powerpc: Optimize memrchr for power8

Vectorized loops are used for sizes greater than 32B to improve
performance over power7 optimization.  This shows as an average
of 25% improvement depending on the position of search
character.  The performance is same for shorter strings.

ChangeLog

@@ -1,3 +1,15 @@
+2017-10-02  Rajalakshmi Srinivasaraghavan  <raji@linux.vnet.ibm.com>
+
+	* sysdeps/powerpc/powerpc64/multiarch/Makefile
+	(sysdep_routines): Add memrchr_power8.
+	* sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c
+	(memrchr): Add __memrchr_power8 to list of memrchr functions.
+	* sysdeps/powerpc/powerpc64/multiarch/memrchr-power8.S:
+	New file.
+	* sysdeps/powerpc/powerpc64/multiarch/memrchr.c
+	(memrchr): Add __memrchr_power8 to ifunc list.
+	* sysdeps/powerpc/powerpc64/power8/memrchr.S: New file.
+
 2017-10-01  H.J. Lu  <hongjiu.lu@intel.com>
 
 	[BZ #18822]

sysdeps/powerpc/powerpc64/multiarch/Makefile

@@ -6,8 +6,9 @@ sysdep_routines += memcpy-power7 memcpy-a2 memcpy-power6 memcpy-cell \
 		   memset-ppc64 memset-power8 \
 		   mempcpy-power7 mempcpy-ppc64 \
 		   memchr-power8 memchr-power7 memchr-ppc64 \
-		   memrchr-power7 memrchr-ppc64 rawmemchr-power7 \
+		   memrchr-power8 memrchr-power7 memrchr-ppc64 \
-		   rawmemchr-ppc64 strlen-power7 strlen-ppc64 \
+		   rawmemchr-power7 rawmemchr-ppc64 \
+		   strlen-power7 strlen-ppc64 \
 		   strnlen-power8 strnlen-power7 strnlen-ppc64 \
 		   strcasecmp-power7 strcasecmp_l-power7 \
 		   strncase-power7 strncase_l-power7 \

sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c

@@ -193,6 +193,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 
   /* Support sysdeps/powerpc/powerpc64/multiarch/memrchr.c.  */
   IFUNC_IMPL (i, name, memrchr,
+	      IFUNC_IMPL_ADD (array, i, memrchr,
+			      hwcap2 & PPC_FEATURE2_ARCH_2_07,
+			      __memrchr_power8)
 	      IFUNC_IMPL_ADD (array, i, memrchr,
 			      hwcap & PPC_FEATURE_HAS_VSX,
 			      __memrchr_power7)

sysdeps/powerpc/powerpc64/multiarch/memrchr-power8.S

@@ -0,0 +1,28 @@
+/* Optimized memrchr implementation for PowerPC64/POWER8.
+   Copyright (C) 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>
+
+#define MEMRCHR __memrchr_power8
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+#undef weak_alias
+#define weak_alias(name,alias)
+
+#include <sysdeps/powerpc/powerpc64/power8/memrchr.S>

sysdeps/powerpc/powerpc64/multiarch/memrchr-ppc64.c

@@ -16,4 +16,16 @@
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
-#include <sysdeps/powerpc/powerpc32/power4/multiarch/memrchr-ppc32.c>
+#include <string.h>
+
+#define MEMRCHR  __memrchr_ppc
+
+#undef weak_alias
+#define weak_alias(a, b)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+extern __typeof (memrchr) __memrchr_ppc attribute_hidden;
+
+#include <string/memrchr.c>

sysdeps/powerpc/powerpc64/power8/memrchr.S

@@ -0,0 +1,321 @@
+/* Optimized memrchr implementation for PowerPC64/POWER8.
+   Copyright (C) 2017 Free Software Foundation, Inc.
+   Contributed by Luis Machado <luisgpm@br.ibm.com>.
+   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>
+
+/* int [r3] memrchr (char *s [r3], int byte [r4], int size [r5])  */
+
+/* TODO: change these to the actual instructions when the minimum required
+   binutils allows it.  */
+#define MTVRD(v, r) .long (0x7c000167 | ((v)<<(32-11)) | ((r)<<(32-16)))
+#define MFVRD(r, v) .long (0x7c000067 | ((v)<<(32-11)) | ((r)<<(32-16)))
+#define VBPERMQ(t, a, b)  .long (0x1000054c \
+				| ((t)<<(32-11)) \
+				| ((a)<<(32-16)) \
+				| ((b)<<(32-21)) )
+#ifndef MEMRCHR
+# define MEMRCHR __memrchr
+#endif
+	.machine  power7
+ENTRY_TOCLESS (MEMRCHR)
+	CALL_MCOUNT 3
+	add	r7, r3, r5      /* Calculate the last acceptable address.  */
+	neg	r0, r7
+	addi	r7, r7, -1
+	mr	r10, r3
+	clrrdi	r6, r7, 7
+	li	r9, 3<<5
+	dcbt	r9, r6, 8       /* Stream hint, decreasing addresses.  */
+
+	/* Replicate BYTE to doubleword.  */
+	insrdi	r4, r4, 8, 48
+	insrdi	r4, r4, 16, 32
+	insrdi	r4, r4, 32, 0
+	li	r6, -8
+	li	r9, -1
+	rlwinm	r0, r0, 3, 26, 28 /* Calculate padding.  */
+	clrrdi	r8, r7, 3
+	srd	r9, r9, r0
+	cmpldi	r5, 32
+	clrrdi	r0, r10, 3
+	ble	L(small_range)
+
+#ifdef __LITTLE_ENDIAN__
+	ldx	r12, 0, r8
+#else
+	ldbrx	r12, 0, r8      /* Load reversed doubleword from memory.  */
+#endif
+	cmpb	r3, r12, r4     /* Check for BYTE in DWORD1.  */
+	and	r3, r3, r9
+	cmpldi	cr7, r3, 0      /* If r3 == 0, no BYTEs have been found.  */
+	bne	cr7, L(done)
+
+	/* Are we now aligned to a quadword boundary?  If so, skip to
+	   the main loop.  Otherwise, go through the alignment code.  */
+	andi.	r12, r8, 15
+	beq	cr0, L(align_qw)
+
+	/* Handle DWORD2 of pair.  */
+#ifdef __LITTLE_ENDIAN__
+	ldx	r12, r8, r6
+#else
+	ldbrx	r12, r8, r6
+#endif
+	addi	r8, r8, -8
+	cmpb	r3, r12, r4
+	cmpldi	cr7, r3, 0
+	bne	cr7, L(done)
+
+	.align	4
+	/* At this point, r8 is 16B aligned.  */
+L(align_qw):
+	sub	r5, r8, r0
+	vspltisb	v0, 0
+	/* Precompute vbpermq constant.  */
+	vspltisb	v10, 3
+	li	r0, 0
+	lvsl	v11, r0, r0
+	vslb	v10, v11, v10
+	MTVRD(v1, r4)
+	vspltb	v1, v1, 7
+	cmpldi	r5, 64
+	ble	L(tail64)
+	/* Are we 64-byte aligned? If so, jump to the vectorized loop.
+	   Note: aligning to 64-byte will necessarily slow down performance for
+	   strings around 64 bytes in length due to the extra comparisons
+	   required to check alignment for the vectorized loop.  This is a
+	   necessary tradeoff we are willing to take in order to speed up the
+	   calculation for larger strings.  */
+	andi.	r11, r8, 63
+	beq	cr0, L(preloop_64B)
+	/* In order to begin the 64B loop, it needs to be 64
+	   bytes aligned.  So read until it is 64B aligned.  */
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	addi	r5, r5, -16
+
+	andi.	r11, r8, 63
+	beq	cr0, L(preloop_64B)
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	addi	r5, r5, -16
+
+	andi.	r11, r8, 63
+	beq	cr0, L(preloop_64B)
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	addi	r5, r5, -16
+	/* At this point it should be 64B aligned.
+	   Prepare for the 64B loop.  */
+L(preloop_64B):
+	cmpldi	r5, 64		/* Check if r5 < 64.  */
+	ble	L(tail64)
+	srdi	r9, r5, 6	/* Number of loop iterations.  */
+	mtctr	r9		/* Setup the counter.  */
+	li	r11, 16		/* Load required offsets.  */
+	li	r9, 32
+	li	r7, 48
+
+	/* Handle r5 > 64.  Loop over the bytes in strides of 64B.  */
+	.align 4
+L(loop):
+	addi	r8, r8, -64	/* Adjust address for the next iteration.  */
+	lvx	v2, 0, r8	/* Load 4 quadwords.  */
+	lvx	v3, r8, r11
+	lvx	v4, v8, r9
+	lvx	v5, v8, r7
+	vcmpequb	v6, v1, v2
+	vcmpequb	v7, v1, v3
+	vcmpequb	v8, v1, v4
+	vcmpequb	v9, v1, v5
+	vor	v11, v6, v7
+	vor	v12, v8, v9
+	vor	v11, v11, v12	/* Compare and merge into one VR for speed.  */
+	vcmpequb.	v11, v0, v11
+	bnl	cr6, L(found)
+	bdnz	L(loop)
+	clrldi	r5, r5, 58
+
+	/* Handle remainder of 64B loop or r5 > 64.  */
+	.align	4
+L(tail64):
+	cmpldi	r5, 0
+	beq	L(null)
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	cmpldi	cr6, r5, 16
+	ble	cr6, L(null)
+	addi	r5, r5, -16
+
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	cmpldi	cr6, r5, 16
+	ble	cr6, L(null)
+	addi	r5, r5, -16
+
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	cmpldi	cr6, r5, 16
+	ble	cr6, L(null)
+	addi	r5, r5, -16
+
+	addi	r8, r8, -16
+	lvx	v4, 0, r8
+	vcmpequb	v6, v1, v4
+	vcmpequb.	v11, v0, v6
+	bnl	cr6, L(found_16B)
+	li	r3, 0
+	blr
+
+	/* Found a match in 64B loop.  */
+	.align	4
+L(found):
+	/* Permute the first bit of each byte into bits 48-63.  */
+	VBPERMQ(v6, v6, v10)
+	VBPERMQ(v7, v7, v10)
+	VBPERMQ(v8, v8, v10)
+	VBPERMQ(v9, v9, v10)
+	/* Shift each component into its correct position for merging.  */
+#ifdef __LITTLE_ENDIAN__
+	vsldoi	v7, v7, v7, 2
+	vsldoi	v8, v8, v8, 4
+	vsldoi	v9, v9, v9, 6
+#else
+	vsldoi	v6, v6, v6, 6
+	vsldoi	v7, v7, v7, 4
+	vsldoi	v8, v8, v8, 2
+#endif
+	/* Merge the results and move to a GPR.  */
+	vor	v11, v6, v7
+	vor	v4, v9, v8
+	vor	v4, v11, v4
+	MFVRD(r5, v4)
+#ifdef __LITTLE_ENDIAN__
+	cntlzd	r6, r5	/* Count leading zeros before the match.  */
+#else
+	addi	r6, r5, -1
+	andc	r6, r6, r5
+	popcntd	r6, r6
+#endif
+	addi	r8, r8, 63
+	sub	r3, r8, r6	/* Compute final address.  */
+	blr
+
+	/* Found a match in last 16 bytes.  */
+	.align	4
+L(found_16B):
+	/* Permute the first bit of each byte into bits 48-63.  */
+	VBPERMQ(v6, v6, v10)
+	/* Shift each component into its correct position for merging.  */
+#ifdef __LITTLE_ENDIAN__
+	vsldoi	v6, v6, v6, 6
+	MFVRD(r7, v6)
+	cntlzd	r6, r7	/* Count leading zeros before the match.  */
+#else
+	MFVRD(r7, v6)
+	addi	r6, r7, -1
+	andc	r6, r6, r7
+	popcntd	r6, r6
+#endif
+	addi	r8, r8, 15
+	sub	r3, r8, r6	/* Compute final address.  */
+	cmpld	r6, r5
+	bltlr
+	li	r3, 0
+	blr
+
+	/* r3 has the output of the cmpb instruction, that is, it contains
+	   0xff in the same position as BYTE in the original
+	   word from the string.  Use that to calculate the pointer.
+	   We need to make sure BYTE is *before* the end of the
+	   range.  */
+L(done):
+	cntlzd	r9, r3	      /* Count leading zeros before the match.  */
+	cmpld	r8, r0         /* Are we on the last word?  */
+	srdi	r6, r9, 3	      /* Convert leading zeros to bytes.  */
+	addi	r0, r6, -7
+	sub	r3, r8, r0
+	cmpld	cr7, r3, r10
+	bnelr
+	bgelr	cr7
+	li	r3, 0
+	blr
+
+	.align	4
+L(null):
+	li	r3, 0
+	blr
+
+/* Deals with size <= 32.  */
+	.align	4
+L(small_range):
+	cmpldi	r5, 0
+	beq	L(null)
+
+#ifdef __LITTLE_ENDIAN__
+	ldx	r12, 0, r8
+#else
+	ldbrx	r12, 0, r8      /* Load reversed doubleword from memory.  */
+#endif
+	cmpb	r3, r12, r4     /* Check for BYTE in DWORD1.  */
+	and	r3, r3, r9
+	cmpldi	cr7, r3, 0
+	bne	cr7, L(done)
+
+	/* Are we done already?  */
+	cmpld	r8, r0
+	addi	r8, r8, -8
+	beqlr
+
+	.align	5
+L(loop_small):
+#ifdef __LITTLE_ENDIAN__
+	ldx	r12, 0, r8
+#else
+	ldbrx	r12, 0, r8
+#endif
+	cmpb	r3, r12, r4
+	cmpld	r8, r0
+	cmpldi	cr7, r3, 0
+	bne	cr7, L(done)
+	addi	r8, r8, -8
+	bne	L(loop_small)
+	blr
+
+END (MEMRCHR)
+weak_alias (__memrchr, memrchr)
+libc_hidden_builtin_def (memrchr)