about summary refs log tree commit diff
path: root/sysdeps/powerpc/powerpc32
diff options
context:
space:
mode:
authorLuis Machado <luisgpm@br.ibm.com>2010-06-14 17:13:24 -0700
committerUlrich Drepper <drepper@redhat.com>2010-06-14 17:13:24 -0700
commit158db1226717c77f30ee2245e164f2e1f7721637 (patch)
tree6d7493e39ee07ca0384041b6455ed036d9f469ae /sysdeps/powerpc/powerpc32
parent4bc93b30337482a0ad1758b4ba069704d09c39da (diff)
downloadglibc-158db1226717c77f30ee2245e164f2e1f7721637.tar.gz
glibc-158db1226717c77f30ee2245e164f2e1f7721637.tar.xz
glibc-158db1226717c77f30ee2245e164f2e1f7721637.zip
power7 string compare optimizations
Diffstat (limited to 'sysdeps/powerpc/powerpc32')
-rw-r--r--sysdeps/powerpc/powerpc32/power7/memcmp.S988
-rw-r--r--sysdeps/powerpc/powerpc32/power7/strncmp.S177
2 files changed, 1165 insertions, 0 deletions
diff --git a/sysdeps/powerpc/powerpc32/power7/memcmp.S b/sysdeps/powerpc/powerpc32/power7/memcmp.S
new file mode 100644
index 0000000000..7d9db515ab
--- /dev/null
+++ b/sysdeps/powerpc/powerpc32/power7/memcmp.S
@@ -0,0 +1,988 @@
+/* Optimized memcmp implementation for POWER7/PowerPC32.
+   Copyright (C) 2010 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA
+   02110-1301 USA.  */
+
+#include <sysdep.h>
+#include <bp-sym.h>
+#include <bp-asm.h>
+
+/* int [r3] memcmp (const char *s1 [r3],
+		    const char *s2 [r4],
+		    size_t size [r5])  */
+
+	.machine power7
+EALIGN (BP_SYM(memcmp),4,0)
+	CALL_MCOUNT
+
+#define rTMP	r0
+#define rRTN	r3
+#define rSTR1	r3	/* first string arg */
+#define rSTR2	r4	/* second string arg */
+#define rN	r5	/* max string length */
+#define rWORD1	r6	/* current word in s1 */
+#define rWORD2	r7	/* current word in s2 */
+#define rWORD3	r8	/* next word in s1 */
+#define rWORD4	r9	/* next word in s2 */
+#define rWORD5	r10	/* next word in s1 */
+#define rWORD6	r11	/* next word in s2 */
+#define rBITDIF	r12	/* bits that differ in s1 & s2 words */
+#define rWORD7	r30	/* next word in s1 */
+#define rWORD8	r31	/* next word in s2 */
+
+	xor	rTMP,rSTR2,rSTR1
+	cmplwi	cr6,rN,0
+	cmplwi	cr1,rN,12
+	clrlwi.	rTMP,rTMP,30
+	clrlwi	rBITDIF,rSTR1,30
+	cmplwi	cr5,rBITDIF,0
+	beq-	cr6,L(zeroLength)
+	dcbt	0,rSTR1
+	dcbt	0,rSTR2
+
+	/* If less than 8 bytes or not aligned, use the unaligned
+	   byte loop.  */
+
+	blt	cr1,L(bytealigned)
+	stwu	1,-64(1)
+	cfi_adjust_cfa_offset(64)
+	stw	r31,48(1)
+	cfi_offset(31,(48-64))
+	stw	r30,44(1)
+	cfi_offset(30,(44-64))
+	bne	L(unaligned)
+/* At this point we know both strings have the same alignment and the
+   compare length is at least 8 bytes.  rBITDIF contains the low order
+   2 bits of rSTR1 and cr5 contains the result of the logical compare
+   of rBITDIF to 0.  If rBITDIF == 0 then we are already word 
+   aligned and can perform the word aligned loop.
+
+   Otherwise we know the two strings have the same alignment (but not
+   yet word aligned).  So we force the string addresses to the next lower
+   word boundary and special case this first word using shift left to
+   eliminate bits preceeding the first byte.  Since we want to join the
+   normal (word aligned) compare loop, starting at the second word,
+   we need to adjust the length (rN) and special case the loop
+   versioning for the first word. This insures that the loop count is
+   correct and the first word (shifted) is in the expected register pair. */
+	.align	4
+L(samealignment):
+	clrrwi	rSTR1,rSTR1,2
+	clrrwi	rSTR2,rSTR2,2
+	beq	cr5,L(Waligned)
+	add	rN,rN,rBITDIF
+	slwi	r11,rBITDIF,3
+	srwi	rTMP,rN,4	/* Divide by 16 */
+	andi.	rBITDIF,rN,12	/* Get the word remainder */
+	lwz	rWORD1,0(rSTR1)
+	lwz	rWORD2,0(rSTR2)
+	cmplwi	cr1,rBITDIF,8
+	cmplwi	cr7,rN,16
+	clrlwi	rN,rN,30
+	beq	L(dPs4)
+	mtctr	rTMP
+	bgt	cr1,L(dPs3)
+	beq	cr1,L(dPs2)
+
+/* Remainder is 4 */
+	.align	3
+L(dsP1):
+	slw	rWORD5,rWORD1,r11
+	slw	rWORD6,rWORD2,r11
+	cmplw	cr5,rWORD5,rWORD6
+	blt	cr7,L(dP1x)
+/* Do something useful in this cycle since we have to branch anyway.  */
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	b	L(dP1e)
+/* Remainder is 8 */
+	.align	4
+L(dPs2):
+	slw	rWORD5,rWORD1,r11
+	slw	rWORD6,rWORD2,r11
+	cmplw	cr6,rWORD5,rWORD6
+	blt	cr7,L(dP2x)
+/* Do something useful in this cycle since we have to branch anyway.  */
+	lwz	rWORD7,4(rSTR1)
+	lwz	rWORD8,4(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	b	L(dP2e)
+/* Remainder is 12 */
+	.align	4
+L(dPs3):
+	slw	rWORD3,rWORD1,r11
+	slw	rWORD4,rWORD2,r11
+	cmplw	cr1,rWORD3,rWORD4
+	b	L(dP3e)
+/* Count is a multiple of 16, remainder is 0 */
+	.align	4
+L(dPs4):
+	mtctr	rTMP
+	slw	rWORD1,rWORD1,r11
+	slw	rWORD2,rWORD2,r11
+	cmplw	cr0,rWORD1,rWORD2
+	b	L(dP4e)
+
+/* At this point we know both strings are word aligned and the
+   compare length is at least 8 bytes.  */
+	.align	4
+L(Waligned):
+	andi.	rBITDIF,rN,12	/* Get the word remainder */
+	srwi	rTMP,rN,4	/* Divide by 16 */
+	cmplwi	cr1,rBITDIF,8
+	cmplwi	cr7,rN,16
+	clrlwi	rN,rN,30
+	beq	L(dP4)
+	bgt	cr1,L(dP3)
+	beq	cr1,L(dP2)
+
+/* Remainder is 4 */
+	.align	4
+L(dP1):
+	mtctr	rTMP
+/* Normally we'd use rWORD7/rWORD8 here, but since we might exit early
+   (8-15 byte compare), we want to use only volatile registers.  This
+   means we can avoid restoring non-volatile registers since we did not
+   change any on the early exit path.  The key here is the non-early
+   exit path only cares about the condition code (cr5), not about which 
+   register pair was used.  */
+	lwz	rWORD5,0(rSTR1)
+	lwz	rWORD6,0(rSTR2)
+	cmplw	cr5,rWORD5,rWORD6
+	blt	cr7,L(dP1x)
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+L(dP1e):
+	lwz	rWORD3,8(rSTR1)
+	lwz	rWORD4,8(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	lwz	rWORD5,12(rSTR1)
+	lwz	rWORD6,12(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr5,L(dLcr5)
+	bne	cr0,L(dLcr0)
+
+	lwzu	rWORD7,16(rSTR1)
+	lwzu	rWORD8,16(rSTR2)
+	bne	cr1,L(dLcr1)
+	cmplw	cr5,rWORD7,rWORD8
+	bdnz	L(dLoop)
+	bne	cr6,L(dLcr6)
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+	.align	3
+L(dP1x):
+	slwi.	r12,rN,3
+	bne	cr5,L(dLcr5)
+	subfic	rN,r12,32	/* Shift count is 32 - (rN * 8).  */
+	lwz	1,0(1)
+	bne	L(d00)
+	li	rRTN,0
+	blr
+
+/* Remainder is 8 */
+	.align	4
+L(dP2):
+	mtctr	rTMP
+	lwz	rWORD5,0(rSTR1)
+	lwz	rWORD6,0(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	blt	cr7,L(dP2x)
+	lwz	rWORD7,4(rSTR1)
+	lwz	rWORD8,4(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+L(dP2e):
+	lwz	rWORD1,8(rSTR1)
+	lwz	rWORD2,8(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	lwz	rWORD3,12(rSTR1)
+	lwz	rWORD4,12(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	addi	rSTR1,rSTR1,4
+	addi	rSTR2,rSTR2,4
+	bne	cr6,L(dLcr6)
+	bne	cr5,L(dLcr5)
+	b	L(dLoop2)
+/* Again we are on a early exit path (16-23 byte compare), we want to
+   only use volatile registers and avoid restoring non-volatile
+   registers.  */
+	.align	4
+L(dP2x):
+	lwz	rWORD3,4(rSTR1)
+	lwz	rWORD4,4(rSTR2)
+	cmplw	cr5,rWORD3,rWORD4
+	slwi.	r12,rN,3
+	bne	cr6,L(dLcr6)
+	addi	rSTR1,rSTR1,4
+	addi	rSTR2,rSTR2,4
+	bne	cr5,L(dLcr5)
+	subfic	rN,r12,32	/* Shift count is 32 - (rN * 8).  */
+	lwz	1,0(1)
+	bne	L(d00)
+	li	rRTN,0
+	blr
+
+/* Remainder is 12 */
+	.align	4
+L(dP3):
+	mtctr	rTMP
+	lwz	rWORD3,0(rSTR1)
+	lwz	rWORD4,0(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+L(dP3e):
+	lwz	rWORD5,4(rSTR1)
+	lwz	rWORD6,4(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	blt	cr7,L(dP3x)
+	lwz	rWORD7,8(rSTR1)
+	lwz	rWORD8,8(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	lwz	rWORD1,12(rSTR1)
+	lwz	rWORD2,12(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	addi	rSTR1,rSTR1,8
+	addi	rSTR2,rSTR2,8
+	bne	cr1,L(dLcr1)
+	bne	cr6,L(dLcr6)
+	b	L(dLoop1)
+/* Again we are on a early exit path (24-31 byte compare), we want to
+   only use volatile registers and avoid restoring non-volatile
+   registers.  */
+	.align	4
+L(dP3x):
+	lwz	rWORD1,8(rSTR1)
+	lwz	rWORD2,8(rSTR2)
+	cmplw	cr5,rWORD1,rWORD2
+	slwi.	r12,rN,3
+	bne	cr1,L(dLcr1)
+	addi	rSTR1,rSTR1,8
+	addi	rSTR2,rSTR2,8
+	bne	cr6,L(dLcr6)
+	subfic	rN,r12,32	/* Shift count is 32 - (rN * 8).  */
+	bne	cr5,L(dLcr5)
+	lwz	1,0(1)
+	bne	L(d00)
+	li	rRTN,0
+	blr
+
+/* Count is a multiple of 16, remainder is 0 */
+	.align	4
+L(dP4):
+	mtctr	rTMP
+	lwz	rWORD1,0(rSTR1)
+	lwz	rWORD2,0(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+L(dP4e):
+	lwz	rWORD3,4(rSTR1)
+	lwz	rWORD4,4(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	lwz	rWORD5,8(rSTR1)
+	lwz	rWORD6,8(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	lwzu	rWORD7,12(rSTR1)
+	lwzu	rWORD8,12(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr0,L(dLcr0)
+	bne	cr1,L(dLcr1)
+	bdz-	L(d24)		/* Adjust CTR as we start with +4 */
+/* This is the primary loop */
+	.align	4
+L(dLoop):
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	bne	cr6,L(dLcr6)
+L(dLoop1):
+	lwz	rWORD3,8(rSTR1)
+	lwz	rWORD4,8(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr5,L(dLcr5)
+L(dLoop2):
+	lwz	rWORD5,12(rSTR1)
+	lwz	rWORD6,12(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr0,L(dLcr0)
+L(dLoop3):
+	lwzu	rWORD7,16(rSTR1)
+	lwzu	rWORD8,16(rSTR2)
+	bne	cr1,L(dLcr1)
+	cmplw	cr0,rWORD1,rWORD2
+	bdnz	L(dLoop)
+
+L(dL4):
+	cmplw	cr1,rWORD3,rWORD4
+	bne	cr6,L(dLcr6)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr5,L(dLcr5)
+	cmplw	cr5,rWORD7,rWORD8
+L(d44):
+	bne	cr0,L(dLcr0)
+L(d34):
+	bne	cr1,L(dLcr1)
+L(d24):
+	bne	cr6,L(dLcr6)
+L(d14):
+	slwi.	r12,rN,3
+	bne	cr5,L(dLcr5) 
+L(d04):
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+	lwz	1,0(1)
+	subfic	rN,r12,32	/* Shift count is 32 - (rN * 8).  */
+	beq	L(zeroLength)
+/* At this point we have a remainder of 1 to 3 bytes to compare.  Since
+   we are aligned it is safe to load the whole word, and use
+   shift right to eliminate bits beyond the compare length. */ 
+L(d00):
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2) 
+	srw	rWORD1,rWORD1,rN
+	srw	rWORD2,rWORD2,rN
+	cmplw	rWORD1,rWORD2
+	li	rRTN,0
+	beqlr
+	li	rRTN,1
+	bgtlr
+	li	rRTN,-1
+	blr
+
+	.align	4
+L(dLcr0):
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+	li	rRTN,1
+	lwz	1,0(1)
+	bgtlr	cr0
+	li	rRTN,-1
+	blr
+	.align	4
+L(dLcr1):
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+	li	rRTN,1
+	lwz	1,0(1)
+	bgtlr	cr1
+	li	rRTN,-1
+	blr
+	.align	4
+L(dLcr6):
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+	li	rRTN,1
+	lwz	1,0(1)
+	bgtlr	cr6
+	li	rRTN,-1
+	blr
+	.align	4
+L(dLcr5):
+	lwz	r30,44(1)
+	lwz	r31,48(1)
+L(dLcr5x):
+	li	rRTN,1
+	lwz	1,0(1)
+	bgtlr	cr5
+	li	rRTN,-1
+	blr
+
+	.align	4
+L(bytealigned):
+	cfi_adjust_cfa_offset(-64)
+	mtctr	rN
+
+/* We need to prime this loop.  This loop is swing modulo scheduled
+   to avoid pipe delays.  The dependent instruction latencies (load to 
+   compare to conditional branch) is 2 to 3 cycles.  In this loop each
+   dispatch group ends in a branch and takes 1 cycle.  Effectively
+   the first iteration of the loop only serves to load operands and 
+   branches based on compares are delayed until the next loop. 
+
+   So we must precondition some registers and condition codes so that
+   we don't exit the loop early on the first iteration.  */
+	lbz	rWORD1,0(rSTR1)
+	lbz	rWORD2,0(rSTR2)
+	bdz	L(b11)
+	cmplw	cr0,rWORD1,rWORD2
+	lbz	rWORD3,1(rSTR1)
+	lbz	rWORD4,1(rSTR2)
+	bdz	L(b12)
+	cmplw	cr1,rWORD3,rWORD4
+	lbzu	rWORD5,2(rSTR1)
+	lbzu	rWORD6,2(rSTR2)
+	bdz	L(b13)
+	.align	4
+L(bLoop):
+	lbzu	rWORD1,1(rSTR1)
+	lbzu	rWORD2,1(rSTR2)
+	bne	cr0,L(bLcr0)
+
+	cmplw	cr6,rWORD5,rWORD6
+	bdz	L(b3i)
+
+	lbzu	rWORD3,1(rSTR1)
+	lbzu	rWORD4,1(rSTR2)
+	bne	cr1,L(bLcr1)
+
+	cmplw	cr0,rWORD1,rWORD2
+	bdz	L(b2i)
+
+	lbzu	rWORD5,1(rSTR1)
+	lbzu	rWORD6,1(rSTR2)
+	bne	cr6,L(bLcr6)
+
+	cmplw	cr1,rWORD3,rWORD4
+	bdnz	L(bLoop)
+
+/* We speculatively loading bytes before we have tested the previous
+   bytes.  But we must avoid overrunning the length (in the ctr) to
+   prevent these speculative loads from causing a segfault.  In this 
+   case the loop will exit early (before the all pending bytes are
+   tested.  In this case we must complete the pending operations
+   before returning.  */
+L(b1i):
+	bne	cr0,L(bLcr0)
+	bne	cr1,L(bLcr1)
+	b	L(bx56)
+	.align	4
+L(b2i):
+	bne	cr6,L(bLcr6)
+	bne	cr0,L(bLcr0)
+	b	L(bx34)
+	.align	4
+L(b3i):
+	bne	cr1,L(bLcr1)
+	bne	cr6,L(bLcr6)
+	b	L(bx12)
+	.align	4
+L(bLcr0):
+	li	rRTN,1
+	bgtlr	cr0
+	li	rRTN,-1
+	blr
+L(bLcr1):
+	li	rRTN,1
+	bgtlr	cr1
+	li	rRTN,-1
+	blr
+L(bLcr6):
+	li	rRTN,1
+	bgtlr	cr6
+	li	rRTN,-1
+	blr
+
+L(b13):
+	bne	cr0,L(bx12)
+	bne	cr1,L(bx34)
+L(bx56):
+	sub	rRTN,rWORD5,rWORD6
+	blr
+	nop
+L(b12):
+	bne	cr0,L(bx12)
+L(bx34):
+	sub	rRTN,rWORD3,rWORD4
+	blr
+
+L(b11):
+L(bx12):
+	sub	rRTN,rWORD1,rWORD2
+	blr
+
+	.align	4 
+L(zeroLengthReturn):
+
+L(zeroLength):
+	li	rRTN,0
+	blr
+
+	cfi_adjust_cfa_offset(64)
+	.align	4
+/* At this point we know the strings have different alignment and the
+   compare length is at least 8 bytes.  rBITDIF contains the low order
+   2 bits of rSTR1 and cr5 contains the result of the logical compare
+   of rBITDIF to 0.  If rBITDIF == 0 then rStr1 is word aligned and can 
+   perform the Wunaligned loop.
+  
+   Otherwise we know that rSTR1 is not aready word aligned yet.
+   So we can force the string addresses to the next lower word
+   boundary and special case this first word using shift left to
+   eliminate bits preceeding the first byte.  Since we want to join the
+   normal (Wualigned) compare loop, starting at the second word,
+   we need to adjust the length (rN) and special case the loop
+   versioning for the first W. This insures that the loop count is
+   correct and the first W (shifted) is in the expected resister pair.  */
+#define rSHL		r29	/* Unaligned shift left count.  */
+#define rSHR		r28	/* Unaligned shift right count.  */
+#define rB		r27	/* Left rotation temp for rWORD2.  */
+#define rD		r26	/* Left rotation temp for rWORD4.  */
+#define rF		r25	/* Left rotation temp for rWORD6.  */
+#define rH		r24	/* Left rotation temp for rWORD8.  */
+#define rA		r0	/* Right rotation temp for rWORD2.  */
+#define rC		r12	/* Right rotation temp for rWORD4.  */
+#define rE		r0	/* Right rotation temp for rWORD6.  */
+#define rG		r12	/* Right rotation temp for rWORD8.  */
+L(unaligned):
+	stw	r29,40(r1)
+	cfi_offset(r29,(40-64))
+	clrlwi	rSHL,rSTR2,30
+	stw	r28,36(r1)
+	cfi_offset(r28,(36-64))
+	beq	cr5,L(Wunaligned)
+	stw	r27,32(r1)
+	cfi_offset(r27,(32-64))
+/* Adjust the logical start of rSTR2 to compensate for the extra bits
+   in the 1st rSTR1 W.  */
+	sub	r27,rSTR2,rBITDIF
+/* But do not attempt to address the W before that W that contains
+   the actual start of rSTR2.  */
+	clrrwi	rSTR2,rSTR2,2
+	stw	r26,28(r1)
+	cfi_offset(r26,(28-64))
+/* Compute the left/right shift counts for the unalign rSTR2,
+   compensating for the logical (W aligned) start of rSTR1.  */ 
+	clrlwi	rSHL,r27,30
+	clrrwi	rSTR1,rSTR1,2
+	stw	r25,24(r1)
+	cfi_offset(r25,(24-64))
+	slwi	rSHL,rSHL,3
+	cmplw	cr5,r27,rSTR2
+	add	rN,rN,rBITDIF
+	slwi	r11,rBITDIF,3
+	stw	r24,20(r1)
+	cfi_offset(r24,(20-64))
+	subfic	rSHR,rSHL,32
+	srwi	rTMP,rN,4	/* Divide by 16 */
+	andi.	rBITDIF,rN,12	/* Get the W remainder */
+/* We normally need to load 2 Ws to start the unaligned rSTR2, but in
+   this special case those bits may be discarded anyway.  Also we
+   must avoid loading a W where none of the bits are part of rSTR2 as
+   this may cross a page boundary and cause a page fault.  */
+	li	rWORD8,0
+	blt	cr5,L(dus0)
+	lwz	rWORD8,0(rSTR2)
+	la	rSTR2,4(rSTR2)
+	slw	rWORD8,rWORD8,rSHL
+
+L(dus0):
+	lwz	rWORD1,0(rSTR1)
+	lwz	rWORD2,0(rSTR2)
+	cmplwi	cr1,rBITDIF,8
+	cmplwi	cr7,rN,16
+	srw	rG,rWORD2,rSHR
+	clrlwi	rN,rN,30
+	beq	L(duPs4)
+	mtctr	rTMP
+	or	rWORD8,rG,rWORD8
+	bgt	cr1,L(duPs3)
+	beq	cr1,L(duPs2)
+
+/* Remainder is 4 */
+	.align	4
+L(dusP1):
+	slw	rB,rWORD2,rSHL
+	slw	rWORD7,rWORD1,r11
+	slw	rWORD8,rWORD8,r11
+	bge	cr7,L(duP1e)
+/* At this point we exit early with the first word compare
+   complete and remainder of 0 to 3 bytes.  See L(du14) for details on
+   how we handle the remaining bytes.  */
+	cmplw	cr5,rWORD7,rWORD8
+	slwi.	rN,rN,3
+	bne	cr5,L(duLcr5)
+	cmplw	cr7,rN,rSHR
+	beq	L(duZeroReturn)
+	li	rA,0
+	ble	cr7,L(dutrim)
+	lwz	rWORD2,4(rSTR2)
+	srw	rA,rWORD2,rSHR
+	b	L(dutrim)
+/* Remainder is 8 */
+	.align	4
+L(duPs2):
+	slw	rH,rWORD2,rSHL
+	slw	rWORD5,rWORD1,r11
+	slw	rWORD6,rWORD8,r11
+	b	L(duP2e)
+/* Remainder is 12 */
+	.align	4
+L(duPs3):
+	slw	rF,rWORD2,rSHL
+	slw	rWORD3,rWORD1,r11
+	slw	rWORD4,rWORD8,r11
+	b	L(duP3e)
+/* Count is a multiple of 16, remainder is 0 */
+	.align	4
+L(duPs4):
+	mtctr	rTMP
+	or	rWORD8,rG,rWORD8
+	slw	rD,rWORD2,rSHL
+	slw	rWORD1,rWORD1,r11
+	slw	rWORD2,rWORD8,r11
+	b	L(duP4e)
+
+/* At this point we know rSTR1 is word aligned and the
+   compare length is at least 8 bytes.  */
+	.align	4
+L(Wunaligned):
+	stw	r27,32(r1)
+	cfi_offset(r27,(32-64))
+	clrrwi	rSTR2,rSTR2,2
+	stw	r26,28(r1)
+	cfi_offset(r26,(28-64))
+	srwi	rTMP,rN,4	/* Divide by 16 */
+	stw	r25,24(r1)
+	cfi_offset(r25,(24-64))
+	andi.	rBITDIF,rN,12	/* Get the W remainder */
+	stw	r24,20(r1)
+	cfi_offset(r24,(24-64))
+	slwi	rSHL,rSHL,3
+	lwz	rWORD6,0(rSTR2)
+	lwzu	rWORD8,4(rSTR2)
+	cmplwi	cr1,rBITDIF,8
+	cmplwi	cr7,rN,16
+	clrlwi	rN,rN,30
+	subfic	rSHR,rSHL,32
+	slw	rH,rWORD6,rSHL
+	beq	L(duP4)
+	mtctr	rTMP
+	bgt	cr1,L(duP3)
+	beq	cr1,L(duP2)
+
+/* Remainder is 4 */
+	.align	4
+L(duP1):
+	srw	rG,rWORD8,rSHR
+	lwz	rWORD7,0(rSTR1)
+	slw	rB,rWORD8,rSHL
+	or	rWORD8,rG,rH
+	blt	cr7,L(duP1x)
+L(duP1e):
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	srw	rA,rWORD2,rSHR
+	slw	rD,rWORD2,rSHL
+	or	rWORD2,rA,rB
+	lwz	rWORD3,8(rSTR1)
+	lwz	rWORD4,8(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	srw	rC,rWORD4,rSHR
+	slw	rF,rWORD4,rSHL
+	bne	cr5,L(duLcr5)
+	or	rWORD4,rC,rD
+	lwz	rWORD5,12(rSTR1)
+	lwz	rWORD6,12(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	srw	rE,rWORD6,rSHR
+	slw	rH,rWORD6,rSHL
+	bne	cr0,L(duLcr0)
+	or	rWORD6,rE,rF
+	cmplw	cr6,rWORD5,rWORD6
+	b	L(duLoop3)
+	.align	4
+/* At this point we exit early with the first word compare
+   complete and remainder of 0 to 3 bytes.  See L(du14) for details on
+   how we handle the remaining bytes.  */
+L(duP1x):
+	cmplw	cr5,rWORD7,rWORD8
+	slwi.	rN,rN,3
+	bne	cr5,L(duLcr5)
+	cmplw	cr7,rN,rSHR
+	beq	L(duZeroReturn)
+	li	rA,0
+	ble	cr7,L(dutrim)
+	ld	rWORD2,8(rSTR2)
+	srw	rA,rWORD2,rSHR
+	b	L(dutrim)
+/* Remainder is 8 */
+	.align	4
+L(duP2):
+	srw	rE,rWORD8,rSHR
+	lwz	rWORD5,0(rSTR1)
+	or	rWORD6,rE,rH
+	slw	rH,rWORD8,rSHL
+L(duP2e):
+	lwz	rWORD7,4(rSTR1)
+	lwz	rWORD8,4(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	srw	rG,rWORD8,rSHR
+	slw	rB,rWORD8,rSHL
+	or	rWORD8,rG,rH
+	blt	cr7,L(duP2x)
+	lwz	rWORD1,8(rSTR1)
+	lwz	rWORD2,8(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr6,L(duLcr6)
+	srw	rA,rWORD2,rSHR
+	slw	rD,rWORD2,rSHL
+	or	rWORD2,rA,rB
+	lwz	rWORD3,12(rSTR1)
+	lwz	rWORD4,12(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	bne	cr5,L(duLcr5)
+	srw	rC,rWORD4,rSHR
+	slw	rF,rWORD4,rSHL
+	or	rWORD4,rC,rD
+	addi	rSTR1,rSTR1,4
+	addi	rSTR2,rSTR2,4
+	cmplw	cr1,rWORD3,rWORD4
+	b	L(duLoop2)
+	.align	4
+L(duP2x):
+	cmplw	cr5,rWORD7,rWORD8
+	addi	rSTR1,rSTR1,4
+	addi	rSTR2,rSTR2,4
+	bne	cr6,L(duLcr6)
+	slwi.	rN,rN,3
+	bne	cr5,L(duLcr5)
+	cmplw	cr7,rN,rSHR
+	beq	L(duZeroReturn)
+	li	rA,0
+	ble	cr7,L(dutrim)
+	lwz	rWORD2,4(rSTR2)
+	srw	rA,rWORD2,rSHR
+	b	L(dutrim)
+
+/* Remainder is 12 */
+	.align	4
+L(duP3):
+	srw	rC,rWORD8,rSHR
+	lwz	rWORD3,0(rSTR1)
+	slw	rF,rWORD8,rSHL
+	or	rWORD4,rC,rH
+L(duP3e):
+	lwz	rWORD5,4(rSTR1)
+	lwz	rWORD6,4(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	srw	rE,rWORD6,rSHR
+	slw	rH,rWORD6,rSHL
+	or	rWORD6,rE,rF
+	lwz	rWORD7,8(rSTR1)
+	lwz	rWORD8,8(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr1,L(duLcr1)
+	srw	rG,rWORD8,rSHR
+	slw	rB,rWORD8,rSHL
+	or	rWORD8,rG,rH
+	blt	cr7,L(duP3x)
+	lwz	rWORD1,12(rSTR1)
+	lwz	rWORD2,12(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr6,L(duLcr6)
+	srw	rA,rWORD2,rSHR
+	slw	rD,rWORD2,rSHL
+	or	rWORD2,rA,rB
+	addi	rSTR1,rSTR1,8
+	addi	rSTR2,rSTR2,8
+	cmplw	cr0,rWORD1,rWORD2
+	b	L(duLoop1)
+	.align	4
+L(duP3x):
+	addi	rSTR1,rSTR1,8
+	addi	rSTR2,rSTR2,8
+	bne	cr1,L(duLcr1)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr6,L(duLcr6)
+	slwi.	rN,rN,3
+	bne	cr5,L(duLcr5)
+	cmplw	cr7,rN,rSHR
+	beq	L(duZeroReturn)
+	li	rA,0
+	ble	cr7,L(dutrim)
+	lwz	rWORD2,4(rSTR2)
+	srw	rA,rWORD2,rSHR
+	b	L(dutrim)
+
+/* Count is a multiple of 16, remainder is 0 */
+	.align	4
+L(duP4):
+	mtctr	rTMP
+	srw	rA,rWORD8,rSHR
+	lwz	rWORD1,0(rSTR1)
+	slw	rD,rWORD8,rSHL
+	or	rWORD2,rA,rH
+L(duP4e):
+	lwz	rWORD3,4(rSTR1)
+	lwz	rWORD4,4(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	srw	rC,rWORD4,rSHR
+	slw	rF,rWORD4,rSHL
+	or	rWORD4,rC,rD
+	lwz	rWORD5,8(rSTR1)
+	lwz	rWORD6,8(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	bne	cr0,L(duLcr0)
+	srw	rE,rWORD6,rSHR
+	slw	rH,rWORD6,rSHL
+	or	rWORD6,rE,rF
+	lwzu	rWORD7,12(rSTR1)
+	lwzu	rWORD8,12(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr1,L(duLcr1)
+	srw	rG,rWORD8,rSHR
+	slw	rB,rWORD8,rSHL
+	or	rWORD8,rG,rH
+	cmplw	cr5,rWORD7,rWORD8
+	bdz	L(du24)		/* Adjust CTR as we start with +4 */
+/* This is the primary loop */
+	.align	4
+L(duLoop):
+	lwz	rWORD1,4(rSTR1)
+	lwz	rWORD2,4(rSTR2)
+	cmplw	cr1,rWORD3,rWORD4
+	bne	cr6,L(duLcr6)
+	srw	rA,rWORD2,rSHR
+	slw	rD,rWORD2,rSHL
+	or	rWORD2,rA,rB
+L(duLoop1):
+	lwz	rWORD3,8(rSTR1)
+	lwz	rWORD4,8(rSTR2)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr5,L(duLcr5)
+	srw	rC,rWORD4,rSHR
+	slw	rF,rWORD4,rSHL
+	or	rWORD4,rC,rD
+L(duLoop2):
+	lwz	rWORD5,12(rSTR1)
+	lwz	rWORD6,12(rSTR2)
+	cmplw	cr5,rWORD7,rWORD8
+	bne	cr0,L(duLcr0)
+	srw	rE,rWORD6,rSHR
+	slw	rH,rWORD6,rSHL
+	or	rWORD6,rE,rF
+L(duLoop3):
+	lwzu	rWORD7,16(rSTR1)
+	lwzu	rWORD8,16(rSTR2)
+	cmplw	cr0,rWORD1,rWORD2
+	bne	cr1,L(duLcr1)
+	srw	rG,rWORD8,rSHR
+	slw	rB,rWORD8,rSHL
+	or	rWORD8,rG,rH
+	bdnz	L(duLoop)
+
+L(duL4):
+	bne	cr1,L(duLcr1)
+	cmplw	cr1,rWORD3,rWORD4
+	bne	cr6,L(duLcr6)
+	cmplw	cr6,rWORD5,rWORD6
+	bne	cr5,L(duLcr5)
+	cmplw	cr5,rWORD7,rWORD8
+L(du44):
+	bne	cr0,L(duLcr0)
+L(du34):
+	bne	cr1,L(duLcr1)
+L(du24):
+	bne	cr6,L(duLcr6)
+L(du14):
+	slwi.	rN,rN,3
+	bne	cr5,L(duLcr5)
+/* At this point we have a remainder of 1 to 3 bytes to compare.  We use
+   shift right to eliminate bits beyond the compare length. 
+
+   However it may not be safe to load rWORD2 which may be beyond the 
+   string length. So we compare the bit length of the remainder to
+   the right shift count (rSHR). If the bit count is less than or equal
+   we do not need to load rWORD2 (all significant bits are already in
+   rB).  */
+	cmplw	cr7,rN,rSHR
+	beq	L(duZeroReturn)
+	li	rA,0
+	ble	cr7,L(dutrim)
+	lwz	rWORD2,4(rSTR2)
+	srw	rA,rWORD2,rSHR
+	.align	4
+L(dutrim):
+	lwz	rWORD1,4(rSTR1)
+	lwz	r31,48(1)
+	subfic	rN,rN,32	/* Shift count is 32 - (rN * 8).  */ 
+	or	rWORD2,rA,rB
+	lwz	r30,44(1)
+	lwz	r29,40(r1)
+	srw	rWORD1,rWORD1,rN
+	srw	rWORD2,rWORD2,rN
+	lwz	r28,36(r1)
+	lwz	r27,32(r1)
+	cmplw	rWORD1,rWORD2
+	li	rRTN,0
+	beq	L(dureturn26)
+	li	rRTN,1
+	bgt	L(dureturn26)
+	li	rRTN,-1
+	b	L(dureturn26)
+	.align	4
+L(duLcr0):
+	lwz	r31,48(1)
+	lwz	r30,44(1)
+	li	rRTN,1
+	bgt	cr0,L(dureturn29)
+	lwz	r29,40(r1)
+	lwz	r28,36(r1)
+	li	rRTN,-1
+	b	L(dureturn27)
+	.align	4
+L(duLcr1):
+	lwz	r31,48(1)
+	lwz	r30,44(1)
+	li	rRTN,1
+	bgt	cr1,L(dureturn29)
+	lwz	r29,40(r1)
+	lwz	r28,36(r1)
+	li	rRTN,-1
+	b	L(dureturn27)
+	.align	4
+L(duLcr6):
+	lwz	r31,48(1)
+	lwz	r30,44(1)
+	li	rRTN,1
+	bgt	cr6,L(dureturn29)
+	lwz	r29,40(r1)
+	lwz	r28,36(r1)
+	li	rRTN,-1
+	b	L(dureturn27)
+	.align	4
+L(duLcr5):
+	lwz	r31,48(1)
+	lwz	r30,44(1)
+	li	rRTN,1
+	bgt	cr5,L(dureturn29)
+	lwz	r29,40(r1)
+	lwz	r28,36(r1)
+	li	rRTN,-1
+	b	L(dureturn27)
+	.align	3
+L(duZeroReturn):
+	li	rRTN,0
+	.align	4
+L(dureturn):
+	lwz	r31,48(1)
+	lwz	r30,44(1)
+L(dureturn29):
+	lwz	r29,40(r1)
+	lwz	r28,36(r1)
+L(dureturn27):
+	lwz	r27,32(r1)
+L(dureturn26):
+	lwz	r26,28(r1)
+L(dureturn25):
+	lwz	r25,24(r1)
+	lwz	r24,20(r1)
+	lwz	1,0(1)
+	blr
+END (BP_SYM (memcmp))
+libc_hidden_builtin_def (memcmp)
+weak_alias (memcmp,bcmp)
diff --git a/sysdeps/powerpc/powerpc32/power7/strncmp.S b/sysdeps/powerpc/powerpc32/power7/strncmp.S
new file mode 100644
index 0000000000..dd0f7099a8
--- /dev/null
+++ b/sysdeps/powerpc/powerpc32/power7/strncmp.S
@@ -0,0 +1,177 @@
+/* Optimized strcmp implementation for POWER7/PowerPC32.
+   Copyright (C) 2010 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA
+   02110-1301 USA.  */
+
+#include <sysdep.h>
+#include <bp-sym.h>
+#include <bp-asm.h>
+
+/* See strlen.s for comments on how the end-of-string testing works.  */
+
+/* int [r3] strncmp (const char *s1 [r3],
+		     const char *s2 [r4],
+		     size_t size [r5])  */
+
+EALIGN (BP_SYM(strncmp),4,0)
+
+#define rTMP	r0
+#define rRTN	r3
+#define rSTR1	r3	/* first string arg */
+#define rSTR2	r4	/* second string arg */
+#define rN	r5	/* max string length */
+/* Note:  The Bounded pointer support in this code is broken.  This code
+   was inherited from PPC32 and and that support was never completed.  
+   Current PPC gcc does not support -fbounds-check or -fbounded-pointers.  */
+#define rWORD1	r6	/* current word in s1 */
+#define rWORD2	r7	/* current word in s2 */
+#define rWORD3	r10
+#define rWORD4	r11
+#define rFEFE	r8	/* constant 0xfefefeff (-0x01010101) */
+#define r7F7F	r9	/* constant 0x7f7f7f7f */
+#define rNEG	r10	/* ~(word in s1 | 0x7f7f7f7f) */
+#define rBITDIF	r11	/* bits that differ in s1 & s2 words */
+
+	dcbt	0,rSTR1
+	or	rTMP,rSTR2,rSTR1
+	lis	r7F7F,0x7f7f
+	dcbt	0,rSTR2
+	clrlwi.	rTMP,rTMP,30
+	cmplwi	cr1,rN,0
+	lis	rFEFE,-0x101
+	bne	L(unaligned)
+/* We are word alligned so set up for two loops.  first a word
+   loop, then fall into the byte loop if any residual.  */
+	srwi.	rTMP,rN,2
+	clrlwi	rN,rN,30
+	addi	rFEFE,rFEFE,-0x101
+	addi	r7F7F,r7F7F,0x7f7f
+	cmplwi	cr1,rN,0
+	beq	L(unaligned)
+
+	mtctr	rTMP
+	lwz	rWORD1,0(rSTR1)
+	lwz	rWORD2,0(rSTR2)
+	b	L(g1)
+
+L(g0):
+	lwzu	rWORD1,4(rSTR1)
+	bne	cr1,L(different)
+	lwzu	rWORD2,4(rSTR2)
+L(g1):	add	rTMP,rFEFE,rWORD1
+	nor	rNEG,r7F7F,rWORD1
+	bdz	L(tail)
+	and.	rTMP,rTMP,rNEG
+	cmpw	cr1,rWORD1,rWORD2
+	beq	L(g0)
+
+/* 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...  */
+
+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)
+	cntlzw	rBITDIF,rBITDIF
+	cntlzw	rNEG,rNEG
+	addi	rNEG,rNEG,7
+	cmpw	cr1,rNEG,rBITDIF
+	sub	rRTN,rWORD1,rWORD2
+	blt	cr1,L(equal)
+	srawi	rRTN,rRTN,31
+	ori	rRTN,rRTN,1
+	blr
+L(equal):
+	li	rRTN,0
+	blr
+
+L(different):
+	lwzu	rWORD1,-4(rSTR1)
+	xor.	rBITDIF,rWORD1,rWORD2
+	sub	rRTN,rWORD1,rWORD2
+	blt	L(highbit)
+	srawi	rRTN,rRTN,31
+	ori	rRTN,rRTN,1
+	blr
+L(highbit):
+	srwi	rWORD2,rWORD2,24
+	srwi	rWORD1,rWORD1,24
+	sub	rRTN,rWORD1,rWORD2
+	blr
+
+
+/* Oh well. In this case, we just do a byte-by-byte comparison.  */
+	.align	4
+L(tail):
+	and.	rTMP,rTMP,rNEG
+	cmpw	cr1,rWORD1,rWORD2
+	bne	L(endstring)
+	addi	rSTR1,rSTR1,4
+	bne	cr1,L(different)
+	addi	rSTR2,rSTR2,4
+	cmplwi	cr1,rN,0
+L(unaligned):
+	mtctr	rN
+	ble	cr1,L(ux)
+L(uz):
+	lbz	rWORD1,0(rSTR1)
+	lbz	rWORD2,0(rSTR2)
+	.align	4
+L(u1):
+	cmpwi	cr1,rWORD1,0
+	bdz	L(u4)
+	cmpw	rWORD1,rWORD2
+	beq	cr1,L(u4)
+	lbzu	rWORD3,1(rSTR1)
+	lbzu	rWORD4,1(rSTR2)
+	bne	L(u4)
+	cmpwi	cr1,rWORD3,0
+	bdz	L(u3)
+	cmpw	rWORD3,rWORD4
+	beq	cr1,L(u3)
+	lbzu	rWORD1,1(rSTR1)
+	lbzu	rWORD2,1(rSTR2)
+	bne	L(u3)
+	cmpwi	cr1,rWORD1,0
+	bdz	L(u4)
+	cmpw	rWORD1,rWORD2
+	beq	cr1,L(u4)
+	lbzu	rWORD3,1(rSTR1)
+	lbzu	rWORD4,1(rSTR2)
+	bne	L(u4)
+	cmpwi	cr1,rWORD3,0
+	bdz	L(u3)
+	cmpw	rWORD3,rWORD4
+	beq	cr1,L(u3)
+	lbzu	rWORD1,1(rSTR1)
+	lbzu	rWORD2,1(rSTR2)
+	beq	L(u1)
+
+L(u3):  sub	rRTN,rWORD3,rWORD4
+	blr
+L(u4):	sub	rRTN,rWORD1,rWORD2
+	blr
+L(ux):
+	li	rRTN,0
+	blr
+END (BP_SYM (strncmp))
+libc_hidden_builtin_def (strncmp)