This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
PowerPC LE strcpy
- From: Alan Modra <amodra at gmail dot com>
- To: libc-alpha at sourceware dot org
- Date: Fri, 9 Aug 2013 14:51:02 +0930
- Subject: PowerPC LE strcpy
The strcpy changes for little-endian are quite straight-forward, just
a matter of rotating the last word differently.
I'll note that the powerpc64 version of stpcpy is just begging to be
converted to use 64-bit loads and stores..
* sysdeps/powerpc/powerpc64/strcpy.S: Add little-endian support.
* sysdeps/powerpc/powerpc32/strcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/stpcpy.S: Likewise.
* sysdeps/powerpc/powerpc32/stpcpy.S: Likewise.
diff --git a/sysdeps/powerpc/powerpc64/strcpy.S b/sysdeps/powerpc/powerpc64/strcpy.S
index 4c6fd3f..a7fd85b 100644
--- a/sysdeps/powerpc/powerpc64/strcpy.S
+++ b/sysdeps/powerpc/powerpc64/strcpy.S
@@ -68,6 +68,32 @@ L(g2): add rTMP, rFEFE, rWORD
mr rALT, rWORD
/* We've hit the end of the string. Do the rest byte-by-byte. */
L(g1):
+#ifdef __LITTLE_ENDIAN__
+ extrdi. rTMP, rALT, 8, 56
+ stb rALT, 8(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 48
+ stb rTMP, 9(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 40
+ stb rTMP, 10(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 32
+ stb rTMP, 11(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 24
+ stb rTMP, 12(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 16
+ stb rTMP, 13(rDEST)
+ beqlr-
+ extrdi. rTMP, rALT, 8, 8
+ stb rTMP, 14(rDEST)
+ beqlr-
+ extrdi rTMP, rALT, 8, 0
+ stb rTMP, 15(rDEST)
+ blr
+#else
extrdi. rTMP, rALT, 8, 0
stb rTMP, 8(rDEST)
beqlr-
@@ -91,6 +117,7 @@ L(g1):
beqlr-
stb rALT, 15(rDEST)
blr
+#endif
/* Oh well. In this case, we just do a byte-by-byte copy. */
.align 4
diff --git a/sysdeps/powerpc/powerpc32/strcpy.S b/sysdeps/powerpc/powerpc32/strcpy.S
index 4ae577d..e938cc4 100644
--- a/sysdeps/powerpc/powerpc32/strcpy.S
+++ b/sysdeps/powerpc/powerpc32/strcpy.S
@@ -62,7 +62,22 @@ L(g2): add rTMP, rFEFE, rWORD
mr rALT, rWORD
/* We've hit the end of the string. Do the rest byte-by-byte. */
-L(g1): rlwinm. rTMP, rALT, 8, 24, 31
+L(g1):
+#ifdef __LITTLE_ENDIAN__
+ rlwinm. rTMP, rALT, 0, 24, 31
+ stb rALT, 4(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 24, 24, 31
+ stb rTMP, 5(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 16, 24, 31
+ stb rTMP, 6(rDEST)
+ beqlr-
+ rlwinm rTMP, rALT, 8, 24, 31
+ stb rTMP, 7(rDEST)
+ blr
+#else
+ rlwinm. rTMP, rALT, 8, 24, 31
stb rTMP, 4(rDEST)
beqlr-
rlwinm. rTMP, rALT, 16, 24, 31
@@ -73,6 +88,7 @@ L(g1): rlwinm. rTMP, rALT, 8, 24, 31
beqlr-
stb rALT, 7(rDEST)
blr
+#endif
/* Oh well. In this case, we just do a byte-by-byte copy. */
.align 4
diff --git a/sysdeps/powerpc/powerpc64/stpcpy.S b/sysdeps/powerpc/powerpc64/stpcpy.S
index 070cd46..c0b3972 100644
--- a/sysdeps/powerpc/powerpc64/stpcpy.S
+++ b/sysdeps/powerpc/powerpc64/stpcpy.S
@@ -62,7 +62,22 @@ L(g2): add rTMP, rFEFE, rWORD
mr rALT, rWORD
/* We've hit the end of the string. Do the rest byte-by-byte. */
-L(g1): rlwinm. rTMP, rALT, 8, 24, 31
+L(g1):
+#ifdef __LITTLE_ENDIAN__
+ rlwinm. rTMP, rALT, 0, 24, 31
+ stbu rALT, 4(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 24, 24, 31
+ stbu rTMP, 1(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 16, 24, 31
+ stbu rTMP, 1(rDEST)
+ beqlr-
+ rlwinm rTMP, rALT, 8, 24, 31
+ stbu rTMP, 1(rDEST)
+ blr
+#else
+ rlwinm. rTMP, rALT, 8, 24, 31
stbu rTMP, 4(rDEST)
beqlr-
rlwinm. rTMP, rALT, 16, 24, 31
@@ -73,6 +88,7 @@ L(g1): rlwinm. rTMP, rALT, 8, 24, 31
beqlr-
stbu rALT, 1(rDEST)
blr
+#endif
/* Oh well. In this case, we just do a byte-by-byte copy. */
.align 4
diff --git a/sysdeps/powerpc/powerpc32/stpcpy.S b/sysdeps/powerpc/powerpc32/stpcpy.S
index 03c6ddd..7e106e0 100644
--- a/sysdeps/powerpc/powerpc32/stpcpy.S
+++ b/sysdeps/powerpc/powerpc32/stpcpy.S
@@ -62,7 +62,22 @@ L(g2): add rTMP, rFEFE, rWORD
mr rALT, rWORD
/* We've hit the end of the string. Do the rest byte-by-byte. */
-L(g1): rlwinm. rTMP, rALT, 8, 24, 31
+L(g1):
+#ifdef __LITTLE_ENDIAN__
+ rlwinm. rTMP, rALT, 0, 24, 31
+ stbu rALT, 4(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 24, 24, 31
+ stbu rTMP, 1(rDEST)
+ beqlr-
+ rlwinm. rTMP, rALT, 16, 24, 31
+ stbu rTMP, 1(rDEST)
+ beqlr-
+ rlwinm rTMP, rALT, 8, 24, 31
+ stbu rTMP, 1(rDEST)
+ blr
+#else
+ rlwinm. rTMP, rALT, 8, 24, 31
stbu rTMP, 4(rDEST)
beqlr-
rlwinm. rTMP, rALT, 16, 24, 31
@@ -73,6 +88,7 @@ L(g1): rlwinm. rTMP, rALT, 8, 24, 31
beqlr-
stbu rALT, 1(rDEST)
blr
+#endif
/* Oh well. In this case, we just do a byte-by-byte copy. */
.align 4
--
Alan Modra
Australia Development Lab, IBM