S/390: Add support for STT_GNU_IFUNC symbols.

Add support for STT_GNU_IFUNC symbols and the new R_390_IRELATIVE
relocation.  Provide optimized version of memcpy, memset, and memcmp
for z10 and z196.

1 files changed, 57 insertions, 33 deletions

diff --git a/sysdeps/s390/s390-32/memcpy.S b/sysdeps/s390/s390-32/memcpy.S
index b2a688ea36..90cc4cbf71 100644
--- a/sysdeps/s390/s390-32/memcpy.S
+++ b/sysdeps/s390/s390-32/memcpy.S
@@ -1,7 +1,6 @@
 /* memcpy - copy a block from source to destination.  S/390 version.
-   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Copyright (C) 2012 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
-   Contributed by Martin Schwidefsky (schwidefsky@de.ibm.com).
 
    The GNU C Library is free software; you can redistribute it and/or
    modify it under the terms of the GNU Lesser General Public
@@ -17,42 +16,67 @@
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
+
+#include "sysdep.h"
+#include "asm-syntax.h"
+
 /* INPUT PARAMETERS
      %r2 = address of destination memory area
      %r3 = address of source memory area
      %r4 = number of bytes to copy.  */
 
-#include "sysdep.h"
-#include "asm-syntax.h"
-
-	.text
+#ifdef USE_MULTIARCH
+ENTRY(memcpy_g5)
+#else
 ENTRY(memcpy)
-	ltr     %r4,%r4
-	jz      .L3
-	ahi     %r4,-1              # length - 1
-	lr      %r1,%r2             # copy destination address
-	lr      %r5,%r4
-	srl     %r5,8
-	ltr     %r5,%r5             # < 256 bytes to move ?
-	jz      .L1
-	chi     %r5,255             # > 1MB to move ?
-	jh      .L4
-.L0:	mvc     0(256,%r1),0(%r3)   # move in 256 byte chunks
-	la      %r1,256(%r1)
-	la      %r3,256(%r3)
-	brct    %r5,.L0
-.L1:	bras    %r5,.L2             # setup base pointer for execute
-	mvc     0(1,%r1),0(%r3)     # instruction for execute
-.L2:	ex      %r4,0(%r5)          # execute mvc with length ((%r4)&255)+1
-.L3:	br      %r14
-	# data copies > 1MB are faster with mvcle.
-.L4:	ahi     %r4,1               # length + 1
-	lr      %r5,%r4	            # source length
-	lr	%r4,%r3             # source address
-	lr	%r3,%r5             # destination length = source length
-.L5:	mvcle	%r2,%r4,0           # thats it, MVCLE is your friend
-	jo	.L5
-	lr      %r2,%r1             # return destination address
-	br	%r14
+#endif
+	.machine "g5"
+        st      %r13,52(%r15)
+        .cfi_offset 13, -44
+        basr    %r13,0
+.L_G5_16:
+        ltr     %r4,%r4
+        je      .L_G5_4
+        ahi     %r4,-1
+        lr      %r5,%r4
+        srl     %r5,8
+        ltr     %r5,%r5
+        lr      %r1,%r2
+        jne     .L_G5_12
+        ex      %r4,.L_G5_17-.L_G5_16(%r13)
+.L_G5_4:
+        l       %r13,52(%r15)
+        br      %r14
+.L_G5_13:
+        chi	%r5,4096             # Switch to mvcle for copies >1MB
+        jh	memcpy_mvcle
+.L_G5_12:
+        mvc     0(256,%r1),0(%r3)
+        la      %r1,256(%r1)
+        la      %r3,256(%r3)
+        brct    %r5,.L_G5_12
+        ex      %r4,.L_G5_17-.L_G5_16(%r13)
+        j       .L_G5_4
+.L_G5_17:
+        mvc     0(1,%r1),0(%r3)
+#ifdef USE_MULTIARCH
+END(memcpy_g5)
+#else
 END(memcpy)
 libc_hidden_builtin_def (memcpy)
+#endif
+
+ENTRY(memcpy_mvcle)
+       # Using as standalone function will result in unexpected
+       # results since the length field is incremented by 1 in order to
+       # compensate the changes already done in the functions above.
+       ahi     %r4,1               # length + 1
+       lr      %r5,%r4             # source length
+       lr      %r4,%r3             # source address
+       lr      %r3,%r5             # destination length = source length
+.L_MVCLE_1:
+       mvcle   %r2,%r4,0           # thats it, MVCLE is your friend
+       jo      .L_MVCLE_1
+       lr      %r2,%r1             # return destination address
+       br      %r14
+END(memcpy_mvcle)