Some POWER specific optimizations

Bug#18842925 : SET THREAD PRIORITY IN INNODB MUTEX SPINLOOP
Like "pause" instruction for hyper-threading at Intel CPUs,
POWER has special instructions only for hinting priority of hardware-threads.

Approved by Sunny in rb#6256

Backport of the 5.7 fix - mysql/mysql-server@c92102a
(excluded cache line size patch)

Suggestion by Stewart Smith

config.h.cmake

@@ -225,6 +225,7 @@
 #cmakedefine HAVE_PREAD 1
 #cmakedefine HAVE_PAUSE_INSTRUCTION 1
 #cmakedefine HAVE_FAKE_PAUSE_INSTRUCTION 1
+#cmakedefine HAVE_HMT_PRIORITY_INSTRUCTION 1
 #cmakedefine HAVE_RDTSCLL 1
 #cmakedefine HAVE_READ_REAL_TIME 1
 #cmakedefine HAVE_PTHREAD_ATTR_CREATE 1

configure.cmake

@@ -855,6 +855,16 @@ IF(NOT CMAKE_CROSSCOMPILING AND NOT MSVC)
     }
    " HAVE_FAKE_PAUSE_INSTRUCTION)
   ENDIF()
+  IF (NOT HAVE_PAUSE_INSTRUCTION)
+    CHECK_C_SOURCE_COMPILES("
+    int main()
+    {
+     __asm__ __volatile__ (\"or 1,1,1\");
+     __asm__ __volatile__ (\"or 2,2,2\");
+     return 0;
+    }
+    " HAVE_HMT_PRIORITY_INSTRUCTION)
+  ENDIF()
 ENDIF()
 
 CHECK_SYMBOL_EXISTS(tcgetattr "termios.h" HAVE_TCGETATTR 1)

storage/innobase/include/ut0ut.h

@@ -92,6 +92,14 @@ struct ut_when_dtor {
 #  define UT_RELAX_CPU() __asm__ __volatile__ ("":::"memory")
 # endif
 
+# if defined(HAVE_HMT_PRIORITY_INSTRUCTION)
+#  define UT_LOW_PRIORITY_CPU() __asm__ __volatile__ ("or 1,1,1")
+#  define UT_RESUME_PRIORITY_CPU() __asm__ __volatile__ ("or 2,2,2")
+# else
+#  define UT_LOW_PRIORITY_CPU() ((void)0)
+#  define UT_RESUME_PRIORITY_CPU() ((void)0)
+# endif
+
 /*********************************************************************//**
 Delays execution for at most max_wait_us microseconds or returns earlier
 if cond becomes true.

storage/innobase/ut/ut0ut.cc

@@ -401,13 +401,17 @@ ut_delay(
 {
 	ulint	i, j;
 
+	UT_LOW_PRIORITY_CPU();
+
 	j = 0;
 
 	for (i = 0; i < delay * 50; i++) {
 		j += i;
 		UT_RELAX_CPU();
 	}
 
+	UT_RESUME_PRIORITY_CPU();
+
 	return(j);
 }
 #endif /* !UNIV_HOTBACKUP */

storage/xtradb/include/ut0ut.h

@@ -89,6 +89,14 @@ struct ut_when_dtor {
 #  define UT_RELAX_CPU() __asm__ __volatile__ ("":::"memory")
 # endif
 
+# if defined(HAVE_HMT_PRIORITY_INSTRUCTION)
+#  define UT_LOW_PRIORITY_CPU() __asm__ __volatile__ ("or 1,1,1")
+#  define UT_RESUME_PRIORITY_CPU() __asm__ __volatile__ ("or 2,2,2")
+# else
+#  define UT_LOW_PRIORITY_CPU() ((void)0)
+#  define UT_RESUME_PRIORITY_CPU() ((void)0)
+# endif
+
 /*********************************************************************//**
 Delays execution for at most max_wait_us microseconds or returns earlier
 if cond becomes true.

storage/xtradb/ut/ut0ut.cc

@@ -402,13 +402,17 @@ ut_delay(
 {
 	ulint	i, j;
 
+	UT_LOW_PRIORITY_CPU();
+
 	j = 0;
 
 	for (i = 0; i < delay * 50; i++) {
 		j += i;
 		UT_RELAX_CPU();
 	}
 
+	UT_RESUME_PRIORITY_CPU();
+
 	return(j);
 }
 #endif /* !UNIV_HOTBACKUP */