PowerPC LE memcpy

http://sourceware.org/ml/libc-alpha/2013-08/msg00103.html

LIttle-endian support for memcpy.  I spent some time cleaning up the
64-bit power7 memcpy, in order to avoid the extra alignment traps
power7 takes for little-endian.  It probably would have been better
to copy the linux kernel version of memcpy.

	* sysdeps/powerpc/powerpc32/power4/memcpy.S: Add little endian support.
	* sysdeps/powerpc/powerpc32/power6/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc32/power7/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc32/power7/mempcpy.S: Likewise.
	* sysdeps/powerpc/powerpc64/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc64/power4/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc64/power6/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc64/power7/memcpy.S: Likewise.
	* sysdeps/powerpc/powerpc64/power7/mempcpy.S: Likewise.  Make better
	use of regs.  Use power7 mtocrf.  Tidy function tails.

1 files changed, 74 insertions, 7 deletions

diff --git a/sysdeps/powerpc/powerpc32/power6/memcpy.S b/sysdeps/powerpc/powerpc32/power6/memcpy.S
index a76f71e04f..f58114a0c5 100644
--- a/sysdeps/powerpc/powerpc32/power6/memcpy.S
+++ b/sysdeps/powerpc/powerpc32/power6/memcpy.S
@@ -219,15 +219,28 @@ L(word_unaligned_short):
     blt   cr6,5f
     srwi  7,6,16
     bgt	  cr6,3f
+#ifdef __LITTLE_ENDIAN__
+    sth   7,0(3)
+#else
     sth   6,0(3)
+#endif
     b     7f
     .align  4
 3:
+#ifdef __LITTLE_ENDIAN__
+    rotlwi 6,6,24
+    stb   6,0(3)
+    sth   7,1(3)
+#else
     stb   7,0(3)
     sth   6,1(3)
+#endif
     b     7f
     .align  4
 5:
+#ifdef __LITTLE_ENDIAN__
+    rotlwi 6,6,8
+#endif
     stb   6,0(3)
 7:
     cmplwi	cr1,10,16
@@ -577,7 +590,11 @@ L(wdu1_32):
     lwz     6,-1(4)
     cmplwi  cr6,31,4
     srwi    8,31,5    /* calculate the 32 byte loop count */
+#ifdef __LITTLE_ENDIAN__
+    srwi    6,6,8
+#else
     slwi    6,6,8
+#endif
     clrlwi  31,31,27   /* The remaining bytes, < 32.  */
     blt     cr5,L(wdu1_32tail)
     mtctr   8
@@ -585,8 +602,12 @@ L(wdu1_32):
 
     lwz   8,3(4)
     lwz   7,4(4)
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,24,32
+#else
 /*  Equivalent to: srwi   8,8,32-8;  or    6,6,8 */
     rlwimi 6,8,8,(32-8),31
+#endif
     b      L(wdu1_loop32x)
     .align  4
 L(wdu1_loop32):
@@ -595,8 +616,12 @@ L(wdu1_loop32):
     lwz   7,4(4)
     stw   10,-8(3)
     stw   11,-4(3)
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,24,32
+#else
 /*  Equivalent to  srwi   8,8,32-8; or    6,6,8 */
     rlwimi 6,8,8,(32-8),31
+#endif
 L(wdu1_loop32x):
     lwz   10,8(4)
     lwz   11,12(4)
@@ -613,7 +638,11 @@ L(wdu1_loop32x):
     stw   6,16(3)
     stw   7,20(3)
     addi  3,3,32
+#ifdef __LITTLE_ENDIAN__
+    srwi  6,8,8
+#else
     slwi  6,8,8
+#endif
     bdnz+ L(wdu1_loop32)
     stw   10,-8(3)
     stw   11,-4(3)
@@ -624,8 +653,12 @@ L(wdu1_32tail):
     blt     cr6,L(wdu_4tail)
     /* calculate and store the final word */
     lwz   8,3(4)
-/*  Equivalent to: srwi   8,8,32-9;  or    6,6,8  */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,24,32
+#else
+/*  Equivalent to: srwi   8,8,32-8;  or    6,6,8  */
     rlwimi 6,8,8,(32-8),31
+#endif
     b     L(wdu_32tailx)
 
 L(wdu2_32):
@@ -633,7 +666,11 @@ L(wdu2_32):
     lwz     6,-2(4)
     cmplwi  cr6,31,4
     srwi    8,31,5    /* calculate the 32 byte loop count */
+#ifdef __LITTLE_ENDIAN__
+    srwi    6,6,16
+#else
     slwi    6,6,16
+#endif
     clrlwi  31,31,27   /* The remaining bytes, < 32.  */
     blt     cr5,L(wdu2_32tail)
     mtctr   8
@@ -641,8 +678,11 @@ L(wdu2_32):
 
     lwz   8,2(4)
     lwz   7,4(4)
-/*  Equivalent to: srwi   8,8,32-8;  or    6,6,8 */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,16,32
+#else
     rlwimi 6,8,16,(32-16),31
+#endif
     b      L(wdu2_loop32x)
     .align  4
 L(wdu2_loop32):
@@ -651,8 +691,11 @@ L(wdu2_loop32):
     lwz   7,4(4)
     stw   10,-8(3)
     stw   11,-4(3)
-/*  Equivalent to  srwi   8,8,32-8; or    6,6,8 */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,16,32
+#else
     rlwimi 6,8,16,(32-16),31
+#endif
 L(wdu2_loop32x):
     lwz   10,8(4)
     lwz   11,12(4)
@@ -670,7 +713,11 @@ L(wdu2_loop32x):
     stw   6,16(3)
     stw   7,20(3)
     addi  3,3,32
+#ifdef __LITTLE_ENDIAN__
+    srwi  6,8,16
+#else
     slwi  6,8,16
+#endif
     bdnz+ L(wdu2_loop32)
     stw   10,-8(3)
     stw   11,-4(3)
@@ -681,8 +728,11 @@ L(wdu2_32tail):
     blt     cr6,L(wdu_4tail)
     /* calculate and store the final word */
     lwz   8,2(4)
-/*  Equivalent to: srwi   8,8,32-9;  or    6,6,8  */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,16,32
+#else
     rlwimi 6,8,16,(32-16),31
+#endif
     b     L(wdu_32tailx)
 
 L(wdu3_32):
@@ -690,7 +740,11 @@ L(wdu3_32):
     lwz     6,-3(4)
     cmplwi  cr6,31,4
     srwi    8,31,5    /* calculate the 32 byte loop count */
+#ifdef __LITTLE_ENDIAN__
+    srwi    6,6,24
+#else
     slwi    6,6,24
+#endif
     clrlwi  31,31,27   /* The remaining bytes, < 32.  */
     blt     cr5,L(wdu3_32tail)
     mtctr   8
@@ -698,8 +752,11 @@ L(wdu3_32):
 
     lwz   8,1(4)
     lwz   7,4(4)
-/*  Equivalent to: srwi   8,8,32-8;  or    6,6,8 */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,8,32
+#else
     rlwimi 6,8,24,(32-24),31
+#endif
     b      L(wdu3_loop32x)
     .align  4
 L(wdu3_loop32):
@@ -708,8 +765,11 @@ L(wdu3_loop32):
     lwz   7,4(4)
     stw   10,-8(3)
     stw   11,-4(3)
-/*  Equivalent to  srwi   8,8,32-8; or    6,6,8 */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,8,32
+#else
     rlwimi 6,8,24,(32-24),31
+#endif
 L(wdu3_loop32x):
     lwz   10,8(4)
     lwz   11,12(4)
@@ -726,7 +786,11 @@ L(wdu3_loop32x):
     stw   6,16(3)
     stw   7,20(3)
     addi  3,3,32
+#ifdef __LITTLE_ENDIAN__
+    srwi  6,8,24
+#else
     slwi  6,8,24
+#endif
     bdnz+ L(wdu3_loop32)
     stw   10,-8(3)
     stw   11,-4(3)
@@ -737,8 +801,11 @@ L(wdu3_32tail):
     blt     cr6,L(wdu_4tail)
     /* calculate and store the final word */
     lwz   8,1(4)
-/*  Equivalent to: srwi   8,8,32-9;  or    6,6,8  */
+#ifdef __LITTLE_ENDIAN__
+    rldimi 6,8,8,32
+#else
     rlwimi 6,8,24,(32-24),31
+#endif
     b     L(wdu_32tailx)
     .align  4
 L(wdu_32tailx):