am 912514ee: am f37bfb32: am 23aeeff5: Merge "Use the correct fuse_init_out structure size."

* commit '912514eeb7bc0e466ecafc5a3d9dda81fc1e5a93':
diff --git a/adb/adb.h b/adb/adb.h
index 4f06800..135999d 100644
--- a/adb/adb.h
+++ b/adb/adb.h
@@ -338,6 +338,9 @@
 int check_header(apacket *p);
 int check_data(apacket *p);
 
+// Define it if you want to dump packets.
+#define DEBUG_PACKETS 0
+
 #if !DEBUG_PACKETS
 #define print_packet(tag,p) do {} while (0)
 #endif
diff --git a/adb/adb_client.c b/adb/adb_client.c
index eb1720d..ac5e15a 100644
--- a/adb/adb_client.c
+++ b/adb/adb_client.c
@@ -279,7 +279,7 @@
 
     fd = _adb_connect(service);
     if(fd == -1) {
-        D("_adb_connect error: %s\n", __adb_error);
+        D("_adb_connect error: %s", __adb_error);
     } else if(fd == -2) {
         fprintf(stderr,"** daemon still not running\n");
     }
diff --git a/adb/adb_trace.h b/adb/adb_trace.h
index 8a5d9f8..b8a2f4c 100644
--- a/adb/adb_trace.h
+++ b/adb/adb_trace.h
@@ -73,8 +73,9 @@
             if (ADB_TRACING) {                         \
                 int save_errno = errno;                \
                 adb_mutex_lock(&D_lock);               \
-                fprintf(stderr, "%s::%s():",           \
-                        __FILE__, __FUNCTION__);       \
+                fprintf(stderr, "%16s: %5d:%5lu | ",   \
+                        __FUNCTION__,                  \
+                        getpid(), adb_thread_id());    \
                 errno = save_errno;                    \
                 fprintf(stderr, __VA_ARGS__ );         \
                 fflush(stderr);                        \
@@ -96,15 +97,16 @@
         } while (0)
 #  define  DD(...)                                     \
         do {                                           \
-          int save_errno = errno;                      \
-          adb_mutex_lock(&D_lock);                     \
-          fprintf(stderr, "%s::%s():",                 \
-                  __FILE__, __FUNCTION__);             \
-          errno = save_errno;                          \
-          fprintf(stderr, __VA_ARGS__ );               \
-          fflush(stderr);                              \
-          adb_mutex_unlock(&D_lock);                   \
-          errno = save_errno;                          \
+            int save_errno = errno;                    \
+            adb_mutex_lock(&D_lock);                   \
+            fprintf(stderr, "%16s: %5d:%5lu | ",       \
+                    __FUNCTION__,                      \
+                    getpid(), adb_thread_id());        \
+            errno = save_errno;                        \
+            fprintf(stderr, __VA_ARGS__ );             \
+            fflush(stderr);                            \
+            adb_mutex_unlock(&D_lock);                 \
+            errno = save_errno;                        \
         } while (0)
 #else
 #  define  D(...)                                      \
diff --git a/adb/fdevent.c b/adb/fdevent.c
index 43e600c..f5ecd14 100644
--- a/adb/fdevent.c
+++ b/adb/fdevent.c
@@ -661,6 +661,8 @@
     if(adb_socketpair(s)) {
         FATAL("cannot create shell-exit socket-pair\n");
     }
+    D("socketpair: (%d,%d)", s[0], s[1]);
+
     SHELL_EXIT_NOTIFY_FD = s[0];
     fdevent *fde;
     fde = fdevent_create(s[1], fdevent_subproc_event_func, NULL);
diff --git a/adb/jdwp_service.c b/adb/jdwp_service.c
index cd62b55..3074e42 100644
--- a/adb/jdwp_service.c
+++ b/adb/jdwp_service.c
@@ -415,6 +415,7 @@
               __FUNCTION__, strerror(errno));
             return -1;
         }
+        D("socketpair: (%d,%d)", fds[0], fds[1]);
 
         proc->out_fds[ proc->out_count ] = fds[1];
         if (++proc->out_count == 1)
diff --git a/adb/services.c b/adb/services.c
index e61371a..bf2f2c3 100644
--- a/adb/services.c
+++ b/adb/services.c
@@ -164,6 +164,7 @@
         printf("cannot create service socket pair\n");
         return -1;
     }
+    D("socketpair: (%d,%d)", s[0], s[1]);
 
     sti = malloc(sizeof(stinfo));
     if(sti == 0) fatal("cannot allocate stinfo");
@@ -264,10 +265,11 @@
 
     // 0 is parent socket, 1 is child socket
     int sv[2];
-    if (unix_socketpair(AF_UNIX, SOCK_STREAM, 0, sv) < 0) {
+    if (adb_socketpair(sv) < 0) {
         printf("[ cannot create socket pair - %s ]\n", strerror(errno));
         return -1;
     }
+    D("socketpair: (%d,%d)", sv[0], sv[1]);
 
     *pid = fork();
     if (*pid < 0) {
diff --git a/adb/sysdeps.h b/adb/sysdeps.h
index ba4306f..11de548 100644
--- a/adb/sysdeps.h
+++ b/adb/sysdeps.h
@@ -76,6 +76,11 @@
     return 0;
 }
 
+static __inline__  unsigned long adb_thread_id()
+{
+    return GetCurrentThreadId();
+}
+
 static __inline__ void  close_on_exec(int  fd)
 {
     /* nothing really */
@@ -515,6 +520,12 @@
 {
     return strtok_r(str, delim, saveptr);
 }
+
+static __inline__ unsigned long adb_thread_id()
+{
+    return (unsigned long)pthread_self();
+}
+
 #undef   strtok_r
 #define  strtok_r  ___xxx_strtok_r
 
diff --git a/adb/transport.c b/adb/transport.c
index f35880c..7db6a47 100644
--- a/adb/transport.c
+++ b/adb/transport.c
@@ -629,7 +629,7 @@
             fatal_errno("cannot open transport socketpair");
         }
 
-        D("transport: %s (%d,%d) starting\n", t->serial, s[0], s[1]);
+        D("transport: %s socketpair: (%d,%d) starting", t->serial, s[0], s[1]);
 
         t->transport_socket = s[0];
         t->fd = s[1];
@@ -673,6 +673,7 @@
     if(adb_socketpair(s)){
         fatal_errno("cannot open transport registration socketpair");
     }
+    D("socketpair: (%d,%d)", s[0], s[1]);
 
     transport_registration_send = s[0];
     transport_registration_recv = s[1];
diff --git a/fastboot/fastboot.c b/fastboot/fastboot.c
index 03b3177..019d16f 100644
--- a/fastboot/fastboot.c
+++ b/fastboot/fastboot.c
@@ -104,7 +104,6 @@
     {"recovery.img", "recovery.sig", "recovery", true},
     {"system.img", "system.sig", "system", false},
     {"vendor.img", "vendor.sig", "vendor", true},
-    {"tos.img", "tos.sig", "tos", true},
 };
 
 void get_my_path(char *path);
@@ -123,8 +122,6 @@
         fn = "system.img";
     } else if(!strcmp(item,"vendor")) {
         fn = "vendor.img";
-    } else if(!strcmp(item,"tos")) {
-        fn = "tos.img";
     } else if(!strcmp(item,"userdata")) {
         fn = "userdata.img";
     } else if(!strcmp(item,"cache")) {
@@ -291,16 +288,15 @@
             "commands:\n"
             "  update <filename>                        reflash device from update.zip\n"
             "  flashall                                 flash boot, system, vendor and if found,\n"
-            "                                           recovery, tos\n"
+            "                                           recovery\n"
             "  flash <partition> [ <filename> ]         write a file to a flash partition\n"
             "  erase <partition>                        erase a flash partition\n"
             "  format[:[<fs type>][:[<size>]] <partition> format a flash partition.\n"
             "                                           Can override the fs type and/or\n"
             "                                           size the bootloader reports.\n"
             "  getvar <variable>                        display a bootloader variable\n"
-            "  boot <kernel> [ <ramdisk> [ <second> ] ] download and boot kernel\n"
-            "  flash:raw boot <kernel> [ <ramdisk> [ <second> ] ] create bootimage and \n"
-            "                                           flash it\n"
+            "  boot <kernel> [ <ramdisk> ]              download and boot kernel\n"
+            "  flash:raw boot <kernel> [ <ramdisk> ]    create bootimage and flash it\n"
             "  devices                                  list all connected devices\n"
             "  continue                                 continue with autoboot\n"
             "  reboot                                   reboot device normally\n"
@@ -328,11 +324,10 @@
 }
 
 void *load_bootable_image(const char *kernel, const char *ramdisk,
-                          const char *secondstage, unsigned *sz,
-                          const char *cmdline)
+                          unsigned *sz, const char *cmdline)
 {
-    void *kdata = 0, *rdata = 0, *sdata = 0;
-    unsigned ksize = 0, rsize = 0, ssize = 0;
+    void *kdata = 0, *rdata = 0;
+    unsigned ksize = 0, rsize = 0;
     void *bdata;
     unsigned bsize;
 
@@ -368,18 +363,10 @@
         }
     }
 
-    if (secondstage) {
-        sdata = load_file(secondstage, &ssize);
-        if(sdata == 0) {
-            fprintf(stderr,"cannot load '%s': %s\n", secondstage, strerror(errno));
-            return  0;
-        }
-    }
-
     fprintf(stderr,"creating boot image...\n");
     bdata = mkbootimg(kdata, ksize, kernel_offset,
                       rdata, rsize, ramdisk_offset,
-                      sdata, ssize, second_offset,
+                      0, 0, second_offset,
                       page_size, base_addr, tags_offset, &bsize);
     if(bdata == 0) {
         fprintf(stderr,"failed to create boot.img\n");
@@ -1157,7 +1144,6 @@
         } else if(!strcmp(*argv, "boot")) {
             char *kname = 0;
             char *rname = 0;
-            char *sname = 0;
             skip(1);
             if (argc > 0) {
                 kname = argv[0];
@@ -1167,11 +1153,7 @@
                 rname = argv[0];
                 skip(1);
             }
-            if (argc > 0) {
-                sname = argv[0];
-                skip(1);
-            }
-            data = load_bootable_image(kname, rname, sname, &sz, cmdline);
+            data = load_bootable_image(kname, rname, &sz, cmdline);
             if (data == 0) return 1;
             fb_queue_download("boot.img", data, sz);
             fb_queue_command("boot", "booting");
@@ -1195,18 +1177,14 @@
             char *pname = argv[1];
             char *kname = argv[2];
             char *rname = 0;
-            char *sname = 0;
             require(3);
-            skip(3);
-            if (argc > 0) {
-                rname = argv[0];
-                skip(1);
+            if(argc > 3) {
+                rname = argv[3];
+                skip(4);
+            } else {
+                skip(3);
             }
-            if (argc > 0) {
-                sname = argv[0];
-                skip(1);
-            }
-            data = load_bootable_image(kname, rname, sname, &sz, cmdline);
+            data = load_bootable_image(kname, rname, &sz, cmdline);
             if (data == 0) die("cannot load bootable image");
             fb_queue_flash(pname, data, sz);
         } else if(!strcmp(*argv, "flashall")) {
diff --git a/init/builtins.c b/init/builtins.c
index de83c93..8dbaab7 100644
--- a/init/builtins.c
+++ b/init/builtins.c
@@ -515,7 +515,12 @@
     pid = fork();
     if (pid > 0) {
         /* Parent.  Wait for the child to return */
-        waitpid(pid, &status, 0);
+        int wp_ret = TEMP_FAILURE_RETRY(waitpid(pid, &status, 0));
+        if (wp_ret < 0) {
+            /* Unexpected error code. We will continue anyway. */
+            NOTICE("waitpid failed rc=%d, errno=%d\n", wp_ret, errno);
+        }
+
         if (WIFEXITED(status)) {
             ret = WEXITSTATUS(status);
         } else {
diff --git a/libcutils/Android.mk b/libcutils/Android.mk
index 5a4be93..469cdcf 100644
--- a/libcutils/Android.mk
+++ b/libcutils/Android.mk
@@ -140,8 +140,8 @@
         arch-x86/android_memset32.S \
 
 LOCAL_SRC_FILES_x86_64 += \
-        arch-x86_64/android_memset16_SSE2-atom.S \
-        arch-x86_64/android_memset32_SSE2-atom.S \
+        arch-x86_64/android_memset16.S \
+        arch-x86_64/android_memset32.S \
 
 LOCAL_CFLAGS_arm += -DHAVE_MEMSET16 -DHAVE_MEMSET32
 LOCAL_CFLAGS_arm64 += -DHAVE_MEMSET16 -DHAVE_MEMSET32
diff --git a/libcutils/arch-x86/android_memset16.S b/libcutils/arch-x86/android_memset16.S
old mode 100644
new mode 100755
index f8b79bd..cb2ff14
--- a/libcutils/arch-x86/android_memset16.S
+++ b/libcutils/arch-x86/android_memset16.S
@@ -13,13 +13,707 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
-# include "cache_wrapper.S"
-# undef __i686
-# define USE_AS_ANDROID
-# define sse2_memset16_atom android_memset16
-# include "sse2-memset16-atom.S"
+#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)
diff --git a/libcutils/arch-x86/android_memset32.S b/libcutils/arch-x86/android_memset32.S
old mode 100644
new mode 100755
index 6249fce..f4326dc
--- a/libcutils/arch-x86/android_memset32.S
+++ b/libcutils/arch-x86/android_memset32.S
@@ -13,13 +13,498 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
-# include "cache_wrapper.S"
-# undef __i686
-# define USE_AS_ANDROID
-# define sse2_memset32_atom android_memset32
-# include "sse2-memset32-atom.S"
+#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)
diff --git a/libcutils/arch-x86/cache_wrapper.S b/libcutils/arch-x86/cache.h
similarity index 95%
rename from libcutils/arch-x86/cache_wrapper.S
rename to libcutils/arch-x86/cache.h
index 9eee25c..1c22fea 100644
--- a/libcutils/arch-x86/cache_wrapper.S
+++ b/libcutils/arch-x86/cache.h
@@ -13,9 +13,6 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
 #if defined(__slm__)
 /* Values are optimized for Silvermont */
diff --git a/libcutils/arch-x86/sse2-memset16-atom.S b/libcutils/arch-x86/sse2-memset16-atom.S
deleted file mode 100755
index c2a762b..0000000
--- a/libcutils/arch-x86/sse2-memset16-atom.S
+++ /dev/null
@@ -1,722 +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.
- */
-/*
- * Contributed by: Intel Corporation
- */
-
-#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
-#else
-# define DEST		PARMS
-# define CHR		DEST+4
-# define LEN		CHR+4
-#endif
-
-#if 1
-# define SETRTNVAL
-#else
-# 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	__i686.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.__i686.get_pc_thunk.bx,"ax",@progbits
-	.globl	__i686.get_pc_thunk.bx
-	.hidden	__i686.get_pc_thunk.bx
-	ALIGN (4)
-	.type	__i686.get_pc_thunk.bx,@function
-__i686.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 (sse2_memset16_atom)
-	ENTRANCE
-
-	movl	LEN(%esp), %ecx
-#ifdef USE_AS_ANDROID
-	shr	$1, %ecx
-#endif
-#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	__i686.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	__i686.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 (sse2_memset16_atom)
diff --git a/libcutils/arch-x86/sse2-memset32-atom.S b/libcutils/arch-x86/sse2-memset32-atom.S
deleted file mode 100755
index 05eb64f..0000000
--- a/libcutils/arch-x86/sse2-memset32-atom.S
+++ /dev/null
@@ -1,513 +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.
- */
-/*
- * Contributed by: Intel Corporation
- */
-
-#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
-#else
-# define DEST		PARMS
-# define DWDS		DEST+4
-# define LEN		DWDS+4
-#endif
-
-#ifdef USE_AS_WMEMSET32
-# define SETRTNVAL	movl DEST(%esp), %eax
-#else
-# define SETRTNVAL
-#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	__i686.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.__i686.get_pc_thunk.bx,"ax",@progbits
-	.globl	__i686.get_pc_thunk.bx
-	.hidden	__i686.get_pc_thunk.bx
-	ALIGN (4)
-	.type	__i686.get_pc_thunk.bx,@function
-__i686.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 (sse2_memset32_atom)
-	ENTRANCE
-
-	movl	LEN(%esp), %ecx
-#ifdef USE_AS_ANDROID
-	shr     $2, %ecx
-#endif
-#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	__i686.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	__i686.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 (sse2_memset32_atom)
diff --git a/libcutils/arch-x86_64/android_memset16_SSE2-atom.S b/libcutils/arch-x86_64/android_memset16.S
similarity index 98%
rename from libcutils/arch-x86_64/android_memset16_SSE2-atom.S
rename to libcutils/arch-x86_64/android_memset16.S
index 48a10ed..cb6d4a3 100644
--- a/libcutils/arch-x86_64/android_memset16_SSE2-atom.S
+++ b/libcutils/arch-x86_64/android_memset16.S
@@ -13,12 +13,13 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
 #include "cache.h"
 
+#ifndef MEMSET
+# define MEMSET		android_memset16
+#endif
+
 #ifndef L
 # define L(label)	.L##label
 #endif
@@ -63,7 +64,7 @@
 
 	.section .text.sse2,"ax",@progbits
 	ALIGN (4)
-ENTRY (android_memset16)	// Address in rdi
+ENTRY (MEMSET)	// Address in rdi
 	shr    $1, %rdx			// Count in rdx
 	movzwl %si, %ecx
 	/* Fill the whole ECX with pattern.  */
@@ -561,4 +562,4 @@
 	movw   %cx, -2(%rdi)
 	ret
 
-END (android_memset16)
+END (MEMSET)
diff --git a/libcutils/arch-x86_64/android_memset32_SSE2-atom.S b/libcutils/arch-x86_64/android_memset32.S
similarity index 98%
rename from libcutils/arch-x86_64/android_memset32_SSE2-atom.S
rename to libcutils/arch-x86_64/android_memset32.S
index 4bdea8e..1514aa2 100644
--- a/libcutils/arch-x86_64/android_memset32_SSE2-atom.S
+++ b/libcutils/arch-x86_64/android_memset32.S
@@ -13,12 +13,13 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
 #include "cache.h"
 
+#ifndef MEMSET
+# define MEMSET		android_memset32
+#endif
+
 #ifndef L
 # define L(label)	.L##label
 #endif
@@ -63,7 +64,7 @@
 
 	.section .text.sse2,"ax",@progbits
 	ALIGN (4)
-ENTRY (android_memset32)	// Address in rdi
+ENTRY (MEMSET)	// Address in rdi
 	shr    $2, %rdx			// Count in rdx
 	movl   %esi, %ecx		// Pattern in ecx
 
@@ -369,4 +370,4 @@
 	movl	%ecx, -4(%rdi)
 	ret
 
-END (android_memset32)
+END (MEMSET)
diff --git a/libcutils/arch-x86_64/cache.h b/libcutils/arch-x86_64/cache.h
index ab5dd2f..f144309 100644
--- a/libcutils/arch-x86_64/cache.h
+++ b/libcutils/arch-x86_64/cache.h
@@ -13,19 +13,10 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-/*
- * Contributed by: Intel Corporation
- */
 
-#if defined(__slm__)
 /* Values are optimized for Silvermont */
 #define SHARED_CACHE_SIZE	(1024*1024)			/* Silvermont L2 Cache */
 #define DATA_CACHE_SIZE		(24*1024)			/* Silvermont L1 Data Cache */
-#else
-/* Values are optimized for Atom */
-#define SHARED_CACHE_SIZE	(512*1024)			/* Atom L2 Cache */
-#define DATA_CACHE_SIZE		(24*1024)			/* Atom L1 Data Cache */
-#endif
 
 #define SHARED_CACHE_SIZE_HALF	(SHARED_CACHE_SIZE / 2)
 #define DATA_CACHE_SIZE_HALF	(DATA_CACHE_SIZE / 2)
diff --git a/libnativebridge/native_bridge.cc b/libnativebridge/native_bridge.cc
index fa80e1e..6602d3f 100644
--- a/libnativebridge/native_bridge.cc
+++ b/libnativebridge/native_bridge.cc
@@ -29,7 +29,6 @@
 enum class NativeBridgeState {
   kNotSetup,                        // Initial state.
   kOpened,                          // After successful dlopen.
-                                    // Temporary meaning: string copied. TODO: remove. b/17440362
   kInitialized,                     // After successful initialization.
   kClosed                           // Closed or errors.
 };
@@ -61,9 +60,6 @@
 // Whether we had an error at some point.
 static bool had_error = false;
 
-// Native bridge filename. TODO: Temporary, remove. b/17440362
-static const char* native_bridge_filename;
-
 // Handle of the loaded library.
 static void* native_bridge_handle = nullptr;
 // Pointer to the callbacks. Available as soon as LoadNativeBridge succeeds, but only initialized
@@ -135,32 +131,28 @@
       state = NativeBridgeState::kClosed;
       had_error = true;
     } else {
-      // Save the name. TODO: Remove this, return to old flow. b/17440362
-      native_bridge_filename = nb_library_filename;
-      runtime_callbacks = runtime_cbs;
-      state = NativeBridgeState::kOpened;
-//       // Try to open the library.
-//       void* handle = dlopen(nb_library_filename, RTLD_LAZY);
-//       if (handle != nullptr) {
-//         callbacks = reinterpret_cast<NativeBridgeCallbacks*>(dlsym(handle,
-//                                                                    kNativeBridgeInterfaceSymbol));
-//         if (callbacks != nullptr) {
-//           // Store the handle for later.
-//           native_bridge_handle = handle;
-//         } else {
-//           dlclose(handle);
-//         }
-//       }
-//
-//       // Two failure conditions: could not find library (dlopen failed), or could not find native
-//       // bridge interface (dlsym failed). Both are an error and close the native bridge.
-//       if (callbacks == nullptr) {
-//         had_error = true;
-//         state = NativeBridgeState::kClosed;
-//       } else {
-//         runtime_callbacks = runtime_cbs;
-//         state = NativeBridgeState::kOpened;
-//       }
+      // Try to open the library.
+      void* handle = dlopen(nb_library_filename, RTLD_LAZY);
+      if (handle != nullptr) {
+        callbacks = reinterpret_cast<NativeBridgeCallbacks*>(dlsym(handle,
+                                                                   kNativeBridgeInterfaceSymbol));
+        if (callbacks != nullptr) {
+          // Store the handle for later.
+          native_bridge_handle = handle;
+        } else {
+          dlclose(handle);
+        }
+      }
+
+      // Two failure conditions: could not find library (dlopen failed), or could not find native
+      // bridge interface (dlsym failed). Both are an error and close the native bridge.
+      if (callbacks == nullptr) {
+        had_error = true;
+        state = NativeBridgeState::kClosed;
+      } else {
+        runtime_callbacks = runtime_cbs;
+        state = NativeBridgeState::kOpened;
+      }
     }
     return state == NativeBridgeState::kOpened;
   }
@@ -171,38 +163,15 @@
   // point we are not multi-threaded, so we do not need locking here.
 
   if (state == NativeBridgeState::kOpened) {
-    // Open and initialize. TODO: Temporary, remove. b/17440362
-    void* handle = dlopen(native_bridge_filename, RTLD_LAZY);
-    if (handle != nullptr) {
-      callbacks = reinterpret_cast<NativeBridgeCallbacks*>(dlsym(handle,
-                                                                 kNativeBridgeInterfaceSymbol));
-      if (callbacks != nullptr) {
-        if (callbacks->initialize(runtime_callbacks)) {
-          state = NativeBridgeState::kInitialized;
-          native_bridge_handle = handle;
-        } else {
-          callbacks = nullptr;
-        }
-      }
-
-      if (callbacks == nullptr) {
-        state = NativeBridgeState::kClosed;
-        had_error = true;
-        dlclose(handle);
-      }
+    // Try to initialize.
+    if (callbacks->initialize(runtime_callbacks)) {
+      state = NativeBridgeState::kInitialized;
     } else {
-      state = NativeBridgeState::kClosed;
+      // Unload the library.
+      dlclose(native_bridge_handle);
       had_error = true;
+      state = NativeBridgeState::kClosed;
     }
-//     // Try to initialize.
-//     if (callbacks->initialize(runtime_callbacks)) {
-//       state = NativeBridgeState::kInitialized;
-//     } else {
-//       // Unload the library.
-//       dlclose(native_bridge_handle);
-//       had_error = true;
-//       state = NativeBridgeState::kClosed;
-//     }
   } else {
     had_error = true;
     state = NativeBridgeState::kClosed;
@@ -216,7 +185,7 @@
   // point we are not multi-threaded, so we do not need locking here.
 
   switch(state) {
-    // case NativeBridgeState::kOpened:  // TODO: Re-add this. b/17440362
+    case NativeBridgeState::kOpened:
     case NativeBridgeState::kInitialized:
       // Unload.
       dlclose(native_bridge_handle);
@@ -227,7 +196,6 @@
       had_error = true;
       break;
 
-    case NativeBridgeState::kOpened:  // TODO: Remove this. b/17440362
     case NativeBridgeState::kClosed:
       // Ignore.
       break;
diff --git a/libnativebridge/tests/ValidNameNativeBridge_test.cpp b/libnativebridge/tests/ValidNameNativeBridge_test.cpp
index b28f5b2..690be4a 100644
--- a/libnativebridge/tests/ValidNameNativeBridge_test.cpp
+++ b/libnativebridge/tests/ValidNameNativeBridge_test.cpp
@@ -27,12 +27,9 @@
     // Now check what happens on LoadNativeBridge.
     EXPECT_EQ(false, NativeBridgeError());
     LoadNativeBridge(kTestName, nullptr);
-    // TODO: Remove this call. b/17440362
-    InitializeNativeBridge();
     // This will lead to an error as the library doesn't exist.
     EXPECT_EQ(true, NativeBridgeError());
-    // TODO: Test again. b/17440362
-//     EXPECT_EQ(false, NativeBridgeAvailable());
+    EXPECT_EQ(false, NativeBridgeAvailable());
 }
 
 }  // namespace android
diff --git a/logcat/tests/logcat_test.cpp b/logcat/tests/logcat_test.cpp
index 85756d5..321eaeb 100644
--- a/logcat/tests/logcat_test.cpp
+++ b/logcat/tests/logcat_test.cpp
@@ -243,7 +243,7 @@
 
     FILE *fp;
     ASSERT_TRUE(NULL != (fp = popen(
-      "logcat -b events -t 100 2>/dev/null",
+      "logcat -v brief -b events -t 100 2>/dev/null",
       "r")));
 
     char buffer[5120];
@@ -275,7 +275,7 @@
 
     // NB: crash log only available in user space
     ASSERT_TRUE(NULL != (fp = popen(
-      "logcat -b radio -b events -b system -b main -g 2>/dev/null",
+      "logcat -v brief -b radio -b events -b system -b main -g 2>/dev/null",
       "r")));
 
     char buffer[5120];
@@ -364,7 +364,7 @@
 
     ASSERT_TRUE(NULL != (fp = popen(
       "( trap exit HUP QUIT INT PIPE KILL ; sleep 6; echo DONE )&"
-      " logcat -b events 2>&1",
+      " logcat -v brief -b events 2>&1",
       "r")));
 
     char buffer[5120];
@@ -433,7 +433,7 @@
 
     ASSERT_TRUE(NULL != (fp = popen(
       "( trap exit HUP QUIT INT PIPE KILL ; sleep 6; echo DONE )&"
-      " logcat -b events -T 5 2>&1",
+      " logcat -v brief -b events -T 5 2>&1",
       "r")));
 
     char buffer[5120];
@@ -503,10 +503,14 @@
             int count = 0;
 
             while (fgets(buffer, sizeof(buffer), fp)) {
-                static const char match[] = "4 log.txt";
+                static const char match_1[] = "4 log.txt";
+                static const char match_2[] = "8 log.txt";
+                static const char match_3[] = "16 log.txt";
                 static const char total[] = "total ";
 
-                if (!strncmp(buffer, match, sizeof(match) - 1)) {
+                if (!strncmp(buffer, match_1, sizeof(match_1) - 1)
+                 || !strncmp(buffer, match_2, sizeof(match_2) - 1)
+                 || !strncmp(buffer, match_3, sizeof(match_3) - 1)) {
                     ++count;
                 } else if (strncmp(buffer, total, sizeof(total) - 1)) {
                     fprintf(stderr, "WARNING: Parse error: %s", buffer);
@@ -542,7 +546,7 @@
     ASSERT_TRUE(NULL != (fp = popen(
       "( trap exit HUP QUIT INT PIPE KILL ; sleep 6; echo DONE )&"
       " logcat -b events -c 2>&1 ;"
-      " logcat -b events 2>&1",
+      " logcat -v brief -b events 2>&1",
       "r")));
 
     char buffer[5120];
diff --git a/rootdir/init.rc b/rootdir/init.rc
index 681324a..09f7297 100644
--- a/rootdir/init.rc
+++ b/rootdir/init.rc
@@ -525,6 +525,7 @@
     onrestart restart zygote
     onrestart restart media
     onrestart restart surfaceflinger
+    onrestart restart inputflinger
     onrestart restart drm
 
 service vold /system/bin/vold
@@ -558,6 +559,12 @@
     group graphics drmrpc
     onrestart restart zygote
 
+service inputflinger /system/bin/inputflinger
+    class main
+    user system
+    group input
+    onrestart restart zygote
+
 service drm /system/bin/drmserver
     class main
     user drm