PowerPC LE strcmp and strncmp

http://sourceware.org/ml/libc-alpha/2013-08/msg00099.html

More little-endian support.  I leave the main strcmp loops unchanged,
(well, except for renumbering rTMP to something other than r0 since
it's needed in an addi insn) and modify the tail for little-endian.

I noticed some of the big-endian tail code was a little untidy so have
cleaned that up too.

	* sysdeps/powerpc/powerpc64/strcmp.S (rTMP2): Define as r0.
	(rTMP): Define as r11.
	(strcmp): Add little-endian support.  Optimise tail.
	* sysdeps/powerpc/powerpc32/strcmp.S: Similarly.
	* sysdeps/powerpc/powerpc64/strncmp.S: Likewise.
	* sysdeps/powerpc/powerpc32/strncmp.S: Likewise.
	* sysdeps/powerpc/powerpc64/power4/strncmp.S: Likewise.
	* sysdeps/powerpc/powerpc32/power4/strncmp.S: Likewise.
	* sysdeps/powerpc/powerpc64/power7/strncmp.S: Likewise.
	* sysdeps/powerpc/powerpc32/power7/strncmp.S: Likewise.

1 files changed, 56 insertions, 9 deletions

diff --git a/sysdeps/powerpc/powerpc64/strcmp.S b/sysdeps/powerpc/powerpc64/strcmp.S
index c9d6dac..7085468 100644
--- a/sysdeps/powerpc/powerpc64/strcmp.S
+++ b/sysdeps/powerpc/powerpc64/strcmp.S
@@ -25,7 +25,7 @@
 EALIGN (strcmp, 4, 0)
 	CALL_MCOUNT 2
 
-#define rTMP	r0
+#define rTMP2	r0
 #define rRTN	r3
 #define rSTR1	r3	/* first string arg */
 #define rSTR2	r4	/* second string arg */
@@ -35,6 +35,7 @@ EALIGN (strcmp, 4, 0)
 #define r7F7F	r8	/* constant 0x7f7f7f7f7f7f7f7f */
 #define rNEG	r9	/* ~(word in s1 | 0x7f7f7f7f7f7f7f7f) */
 #define rBITDIF	r10	/* bits that differ in s1 & s2 words */
+#define rTMP	r11
 
 	dcbt	0,rSTR1
 	or	rTMP, rSTR2, rSTR1
@@ -58,19 +59,66 @@ L(g0):	ldu	rWORD1, 8(rSTR1)
 	ldu	rWORD2, 8(rSTR2)
 L(g1):	add	rTMP, rFEFE, rWORD1
 	nor	rNEG, r7F7F, rWORD1
-
 	and.	rTMP, rTMP, rNEG
 	cmpd	cr1, rWORD1, rWORD2
 	beq+	L(g0)
-L(endstring):
+
 /* OK. We've hit the end of the string. We need to be careful that
    we don't compare two strings as different because of gunk beyond
    the end of the strings...  */
+#ifdef __LITTLE_ENDIAN__
+L(endstring):
+	addi    rTMP2, rTMP, -1
+	beq	cr1, L(equal)
+	andc    rTMP2, rTMP2, rTMP
+	rldimi	rTMP2, rTMP2, 1, 0
+	and	rWORD2, rWORD2, rTMP2	/* Mask off gunk.  */
+	and	rWORD1, rWORD1, rTMP2
+	cmpd	cr1, rWORD1, rWORD2
+	beq	cr1, L(equal)
+	xor	rBITDIF, rWORD1, rWORD2	/* rBITDIF has bits that differ.  */
+	neg	rNEG, rBITDIF
+	and	rNEG, rNEG, rBITDIF	/* rNEG has LS bit that differs.  */
+	cntlzd	rNEG, rNEG		/* bitcount of the bit.  */
+	andi.	rNEG, rNEG, 56		/* bitcount to LS byte that differs. */
+	sld	rWORD1, rWORD1, rNEG	/* shift left to clear MS bytes.  */
+	sld	rWORD2, rWORD2, rNEG
+	xor.	rBITDIF, rWORD1, rWORD2
+	sub	rRTN, rWORD1, rWORD2
+	blt-	L(highbit)
+	sradi	rRTN, rRTN, 63		/* must return an int.  */
+	ori	rRTN, rRTN, 1
+	blr
+L(equal):
+	li	rRTN, 0
+	blr
+
+L(different):
+	ld	rWORD1, -8(rSTR1)
+	xor	rBITDIF, rWORD1, rWORD2	/* rBITDIF has bits that differ.  */
+	neg	rNEG, rBITDIF
+	and	rNEG, rNEG, rBITDIF	/* rNEG has LS bit that differs.  */
+	cntlzd	rNEG, rNEG		/* bitcount of the bit.  */
+	andi.	rNEG, rNEG, 56		/* bitcount to LS byte that differs. */
+	sld	rWORD1, rWORD1, rNEG	/* shift left to clear MS bytes.  */
+	sld	rWORD2, rWORD2, rNEG
+	xor.	rBITDIF, rWORD1, rWORD2
+	sub	rRTN, rWORD1, rWORD2
+	blt-	L(highbit)
+	sradi	rRTN, rRTN, 63
+	ori	rRTN, rRTN, 1
+	blr
+L(highbit):
+	sradi	rRTN, rWORD2, 63
+	ori	rRTN, rRTN, 1
+	blr
+
+#else
+L(endstring):
 	and	rTMP, r7F7F, rWORD1
 	beq	cr1, L(equal)
 	add	rTMP, rTMP, r7F7F
 	xor.	rBITDIF, rWORD1, rWORD2
-
 	andc	rNEG, rNEG, rTMP
 	blt-	L(highbit)
 	cntlzd	rBITDIF, rBITDIF
@@ -79,7 +127,7 @@ L(endstring):
 	cmpd	cr1, rNEG, rBITDIF
 	sub	rRTN, rWORD1, rWORD2
 	blt-	cr1, L(equal)
-	sradi	rRTN, rRTN, 63
+	sradi	rRTN, rRTN, 63		/* must return an int.  */
 	ori	rRTN, rRTN, 1
 	blr
 L(equal):
@@ -95,11 +143,10 @@ L(different):
 	ori	rRTN, rRTN, 1
 	blr
 L(highbit):
-	srdi	rWORD2, rWORD2, 56
-	srdi	rWORD1, rWORD1, 56
-	sub	rRTN, rWORD1, rWORD2
+	sradi	rRTN, rWORD2, 63
+	ori	rRTN, rRTN, 1
 	blr
-
+#endif
 
 /* Oh well.  In this case, we just do a byte-by-byte comparison.  */
 	.align 4