush all pending memory write invalidates.  */
-
-#if (KMP_ARCH_X86 || KMP_ARCH_X86_64)
-#if KMP_MIC
-// fence-style instructions do not exist, but lock; xaddl $0,(%rsp) can be used.
-// We shouldn't need it, though, since the ABI rules require that
-// * If the compiler generates NGO stores it also generates the fence
-// * If users hand-code NGO stores they should insert the fence
-// therefore no incomplete unordered stores should be visible.
-#else
-  // C74404
-  // This is to address non-temporal store instructions (sfence needed).
-  // The clflush instruction is addressed either (mfence needed).
-  // Probably the non-temporal load monvtdqa instruction should also be
-  // addressed.
-  // mfence is a SSE2 instruction. Do not execute it if CPU is not SSE2.
-  if (!__kmp_cpuinfo.initialized) {
-    __kmp_query_cpuid(&__kmp_cpuinfo);
-  }
-  if (!__kmp_cpuinfo.flags.sse2) {
-    // CPU cannot execute SSE2 instructions.
-  } else {
-#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
-    _mm_mfence();
-#elif KMP_COMPILER_MSVC
-    MemoryBarrier();
-#else
-    __sync_synchronize();
-#endif // KMP_COMPILER_ICC || KMP_COMPILER_ICX
-  }
-#endif // KMP_MIC
-#elif (KMP_ARCH_ARM || KMP_ARCH_AARCH64 || KMP_ARCH_MIPS || KMP_ARCH_MIPS64 || \
-       KMP_ARCH_RISCV64)
-// Nothing to see here move along
-#elif KMP_ARCH_PPC64
-// Nothing needed here (we have a real MB above).
-#else
-#error Unknown or unsupported architecture
-#endif
+  KMP_MFENCE(); /* Flush all pending memory write invalidates.  */
 
 #if OMPT_SUPPORT && OMPT_OPTIONAL
   if (ompt_enabled.ompt_callback_flush) {
     ompt_callbacks.ompt_callback(ompt_callback_flush)(
         __ompt_get_thread_data_internal(), OMPT_GET_RETURN_ADDRESS(0));
   }
 #endif
 }

openmp/runtime/src/kmp_os.h

 #endif
 #endif
 
 #ifndef KMP_MB
 #define KMP_MB() /* nothing to do */
 #endif
 
 #if KMP_ARCH_X86 || KMP_ARCH_X86_64
+#if KMP_MIC
+// fence-style instructions do not exist, but lock; xaddl $0,(%rsp) can be used.
+// We shouldn't need it, though, since the ABI rules require that
+// * If the compiler generates NGO stores it also generates the fence
+// * If users hand-code NGO stores they should insert the fence
+// therefore no incomplete unordered stores should be visible.
+#define KMP_MFENCE() /* Nothing */
+#define KMP_SFENCE() /* Nothing */
+#else
 #if KMP_COMPILER_ICC || KMP_COMPILER_ICX
 #define KMP_MFENCE_() _mm_mfence()
 #define KMP_SFENCE_() _mm_sfence()
 #elif KMP_COMPILER_MSVC
 #define KMP_MFENCE_() MemoryBarrier()
 #define KMP_SFENCE_() MemoryBarrier()
 #else
 #define KMP_MFENCE_() __sync_synchronize()
 #define KMP_SFENCE_() __sync_synchronize()
 #endif
 #define KMP_MFENCE()                                                           \
   if (UNLIKELY(!__kmp_cpuinfo.initialized)) {                                  \
     __kmp_query_cpuid(&__kmp_cpuinfo);                                         \
   }                                                                            \
   if (__kmp_cpuinfo.flags.sse2) {                                              \
     KMP_MFENCE_();                                                             \
   }
 #define KMP_SFENCE() KMP_SFENCE_()
+#endif
 #else
 #define KMP_MFENCE() KMP_MB()
 #define KMP_SFENCE() KMP_MB()
 #endif
 
 #ifndef KMP_IMB
 #define KMP_IMB() /* nothing to do */
 #endif