x86: Reduce code size of mem{move|pcpy|cpy}-ssse3

The goal is to remove most SSSE3 function as SSE4, AVX2, and EVEX are
generally preferable. memcpy/memmove is one exception where avoiding
unaligned loads with `palignr` is important for some targets.

This commit replaces memmove-ssse3 with a better optimized are lower
code footprint verion. As well it aliases memcpy to memmove.

Aside from this function all other SSSE3 functions should be safe to
remove.

The performance is not changed drastically although shows overall
improvements without any major regressions or gains.

bench-memcpy geometric_mean(N=50) New / Original: 0.957

bench-memcpy-random geometric_mean(N=50) New / Original: 0.912

bench-memcpy-large geometric_mean(N=50) New / Original: 0.892

Benchmarks where run on Zhaoxin KX-6840@2000MHz See attached numbers
for all results.

More important this saves 7246 bytes of code size in memmove an
additional 10741 bytes by reusing memmove code for memcpy (total 17987
bytes saves). As well an additional 896 bytes of rodata for the jump
table entries.

3 files changed, 380 insertions, 3156 deletions

diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
index 303fb5d734..e7ea963fc0 100644
--- a/sysdeps/x86_64/multiarch/Makefile
+++ b/sysdeps/x86_64/multiarch/Makefile
@@ -16,7 +16,6 @@ sysdep_routines += \
   memcmpeq-avx2-rtm \
   memcmpeq-evex \
   memcmpeq-sse2 \
-  memcpy-ssse3 \
   memmove-avx-unaligned-erms \
   memmove-avx-unaligned-erms-rtm \
   memmove-avx512-no-vzeroupper \
diff --git a/sysdeps/x86_64/multiarch/memcpy-ssse3.S b/sysdeps/x86_64/multiarch/memcpy-ssse3.S
deleted file mode 100644
index 65644d3a09..0000000000
--- a/sysdeps/x86_64/multiarch/memcpy-ssse3.S
+++ /dev/null
@@ -1,3151 +0,0 @@
-/* memcpy with SSSE3
-   Copyright (C) 2010-2022 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
-   <https://www.gnu.org/licenses/>.  */
-
-#include <sysdep.h>
-
-#if IS_IN (libc)
-
-#include "asm-syntax.h"
-
-#ifndef MEMCPY
-# define MEMCPY		__memcpy_ssse3
-# define MEMCPY_CHK	__memcpy_chk_ssse3
-# define MEMPCPY	__mempcpy_ssse3
-# define MEMPCPY_CHK	__mempcpy_chk_ssse3
-#endif
-
-#define JMPTBL(I, B)	I - B
-
-/* Branch to an entry in a jump table.  TABLE is a jump table with
-   relative offsets.  INDEX is a register contains the index into the
-   jump table.  SCALE is the scale of INDEX.  */
-#define BRANCH_TO_JMPTBL_ENTRY(TABLE, INDEX, SCALE)		\
-  lea		TABLE(%rip), %r11;				\
-  movslq	(%r11, INDEX, SCALE), INDEX;			\
-  lea		(%r11, INDEX), INDEX;				\
-  _CET_NOTRACK jmp *INDEX;					\
-  ud2
-
-	.section .text.ssse3,"ax",@progbits
-#if !defined USE_AS_MEMPCPY && !defined USE_AS_MEMMOVE
-ENTRY (MEMPCPY_CHK)
-	cmp	%RDX_LP, %RCX_LP
-	jb	HIDDEN_JUMPTARGET (__chk_fail)
-END (MEMPCPY_CHK)
-
-ENTRY (MEMPCPY)
-	mov	%RDI_LP, %RAX_LP
-	add	%RDX_LP, %RAX_LP
-	jmp	L(start)
-END (MEMPCPY)
-#endif
-
-#if !defined USE_AS_BCOPY
-ENTRY (MEMCPY_CHK)
-	cmp	%RDX_LP, %RCX_LP
-	jb	HIDDEN_JUMPTARGET (__chk_fail)
-END (MEMCPY_CHK)
-#endif
-
-ENTRY (MEMCPY)
-	mov	%RDI_LP, %RAX_LP
-#ifdef USE_AS_MEMPCPY
-	add	%RDX_LP, %RAX_LP
-#endif
-
-#ifdef __ILP32__
-	/* Clear the upper 32 bits.  */
-	mov	%edx, %edx
-#endif
-
-#ifdef USE_AS_MEMMOVE
-	cmp	%rsi, %rdi
-	jb	L(copy_forward)
-	je	L(write_0bytes)
-	cmp	$79, %rdx
-	jbe	L(copy_forward)
-	jmp	L(copy_backward)
-L(copy_forward):
-#endif
-L(start):
-	cmp	$79, %rdx
-	lea     L(table_less_80bytes)(%rip), %r11
-	ja	L(80bytesormore)
-	movslq	(%r11, %rdx, 4), %r9
-	add	%rdx, %rsi
-	add	%rdx, %rdi
-	add	%r11, %r9
-	_CET_NOTRACK jmp *%r9
-	ud2
-
-	.p2align 4
-L(80bytesormore):
-#ifndef USE_AS_MEMMOVE
-	cmp	%dil, %sil
-	jle	L(copy_backward)
-#endif
-
-	movdqu	(%rsi), %xmm0
-	mov	%rdi, %rcx
-	and	$-16, %rdi
-	add	$16, %rdi
-	mov	%rcx, %r8
-	sub	%rdi, %rcx
-	add	%rcx, %rdx
-	sub	%rcx, %rsi
-
-#ifdef SHARED_CACHE_SIZE_HALF
-	mov	$SHARED_CACHE_SIZE_HALF, %RCX_LP
-#else
-	mov	__x86_shared_cache_size_half(%rip), %RCX_LP
-#endif
-	cmp	%rcx, %rdx
-	mov	%rsi, %r9
-	ja	L(large_page_fwd)
-	and	$0xf, %r9
-	jz	L(shl_0)
-#ifdef DATA_CACHE_SIZE_HALF
-	mov	$DATA_CACHE_SIZE_HALF, %RCX_LP
-#else
-	mov	__x86_data_cache_size_half(%rip), %RCX_LP
-#endif
-	BRANCH_TO_JMPTBL_ENTRY (L(shl_table), %r9, 4)
-
-	.p2align 4
-L(copy_backward):
-	movdqu	-16(%rsi, %rdx), %xmm0
-	add	%rdx, %rsi
-	lea	-16(%rdi, %rdx), %r8
-	add	%rdx, %rdi
-
-	mov	%rdi, %rcx
-	and	$0xf, %rcx
-	xor	%rcx, %rdi
-	sub	%rcx, %rdx
-	sub	%rcx, %rsi
-
-#ifdef SHARED_CACHE_SIZE_HALF
-	mov	$SHARED_CACHE_SIZE_HALF, %RCX_LP
-#else
-	mov	__x86_shared_cache_size_half(%rip), %RCX_LP
-#endif
-
-	cmp	%rcx, %rdx
-	mov	%rsi, %r9
-	ja	L(large_page_bwd)
-	and	$0xf, %r9
-	jz	L(shl_0_bwd)
-#ifdef DATA_CACHE_SIZE_HALF
-	mov	$DATA_CACHE_SIZE_HALF, %RCX_LP
-#else
-	mov	__x86_data_cache_size_half(%rip), %RCX_LP
-#endif
-	BRANCH_TO_JMPTBL_ENTRY (L(shl_table_bwd), %r9, 4)
-
-	.p2align 4
-L(shl_0):
-	sub	$16, %rdx
-	movdqa	(%rsi), %xmm1
-	add	$16, %rsi
-	movdqa	%xmm1, (%rdi)
-	add	$16, %rdi
-	cmp	$128, %rdx
-	movdqu	%xmm0, (%r8)
-	ja	L(shl_0_gobble)
-	cmp	$64, %rdx
-	jb	L(shl_0_less_64bytes)
-	movaps	(%rsi), %xmm4
-	movaps	16(%rsi), %xmm1
-	movaps	32(%rsi), %xmm2
-	movaps	48(%rsi), %xmm3
-	movaps	%xmm4, (%rdi)
-	movaps	%xmm1, 16(%rdi)
-	movaps	%xmm2, 32(%rdi)
-	movaps	%xmm3, 48(%rdi)
-	sub	$64, %rdx
-	add	$64, %rsi
-	add	$64, %rdi
-L(shl_0_less_64bytes):
-	add	%rdx, %rsi
-	add	%rdx, %rdi
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_0_gobble):
-#ifdef DATA_CACHE_SIZE_HALF
-	cmp	$DATA_CACHE_SIZE_HALF, %RDX_LP
-#else
-	cmp	__x86_data_cache_size_half(%rip), %RDX_LP
-#endif
-	lea	-128(%rdx), %rdx
-	jae	L(shl_0_gobble_mem_loop)
-L(shl_0_gobble_cache_loop):
-	movdqa	(%rsi), %xmm4
-	movaps	0x10(%rsi), %xmm1
-	movaps	0x20(%rsi), %xmm2
-	movaps	0x30(%rsi), %xmm3
-
-	movdqa	%xmm4, (%rdi)
-	movaps	%xmm1, 0x10(%rdi)
-	movaps	%xmm2, 0x20(%rdi)
-	movaps	%xmm3, 0x30(%rdi)
-
-	sub	$128, %rdx
-	movaps	0x40(%rsi), %xmm4
-	movaps	0x50(%rsi), %xmm5
-	movaps	0x60(%rsi), %xmm6
-	movaps	0x70(%rsi), %xmm7
-	lea	0x80(%rsi), %rsi
-	movaps	%xmm4, 0x40(%rdi)
-	movaps	%xmm5, 0x50(%rdi)
-	movaps	%xmm6, 0x60(%rdi)
-	movaps	%xmm7, 0x70(%rdi)
-	lea	0x80(%rdi), %rdi
-
-	jae	L(shl_0_gobble_cache_loop)
-	cmp	$-0x40, %rdx
-	lea	0x80(%rdx), %rdx
-	jl	L(shl_0_cache_less_64bytes)
-
-	movdqa	(%rsi), %xmm4
-	sub	$0x40, %rdx
-	movdqa	0x10(%rsi), %xmm1
-
-	movdqa	%xmm4, (%rdi)
-	movdqa	%xmm1, 0x10(%rdi)
-
-	movdqa	0x20(%rsi), %xmm4
-	movdqa	0x30(%rsi), %xmm1
-	add	$0x40, %rsi
-
-	movdqa	%xmm4, 0x20(%rdi)
-	movdqa	%xmm1, 0x30(%rdi)
-	add	$0x40, %rdi
-L(shl_0_cache_less_64bytes):
-	add	%rdx, %rsi
-	add	%rdx, %rdi
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_0_gobble_mem_loop):
-	prefetcht0 0x1c0(%rsi)
-	prefetcht0 0x280(%rsi)
-
-	movdqa	(%rsi), %xmm0
-	movdqa	0x10(%rsi), %xmm1
-	movdqa	0x20(%rsi), %xmm2
-	movdqa	0x30(%rsi), %xmm3
-	movdqa	0x40(%rsi), %xmm4
-	movdqa	0x50(%rsi), %xmm5
-	movdqa	0x60(%rsi), %xmm6
-	movdqa	0x70(%rsi), %xmm7
-	lea	0x80(%rsi), %rsi
-	sub	$0x80, %rdx
-	movdqa	%xmm0, (%rdi)
-	movdqa	%xmm1, 0x10(%rdi)
-	movdqa	%xmm2, 0x20(%rdi)
-	movdqa	%xmm3, 0x30(%rdi)
-	movdqa	%xmm4, 0x40(%rdi)
-	movdqa	%xmm5, 0x50(%rdi)
-	movdqa	%xmm6, 0x60(%rdi)
-	movdqa	%xmm7, 0x70(%rdi)
-	lea	0x80(%rdi), %rdi
-
-	jae	L(shl_0_gobble_mem_loop)
-	cmp	$-0x40, %rdx
-	lea	0x80(%rdx), %rdx
-	jl	L(shl_0_mem_less_64bytes)
-
-	movdqa	(%rsi), %xmm0
-	sub	$0x40, %rdx
-	movdqa	0x10(%rsi), %xmm1
-
-	movdqa	%xmm0, (%rdi)
-	movdqa	%xmm1, 0x10(%rdi)
-
-	movdqa	0x20(%rsi), %xmm0
-	movdqa	0x30(%rsi), %xmm1
-	add	$0x40, %rsi
-
-	movdqa	%xmm0, 0x20(%rdi)
-	movdqa	%xmm1, 0x30(%rdi)
-	add	$0x40, %rdi
-L(shl_0_mem_less_64bytes):
-	cmp	$0x20, %rdx
-	jb	L(shl_0_mem_less_32bytes)
-	movdqa	(%rsi), %xmm0
-	sub	$0x20, %rdx
-	movdqa	0x10(%rsi), %xmm1
-	add	$0x20, %rsi
-	movdqa	%xmm0, (%rdi)
-	movdqa	%xmm1, 0x10(%rdi)
-	add	$0x20, %rdi
-L(shl_0_mem_less_32bytes):
-	add	%rdx, %rdi
-	add	%rdx, %rsi
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_0_bwd):
-	sub	$16, %rdx
-	movdqa	-0x10(%rsi), %xmm1
-	sub	$16, %rsi
-	movdqa	%xmm1, -0x10(%rdi)
-	sub	$16, %rdi
-	cmp	$0x80, %rdx
-	movdqu	%xmm0, (%r8)
-	ja	L(shl_0_gobble_bwd)
-	cmp	$64, %rdx
-	jb	L(shl_0_less_64bytes_bwd)
-	movaps	-0x10(%rsi), %xmm0
-	movaps	-0x20(%rsi), %xmm1
-	movaps	-0x30(%rsi), %xmm2
-	movaps	-0x40(%rsi), %xmm3
-	movaps	%xmm0, -0x10(%rdi)
-	movaps	%xmm1, -0x20(%rdi)
-	movaps	%xmm2, -0x30(%rdi)
-	movaps	%xmm3, -0x40(%rdi)
-	sub	$64, %rdx
-	sub	$0x40, %rsi
-	sub	$0x40, %rdi
-L(shl_0_less_64bytes_bwd):
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_0_gobble_bwd):
-#ifdef DATA_CACHE_SIZE_HALF
-	cmp	$DATA_CACHE_SIZE_HALF, %RDX_LP
-#else
-	cmp	__x86_data_cache_size_half(%rip), %RDX_LP
-#endif
-	lea	-128(%rdx), %rdx
-	jae	L(shl_0_gobble_mem_bwd_loop)
-L(shl_0_gobble_bwd_loop):
-	movdqa	-0x10(%rsi), %xmm0
-	movaps	-0x20(%rsi), %xmm1
-	movaps	-0x30(%rsi), %xmm2
-	movaps	-0x40(%rsi), %xmm3
-
-	movdqa	%xmm0, -0x10(%rdi)
-	movaps	%xmm1, -0x20(%rdi)
-	movaps	%xmm2, -0x30(%rdi)
-	movaps	%xmm3, -0x40(%rdi)
-
-	sub	$0x80, %rdx
-	movaps	-0x50(%rsi), %xmm4
-	movaps	-0x60(%rsi), %xmm5
-	movaps	-0x70(%rsi), %xmm6
-	movaps	-0x80(%rsi), %xmm7
-	lea	-0x80(%rsi), %rsi
-	movaps	%xmm4, -0x50(%rdi)
-	movaps	%xmm5, -0x60(%rdi)
-	movaps	%xmm6, -0x70(%rdi)
-	movaps	%xmm7, -0x80(%rdi)
-	lea	-0x80(%rdi), %rdi
-
-	jae	L(shl_0_gobble_bwd_loop)
-	cmp	$-0x40, %rdx
-	lea	0x80(%rdx), %rdx
-	jl	L(shl_0_gobble_bwd_less_64bytes)
-
-	movdqa	-0x10(%rsi), %xmm0
-	sub	$0x40, %rdx
-	movdqa	-0x20(%rsi), %xmm1
-
-	movdqa	%xmm0, -0x10(%rdi)
-	movdqa	%xmm1, -0x20(%rdi)
-
-	movdqa	-0x30(%rsi), %xmm0
-	movdqa	-0x40(%rsi), %xmm1
-	sub	$0x40, %rsi
-
-	movdqa	%xmm0, -0x30(%rdi)
-	movdqa	%xmm1, -0x40(%rdi)
-	sub	$0x40, %rdi
-L(shl_0_gobble_bwd_less_64bytes):
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_0_gobble_mem_bwd_loop):
-	prefetcht0 -0x1c0(%rsi)
-	prefetcht0 -0x280(%rsi)
-	movdqa	-0x10(%rsi), %xmm0
-	movdqa	-0x20(%rsi), %xmm1
-	movdqa	-0x30(%rsi), %xmm2
-	movdqa	-0x40(%rsi), %xmm3
-	movdqa	-0x50(%rsi), %xmm4
-	movdqa	-0x60(%rsi), %xmm5
-	movdqa	-0x70(%rsi), %xmm6
-	movdqa	-0x80(%rsi), %xmm7
-	lea	-0x80(%rsi), %rsi
-	sub	$0x80, %rdx
-	movdqa	%xmm0, -0x10(%rdi)
-	movdqa	%xmm1, -0x20(%rdi)
-	movdqa	%xmm2, -0x30(%rdi)
-	movdqa	%xmm3, -0x40(%rdi)
-	movdqa	%xmm4, -0x50(%rdi)
-	movdqa	%xmm5, -0x60(%rdi)
-	movdqa	%xmm6, -0x70(%rdi)
-	movdqa	%xmm7, -0x80(%rdi)
-	lea	-0x80(%rdi), %rdi
-
-	jae	L(shl_0_gobble_mem_bwd_loop)
-	cmp	$-0x40, %rdx
-	lea	0x80(%rdx), %rdx
-	jl	L(shl_0_mem_bwd_less_64bytes)
-
-	movdqa	-0x10(%rsi), %xmm0
-	sub	$0x40, %rdx
-	movdqa	-0x20(%rsi), %xmm1
-
-	movdqa	%xmm0, -0x10(%rdi)
-	movdqa	%xmm1, -0x20(%rdi)
-
-	movdqa	-0x30(%rsi), %xmm0
-	movdqa	-0x40(%rsi), %xmm1
-	sub	$0x40, %rsi
-
-	movdqa	%xmm0, -0x30(%rdi)
-	movdqa	%xmm1, -0x40(%rdi)
-	sub	$0x40, %rdi
-L(shl_0_mem_bwd_less_64bytes):
-	cmp	$0x20, %rdx
-	jb	L(shl_0_mem_bwd_less_32bytes)
-	movdqa	-0x10(%rsi), %xmm0
-	sub	$0x20, %rdx
-	movdqa	-0x20(%rsi), %xmm1
-	sub	$0x20, %rsi
-	movdqa	%xmm0, -0x10(%rdi)
-	movdqa	%xmm1, -0x20(%rdi)
-	sub	$0x20, %rdi
-L(shl_0_mem_bwd_less_32bytes):
-	BRANCH_TO_JMPTBL_ENTRY (L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_1):
-	lea	(L(shl_1_loop_L1)-L(shl_1))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x01(%rsi), %xmm1
-	jb	L(L1_fwd)
-	lea	(L(shl_1_loop_L2)-L(shl_1_loop_L1))(%r9), %r9
-L(L1_fwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_1_loop_L2):
-	prefetchnta 0x1c0(%rsi)
-L(shl_1_loop_L1):
-	sub	$64, %rdx
-	movaps	0x0f(%rsi), %xmm2
-	movaps	0x1f(%rsi), %xmm3
-	movaps	0x2f(%rsi), %xmm4
-	movaps	0x3f(%rsi), %xmm5
-	movdqa	%xmm5, %xmm6
-	palignr	$1, %xmm4, %xmm5
-	lea	64(%rsi), %rsi
-	palignr	$1, %xmm3, %xmm4
-	palignr	$1, %xmm2, %xmm3
-	lea	64(%rdi), %rdi
-	palignr	$1, %xmm1, %xmm2
-	movdqa	%xmm6, %xmm1
-	movdqa	%xmm2, -0x40(%rdi)
-	movaps	%xmm3, -0x30(%rdi)
-	jb	L(shl_1_end)
-	movaps	%xmm4, -0x20(%rdi)
-	movaps	%xmm5, -0x10(%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_1_end):
-	movaps	%xmm4, -0x20(%rdi)
-	lea	64(%rdx), %rdx
-	movaps	%xmm5, -0x10(%rdi)
-	add	%rdx, %rdi
-	movdqu	%xmm0, (%r8)
-	add	%rdx, %rsi
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_1_bwd):
-	lea	(L(shl_1_bwd_loop_L1)-L(shl_1_bwd))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x01(%rsi), %xmm1
-	jb	L(L1_bwd)
-	lea	(L(shl_1_bwd_loop_L2)-L(shl_1_bwd_loop_L1))(%r9), %r9
-L(L1_bwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_1_bwd_loop_L2):
-	prefetchnta -0x1c0(%rsi)
-L(shl_1_bwd_loop_L1):
-	movaps	-0x11(%rsi), %xmm2
-	sub	$0x40, %rdx
-	movaps	-0x21(%rsi), %xmm3
-	movaps	-0x31(%rsi), %xmm4
-	movaps	-0x41(%rsi), %xmm5
-	lea	-0x40(%rsi), %rsi
-	palignr	$1, %xmm2, %xmm1
-	palignr	$1, %xmm3, %xmm2
-	palignr	$1, %xmm4, %xmm3
-	palignr	$1, %xmm5, %xmm4
-
-	movaps	%xmm1, -0x10(%rdi)
-	movaps	%xmm5, %xmm1
-
-	movaps	%xmm2, -0x20(%rdi)
-	lea	-0x40(%rdi), %rdi
-
-	movaps	%xmm3, 0x10(%rdi)
-	jb	L(shl_1_bwd_end)
-	movaps	%xmm4, (%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_1_bwd_end):
-	movaps	%xmm4, (%rdi)
-	lea	64(%rdx), %rdx
-	movdqu	%xmm0, (%r8)
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_2):
-	lea	(L(shl_2_loop_L1)-L(shl_2))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x02(%rsi), %xmm1
-	jb	L(L2_fwd)
-	lea	(L(shl_2_loop_L2)-L(shl_2_loop_L1))(%r9), %r9
-L(L2_fwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_2_loop_L2):
-	prefetchnta 0x1c0(%rsi)
-L(shl_2_loop_L1):
-	sub	$64, %rdx
-	movaps	0x0e(%rsi), %xmm2
-	movaps	0x1e(%rsi), %xmm3
-	movaps	0x2e(%rsi), %xmm4
-	movaps	0x3e(%rsi), %xmm5
-	movdqa	%xmm5, %xmm6
-	palignr	$2, %xmm4, %xmm5
-	lea	64(%rsi), %rsi
-	palignr	$2, %xmm3, %xmm4
-	palignr	$2, %xmm2, %xmm3
-	lea	64(%rdi), %rdi
-	palignr	$2, %xmm1, %xmm2
-	movdqa	%xmm6, %xmm1
-	movdqa	%xmm2, -0x40(%rdi)
-	movaps	%xmm3, -0x30(%rdi)
-	jb	L(shl_2_end)
-	movaps	%xmm4, -0x20(%rdi)
-	movaps	%xmm5, -0x10(%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_2_end):
-	movaps	%xmm4, -0x20(%rdi)
-	lea	64(%rdx), %rdx
-	movaps	%xmm5, -0x10(%rdi)
-	add	%rdx, %rdi
-	movdqu	%xmm0, (%r8)
-	add	%rdx, %rsi
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_2_bwd):
-	lea	(L(shl_2_bwd_loop_L1)-L(shl_2_bwd))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x02(%rsi), %xmm1
-	jb	L(L2_bwd)
-	lea	(L(shl_2_bwd_loop_L2)-L(shl_2_bwd_loop_L1))(%r9), %r9
-L(L2_bwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_2_bwd_loop_L2):
-	prefetchnta -0x1c0(%rsi)
-L(shl_2_bwd_loop_L1):
-	movaps	-0x12(%rsi), %xmm2
-	sub	$0x40, %rdx
-	movaps	-0x22(%rsi), %xmm3
-	movaps	-0x32(%rsi), %xmm4
-	movaps	-0x42(%rsi), %xmm5
-	lea	-0x40(%rsi), %rsi
-	palignr	$2, %xmm2, %xmm1
-	palignr	$2, %xmm3, %xmm2
-	palignr	$2, %xmm4, %xmm3
-	palignr	$2, %xmm5, %xmm4
-
-	movaps	%xmm1, -0x10(%rdi)
-	movaps	%xmm5, %xmm1
-
-	movaps	%xmm2, -0x20(%rdi)
-	lea	-0x40(%rdi), %rdi
-
-	movaps	%xmm3, 0x10(%rdi)
-	jb	L(shl_2_bwd_end)
-	movaps	%xmm4, (%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_2_bwd_end):
-	movaps	%xmm4, (%rdi)
-	lea	64(%rdx), %rdx
-	movdqu	%xmm0, (%r8)
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_3):
-	lea	(L(shl_3_loop_L1)-L(shl_3))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x03(%rsi), %xmm1
-	jb	L(L3_fwd)
-	lea	(L(shl_3_loop_L2)-L(shl_3_loop_L1))(%r9), %r9
-L(L3_fwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_3_loop_L2):
-	prefetchnta 0x1c0(%rsi)
-L(shl_3_loop_L1):
-	sub	$64, %rdx
-	movaps	0x0d(%rsi), %xmm2
-	movaps	0x1d(%rsi), %xmm3
-	movaps	0x2d(%rsi), %xmm4
-	movaps	0x3d(%rsi), %xmm5
-	movdqa	%xmm5, %xmm6
-	palignr	$3, %xmm4, %xmm5
-	lea	64(%rsi), %rsi
-	palignr	$3, %xmm3, %xmm4
-	palignr	$3, %xmm2, %xmm3
-	lea	64(%rdi), %rdi
-	palignr	$3, %xmm1, %xmm2
-	movdqa	%xmm6, %xmm1
-	movdqa	%xmm2, -0x40(%rdi)
-	movaps	%xmm3, -0x30(%rdi)
-	jb	L(shl_3_end)
-	movaps	%xmm4, -0x20(%rdi)
-	movaps	%xmm5, -0x10(%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_3_end):
-	movaps	%xmm4, -0x20(%rdi)
-	lea	64(%rdx), %rdx
-	movaps	%xmm5, -0x10(%rdi)
-	add	%rdx, %rdi
-	movdqu	%xmm0, (%r8)
-	add	%rdx, %rsi
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_3_bwd):
-	lea	(L(shl_3_bwd_loop_L1)-L(shl_3_bwd))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x03(%rsi), %xmm1
-	jb	L(L3_bwd)
-	lea	(L(shl_3_bwd_loop_L2)-L(shl_3_bwd_loop_L1))(%r9), %r9
-L(L3_bwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_3_bwd_loop_L2):
-	prefetchnta -0x1c0(%rsi)
-L(shl_3_bwd_loop_L1):
-	movaps	-0x13(%rsi), %xmm2
-	sub	$0x40, %rdx
-	movaps	-0x23(%rsi), %xmm3
-	movaps	-0x33(%rsi), %xmm4
-	movaps	-0x43(%rsi), %xmm5
-	lea	-0x40(%rsi), %rsi
-	palignr	$3, %xmm2, %xmm1
-	palignr	$3, %xmm3, %xmm2
-	palignr	$3, %xmm4, %xmm3
-	palignr	$3, %xmm5, %xmm4
-
-	movaps	%xmm1, -0x10(%rdi)
-	movaps	%xmm5, %xmm1
-
-	movaps	%xmm2, -0x20(%rdi)
-	lea	-0x40(%rdi), %rdi
-
-	movaps	%xmm3, 0x10(%rdi)
-	jb	L(shl_3_bwd_end)
-	movaps	%xmm4, (%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_3_bwd_end):
-	movaps	%xmm4, (%rdi)
-	lea	64(%rdx), %rdx
-	movdqu	%xmm0, (%r8)
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_4):
-	lea	(L(shl_4_loop_L1)-L(shl_4))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x04(%rsi), %xmm1
-	jb	L(L4_fwd)
-	lea	(L(shl_4_loop_L2)-L(shl_4_loop_L1))(%r9), %r9
-L(L4_fwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_4_loop_L2):
-	prefetchnta 0x1c0(%rsi)
-L(shl_4_loop_L1):
-	sub	$64, %rdx
-	movaps	0x0c(%rsi), %xmm2
-	movaps	0x1c(%rsi), %xmm3
-	movaps	0x2c(%rsi), %xmm4
-	movaps	0x3c(%rsi), %xmm5
-	movdqa	%xmm5, %xmm6
-	palignr	$4, %xmm4, %xmm5
-	lea	64(%rsi), %rsi
-	palignr	$4, %xmm3, %xmm4
-	palignr	$4, %xmm2, %xmm3
-	lea	64(%rdi), %rdi
-	palignr	$4, %xmm1, %xmm2
-	movdqa	%xmm6, %xmm1
-	movdqa	%xmm2, -0x40(%rdi)
-	movaps	%xmm3, -0x30(%rdi)
-	jb	L(shl_4_end)
-	movaps	%xmm4, -0x20(%rdi)
-	movaps	%xmm5, -0x10(%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_4_end):
-	movaps	%xmm4, -0x20(%rdi)
-	lea	64(%rdx), %rdx
-	movaps	%xmm5, -0x10(%rdi)
-	add	%rdx, %rdi
-	movdqu	%xmm0, (%r8)
-	add	%rdx, %rsi
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_4_bwd):
-	lea	(L(shl_4_bwd_loop_L1)-L(shl_4_bwd))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x04(%rsi), %xmm1
-	jb	L(L4_bwd)
-	lea	(L(shl_4_bwd_loop_L2)-L(shl_4_bwd_loop_L1))(%r9), %r9
-L(L4_bwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_4_bwd_loop_L2):
-	prefetchnta -0x1c0(%rsi)
-L(shl_4_bwd_loop_L1):
-	movaps	-0x14(%rsi), %xmm2
-	sub	$0x40, %rdx
-	movaps	-0x24(%rsi), %xmm3
-	movaps	-0x34(%rsi), %xmm4
-	movaps	-0x44(%rsi), %xmm5
-	lea	-0x40(%rsi), %rsi
-	palignr	$4, %xmm2, %xmm1
-	palignr	$4, %xmm3, %xmm2
-	palignr	$4, %xmm4, %xmm3
-	palignr	$4, %xmm5, %xmm4
-
-	movaps	%xmm1, -0x10(%rdi)
-	movaps	%xmm5, %xmm1
-
-	movaps	%xmm2, -0x20(%rdi)
-	lea	-0x40(%rdi), %rdi
-
-	movaps	%xmm3, 0x10(%rdi)
-	jb	L(shl_4_bwd_end)
-	movaps	%xmm4, (%rdi)
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_4_bwd_end):
-	movaps	%xmm4, (%rdi)
-	lea	64(%rdx), %rdx
-	movdqu	%xmm0, (%r8)
-	BRANCH_TO_JMPTBL_ENTRY(L(table_less_80bytes), %rdx, 4)
-
-	.p2align 4
-L(shl_5):
-	lea	(L(shl_5_loop_L1)-L(shl_5))(%r9), %r9
-	cmp	%rcx, %rdx
-	movaps	-0x05(%rsi), %xmm1
-	jb	L(L5_fwd)
-	lea	(L(shl_5_loop_L2)-L(shl_5_loop_L1))(%r9), %r9
-L(L5_fwd):
-	lea	-64(%rdx), %rdx
-	_CET_NOTRACK jmp *%r9
-	ud2
-L(shl_5_loop_L2):
-	prefetchnta 0x1c0(%rsi)
-L(shl_5_loop_L1):
-	sub	$64, %rdx
-	movaps	0x0b(%rsi), %xmm2
-	movaps	0x1b(%rsi), %xmm3
-	movaps	0x2b(%rsi), %xmm4
-	movaps	0x3b(%rsi), %xmm5
-	movdqa	%xmm5, %xmm6
-	palignr	$5, %xmm4, %xmm5
-	lea	64(%rsi), %rsi
-	palignr	$5, %xmm3, %xmm4
-	palignr	$5, %xmm2, %xmm3
-	lea	64(%rdi), %rdi
-	palignr	$5, %xmm1, %xmm2
-	movdqa	%xmm6, %xmm1
-	movdqa	%xmm2, -0x40(%rdi)
-	movaps	%xmm3, -0x30(%rdi)
-	jb	L(shl_5_end)