about summary refs log tree commit diff
path: root/sysdeps/powerpc/powerpc64/multiarch
diff options
context:
space:
mode:
authorPaul E. Murphy <murphyp@linux.vnet.ibm.com>2020-05-18 11:16:06 -0500
committerPaul E. Murphy <murphyp@linux.vnet.ibm.com>2020-06-05 15:30:00 -0500
commita23bd00f9d810c28d9e83ce1d7cf53968375937d (patch)
treead8b0472058d43b628bb9882d999fa3b3514cd7c /sysdeps/powerpc/powerpc64/multiarch
parent6ef422750985f7e60a8d480f07ecda59e0311fdf (diff)
downloadglibc-a23bd00f9d810c28d9e83ce1d7cf53968375937d.tar.gz
glibc-a23bd00f9d810c28d9e83ce1d7cf53968375937d.tar.xz
glibc-a23bd00f9d810c28d9e83ce1d7cf53968375937d.zip
powerpc64le: add optimized strlen for P9
This started as a trivial change to Anton's rawmemchr.  I got
carried away.  This is a hybrid between P8's asympotically
faster 64B checks with extremely efficient small string checks
e.g <64B (and sometimes a little bit more depending on alignment).

The second trick is to align to 64B by running a 48B checking loop
16B at a time until we naturally align to 64B (i.e checking 48/96/144
bytes/iteration based on the alignment after the first 5 comparisons).
This allieviates the need to check page boundaries.

Finally, explicly use the P7 strlen with the runtime loader when building
P9.  We need to be cautious about vector/vsx extensions here on P9 only
builds.
Diffstat (limited to 'sysdeps/powerpc/powerpc64/multiarch')
-rw-r--r--sysdeps/powerpc/powerpc64/multiarch/Makefile2
-rw-r--r--sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c4
-rw-r--r--sysdeps/powerpc/powerpc64/multiarch/strlen-power9.S2
-rw-r--r--sysdeps/powerpc/powerpc64/multiarch/strlen.c5
4 files changed, 12 insertions, 1 deletions
diff --git a/sysdeps/powerpc/powerpc64/multiarch/Makefile b/sysdeps/powerpc/powerpc64/multiarch/Makefile
index fc2268f6b5..19acb6c64a 100644
--- a/sysdeps/powerpc/powerpc64/multiarch/Makefile
+++ b/sysdeps/powerpc/powerpc64/multiarch/Makefile
@@ -33,7 +33,7 @@ sysdep_routines += memcpy-power8-cached memcpy-power7 memcpy-a2 memcpy-power6 \
 
 ifneq (,$(filter %le,$(config-machine)))
 sysdep_routines += strcmp-power9 strncmp-power9 strcpy-power9 stpcpy-power9 \
-		   rawmemchr-power9
+		   rawmemchr-power9 strlen-power9
 endif
 CFLAGS-strncase-power7.c += -mcpu=power7 -funroll-loops
 CFLAGS-strncase_l-power7.c += -mcpu=power7 -funroll-loops
diff --git a/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c b/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c
index 59a227ee22..ea10b00417 100644
--- a/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c
@@ -111,6 +111,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 
   /* Support sysdeps/powerpc/powerpc64/multiarch/strlen.c.  */
   IFUNC_IMPL (i, name, strlen,
+#ifdef __LITTLE_ENDIAN__
+	      IFUNC_IMPL_ADD (array, i, strcpy, hwcap2 & PPC_FEATURE2_ARCH_3_00,
+			      __strlen_power9)
+#endif
 	      IFUNC_IMPL_ADD (array, i, strlen, hwcap2 & PPC_FEATURE2_ARCH_2_07,
 			      __strlen_power8)
 	      IFUNC_IMPL_ADD (array, i, strlen, hwcap & PPC_FEATURE_HAS_VSX,
diff --git a/sysdeps/powerpc/powerpc64/multiarch/strlen-power9.S b/sysdeps/powerpc/powerpc64/multiarch/strlen-power9.S
new file mode 100644
index 0000000000..68c8d54b5f
--- /dev/null
+++ b/sysdeps/powerpc/powerpc64/multiarch/strlen-power9.S
@@ -0,0 +1,2 @@
+#define STRLEN __strlen_power9
+#include <sysdeps/powerpc/powerpc64/le/power9/strlen.S>
diff --git a/sysdeps/powerpc/powerpc64/multiarch/strlen.c b/sysdeps/powerpc/powerpc64/multiarch/strlen.c
index e587554221..cd9dc78a7c 100644
--- a/sysdeps/powerpc/powerpc64/multiarch/strlen.c
+++ b/sysdeps/powerpc/powerpc64/multiarch/strlen.c
@@ -30,8 +30,13 @@ extern __typeof (__redirect_strlen) __libc_strlen;
 extern __typeof (__redirect_strlen) __strlen_ppc attribute_hidden;
 extern __typeof (__redirect_strlen) __strlen_power7 attribute_hidden;
 extern __typeof (__redirect_strlen) __strlen_power8 attribute_hidden;
+extern __typeof (__redirect_strlen) __strlen_power9 attribute_hidden;
 
 libc_ifunc (__libc_strlen,
+# ifdef __LITTLE_ENDIAN__
+	  (hwcap2 & PPC_FEATURE2_ARCH_3_00)
+	  ? __strlen_power9 :
+# endif
 	    (hwcap2 & PPC_FEATURE2_ARCH_2_07)
 	    ? __strlen_power8 :
 	      (hwcap & PPC_FEATURE_HAS_VSX)