about summary refs log tree commit diff
path: root/REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S
diff options
context:
space:
mode:
authorZack Weinberg <zackw@panix.com>2017-06-08 15:39:03 -0400
committerZack Weinberg <zackw@panix.com>2017-06-08 15:39:03 -0400
commit5046dbb4a7eba5eccfd258f92f4735c9ffc8d069 (patch)
tree4470480d904b65cf14ca524f96f79eca818c3eaf /REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S
parent199fc19d3aaaf57944ef036e15904febe877fc93 (diff)
downloadglibc-zack/build-layout-experiment.tar.gz
glibc-zack/build-layout-experiment.tar.xz
glibc-zack/build-layout-experiment.zip
Prepare for radical source tree reorganization. zack/build-layout-experiment
All top-level files and directories are moved into a temporary storage
directory, REORG.TODO, except for files that will certainly still
exist in their current form at top level when we're done (COPYING,
COPYING.LIB, LICENSES, NEWS, README), all old ChangeLog files (which
are moved to the new directory OldChangeLogs, instead), and the
generated file INSTALL (which is just deleted; in the new order, there
will be no generated files checked into version control).
Diffstat (limited to 'REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S')
-rw-r--r--REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S150
1 files changed, 150 insertions, 0 deletions
diff --git a/REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S b/REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S
new file mode 100644
index 0000000000..52c4bbc1f2
--- /dev/null
+++ b/REORG.TODO/sysdeps/powerpc/powerpc32/strcmp.S
@@ -0,0 +1,150 @@
+/* Optimized strcmp implementation for PowerPC.
+   Copyright (C) 1997-2017 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, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+
+/* See strlen.s for comments on how the end-of-string testing works.  */
+
+/* int [r3] strcmp (const char *s1 [r3], const char *s2 [r4])  */
+
+EALIGN (strcmp, 4, 0)
+
+#define rTMP2	r0
+#define rRTN	r3
+#define rSTR1	r3	/* first string arg */
+#define rSTR2	r4	/* second string arg */
+#define rWORD1	r5	/* current word in s1 */
+#define rWORD2	r6	/* current word in s2 */
+#define rFEFE	r7	/* constant 0xfefefeff (-0x01010101) */
+#define r7F7F	r8	/* constant 0x7f7f7f7f */
+#define rNEG	r9	/* ~(word in s1 | 0x7f7f7f7f) */
+#define rBITDIF	r10	/* bits that differ in s1 & s2 words */
+#define rTMP	r11
+
+
+	or	rTMP, rSTR2, rSTR1
+	clrlwi.	rTMP, rTMP, 30
+	lis	rFEFE, -0x101
+	bne	L(unaligned)
+
+	lwz	rWORD1, 0(rSTR1)
+	lwz	rWORD2, 0(rSTR2)
+	lis	r7F7F, 0x7f7f
+	addi	rFEFE, rFEFE, -0x101
+	addi	r7F7F, r7F7F, 0x7f7f
+	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
+	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...  */
+#ifdef __LITTLE_ENDIAN__
+L(endstring):
+	addi    rTMP2, rTMP, -1
+	andc    rTMP2, rTMP2, rTMP
+	rlwimi	rTMP2, rTMP2, 1, 0, 30
+	and	rWORD2, rWORD2, rTMP2		/* Mask off gunk.  */
+	and	rWORD1, rWORD1, rTMP2
+	rlwinm	rTMP2, rWORD2, 8, 0xffffffff	/* Byte reverse word.  */
+	rlwinm	rTMP, rWORD1, 8, 0xffffffff
+	rlwimi	rTMP2, rWORD2, 24, 0, 7
+	rlwimi	rTMP, rWORD1, 24, 0, 7
+	rlwimi	rTMP2, rWORD2, 24, 16, 23
+	rlwimi	rTMP, rWORD1, 24, 16, 23
+	xor.	rBITDIF, rTMP, rTMP2
+	sub	rRTN, rTMP, rTMP2
+	bgelr+
+	ori	rRTN, rTMP2, 1
+	blr
+
+L(different):
+	lwz	rWORD1, -4(rSTR1)
+	rlwinm	rTMP2, rWORD2, 8, 0xffffffff	/* Byte reverse word.  */
+	rlwinm	rTMP, rWORD1, 8, 0xffffffff
+	rlwimi	rTMP2, rWORD2, 24, 0, 7
+	rlwimi	rTMP, rWORD1, 24, 0, 7
+	rlwimi	rTMP2, rWORD2, 24, 16, 23
+	rlwimi	rTMP, rWORD1, 24, 16, 23
+	xor.	rBITDIF, rTMP, rTMP2
+	sub	rRTN, rTMP, rTMP2
+	bgelr+
+	ori	rRTN, rTMP2, 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)
+	cntlzw	rBITDIF, rBITDIF
+	cntlzw	rNEG, rNEG
+	addi	rNEG, rNEG, 7
+	cmpw	cr1, rNEG, rBITDIF
+	sub	rRTN, rWORD1, rWORD2
+	bgelr+	cr1
+L(equal):
+	li	rRTN, 0
+	blr
+
+L(different):
+	lwz	rWORD1, -4(rSTR1)
+	xor.	rBITDIF, rWORD1, rWORD2
+	sub	rRTN, rWORD1, rWORD2
+	bgelr+
+L(highbit):
+	ori	rRTN, rWORD2, 1
+	blr
+#endif
+
+/* Oh well.  In this case, we just do a byte-by-byte comparison.  */
+	.align 4
+L(unaligned):
+	lbz	rWORD1, 0(rSTR1)
+	lbz	rWORD2, 0(rSTR2)
+	b	L(u1)
+
+L(u0):	lbzu	rWORD1, 1(rSTR1)
+	bne-	L(u4)
+	lbzu	rWORD2, 1(rSTR2)
+L(u1):	cmpwi	cr1, rWORD1, 0
+	beq-	cr1, L(u3)
+	cmpw	rWORD1, rWORD2
+	bne-	L(u3)
+	lbzu	rWORD1, 1(rSTR1)
+	lbzu	rWORD2, 1(rSTR2)
+	cmpwi	cr1, rWORD1, 0
+	cmpw	rWORD1, rWORD2
+	bne+	cr1, L(u0)
+L(u3):	sub	rRTN, rWORD1, rWORD2
+	blr
+L(u4):	lbz	rWORD1, -1(rSTR1)
+	sub	rRTN, rWORD1, rWORD2
+	blr
+END (strcmp)
+libc_hidden_builtin_def (strcmp)