openkylin
/
platform_system_core
mirror of https://gitee.com/openkylin/platform_system_core.git
9
0
Fork 0
Code Issues Projects Releases Wiki Activity

Remove unused android_memset16()/android_memset32(). 

Pixelflinger is gone.

Test: treehugger
Change-Id: I6954ae6a860102be0f6d76f33e76ed3210e5c152
This commit is contained in:
Elliott Hughes 2019-11-08 15:59:30 -08:00
parent 6188334d62
commit 7b8fcfea0e
10 changed files with 1 additions and 2793 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
libcutils/Android.bp
View File

@ -114,36 +114,17 @@ cc_library {
},
android_arm: {
srcs: ["arch-arm/memset32.S"],
sanitize: {
misc_undefined: ["integer"],
},
},
android_arm64: {
srcs: ["arch-arm64/android_memset.S"],
sanitize: {
misc_undefined: ["integer"],
},
},
android_mips: {
srcs: ["arch-mips/android_memset.c"],
sanitize: {
misc_undefined: ["integer"],
},
},
android_mips64: {
srcs: ["arch-mips/android_memset.c"],
sanitize: {
misc_undefined: ["integer"],
},
},
android_x86: {
srcs: [
"arch-x86/android_memset16.S",
"arch-x86/android_memset32.S",
],
// TODO: This is to work around b/29412086.
// Remove once __mulodi4 is available and move the "sanitize" block
// to the android target.
@ -153,10 +134,6 @@ cc_library {
},
android_x86_64: {
srcs: [
"arch-x86_64/android_memset16.S",
"arch-x86_64/android_memset32.S",
],
sanitize: {
misc_undefined: ["integer"],
},
@ -206,7 +183,6 @@ cc_defaults {
"android_get_control_socket_test.cpp",
"ashmem_test.cpp",
"fs_config_test.cpp",
"memset_test.cpp",
"multiuser_test.cpp",
"properties_test.cpp",
"sched_policy_test.cpp",

100
libcutils/arch-arm/memset32.S
View File

@ -1,100 +0,0 @@
/*
* Copyright (C) 2006 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* memset32.S
*
*/
.syntax unified
.text
.align
.global android_memset32
.type android_memset32, %function
.global android_memset16
.type android_memset16, %function
/*
* Optimized memset32 and memset16 for ARM.
*
* void android_memset16(uint16_t* dst, uint16_t value, size_t size);
* void android_memset32(uint32_t* dst, uint32_t value, size_t size);
*
*/
android_memset16:
.fnstart
cmp r2, #1
bxle lr
/* expand the data to 32 bits */
mov r1, r1, lsl #16
orr r1, r1, r1, lsr #16
/* align to 32 bits */
tst r0, #2
strhne r1, [r0], #2
subne r2, r2, #2
.fnend
android_memset32:
.fnstart
.cfi_startproc
str lr, [sp, #-4]!
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
/* align the destination to a cache-line */
mov r12, r1
mov lr, r1
rsb r3, r0, #0
ands r3, r3, #0x1C
beq .Laligned32
cmp r3, r2
andhi r3, r2, #0x1C
sub r2, r2, r3
/* conditionally writes 0 to 7 words (length in r3) */
movs r3, r3, lsl #28
stmiacs r0!, {r1, lr}
stmiacs r0!, {r1, lr}
stmiami r0!, {r1, lr}
movs r3, r3, lsl #2
strcs r1, [r0], #4
.Laligned32:
mov r3, r1
1: subs r2, r2, #32
stmiahs r0!, {r1,r3,r12,lr}
stmiahs r0!, {r1,r3,r12,lr}
bhs 1b
add r2, r2, #32
/* conditionally stores 0 to 30 bytes */
movs r2, r2, lsl #28
stmiacs r0!, {r1,r3,r12,lr}
stmiami r0!, {r1,lr}
movs r2, r2, lsl #2
strcs r1, [r0], #4
strhmi lr, [r0], #2
ldr lr, [sp], #4
.cfi_def_cfa_offset 0
.cfi_restore lr
bx lr
.cfi_endproc
.fnend

211
libcutils/arch-arm64/android_memset.S
View File

@ -1,211 +0,0 @@
/* Copyright (c) 2012, Linaro Limited
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the Linaro nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Assumptions:
*
* ARMv8-a, AArch64
* Unaligned accesses
*
*/
/* By default we assume that the DC instruction can be used to zero
data blocks more efficiently. In some circumstances this might be
unsafe, for example in an asymmetric multiprocessor environment with
different DC clear lengths (neither the upper nor lower lengths are
safe to use). */
#define dst x0
#define count x2
#define tmp1 x3
#define tmp1w w3
#define tmp2 x4
#define tmp2w w4
#define zva_len_x x5
#define zva_len w5
#define zva_bits_x x6
#define A_l x1
#define A_lw w1
#define tmp3w w9
#define ENTRY(f) \
.text; \
.globl f; \
.align 0; \
.type f, %function; \
f: \
.cfi_startproc \
#define END(f) \
.cfi_endproc; \
.size f, .-f; \
ENTRY(android_memset16)
ands A_lw, A_lw, #0xffff
b.eq .Lzero_mem
orr A_lw, A_lw, A_lw, lsl #16
b .Lexpand_to_64
END(android_memset16)
ENTRY(android_memset32)
cmp A_lw, #0
b.eq .Lzero_mem
.Lexpand_to_64:
orr A_l, A_l, A_l, lsl #32
.Ltail_maybe_long:
cmp count, #64
b.ge .Lnot_short
.Ltail_maybe_tiny:
cmp count, #15
b.le .Ltail15tiny
.Ltail63:
ands tmp1, count, #0x30
b.eq .Ltail15
add dst, dst, tmp1
cmp tmp1w, #0x20
b.eq 1f
b.lt 2f
stp A_l, A_l, [dst, #-48]
1:
stp A_l, A_l, [dst, #-32]
2:
stp A_l, A_l, [dst, #-16]
.Ltail15:
and count, count, #15
add dst, dst, count
stp A_l, A_l, [dst, #-16] /* Repeat some/all of last store. */
ret
.Ltail15tiny:
/* Set up to 15 bytes. Does not assume earlier memory
being set. */
tbz count, #3, 1f
str A_l, [dst], #8
1:
tbz count, #2, 1f
str A_lw, [dst], #4
1:
tbz count, #1, 1f
strh A_lw, [dst], #2
1:
ret
/* Critical loop. Start at a new cache line boundary. Assuming
* 64 bytes per line, this ensures the entire loop is in one line. */
.p2align 6
.Lnot_short:
neg tmp2, dst
ands tmp2, tmp2, #15
b.eq 2f
/* Bring DST to 128-bit (16-byte) alignment. We know that there's
* more than that to set, so we simply store 16 bytes and advance by
* the amount required to reach alignment. */
sub count, count, tmp2
stp A_l, A_l, [dst]
add dst, dst, tmp2
/* There may be less than 63 bytes to go now. */
cmp count, #63
b.le .Ltail63
2:
sub dst, dst, #16 /* Pre-bias. */
sub count, count, #64
1:
stp A_l, A_l, [dst, #16]
stp A_l, A_l, [dst, #32]
stp A_l, A_l, [dst, #48]
stp A_l, A_l, [dst, #64]!
subs count, count, #64
b.ge 1b
tst count, #0x3f
add dst, dst, #16
b.ne .Ltail63
ret
/* For zeroing memory, check to see if we can use the ZVA feature to
* zero entire 'cache' lines. */
.Lzero_mem:
mov A_l, #0
cmp count, #63
b.le .Ltail_maybe_tiny
neg tmp2, dst
ands tmp2, tmp2, #15
b.eq 1f
sub count, count, tmp2
stp A_l, A_l, [dst]
add dst, dst, tmp2
cmp count, #63
b.le .Ltail63
1:
/* For zeroing small amounts of memory, it's not worth setting up
* the line-clear code. */
cmp count, #128
b.lt .Lnot_short
mrs tmp1, dczid_el0
tbnz tmp1, #4, .Lnot_short
mov tmp3w, #4
and zva_len, tmp1w, #15 /* Safety: other bits reserved. */
lsl zva_len, tmp3w, zva_len
.Lzero_by_line:
/* Compute how far we need to go to become suitably aligned. We're
* already at quad-word alignment. */
cmp count, zva_len_x
b.lt .Lnot_short /* Not enough to reach alignment. */
sub zva_bits_x, zva_len_x, #1
neg tmp2, dst
ands tmp2, tmp2, zva_bits_x
b.eq 1f /* Already aligned. */
/* Not aligned, check that there's enough to copy after alignment. */
sub tmp1, count, tmp2
cmp tmp1, #64
ccmp tmp1, zva_len_x, #8, ge /* NZCV=0b1000 */
b.lt .Lnot_short
/* We know that there's at least 64 bytes to zero and that it's safe
* to overrun by 64 bytes. */
mov count, tmp1
2:
stp A_l, A_l, [dst]
stp A_l, A_l, [dst, #16]
stp A_l, A_l, [dst, #32]
subs tmp2, tmp2, #64
stp A_l, A_l, [dst, #48]
add dst, dst, #64
b.ge 2b
/* We've overrun a bit, so adjust dst downwards. */
add dst, dst, tmp2
1:
sub count, count, zva_len_x
3:
dc zva, dst
add dst, dst, zva_len_x
subs count, count, zva_len_x
b.ge 3b
ands count, count, zva_bits_x
b.ne .Ltail_maybe_long
ret
END(android_memset32)

100
libcutils/arch-mips/android_memset.c
View File

@ -1,100 +0,0 @@
/*
* Copyright (C) 2015 The Android Open Source Project
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
* OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/* generic C version for any machine */
#include <cutils/memory.h>
#ifdef __clang__
__attribute__((no_sanitize("integer")))
#endif
void android_memset16(uint16_t* dst, uint16_t value, size_t size)
{
/* optimized version of
size >>= 1;
while (size--)
*dst++ = value;
*/
size >>= 1;
if (((uintptr_t)dst & 2) && size) {
/* fill unpaired first elem separately */
*dst++ = value;
size--;
}
/* dst is now 32-bit-aligned */
/* fill body with 32-bit pairs */
uint32_t value32 = (((uint32_t)value) << 16) | ((uint32_t)value);
android_memset32((uint32_t*) dst, value32, size<<1);
if (size & 1) {
dst[size-1] = value; /* fill unpaired last elem */
}
}
#ifdef __clang__
__attribute__((no_sanitize("integer")))
#endif
void android_memset32(uint32_t* dst, uint32_t value, size_t size)
{
/* optimized version of
size >>= 2;
while (size--)
*dst++ = value;
*/
size >>= 2;
if (((uintptr_t)dst & 4) && size) {
/* fill unpaired first 32-bit elem separately */
*dst++ = value;
size--;
}
/* dst is now 64-bit aligned */
/* fill body with 64-bit pairs */
uint64_t value64 = (((uint64_t)value) << 32) | ((uint64_t)value);
uint64_t* dst64 = (uint64_t*)dst;
while (size >= 12) {
dst64[0] = value64;
dst64[1] = value64;
dst64[2] = value64;
dst64[3] = value64;
dst64[4] = value64;
dst64[5] = value64;
size -= 12;
dst64 += 6;
}
/* fill remainder with original 32-bit single-elem loop */
dst = (uint32_t*) dst64;
while (size != 0) {
size--;
*dst++ = value;
}
}

719
libcutils/arch-x86/android_memset16.S
View File

@ -1,719 +0,0 @@
/*
* Copyright (C) 2010 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "cache.h"
#ifndef MEMSET
# define MEMSET android_memset16
#endif
#ifndef L
# define L(label) .L##label
#endif
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
#ifndef cfi_startproc
# define cfi_startproc .cfi_startproc
#endif
#ifndef cfi_endproc
# define cfi_endproc .cfi_endproc
#endif
#ifndef cfi_rel_offset
# define cfi_rel_offset(reg, off) .cfi_rel_offset reg, off
#endif
#ifndef cfi_restore
# define cfi_restore(reg) .cfi_restore reg
#endif
#ifndef cfi_adjust_cfa_offset
# define cfi_adjust_cfa_offset(off) .cfi_adjust_cfa_offset off
#endif
#ifndef ENTRY
# define ENTRY(name) \
.type name, @function; \
.globl name; \
.p2align 4; \
name: \
cfi_startproc
#endif
#ifndef END
# define END(name) \
cfi_endproc; \
.size name, .-name
#endif
#define CFI_PUSH(REG) \
cfi_adjust_cfa_offset (4); \
cfi_rel_offset (REG, 0)
#define CFI_POP(REG) \
cfi_adjust_cfa_offset (-4); \
cfi_restore (REG)
#define PUSH(REG) pushl REG; CFI_PUSH (REG)
#define POP(REG) popl REG; CFI_POP (REG)
#ifdef USE_AS_BZERO16
# define DEST PARMS
# define LEN DEST+4
# define SETRTNVAL
#else
# define DEST PARMS
# define CHR DEST+4
# define LEN CHR+4
# define SETRTNVAL movl DEST(%esp), %eax
#endif
#if (defined SHARED || defined __PIC__)
# define ENTRANCE PUSH (%ebx);
# define RETURN_END POP (%ebx); ret
# define RETURN RETURN_END; CFI_PUSH (%ebx)
# define PARMS 8 /* Preserve EBX. */
# define JMPTBL(I, B) I - B
/* Load an entry in a jump table into EBX and branch to it. TABLE is a
jump table with relative offsets. */
# define BRANCH_TO_JMPTBL_ENTRY(TABLE) \
/* We first load PC into EBX. */ \
call __x86.get_pc_thunk.bx; \
/* Get the address of the jump table. */ \
add $(TABLE - .), %ebx; \
/* Get the entry and convert the relative offset to the \
absolute address. */ \
add (%ebx,%ecx,4), %ebx; \
/* We loaded the jump table and adjuested EDX. Go. */ \
jmp *%ebx
.section .gnu.linkonce.t.__x86.get_pc_thunk.bx,"ax",@progbits
.globl __x86.get_pc_thunk.bx
.hidden __x86.get_pc_thunk.bx
ALIGN (4)
.type __x86.get_pc_thunk.bx,@function
__x86.get_pc_thunk.bx:
movl (%esp), %ebx
ret
#else
# define ENTRANCE
# define RETURN_END ret
# define RETURN RETURN_END
# define PARMS 4
# define JMPTBL(I, B) I
/* Branch to an entry in a jump table. TABLE is a jump table with
absolute offsets. */
# define BRANCH_TO_JMPTBL_ENTRY(TABLE) \
jmp *TABLE(,%ecx,4)
#endif
.section .text.sse2,"ax",@progbits
ALIGN (4)
ENTRY (MEMSET)
ENTRANCE
movl LEN(%esp), %ecx
shr $1, %ecx
#ifdef USE_AS_BZERO16
xor %eax, %eax
#else
movzwl CHR(%esp), %eax
mov %eax, %edx
shl $16, %eax
or %edx, %eax
#endif
movl DEST(%esp), %edx
cmp $32, %ecx
jae L(32wordsormore)
L(write_less32words):
lea (%edx, %ecx, 2), %edx
BRANCH_TO_JMPTBL_ENTRY (L(table_less32words))
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_less32words):
.int JMPTBL (L(write_0words), L(table_less32words))
.int JMPTBL (L(write_1words), L(table_less32words))
.int JMPTBL (L(write_2words), L(table_less32words))
.int JMPTBL (L(write_3words), L(table_less32words))
.int JMPTBL (L(write_4words), L(table_less32words))
.int JMPTBL (L(write_5words), L(table_less32words))
.int JMPTBL (L(write_6words), L(table_less32words))
.int JMPTBL (L(write_7words), L(table_less32words))
.int JMPTBL (L(write_8words), L(table_less32words))
.int JMPTBL (L(write_9words), L(table_less32words))
.int JMPTBL (L(write_10words), L(table_less32words))
.int JMPTBL (L(write_11words), L(table_less32words))
.int JMPTBL (L(write_12words), L(table_less32words))
.int JMPTBL (L(write_13words), L(table_less32words))
.int JMPTBL (L(write_14words), L(table_less32words))
.int JMPTBL (L(write_15words), L(table_less32words))
.int JMPTBL (L(write_16words), L(table_less32words))
.int JMPTBL (L(write_17words), L(table_less32words))
.int JMPTBL (L(write_18words), L(table_less32words))
.int JMPTBL (L(write_19words), L(table_less32words))
.int JMPTBL (L(write_20words), L(table_less32words))
.int JMPTBL (L(write_21words), L(table_less32words))
.int JMPTBL (L(write_22words), L(table_less32words))
.int JMPTBL (L(write_23words), L(table_less32words))
.int JMPTBL (L(write_24words), L(table_less32words))
.int JMPTBL (L(write_25words), L(table_less32words))
.int JMPTBL (L(write_26words), L(table_less32words))
.int JMPTBL (L(write_27words), L(table_less32words))
.int JMPTBL (L(write_28words), L(table_less32words))
.int JMPTBL (L(write_29words), L(table_less32words))
.int JMPTBL (L(write_30words), L(table_less32words))
.int JMPTBL (L(write_31words), L(table_less32words))
.popsection
ALIGN (4)
L(write_28words):
movl %eax, -56(%edx)
movl %eax, -52(%edx)
L(write_24words):
movl %eax, -48(%edx)
movl %eax, -44(%edx)
L(write_20words):
movl %eax, -40(%edx)
movl %eax, -36(%edx)
L(write_16words):
movl %eax, -32(%edx)
movl %eax, -28(%edx)
L(write_12words):
movl %eax, -24(%edx)
movl %eax, -20(%edx)
L(write_8words):
movl %eax, -16(%edx)
movl %eax, -12(%edx)
L(write_4words):
movl %eax, -8(%edx)
movl %eax, -4(%edx)
L(write_0words):
SETRTNVAL
RETURN
ALIGN (4)
L(write_29words):
movl %eax, -58(%edx)
movl %eax, -54(%edx)
L(write_25words):
movl %eax, -50(%edx)
movl %eax, -46(%edx)
L(write_21words):
movl %eax, -42(%edx)
movl %eax, -38(%edx)
L(write_17words):
movl %eax, -34(%edx)
movl %eax, -30(%edx)
L(write_13words):
movl %eax, -26(%edx)
movl %eax, -22(%edx)
L(write_9words):
movl %eax, -18(%edx)
movl %eax, -14(%edx)
L(write_5words):
movl %eax, -10(%edx)
movl %eax, -6(%edx)
L(write_1words):
mov %ax, -2(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(write_30words):
movl %eax, -60(%edx)
movl %eax, -56(%edx)
L(write_26words):
movl %eax, -52(%edx)
movl %eax, -48(%edx)
L(write_22words):
movl %eax, -44(%edx)
movl %eax, -40(%edx)
L(write_18words):
movl %eax, -36(%edx)
movl %eax, -32(%edx)
L(write_14words):
movl %eax, -28(%edx)
movl %eax, -24(%edx)
L(write_10words):
movl %eax, -20(%edx)
movl %eax, -16(%edx)
L(write_6words):
movl %eax, -12(%edx)
movl %eax, -8(%edx)
L(write_2words):
movl %eax, -4(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(write_31words):
movl %eax, -62(%edx)
movl %eax, -58(%edx)
L(write_27words):
movl %eax, -54(%edx)
movl %eax, -50(%edx)
L(write_23words):
movl %eax, -46(%edx)
movl %eax, -42(%edx)
L(write_19words):
movl %eax, -38(%edx)
movl %eax, -34(%edx)
L(write_15words):
movl %eax, -30(%edx)
movl %eax, -26(%edx)
L(write_11words):
movl %eax, -22(%edx)
movl %eax, -18(%edx)
L(write_7words):
movl %eax, -14(%edx)
movl %eax, -10(%edx)
L(write_3words):
movl %eax, -6(%edx)
movw %ax, -2(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(32wordsormore):
shl $1, %ecx
test $0x01, %edx
jz L(aligned2bytes)
mov %eax, (%edx)
mov %eax, -4(%edx, %ecx)
sub $2, %ecx
add $1, %edx
rol $8, %eax
L(aligned2bytes):
#ifdef USE_AS_BZERO16
pxor %xmm0, %xmm0
#else
movd %eax, %xmm0
pshufd $0, %xmm0, %xmm0
#endif
testl $0xf, %edx
jz L(aligned_16)
/* ECX > 32 and EDX is not 16 byte aligned. */
L(not_aligned_16):
movdqu %xmm0, (%edx)
movl %edx, %eax
and $-16, %edx
add $16, %edx
sub %edx, %eax
add %eax, %ecx
movd %xmm0, %eax
ALIGN (4)
L(aligned_16):
cmp $128, %ecx
jae L(128bytesormore)
L(aligned_16_less128bytes):
add %ecx, %edx
shr $1, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
ALIGN (4)
L(128bytesormore):
#ifdef SHARED_CACHE_SIZE
PUSH (%ebx)
mov $SHARED_CACHE_SIZE, %ebx
#else
# if (defined SHARED || defined __PIC__)
call __x86.get_pc_thunk.bx
add $_GLOBAL_OFFSET_TABLE_, %ebx
mov __x86_shared_cache_size@GOTOFF(%ebx), %ebx
# else
PUSH (%ebx)
mov __x86_shared_cache_size, %ebx
# endif
#endif
cmp %ebx, %ecx
jae L(128bytesormore_nt_start)
#ifdef DATA_CACHE_SIZE
POP (%ebx)
# define RESTORE_EBX_STATE CFI_PUSH (%ebx)
cmp $DATA_CACHE_SIZE, %ecx
#else
# if (defined SHARED || defined __PIC__)
# define RESTORE_EBX_STATE
call __x86.get_pc_thunk.bx
add $_GLOBAL_OFFSET_TABLE_, %ebx
cmp __x86_data_cache_size@GOTOFF(%ebx), %ecx
# else
POP (%ebx)
# define RESTORE_EBX_STATE CFI_PUSH (%ebx)
cmp __x86_data_cache_size, %ecx
# endif
#endif
jae L(128bytes_L2_normal)
subl $128, %ecx
L(128bytesormore_normal):
sub $128, %ecx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
lea 128(%edx), %edx
jb L(128bytesless_normal)
sub $128, %ecx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
lea 128(%edx), %edx
jae L(128bytesormore_normal)
L(128bytesless_normal):
lea 128(%ecx), %ecx
add %ecx, %edx
shr $1, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
ALIGN (4)
L(128bytes_L2_normal):
prefetcht0 0x380(%edx)
prefetcht0 0x3c0(%edx)
sub $128, %ecx
movdqa %xmm0, (%edx)
movaps %xmm0, 0x10(%edx)
movaps %xmm0, 0x20(%edx)
movaps %xmm0, 0x30(%edx)
movaps %xmm0, 0x40(%edx)
movaps %xmm0, 0x50(%edx)
movaps %xmm0, 0x60(%edx)
movaps %xmm0, 0x70(%edx)
add $128, %edx
cmp $128, %ecx
jae L(128bytes_L2_normal)
L(128bytesless_L2_normal):
add %ecx, %edx
shr $1, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
RESTORE_EBX_STATE
L(128bytesormore_nt_start):
sub %ebx, %ecx
mov %ebx, %eax
and $0x7f, %eax
add %eax, %ecx
movd %xmm0, %eax
ALIGN (4)
L(128bytesormore_shared_cache_loop):
prefetcht0 0x3c0(%edx)
prefetcht0 0x380(%edx)
sub $0x80, %ebx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
add $0x80, %edx
cmp $0x80, %ebx
jae L(128bytesormore_shared_cache_loop)
cmp $0x80, %ecx
jb L(shared_cache_loop_end)
ALIGN (4)
L(128bytesormore_nt):
sub $0x80, %ecx
movntdq %xmm0, (%edx)
movntdq %xmm0, 0x10(%edx)
movntdq %xmm0, 0x20(%edx)
movntdq %xmm0, 0x30(%edx)
movntdq %xmm0, 0x40(%edx)
movntdq %xmm0, 0x50(%edx)
movntdq %xmm0, 0x60(%edx)
movntdq %xmm0, 0x70(%edx)
add $0x80, %edx
cmp $0x80, %ecx
jae L(128bytesormore_nt)
sfence
L(shared_cache_loop_end):
#if defined DATA_CACHE_SIZE || !(defined SHARED || defined __PIC__)
POP (%ebx)
#endif
add %ecx, %edx
shr $1, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_16_128bytes):
.int JMPTBL (L(aligned_16_0bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_2bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_4bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_6bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_8bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_10bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_12bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_14bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_16bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_18bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_20bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_22bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_24bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_26bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_28bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_30bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_32bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_34bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_36bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_38bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_40bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_42bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_44bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_46bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_48bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_50bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_52bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_54bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_56bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_58bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_60bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_62bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_64bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_66bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_68bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_70bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_72bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_74bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_76bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_78bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_80bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_82bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_84bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_86bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_88bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_90bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_92bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_94bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_96bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_98bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_100bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_102bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_104bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_106bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_108bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_110bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_112bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_114bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_116bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_118bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_120bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_122bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_124bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_126bytes), L(table_16_128bytes))
.popsection
ALIGN (4)
L(aligned_16_112bytes):
movdqa %xmm0, -112(%edx)
L(aligned_16_96bytes):
movdqa %xmm0, -96(%edx)
L(aligned_16_80bytes):
movdqa %xmm0, -80(%edx)
L(aligned_16_64bytes):
movdqa %xmm0, -64(%edx)
L(aligned_16_48bytes):
movdqa %xmm0, -48(%edx)
L(aligned_16_32bytes):
movdqa %xmm0, -32(%edx)
L(aligned_16_16bytes):
movdqa %xmm0, -16(%edx)
L(aligned_16_0bytes):
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_114bytes):
movdqa %xmm0, -114(%edx)
L(aligned_16_98bytes):
movdqa %xmm0, -98(%edx)
L(aligned_16_82bytes):
movdqa %xmm0, -82(%edx)
L(aligned_16_66bytes):
movdqa %xmm0, -66(%edx)
L(aligned_16_50bytes):
movdqa %xmm0, -50(%edx)
L(aligned_16_34bytes):
movdqa %xmm0, -34(%edx)
L(aligned_16_18bytes):
movdqa %xmm0, -18(%edx)
L(aligned_16_2bytes):
movw %ax, -2(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_116bytes):
movdqa %xmm0, -116(%edx)
L(aligned_16_100bytes):
movdqa %xmm0, -100(%edx)
L(aligned_16_84bytes):
movdqa %xmm0, -84(%edx)
L(aligned_16_68bytes):
movdqa %xmm0, -68(%edx)
L(aligned_16_52bytes):
movdqa %xmm0, -52(%edx)
L(aligned_16_36bytes):
movdqa %xmm0, -36(%edx)
L(aligned_16_20bytes):
movdqa %xmm0, -20(%edx)
L(aligned_16_4bytes):
movl %eax, -4(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_118bytes):
movdqa %xmm0, -118(%edx)
L(aligned_16_102bytes):
movdqa %xmm0, -102(%edx)
L(aligned_16_86bytes):
movdqa %xmm0, -86(%edx)
L(aligned_16_70bytes):
movdqa %xmm0, -70(%edx)
L(aligned_16_54bytes):
movdqa %xmm0, -54(%edx)
L(aligned_16_38bytes):
movdqa %xmm0, -38(%edx)
L(aligned_16_22bytes):
movdqa %xmm0, -22(%edx)
L(aligned_16_6bytes):
movl %eax, -6(%edx)
movw %ax, -2(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_120bytes):
movdqa %xmm0, -120(%edx)
L(aligned_16_104bytes):
movdqa %xmm0, -104(%edx)
L(aligned_16_88bytes):
movdqa %xmm0, -88(%edx)
L(aligned_16_72bytes):
movdqa %xmm0, -72(%edx)
L(aligned_16_56bytes):
movdqa %xmm0, -56(%edx)
L(aligned_16_40bytes):
movdqa %xmm0, -40(%edx)
L(aligned_16_24bytes):
movdqa %xmm0, -24(%edx)
L(aligned_16_8bytes):
movq %xmm0, -8(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_122bytes):
movdqa %xmm0, -122(%edx)
L(aligned_16_106bytes):
movdqa %xmm0, -106(%edx)
L(aligned_16_90bytes):
movdqa %xmm0, -90(%edx)
L(aligned_16_74bytes):
movdqa %xmm0, -74(%edx)
L(aligned_16_58bytes):
movdqa %xmm0, -58(%edx)
L(aligned_16_42bytes):
movdqa %xmm0, -42(%edx)
L(aligned_16_26bytes):
movdqa %xmm0, -26(%edx)
L(aligned_16_10bytes):
movq %xmm0, -10(%edx)
movw %ax, -2(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_124bytes):
movdqa %xmm0, -124(%edx)
L(aligned_16_108bytes):
movdqa %xmm0, -108(%edx)
L(aligned_16_92bytes):
movdqa %xmm0, -92(%edx)
L(aligned_16_76bytes):
movdqa %xmm0, -76(%edx)
L(aligned_16_60bytes):
movdqa %xmm0, -60(%edx)
L(aligned_16_44bytes):
movdqa %xmm0, -44(%edx)
L(aligned_16_28bytes):
movdqa %xmm0, -28(%edx)
L(aligned_16_12bytes):
movq %xmm0, -12(%edx)
movl %eax, -4(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_126bytes):
movdqa %xmm0, -126(%edx)
L(aligned_16_110bytes):
movdqa %xmm0, -110(%edx)
L(aligned_16_94bytes):
movdqa %xmm0, -94(%edx)
L(aligned_16_78bytes):
movdqa %xmm0, -78(%edx)
L(aligned_16_62bytes):
movdqa %xmm0, -62(%edx)
L(aligned_16_46bytes):
movdqa %xmm0, -46(%edx)
L(aligned_16_30bytes):
movdqa %xmm0, -30(%edx)
L(aligned_16_14bytes):
movq %xmm0, -14(%edx)
movl %eax, -6(%edx)
movw %ax, -2(%edx)
SETRTNVAL
RETURN
END (MEMSET)

510
libcutils/arch-x86/android_memset32.S
View File

@ -1,510 +0,0 @@
/*
* Copyright (C) 2010 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "cache.h"
#ifndef MEMSET
# define MEMSET android_memset32
#endif
#ifndef L
# define L(label) .L##label
#endif
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
#ifndef cfi_startproc
# define cfi_startproc .cfi_startproc
#endif
#ifndef cfi_endproc
# define cfi_endproc .cfi_endproc
#endif
#ifndef cfi_rel_offset
# define cfi_rel_offset(reg, off) .cfi_rel_offset reg, off
#endif
#ifndef cfi_restore
# define cfi_restore(reg) .cfi_restore reg
#endif
#ifndef cfi_adjust_cfa_offset
# define cfi_adjust_cfa_offset(off) .cfi_adjust_cfa_offset off
#endif
#ifndef ENTRY
# define ENTRY(name) \
.type name, @function; \
.globl name; \
.p2align 4; \
name: \
cfi_startproc
#endif
#ifndef END
# define END(name) \
cfi_endproc; \
.size name, .-name
#endif
#define CFI_PUSH(REG) \
cfi_adjust_cfa_offset (4); \
cfi_rel_offset (REG, 0)
#define CFI_POP(REG) \
cfi_adjust_cfa_offset (-4); \
cfi_restore (REG)
#define PUSH(REG) pushl REG; CFI_PUSH (REG)
#define POP(REG) popl REG; CFI_POP (REG)
#ifdef USE_AS_BZERO32
# define DEST PARMS
# define LEN DEST+4
# define SETRTNVAL
#else
# define DEST PARMS
# define DWDS DEST+4
# define LEN DWDS+4
# define SETRTNVAL movl DEST(%esp), %eax
#endif
#if (defined SHARED || defined __PIC__)
# define ENTRANCE PUSH (%ebx);
# define RETURN_END POP (%ebx); ret
# define RETURN RETURN_END; CFI_PUSH (%ebx)
# define PARMS 8 /* Preserve EBX. */
# define JMPTBL(I, B) I - B
/* Load an entry in a jump table into EBX and branch to it. TABLE is a
jump table with relative offsets. */
# define BRANCH_TO_JMPTBL_ENTRY(TABLE) \
/* We first load PC into EBX. */ \
call __x86.get_pc_thunk.bx; \
/* Get the address of the jump table. */ \
add $(TABLE - .), %ebx; \
/* Get the entry and convert the relative offset to the \
absolute address. */ \
add (%ebx,%ecx,4), %ebx; \
/* We loaded the jump table and adjuested EDX. Go. */ \
jmp *%ebx
.section .gnu.linkonce.t.__x86.get_pc_thunk.bx,"ax",@progbits
.globl __x86.get_pc_thunk.bx
.hidden __x86.get_pc_thunk.bx
ALIGN (4)
.type __x86.get_pc_thunk.bx,@function
__x86.get_pc_thunk.bx:
movl (%esp), %ebx
ret
#else
# define ENTRANCE
# define RETURN_END ret
# define RETURN RETURN_END
# define PARMS 4
# define JMPTBL(I, B) I
/* Branch to an entry in a jump table. TABLE is a jump table with
absolute offsets. */
# define BRANCH_TO_JMPTBL_ENTRY(TABLE) \
jmp *TABLE(,%ecx,4)
#endif
.section .text.sse2,"ax",@progbits
ALIGN (4)
ENTRY (MEMSET)
ENTRANCE
movl LEN(%esp), %ecx
shr $2, %ecx
#ifdef USE_AS_BZERO32
xor %eax, %eax
#else
mov DWDS(%esp), %eax
mov %eax, %edx
#endif
movl DEST(%esp), %edx
cmp $16, %ecx
jae L(16dbwordsormore)
L(write_less16dbwords):
lea (%edx, %ecx, 4), %edx
BRANCH_TO_JMPTBL_ENTRY (L(table_less16dbwords))
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_less16dbwords):
.int JMPTBL (L(write_0dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_1dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_2dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_3dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_4dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_5dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_6dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_7dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_8dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_9dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_10dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_11dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_12dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_13dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_14dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_15dbwords), L(table_less16dbwords))
.popsection
ALIGN (4)
L(write_15dbwords):
movl %eax, -60(%edx)
L(write_14dbwords):
movl %eax, -56(%edx)
L(write_13dbwords):
movl %eax, -52(%edx)
L(write_12dbwords):
movl %eax, -48(%edx)
L(write_11dbwords):
movl %eax, -44(%edx)
L(write_10dbwords):
movl %eax, -40(%edx)
L(write_9dbwords):
movl %eax, -36(%edx)
L(write_8dbwords):
movl %eax, -32(%edx)
L(write_7dbwords):
movl %eax, -28(%edx)
L(write_6dbwords):
movl %eax, -24(%edx)
L(write_5dbwords):
movl %eax, -20(%edx)
L(write_4dbwords):
movl %eax, -16(%edx)
L(write_3dbwords):
movl %eax, -12(%edx)
L(write_2dbwords):
movl %eax, -8(%edx)
L(write_1dbwords):
movl %eax, -4(%edx)
L(write_0dbwords):
SETRTNVAL
RETURN
ALIGN (4)
L(16dbwordsormore):
test $3, %edx
jz L(aligned4bytes)
mov %eax, (%edx)
mov %eax, -4(%edx, %ecx, 4)
sub $1, %ecx
rol $24, %eax
add $1, %edx
test $3, %edx
jz L(aligned4bytes)
ror $8, %eax
add $1, %edx
test $3, %edx
jz L(aligned4bytes)
ror $8, %eax
add $1, %edx
L(aligned4bytes):
shl $2, %ecx
#ifdef USE_AS_BZERO32
pxor %xmm0, %xmm0
#else
movd %eax, %xmm0
pshufd $0, %xmm0, %xmm0
#endif
testl $0xf, %edx
jz L(aligned_16)
/* ECX > 32 and EDX is not 16 byte aligned. */
L(not_aligned_16):
movdqu %xmm0, (%edx)
movl %edx, %eax
and $-16, %edx
add $16, %edx
sub %edx, %eax
add %eax, %ecx
movd %xmm0, %eax
ALIGN (4)
L(aligned_16):
cmp $128, %ecx
jae L(128bytesormore)
L(aligned_16_less128bytes):
add %ecx, %edx
shr $2, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
ALIGN (4)
L(128bytesormore):
#ifdef SHARED_CACHE_SIZE
PUSH (%ebx)
mov $SHARED_CACHE_SIZE, %ebx
#else
# if (defined SHARED || defined __PIC__)
call __x86.get_pc_thunk.bx
add $_GLOBAL_OFFSET_TABLE_, %ebx
mov __x86_shared_cache_size@GOTOFF(%ebx), %ebx
# else
PUSH (%ebx)
mov __x86_shared_cache_size, %ebx
# endif
#endif
cmp %ebx, %ecx
jae L(128bytesormore_nt_start)
#ifdef DATA_CACHE_SIZE
POP (%ebx)
# define RESTORE_EBX_STATE CFI_PUSH (%ebx)
cmp $DATA_CACHE_SIZE, %ecx
#else
# if (defined SHARED || defined __PIC__)
# define RESTORE_EBX_STATE
call __x86.get_pc_thunk.bx
add $_GLOBAL_OFFSET_TABLE_, %ebx
cmp __x86_data_cache_size@GOTOFF(%ebx), %ecx
# else
POP (%ebx)
# define RESTORE_EBX_STATE CFI_PUSH (%ebx)
cmp __x86_data_cache_size, %ecx
# endif
#endif
jae L(128bytes_L2_normal)
subl $128, %ecx
L(128bytesormore_normal):
sub $128, %ecx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
lea 128(%edx), %edx
jb L(128bytesless_normal)
sub $128, %ecx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
lea 128(%edx), %edx
jae L(128bytesormore_normal)
L(128bytesless_normal):
lea 128(%ecx), %ecx
add %ecx, %edx
shr $2, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
ALIGN (4)
L(128bytes_L2_normal):
prefetcht0 0x380(%edx)
prefetcht0 0x3c0(%edx)
sub $128, %ecx
movdqa %xmm0, (%edx)
movaps %xmm0, 0x10(%edx)
movaps %xmm0, 0x20(%edx)
movaps %xmm0, 0x30(%edx)
movaps %xmm0, 0x40(%edx)
movaps %xmm0, 0x50(%edx)
movaps %xmm0, 0x60(%edx)
movaps %xmm0, 0x70(%edx)
add $128, %edx
cmp $128, %ecx
jae L(128bytes_L2_normal)
L(128bytesless_L2_normal):
add %ecx, %edx
shr $2, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
RESTORE_EBX_STATE
L(128bytesormore_nt_start):
sub %ebx, %ecx
mov %ebx, %eax
and $0x7f, %eax
add %eax, %ecx
movd %xmm0, %eax
ALIGN (4)
L(128bytesormore_shared_cache_loop):
prefetcht0 0x3c0(%edx)
prefetcht0 0x380(%edx)
sub $0x80, %ebx
movdqa %xmm0, (%edx)
movdqa %xmm0, 0x10(%edx)
movdqa %xmm0, 0x20(%edx)
movdqa %xmm0, 0x30(%edx)
movdqa %xmm0, 0x40(%edx)
movdqa %xmm0, 0x50(%edx)
movdqa %xmm0, 0x60(%edx)
movdqa %xmm0, 0x70(%edx)
add $0x80, %edx
cmp $0x80, %ebx
jae L(128bytesormore_shared_cache_loop)
cmp $0x80, %ecx
jb L(shared_cache_loop_end)
ALIGN (4)
L(128bytesormore_nt):
sub $0x80, %ecx
movntdq %xmm0, (%edx)
movntdq %xmm0, 0x10(%edx)
movntdq %xmm0, 0x20(%edx)
movntdq %xmm0, 0x30(%edx)
movntdq %xmm0, 0x40(%edx)
movntdq %xmm0, 0x50(%edx)
movntdq %xmm0, 0x60(%edx)
movntdq %xmm0, 0x70(%edx)
add $0x80, %edx
cmp $0x80, %ecx
jae L(128bytesormore_nt)
sfence
L(shared_cache_loop_end):
#if defined DATA_CACHE_SIZE || !(defined SHARED || defined __PIC__)
POP (%ebx)
#endif
add %ecx, %edx
shr $2, %ecx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_16_128bytes):
.int JMPTBL (L(aligned_16_0bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_4bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_8bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_12bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_16bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_20bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_24bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_28bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_32bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_36bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_40bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_44bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_48bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_52bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_56bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_60bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_64bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_68bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_72bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_76bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_80bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_84bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_88bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_92bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_96bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_100bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_104bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_108bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_112bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_116bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_120bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_124bytes), L(table_16_128bytes))
.popsection
ALIGN (4)
L(aligned_16_112bytes):
movdqa %xmm0, -112(%edx)
L(aligned_16_96bytes):
movdqa %xmm0, -96(%edx)
L(aligned_16_80bytes):
movdqa %xmm0, -80(%edx)
L(aligned_16_64bytes):
movdqa %xmm0, -64(%edx)
L(aligned_16_48bytes):
movdqa %xmm0, -48(%edx)
L(aligned_16_32bytes):
movdqa %xmm0, -32(%edx)
L(aligned_16_16bytes):
movdqa %xmm0, -16(%edx)
L(aligned_16_0bytes):
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_116bytes):
movdqa %xmm0, -116(%edx)
L(aligned_16_100bytes):
movdqa %xmm0, -100(%edx)
L(aligned_16_84bytes):
movdqa %xmm0, -84(%edx)
L(aligned_16_68bytes):
movdqa %xmm0, -68(%edx)
L(aligned_16_52bytes):
movdqa %xmm0, -52(%edx)
L(aligned_16_36bytes):
movdqa %xmm0, -36(%edx)
L(aligned_16_20bytes):
movdqa %xmm0, -20(%edx)
L(aligned_16_4bytes):
movl %eax, -4(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_120bytes):
movdqa %xmm0, -120(%edx)
L(aligned_16_104bytes):
movdqa %xmm0, -104(%edx)
L(aligned_16_88bytes):
movdqa %xmm0, -88(%edx)
L(aligned_16_72bytes):
movdqa %xmm0, -72(%edx)
L(aligned_16_56bytes):
movdqa %xmm0, -56(%edx)
L(aligned_16_40bytes):
movdqa %xmm0, -40(%edx)
L(aligned_16_24bytes):
movdqa %xmm0, -24(%edx)
L(aligned_16_8bytes):
movq %xmm0, -8(%edx)
SETRTNVAL
RETURN
ALIGN (4)
L(aligned_16_124bytes):
movdqa %xmm0, -124(%edx)
L(aligned_16_108bytes):
movdqa %xmm0, -108(%edx)
L(aligned_16_92bytes):
movdqa %xmm0, -92(%edx)
L(aligned_16_76bytes):
movdqa %xmm0, -76(%edx)
L(aligned_16_60bytes):
movdqa %xmm0, -60(%edx)
L(aligned_16_44bytes):
movdqa %xmm0, -44(%edx)
L(aligned_16_28bytes):
movdqa %xmm0, -28(%edx)
L(aligned_16_12bytes):
movq %xmm0, -12(%edx)
movl %eax, -4(%edx)
SETRTNVAL
RETURN
END (MEMSET)

565
libcutils/arch-x86_64/android_memset16.S
View File

@ -1,565 +0,0 @@
/*
* Copyright (C) 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "cache.h"
#ifndef MEMSET
# define MEMSET android_memset16
#endif
#ifndef L
# define L(label) .L##label
#endif
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
#ifndef cfi_startproc
# define cfi_startproc .cfi_startproc
#endif
#ifndef cfi_endproc
# define cfi_endproc .cfi_endproc
#endif
#ifndef ENTRY
# define ENTRY(name) \
.type name, @function; \
.globl name; \
.p2align 4; \
name: \
cfi_startproc
#endif
#ifndef END
# define END(name) \
cfi_endproc; \
.size name, .-name
#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; \
jmp *INDEX
.section .text.sse2,"ax",@progbits
ALIGN (4)
ENTRY (MEMSET) // Address in rdi
shr $1, %rdx // Count in rdx
movzwl %si, %ecx
/* Fill the whole ECX with pattern. */
shl $16, %esi
or %esi, %ecx // Pattern in ecx
cmp $32, %rdx
jae L(32wordsormore)
L(write_less32words):
lea (%rdi, %rdx, 2), %rdi
BRANCH_TO_JMPTBL_ENTRY (L(table_less32words), %rdx, 4)
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_less32words):
.int JMPTBL (L(write_0words), L(table_less32words))
.int JMPTBL (L(write_1words), L(table_less32words))
.int JMPTBL (L(write_2words), L(table_less32words))
.int JMPTBL (L(write_3words), L(table_less32words))
.int JMPTBL (L(write_4words), L(table_less32words))
.int JMPTBL (L(write_5words), L(table_less32words))
.int JMPTBL (L(write_6words), L(table_less32words))
.int JMPTBL (L(write_7words), L(table_less32words))
.int JMPTBL (L(write_8words), L(table_less32words))
.int JMPTBL (L(write_9words), L(table_less32words))
.int JMPTBL (L(write_10words), L(table_less32words))
.int JMPTBL (L(write_11words), L(table_less32words))
.int JMPTBL (L(write_12words), L(table_less32words))
.int JMPTBL (L(write_13words), L(table_less32words))
.int JMPTBL (L(write_14words), L(table_less32words))
.int JMPTBL (L(write_15words), L(table_less32words))
.int JMPTBL (L(write_16words), L(table_less32words))
.int JMPTBL (L(write_17words), L(table_less32words))
.int JMPTBL (L(write_18words), L(table_less32words))
.int JMPTBL (L(write_19words), L(table_less32words))
.int JMPTBL (L(write_20words), L(table_less32words))
.int JMPTBL (L(write_21words), L(table_less32words))
.int JMPTBL (L(write_22words), L(table_less32words))
.int JMPTBL (L(write_23words), L(table_less32words))
.int JMPTBL (L(write_24words), L(table_less32words))
.int JMPTBL (L(write_25words), L(table_less32words))
.int JMPTBL (L(write_26words), L(table_less32words))
.int JMPTBL (L(write_27words), L(table_less32words))
.int JMPTBL (L(write_28words), L(table_less32words))
.int JMPTBL (L(write_29words), L(table_less32words))
.int JMPTBL (L(write_30words), L(table_less32words))
.int JMPTBL (L(write_31words), L(table_less32words))
.popsection
ALIGN (4)
L(write_28words):
movl %ecx, -56(%rdi)
movl %ecx, -52(%rdi)
L(write_24words):
movl %ecx, -48(%rdi)
movl %ecx, -44(%rdi)
L(write_20words):
movl %ecx, -40(%rdi)
movl %ecx, -36(%rdi)
L(write_16words):
movl %ecx, -32(%rdi)
movl %ecx, -28(%rdi)
L(write_12words):
movl %ecx, -24(%rdi)
movl %ecx, -20(%rdi)
L(write_8words):
movl %ecx, -16(%rdi)
movl %ecx, -12(%rdi)
L(write_4words):
movl %ecx, -8(%rdi)
movl %ecx, -4(%rdi)
L(write_0words):
ret
ALIGN (4)
L(write_29words):
movl %ecx, -58(%rdi)
movl %ecx, -54(%rdi)
L(write_25words):
movl %ecx, -50(%rdi)
movl %ecx, -46(%rdi)
L(write_21words):
movl %ecx, -42(%rdi)
movl %ecx, -38(%rdi)
L(write_17words):
movl %ecx, -34(%rdi)
movl %ecx, -30(%rdi)
L(write_13words):
movl %ecx, -26(%rdi)
movl %ecx, -22(%rdi)
L(write_9words):
movl %ecx, -18(%rdi)
movl %ecx, -14(%rdi)
L(write_5words):
movl %ecx, -10(%rdi)
movl %ecx, -6(%rdi)
L(write_1words):
mov %cx, -2(%rdi)
ret
ALIGN (4)
L(write_30words):
movl %ecx, -60(%rdi)
movl %ecx, -56(%rdi)
L(write_26words):
movl %ecx, -52(%rdi)
movl %ecx, -48(%rdi)
L(write_22words):
movl %ecx, -44(%rdi)
movl %ecx, -40(%rdi)
L(write_18words):
movl %ecx, -36(%rdi)
movl %ecx, -32(%rdi)
L(write_14words):
movl %ecx, -28(%rdi)
movl %ecx, -24(%rdi)
L(write_10words):
movl %ecx, -20(%rdi)
movl %ecx, -16(%rdi)
L(write_6words):
movl %ecx, -12(%rdi)
movl %ecx, -8(%rdi)
L(write_2words):
movl %ecx, -4(%rdi)
ret
ALIGN (4)
L(write_31words):
movl %ecx, -62(%rdi)
movl %ecx, -58(%rdi)
L(write_27words):
movl %ecx, -54(%rdi)
movl %ecx, -50(%rdi)
L(write_23words):
movl %ecx, -46(%rdi)
movl %ecx, -42(%rdi)
L(write_19words):
movl %ecx, -38(%rdi)
movl %ecx, -34(%rdi)
L(write_15words):
movl %ecx, -30(%rdi)
movl %ecx, -26(%rdi)
L(write_11words):
movl %ecx, -22(%rdi)
movl %ecx, -18(%rdi)
L(write_7words):
movl %ecx, -14(%rdi)
movl %ecx, -10(%rdi)
L(write_3words):
movl %ecx, -6(%rdi)
movw %cx, -2(%rdi)
ret
ALIGN (4)
L(32wordsormore):
shl $1, %rdx
test $0x01, %edi
jz L(aligned2bytes)
mov %ecx, (%rdi)
mov %ecx, -4(%rdi, %rdx)
sub $2, %rdx
add $1, %rdi
rol $8, %ecx
L(aligned2bytes):
/* Fill xmm0 with the pattern. */
movd %ecx, %xmm0
pshufd $0, %xmm0, %xmm0
testl $0xf, %edi
jz L(aligned_16)
/* RDX > 32 and RDI is not 16 byte aligned. */
movdqu %xmm0, (%rdi)
mov %rdi, %rsi
and $-16, %rdi
add $16, %rdi
sub %rdi, %rsi
add %rsi, %rdx
ALIGN (4)
L(aligned_16):
cmp $128, %rdx
jge L(128bytesormore)
L(aligned_16_less128bytes):
add %rdx, %rdi
shr $1, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
ALIGN (4)
L(128bytesormore):
cmp $SHARED_CACHE_SIZE, %rdx
jg L(128bytesormore_nt)
L(128bytesormore_normal):
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jge L(128bytesormore_normal)
L(128bytesless_normal):
add %rdx, %rdi
shr $1, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
ALIGN (4)
L(128bytesormore_nt):
sub $128, %rdx
movntdq %xmm0, (%rdi)
movntdq %xmm0, 0x10(%rdi)
movntdq %xmm0, 0x20(%rdi)
movntdq %xmm0, 0x30(%rdi)
movntdq %xmm0, 0x40(%rdi)
movntdq %xmm0, 0x50(%rdi)
movntdq %xmm0, 0x60(%rdi)
movntdq %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jge L(128bytesormore_nt)
sfence
add %rdx, %rdi
shr $1, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_16_128bytes):
.int JMPTBL (L(aligned_16_0bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_2bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_4bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_6bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_8bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_10bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_12bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_14bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_16bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_18bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_20bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_22bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_24bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_26bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_28bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_30bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_32bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_34bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_36bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_38bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_40bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_42bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_44bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_46bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_48bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_50bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_52bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_54bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_56bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_58bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_60bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_62bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_64bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_66bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_68bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_70bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_72bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_74bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_76bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_78bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_80bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_82bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_84bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_86bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_88bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_90bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_92bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_94bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_96bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_98bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_100bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_102bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_104bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_106bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_108bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_110bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_112bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_114bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_116bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_118bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_120bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_122bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_124bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_126bytes), L(table_16_128bytes))
.popsection
ALIGN (4)
L(aligned_16_112bytes):
movdqa %xmm0, -112(%rdi)
L(aligned_16_96bytes):
movdqa %xmm0, -96(%rdi)
L(aligned_16_80bytes):
movdqa %xmm0, -80(%rdi)
L(aligned_16_64bytes):
movdqa %xmm0, -64(%rdi)
L(aligned_16_48bytes):
movdqa %xmm0, -48(%rdi)
L(aligned_16_32bytes):
movdqa %xmm0, -32(%rdi)
L(aligned_16_16bytes):
movdqa %xmm0, -16(%rdi)
L(aligned_16_0bytes):
ret
ALIGN (4)
L(aligned_16_114bytes):
movdqa %xmm0, -114(%rdi)
L(aligned_16_98bytes):
movdqa %xmm0, -98(%rdi)
L(aligned_16_82bytes):
movdqa %xmm0, -82(%rdi)
L(aligned_16_66bytes):
movdqa %xmm0, -66(%rdi)
L(aligned_16_50bytes):
movdqa %xmm0, -50(%rdi)
L(aligned_16_34bytes):
movdqa %xmm0, -34(%rdi)
L(aligned_16_18bytes):
movdqa %xmm0, -18(%rdi)
L(aligned_16_2bytes):
movw %cx, -2(%rdi)
ret
ALIGN (4)
L(aligned_16_116bytes):
movdqa %xmm0, -116(%rdi)
L(aligned_16_100bytes):
movdqa %xmm0, -100(%rdi)
L(aligned_16_84bytes):
movdqa %xmm0, -84(%rdi)
L(aligned_16_68bytes):
movdqa %xmm0, -68(%rdi)
L(aligned_16_52bytes):
movdqa %xmm0, -52(%rdi)
L(aligned_16_36bytes):
movdqa %xmm0, -36(%rdi)
L(aligned_16_20bytes):
movdqa %xmm0, -20(%rdi)
L(aligned_16_4bytes):
movl %ecx, -4(%rdi)
ret
ALIGN (4)
L(aligned_16_118bytes):
movdqa %xmm0, -118(%rdi)
L(aligned_16_102bytes):
movdqa %xmm0, -102(%rdi)
L(aligned_16_86bytes):
movdqa %xmm0, -86(%rdi)
L(aligned_16_70bytes):
movdqa %xmm0, -70(%rdi)
L(aligned_16_54bytes):
movdqa %xmm0, -54(%rdi)
L(aligned_16_38bytes):
movdqa %xmm0, -38(%rdi)
L(aligned_16_22bytes):
movdqa %xmm0, -22(%rdi)
L(aligned_16_6bytes):
movl %ecx, -6(%rdi)
movw %cx, -2(%rdi)
ret
ALIGN (4)
L(aligned_16_120bytes):
movdqa %xmm0, -120(%rdi)
L(aligned_16_104bytes):
movdqa %xmm0, -104(%rdi)
L(aligned_16_88bytes):
movdqa %xmm0, -88(%rdi)
L(aligned_16_72bytes):
movdqa %xmm0, -72(%rdi)
L(aligned_16_56bytes):
movdqa %xmm0, -56(%rdi)
L(aligned_16_40bytes):
movdqa %xmm0, -40(%rdi)
L(aligned_16_24bytes):
movdqa %xmm0, -24(%rdi)
L(aligned_16_8bytes):
movq %xmm0, -8(%rdi)
ret
ALIGN (4)
L(aligned_16_122bytes):
movdqa %xmm0, -122(%rdi)
L(aligned_16_106bytes):
movdqa %xmm0, -106(%rdi)
L(aligned_16_90bytes):
movdqa %xmm0, -90(%rdi)
L(aligned_16_74bytes):
movdqa %xmm0, -74(%rdi)
L(aligned_16_58bytes):
movdqa %xmm0, -58(%rdi)
L(aligned_16_42bytes):
movdqa %xmm0, -42(%rdi)
L(aligned_16_26bytes):
movdqa %xmm0, -26(%rdi)
L(aligned_16_10bytes):
movq %xmm0, -10(%rdi)
movw %cx, -2(%rdi)
ret
ALIGN (4)
L(aligned_16_124bytes):
movdqa %xmm0, -124(%rdi)
L(aligned_16_108bytes):
movdqa %xmm0, -108(%rdi)
L(aligned_16_92bytes):
movdqa %xmm0, -92(%rdi)
L(aligned_16_76bytes):
movdqa %xmm0, -76(%rdi)
L(aligned_16_60bytes):
movdqa %xmm0, -60(%rdi)
L(aligned_16_44bytes):
movdqa %xmm0, -44(%rdi)
L(aligned_16_28bytes):
movdqa %xmm0, -28(%rdi)
L(aligned_16_12bytes):
movq %xmm0, -12(%rdi)
movl %ecx, -4(%rdi)
ret
ALIGN (4)
L(aligned_16_126bytes):
movdqa %xmm0, -126(%rdi)
L(aligned_16_110bytes):
movdqa %xmm0, -110(%rdi)
L(aligned_16_94bytes):
movdqa %xmm0, -94(%rdi)
L(aligned_16_78bytes):
movdqa %xmm0, -78(%rdi)
L(aligned_16_62bytes):
movdqa %xmm0, -62(%rdi)
L(aligned_16_46bytes):
movdqa %xmm0, -46(%rdi)
L(aligned_16_30bytes):
movdqa %xmm0, -30(%rdi)
L(aligned_16_14bytes):
movq %xmm0, -14(%rdi)
movl %ecx, -6(%rdi)
movw %cx, -2(%rdi)
ret
END (MEMSET)

373
libcutils/arch-x86_64/android_memset32.S
View File

@ -1,373 +0,0 @@
/*
* Copyright (C) 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "cache.h"
#ifndef MEMSET
# define MEMSET android_memset32
#endif
#ifndef L
# define L(label) .L##label
#endif
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
#ifndef cfi_startproc
# define cfi_startproc .cfi_startproc
#endif
#ifndef cfi_endproc
# define cfi_endproc .cfi_endproc
#endif
#ifndef ENTRY
# define ENTRY(name) \
.type name, @function; \
.globl name; \
.p2align 4; \
name: \
cfi_startproc
#endif
#ifndef END
# define END(name) \
cfi_endproc; \
.size name, .-name
#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; \
jmp *INDEX
.section .text.sse2,"ax",@progbits
ALIGN (4)
ENTRY (MEMSET) // Address in rdi
shr $2, %rdx // Count in rdx
movl %esi, %ecx // Pattern in ecx
cmp $16, %rdx
jae L(16dbwordsormore)
L(write_less16dbwords):
lea (%rdi, %rdx, 4), %rdi
BRANCH_TO_JMPTBL_ENTRY (L(table_less16dbwords), %rdx, 4)
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_less16dbwords):
.int JMPTBL (L(write_0dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_1dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_2dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_3dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_4dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_5dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_6dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_7dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_8dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_9dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_10dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_11dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_12dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_13dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_14dbwords), L(table_less16dbwords))
.int JMPTBL (L(write_15dbwords), L(table_less16dbwords))
.popsection
ALIGN (4)
L(write_15dbwords):
movl %ecx, -60(%rdi)
L(write_14dbwords):
movl %ecx, -56(%rdi)
L(write_13dbwords):
movl %ecx, -52(%rdi)
L(write_12dbwords):
movl %ecx, -48(%rdi)
L(write_11dbwords):
movl %ecx, -44(%rdi)
L(write_10dbwords):
movl %ecx, -40(%rdi)
L(write_9dbwords):
movl %ecx, -36(%rdi)
L(write_8dbwords):
movl %ecx, -32(%rdi)
L(write_7dbwords):
movl %ecx, -28(%rdi)
L(write_6dbwords):
movl %ecx, -24(%rdi)
L(write_5dbwords):
movl %ecx, -20(%rdi)
L(write_4dbwords):
movl %ecx, -16(%rdi)
L(write_3dbwords):
movl %ecx, -12(%rdi)
L(write_2dbwords):
movl %ecx, -8(%rdi)
L(write_1dbwords):
movl %ecx, -4(%rdi)
L(write_0dbwords):
ret
ALIGN (4)
L(16dbwordsormore):
test $3, %edi
jz L(aligned4bytes)
mov %ecx, (%rdi)
mov %ecx, -4(%rdi, %rdx, 4)
sub $1, %rdx
rol $24, %ecx
add $1, %rdi
test $3, %edi
jz L(aligned4bytes)
ror $8, %ecx
add $1, %rdi
test $3, %edi
jz L(aligned4bytes)
ror $8, %ecx
add $1, %rdi
L(aligned4bytes):
shl $2, %rdx
/* Fill xmm0 with the pattern. */
movd %ecx, %xmm0
pshufd $0, %xmm0, %xmm0
testl $0xf, %edi
jz L(aligned_16)
/* RDX > 32 and RDI is not 16 byte aligned. */
movdqu %xmm0, (%rdi)
mov %rdi, %rsi
and $-16, %rdi
add $16, %rdi
sub %rdi, %rsi
add %rsi, %rdx
ALIGN (4)
L(aligned_16):
cmp $128, %rdx
jge L(128bytesormore)
L(aligned_16_less128bytes):
add %rdx, %rdi
shr $2, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
ALIGN (4)
L(128bytesormore):
cmp $SHARED_CACHE_SIZE, %rdx
jg L(128bytesormore_nt)
L(128bytesormore_normal):
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jl L(128bytesless_normal)
sub $128, %rdx
movdqa %xmm0, (%rdi)
movdqa %xmm0, 0x10(%rdi)
movdqa %xmm0, 0x20(%rdi)
movdqa %xmm0, 0x30(%rdi)
movdqa %xmm0, 0x40(%rdi)
movdqa %xmm0, 0x50(%rdi)
movdqa %xmm0, 0x60(%rdi)
movdqa %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jge L(128bytesormore_normal)
L(128bytesless_normal):
add %rdx, %rdi
shr $2, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
ALIGN (4)
L(128bytesormore_nt):
sub $128, %rdx
movntdq %xmm0, (%rdi)
movntdq %xmm0, 0x10(%rdi)
movntdq %xmm0, 0x20(%rdi)
movntdq %xmm0, 0x30(%rdi)
movntdq %xmm0, 0x40(%rdi)
movntdq %xmm0, 0x50(%rdi)
movntdq %xmm0, 0x60(%rdi)
movntdq %xmm0, 0x70(%rdi)
lea 128(%rdi), %rdi
cmp $128, %rdx
jge L(128bytesormore_nt)
sfence
add %rdx, %rdi
shr $2, %rdx
BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes), %rdx, 4)
.pushsection .rodata.sse2,"a",@progbits
ALIGN (2)
L(table_16_128bytes):
.int JMPTBL (L(aligned_16_0bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_4bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_8bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_12bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_16bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_20bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_24bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_28bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_32bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_36bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_40bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_44bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_48bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_52bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_56bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_60bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_64bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_68bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_72bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_76bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_80bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_84bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_88bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_92bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_96bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_100bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_104bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_108bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_112bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_116bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_120bytes), L(table_16_128bytes))
.int JMPTBL (L(aligned_16_124bytes), L(table_16_128bytes))
.popsection
ALIGN (4)
L(aligned_16_112bytes):
movdqa %xmm0, -112(%rdi)
L(aligned_16_96bytes):
movdqa %xmm0, -96(%rdi)
L(aligned_16_80bytes):
movdqa %xmm0, -80(%rdi)
L(aligned_16_64bytes):
movdqa %xmm0, -64(%rdi)
L(aligned_16_48bytes):
movdqa %xmm0, -48(%rdi)
L(aligned_16_32bytes):
movdqa %xmm0, -32(%rdi)
L(aligned_16_16bytes):
movdqa %xmm0, -16(%rdi)
L(aligned_16_0bytes):
ret
ALIGN (4)
L(aligned_16_116bytes):
movdqa %xmm0, -116(%rdi)
L(aligned_16_100bytes):
movdqa %xmm0, -100(%rdi)
L(aligned_16_84bytes):
movdqa %xmm0, -84(%rdi)
L(aligned_16_68bytes):
movdqa %xmm0, -68(%rdi)
L(aligned_16_52bytes):
movdqa %xmm0, -52(%rdi)
L(aligned_16_36bytes):
movdqa %xmm0, -36(%rdi)
L(aligned_16_20bytes):
movdqa %xmm0, -20(%rdi)
L(aligned_16_4bytes):
movl %ecx, -4(%rdi)
ret
ALIGN (4)
L(aligned_16_120bytes):
movdqa %xmm0, -120(%rdi)
L(aligned_16_104bytes):
movdqa %xmm0, -104(%rdi)
L(aligned_16_88bytes):
movdqa %xmm0, -88(%rdi)
L(aligned_16_72bytes):
movdqa %xmm0, -72(%rdi)
L(aligned_16_56bytes):
movdqa %xmm0, -56(%rdi)
L(aligned_16_40bytes):
movdqa %xmm0, -40(%rdi)
L(aligned_16_24bytes):
movdqa %xmm0, -24(%rdi)
L(aligned_16_8bytes):
movq %xmm0, -8(%rdi)
ret
ALIGN (4)
L(aligned_16_124bytes):
movdqa %xmm0, -124(%rdi)
L(aligned_16_108bytes):
movdqa %xmm0, -108(%rdi)
L(aligned_16_92bytes):
movdqa %xmm0, -92(%rdi)
L(aligned_16_76bytes):
movdqa %xmm0, -76(%rdi)
L(aligned_16_60bytes):
movdqa %xmm0, -60(%rdi)
L(aligned_16_44bytes):
movdqa %xmm0, -44(%rdi)
L(aligned_16_28bytes):
movdqa %xmm0, -28(%rdi)
L(aligned_16_12bytes):
movq %xmm0, -12(%rdi)
movl %ecx, -4(%rdi)
ret
END (MEMSET)

11
libcutils/include/cutils/memory.h
View File

@ -14,8 +14,7 @@
* limitations under the License.
*/
#ifndef ANDROID_CUTILS_MEMORY_H
#define ANDROID_CUTILS_MEMORY_H
#pragma once
#include <stdint.h>
#include <sys/types.h>
@ -24,12 +23,6 @@
extern "C" {
#endif
/* size is given in bytes and must be multiple of 2 */
void android_memset16(uint16_t* dst, uint16_t value, size_t size);
/* size is given in bytes and must be multiple of 4 */
void android_memset32(uint32_t* dst, uint32_t value, size_t size);
#if defined(__GLIBC__) || defined(_WIN32)
/* Declaration of strlcpy() for platforms that don't already have it. */
size_t strlcpy(char *dst, const char *src, size_t size);
@ -38,5 +31,3 @@ size_t strlcpy(char *dst, const char *src, size_t size);
#ifdef __cplusplus
} // extern "C"
#endif
#endif // ANDROID_CUTILS_MEMORY_H

181
libcutils/memset_test.cpp
View File

@ -1,181 +0,0 @@
/*
* Copyright (C) 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>
#include <sys/types.h>
#include <memory>
#include <cutils/memory.h>
#include <gtest/gtest.h>
#define FENCEPOST_LENGTH 8
#define MAX_TEST_SIZE (64*1024)
// Choose values that have no repeating byte values.
#define MEMSET16_PATTERN 0xb139
#define MEMSET32_PATTERN 0x48193a27
enum test_e {
MEMSET16 = 0,
MEMSET32,
};
static int g_memset16_aligns[][2] = {
{ 2, 0 },
{ 4, 0 },
{ 8, 0 },
{ 16, 0 },
{ 32, 0 },
{ 64, 0 },
{ 128, 0 },
{ 4, 2 },
{ 8, 2 },
{ 8, 4 },
{ 8, 6 },
{ 128, 2 },
{ 128, 4 },
{ 128, 6 },
{ 128, 8 },
{ 128, 10 },
{ 128, 12 },
{ 128, 14 },
{ 128, 16 },
};
static int g_memset32_aligns[][2] = {
{ 4, 0 },
{ 8, 0 },
{ 16, 0 },
{ 32, 0 },
{ 64, 0 },
{ 128, 0 },
{ 8, 4 },
{ 128, 4 },
{ 128, 8 },
{ 128, 12 },
{ 128, 16 },
};
static size_t GetIncrement(size_t len, size_t min_incr) {
if (len >= 4096) {
return 1024;
} else if (len >= 1024) {
return 256;
}
return min_incr;
}
// Return a pointer into the current buffer with the specified alignment.
static void *GetAlignedPtr(void *orig_ptr, int alignment, int or_mask) {
uint64_t ptr = reinterpret_cast<uint64_t>(orig_ptr);
if (alignment > 0) {
// When setting the alignment, set it to exactly the alignment chosen.
// The pointer returned will be guaranteed not to be aligned to anything
// more than that.
ptr += alignment - (ptr & (alignment - 1));
ptr |= alignment | or_mask;
}
return reinterpret_cast<void*>(ptr);
}
static void SetFencepost(uint8_t *buffer) {
for (int i = 0; i < FENCEPOST_LENGTH; i += 2) {
buffer[i] = 0xde;
buffer[i+1] = 0xad;
}
}
static void VerifyFencepost(uint8_t *buffer) {
for (int i = 0; i < FENCEPOST_LENGTH; i += 2) {
if (buffer[i] != 0xde || buffer[i+1] != 0xad) {
uint8_t expected_value;
if (buffer[i] == 0xde) {
i++;
expected_value = 0xad;
} else {
expected_value = 0xde;
}
ASSERT_EQ(expected_value, buffer[i]);
}
}
}
void RunMemsetTests(test_e test_type, uint32_t value, int align[][2], size_t num_aligns) {
size_t min_incr = 4;
if (test_type == MEMSET16) {
min_incr = 2;
value |= value << 16;
}
std::unique_ptr<uint32_t[]> expected_buf(new uint32_t[MAX_TEST_SIZE/sizeof(uint32_t)]);
for (size_t i = 0; i < MAX_TEST_SIZE/sizeof(uint32_t); i++) {
expected_buf[i] = value;
}
// Allocate one large buffer with lots of extra space so that we can
// guarantee that all possible alignments will fit.
std::unique_ptr<uint8_t[]> buf(new uint8_t[3*MAX_TEST_SIZE]);
uint8_t *buf_align;
for (size_t i = 0; i < num_aligns; i++) {
size_t incr = min_incr;
for (size_t len = incr; len <= MAX_TEST_SIZE; len += incr) {
incr = GetIncrement(len, min_incr);
buf_align = reinterpret_cast<uint8_t*>(GetAlignedPtr(
buf.get()+FENCEPOST_LENGTH, align[i][0], align[i][1]));
SetFencepost(&buf_align[-FENCEPOST_LENGTH]);
SetFencepost(&buf_align[len]);
memset(buf_align, 0xff, len);
if (test_type == MEMSET16) {
android_memset16(reinterpret_cast<uint16_t*>(buf_align), value, len);
} else {
android_memset32(reinterpret_cast<uint32_t*>(buf_align), value, len);
}
ASSERT_EQ(0, memcmp(expected_buf.get(), buf_align, len))
<< "Failed size " << len << " align " << align[i][0] << " " << align[i][1] << "\n";
VerifyFencepost(&buf_align[-FENCEPOST_LENGTH]);
VerifyFencepost(&buf_align[len]);
}
}
}
TEST(libcutils, android_memset16_non_zero) {
RunMemsetTests(MEMSET16, MEMSET16_PATTERN, g_memset16_aligns, sizeof(g_memset16_aligns)/sizeof(int[2]));
}
TEST(libcutils, android_memset16_zero) {
RunMemsetTests(MEMSET16, 0, g_memset16_aligns, sizeof(g_memset16_aligns)/sizeof(int[2]));
}
TEST(libcutils, android_memset32_non_zero) {
RunMemsetTests(MEMSET32, MEMSET32_PATTERN, g_memset32_aligns, sizeof(g_memset32_aligns)/sizeof(int[2]));
}
TEST(libcutils, android_memset32_zero) {
RunMemsetTests(MEMSET32, 0, g_memset32_aligns, sizeof(g_memset32_aligns)/sizeof(int[2]));
}