Merge pull request #99 from stressapptest/new-asm

New asm

7 files changed, 116 insertions, 142 deletions

diff --git a/Android.bp b/Android.bp
new file mode 100644
index 0000000..0ad0558
--- /dev/null
+++ b/Android.bp
@@ -0,0 +1,75 @@
+package {
+    default_applicable_licenses: ["external_stressapptest_license"],
+}
+// Added automatically by a large-scale-change that took the approach of
+// 'apply every license found to every target'. While this makes sure we respect
+// every license restriction, it may not be entirely correct.
+//
+// e.g. GPL in an MIT project might only apply to the contrib/ directory.
+//
+// Please consider splitting the single license below into multiple licenses,
+// taking care not to lose any license_kind information, and overriding the
+// default license using the 'licenses: [...]' property on targets as needed.
+//
+// For unused files, consider creating a 'filegroup' with "//visibility:private"
+// to attach the license to, and including a comment whether the files may be
+// used in the current project.
+//
+// large-scale-change included anything that looked like it might be a license
+// text as a license_text. e.g. LICENSE, NOTICE, COPYING etc.
+//
+// Please consider removing redundant or irrelevant files from 'license_text:'.
+// http://go/android-license-faq
+license {
+    name: "external_stressapptest_license",
+    visibility: [":__subpackages__"],
+    license_kinds: [
+        "SPDX-license-identifier-Apache-2.0",
+        "SPDX-license-identifier-MIT",
+        "legacy_notice",
+    ],
+    license_text: [
+        "COPYING",
+        "NOTICE",
+    ],
+}
+cc_binary {
+    name: "stressapptest",
+    srcs: [
+        "src/main.cc",
+        "src/adler32memcpy.cc",
+        "src/disk_blocks.cc",
+        "src/error_diag.cc",
+        "src/finelock_queue.cc",
+        "src/logger.cc",
+        "src/os.cc",
+        "src/os_factory.cc",
+        "src/pattern.cc",
+        "src/queue.cc",
+        "src/sat.cc",
+        "src/sat_factory.cc",
+        "src/worker.cc",
+    ],
+    arch: {
+        arm: { cflags: ["-DSTRESSAPPTEST_CPU_ARMV7A"] },
+        arm64: { cflags: ["-DSTRESSAPPTEST_CPU_AARCH64"] },
+        x86: {
+            enabled: false,
+        },
+        x86_64: {
+            enabled: false,
+        },
+    },
+    cflags: [
+        "-DHAVE_CONFIG_H",
+        "-DANDROID",
+        "-DNDEBUG",
+        "-UDEBUG",
+        "-DCHECKOPTS",
+        "-Wall",
+        "-Werror",
+        "-Wno-unused-parameter",
+        "-Wno-#warnings",
+        "-Wno-implicit-fallthrough",
+    ],
+}
diff --git a/Android.mk b/Android.mk
deleted file mode 100644
index db434cf..0000000
--- a/Android.mk
+++ /dev/null
@@ -1,29 +0,0 @@
-LOCAL_PATH := $(call my-dir)
-
-include $(CLEAR_VARS)
-
-LOCAL_SRC_FILES := \
-	src/main.cc \
-	src/adler32memcpy.cc \
-	src/disk_blocks.cc \
-	src/error_diag.cc \
-	src/finelock_queue.cc \
-	src/logger.cc \
-	src/os.cc \
-	src/os_factory.cc \
-	src/pattern.cc \
-	src/queue.cc \
-	src/sat.cc \
-	src/sat_factory.cc \
-	src/worker.cc
-
-# Build 64 bit by default
-# LOCAL_MULTILIB := 32
-LOCAL_MODULE:= stressapptest
-LOCAL_MODULE_TAGS := optional
-
-LOCAL_CFLAGS := -DHAVE_CONFIG_H -DANDROID -DNDEBUG -UDEBUG -DCHECKOPTS
-
-LOCAL_CPP_EXTENSION := .cc
-
-include $(BUILD_EXECUTABLE)
diff --git a/src/adler32memcpy.cc b/src/adler32memcpy.cc
index 264f51a..d67a362 100644
--- a/src/adler32memcpy.cc
+++ b/src/adler32memcpy.cc
@@ -403,7 +403,7 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
 #if defined(STRESSAPPTEST_CPU_AARCH64)
 #define src_r "x3"
 #define dst_r "x4"
-#define blocks_r "x5"
+#define blocks_r "w5"
 #define crc_r "x6"
 #else
 #define src_r "r3"
@@ -433,7 +433,7 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
       "mov " src_r ", %[src];\n"
       "mov " dst_r ", %[dst];\n"
       "mov " crc_r ", %[crc];\n"
-      "mov " blocks_r ", %[blocks];\n"
+      "mov " blocks_r ", %w[blocks];\n"
 
       // Loop over block count.
       "cmp " blocks_r ", #0;\n"   // Compare counter to zero.
@@ -448,8 +448,8 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
       "prfm pldl1strm, [" src_r ", #256];\n"
 
       // Init checksum
-      "ld1.2d {v0}, [" crc_r "];\n"
-      "dup.2d v1, xzr;\n"
+      "ld1 {v0.2d}, [" crc_r "];\n"
+      "dup v1.2d, xzr;\n"
 
       // Start of the loop which copies 64 bytes from source to dst each time.
       "TOP:\n"
@@ -462,15 +462,15 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
       // We are using 2 words out of 4 words in each qX register,
       // word index 0 and word index 2. We'll swizzle them in a bit.
       // Copy it.
-      "ld1.2d {v8, v9, v10, v11}, [" src_r "], #64;\n"
-      "st1.2d {v8, v9, v10, v11}, [" dst_r "], #64;\n"
+      "ld1 {v8.2d, v9.2d, v10.2d, v11.2d}, [" src_r "], #64;\n"
+      "st1 {v8.2d, v9.2d, v10.2d, v11.2d}, [" dst_r "], #64;\n"
 
       // Arrange it.
-      "dup.4s v12, wzr; \n"
-      "dup.4s v13, wzr; \n"
-      "dup.4s v14, wzr; \n"
-      "dup.4s v15, wzr; \n"
-      "dup.4s v16, wzr; \n"
+      "dup v12.4s, wzr; \n"
+      "dup v13.4s, wzr; \n"
+      "dup v14.4s, wzr; \n"
+      "dup v15.4s, wzr; \n"
+      "dup v16.4s, wzr; \n"
       // This exchenges words 1,3 in the filled registers with
       // words 0,2 in the empty registers.
       "trn1 v12.4s, v8.4s, v12.4s; \n"
@@ -487,23 +487,23 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
       // Overflow can occur only if there are more
       // than 2^16 additions => more than 2^17 words => more than 2^19 bytes so
       // if size_in_bytes > 2^19 than overflow occurs.
-      "add.2d  v0, v0, v12;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v8;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v13;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v9;\n"
-      "add.2d  v1, v1, v0;\n"
-
-      "add.2d  v0, v0, v14;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v10;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v15;\n"
-      "add.2d  v1, v1, v0;\n"
-      "add.2d  v0, v0, v11;\n"
-      "add.2d  v1, v1, v0;\n"
+      "add v0.2d, v0.2d, v12.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v8.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v13.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v9.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+
+      "add v0.2d, v0.2d, v14.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v10.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v15.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
+      "add v0.2d, v0.2d, v11.2d;\n"
+      "add v1.2d, v1.2d, v0.2d;\n"
 
       // Increment counter and loop.
       "sub " blocks_r ", " blocks_r ", #1;\n"
@@ -516,7 +516,7 @@ bool AdlerMemcpyAsm(uint64 *dstmem64, uint64 *srcmem64,
       // 64 bit numbers and have to be converted to 64 bit numbers)
       // seems like Adler128 (since size of each part is 4 byte rather than
       // 1 byte).
-      "st1.2d {v0, v1}, [" crc_r "];   \n"
+      "st1 {v0.2d, v1.2d}, [" crc_r "];   \n"
 
       // Output registers.
       :
diff --git a/src/os.cc b/src/os.cc
index 5b311a8..1928e0a 100644
--- a/src/os.cc
+++ b/src/os.cc
@@ -212,7 +212,6 @@ void OsLayer::GetFeatures() {
   // TODO(nsanders): add detect from /proc/cpuinfo or /proc/self/auxv.
   // For now assume neon and don't run -W if you don't have it.
   has_vector_ = true; // NEON.
-#warning "Unsupported CPU type ARMV7A: unable to determine feature set."
 #else
 #warning "Unsupported CPU type: unable to determine feature set."
 #endif
diff --git a/src/os.h b/src/os.h
index 06987f9..b189323 100644
--- a/src/os.h
+++ b/src/os.h
@@ -355,6 +355,7 @@ class OsLayer {
   virtual bool AdlerMemcpyWarm(uint64 *dstmem, uint64 *srcmem,
                                unsigned int size_in_bytes,
                                AdlerChecksum *checksum);
+  bool has_vector() const { return has_vector_; }
 
   // Store a callback to use to print
   // app-specific info about the last error location.
diff --git a/src/stressapptest_config_android.h b/src/stressapptest_config_android.h
index c3bf78f..3a6fcc8 100644
--- a/src/stressapptest_config_android.h
+++ b/src/stressapptest_config_android.h
@@ -1,222 +1,152 @@
 /* src/stressapptest_config.h.  Generated from stressapptest_config.h.in by configure.  */
 /* src/stressapptest_config.h.in.  Generated from configure.ac by autoheader.  */
-
 /* Define to 1 if the `closedir' function returns void instead of `int'. */
 #define CLOSEDIR_VOID 1
-
 /* Define to 1 if you have the <arpa/inet.h> header file. */
 #define HAVE_ARPA_INET_H 1
-
 /* Define to 1 if you have the declaration of `strerror_r', and to 0 if you
    don't. */
 #define HAVE_DECL_STRERROR_R 1
-
 /* Define to 1 if you have the <dirent.h> header file, and it defines `DIR'.
    */
 #define HAVE_DIRENT_H 1
-
 /* Define to 1 if you don't have `vprintf' but do have `_doprnt.' */
 /* #undef HAVE_DOPRNT */
-
 /* Define to 1 if you have the <fcntl.h> header file. */
 #define HAVE_FCNTL_H 1
-
 /* Define to 1 if you have the `ftruncate' function. */
 #define HAVE_FTRUNCATE 1
-
 /* Define to 1 if you have the `gettimeofday' function. */
 #define HAVE_GETTIMEOFDAY 1
-
 /* Define to 1 if you have the <inttypes.h> header file. */
 #define HAVE_INTTYPES_H 1
-
 /* Define to 1 if you have the <libaio.h> header file. */
 /* #undef HAVE_LIBAIO_H */
-
 /* Define to 1 if you have the <memory.h> header file. */
 #define HAVE_MEMORY_H 1
-
 /* Define to 1 if you have the `memset' function. */
 #define HAVE_MEMSET 1
-
 /* Define to 1 if you have the `mmap64' function. */
 /* #undef HAVE_MMAP64 */
-
 /* Define to 1 if you have the `munmap' function. */
 #define HAVE_MUNMAP 1
-
 /* Define to 1 if you have the <ndir.h> header file, and it defines `DIR'. */
 /* #undef HAVE_NDIR_H */
-
 /* Define to 1 if you have the <netdb.h> header file. */
 #define HAVE_NETDB_H 1
-
 /* Define to 1 if you have the `posix_memalign' function. */
 #define HAVE_POSIX_MEMALIGN 1
-
+/* Define to 1 if the system has `pthread_barrier'. */
+#define HAVE_PTHREAD_BARRIERS 1
 /* Define to 1 if you have the <pthread.h> header file. */
 #define HAVE_PTHREAD_H 1
-
 /* Define to 1 if you have the `rand_r' function. */
 /* #undef HAVE_RAND_R */
-
 /* Define to 1 if you have the `sched_getaffinity' function. */
 #define HAVE_SCHED_GETAFFINITY 1
-
 /* Define to 1 if you have the `select' function. */
 #define HAVE_SELECT 1
-
 /* Define to 1 if you have the `socket' function. */
 #define HAVE_SOCKET 1
-
 /* Define to 1 if stdbool.h conforms to C99. */
-/* #undef HAVE_STDBOOL_H */
-
+#define HAVE_STDBOOL_H 1
 /* Define to 1 if you have the <stdint.h> header file. */
 #define HAVE_STDINT_H 1
-
 /* Define to 1 if you have the <stdlib.h> header file. */
 #define HAVE_STDLIB_H 1
-
 /* Define to 1 if you have the `strerror_r' function. */
 #define HAVE_STRERROR_R 1
-
 /* Define to 1 if you have the <strings.h> header file. */
 #define HAVE_STRINGS_H 1
-
 /* Define to 1 if you have the <string.h> header file. */
 #define HAVE_STRING_H 1
-
 /* Define to 1 if you have the `strtol' function. */
 #define HAVE_STRTOL 1
-
 /* Define to 1 if you have the `strtoull' function. */
 #define HAVE_STRTOULL 1
-
 /* Define to 1 if you have the <sys/dir.h> header file, and it defines `DIR'.
    */
 /* #undef HAVE_SYS_DIR_H */
-
 /* Define to 1 if you have the <sys/ioctl.h> header file. */
 #define HAVE_SYS_IOCTL_H 1
-
 /* Define to 1 if you have the <sys/ndir.h> header file, and it defines `DIR'.
    */
 /* #undef HAVE_SYS_NDIR_H */
-
 /* Define to 1 if you have the <sys/select.h> header file. */
 #define HAVE_SYS_SELECT_H 1
-
 /* Define to 1 if you have the <sys/shm.h> header file. */
 /* #undef HAVE_SYS_SHM_H */
-
 /* Define to 1 if you have the <sys/socket.h> header file. */
 #define HAVE_SYS_SOCKET_H 1
-
 /* Define to 1 if you have the <sys/stat.h> header file. */
 #define HAVE_SYS_STAT_H 1
-
 /* Define to 1 if you have the <sys/time.h> header file. */
 #define HAVE_SYS_TIME_H 1
-
 /* Define to 1 if you have the <sys/types.h> header file. */
 #define HAVE_SYS_TYPES_H 1
-
 /* Define to 1 if you have the <unistd.h> header file. */
 #define HAVE_UNISTD_H 1
-
 /* Define to 1 if you have the `vprintf' function. */
 #define HAVE_VPRINTF 1
-
 /* Define to 1 if the system has the type `_Bool'. */
 #define HAVE__BOOL 1
-
 /* Name of package */
 #define PACKAGE "stressapptest"
-
 /* Define to the address where bug reports for this package should be sent. */
 #define PACKAGE_BUGREPORT "opensource@google.com"
-
 /* Define to the full name of this package. */
 #define PACKAGE_NAME "stressapptest"
-
 /* Define to the full name and version of this package. */
-#define PACKAGE_STRING "stressapptest 1.0.4_autoconf"
-
+#define PACKAGE_STRING "stressapptest 1.0.9_autoconf"
 /* Define to the one symbol short name of this package. */
 #define PACKAGE_TARNAME "stressapptest"
-
 /* Define to the home page for this package. */
 #define PACKAGE_URL ""
-
 /* Define to the version of this package. */
-#define PACKAGE_VERSION "1.0.4_autoconf"
-
+#define PACKAGE_VERSION "1.0.9_autoconf"
 /* Define as the return type of signal handlers (`int' or `void'). */
 #define RETSIGTYPE void
-
 /* Define to the type of arg 1 for `select'. */
 #define SELECT_TYPE_ARG1 int
-
 /* Define to the type of args 2, 3 and 4 for `select'. */
 #define SELECT_TYPE_ARG234 (fd_set *)
-
 /* Define to the type of arg 5 for `select'. */
 #define SELECT_TYPE_ARG5 (struct timeval *)
-
 /* Define to 1 if you have the ANSI C header files. */
 #define STDC_HEADERS 1
-
 /* Define to 1 if strerror_r returns char *. */
 /* #undef STRERROR_R_CHAR_P */
-
 /* Defined if the target CPU is armv7a */
-#define STRESSAPPTEST_CPU_ARMV7A /**/
-
+/* #define STRESSAPPTEST_CPU_ARMV7A */
 /* Defined if the target CPU is i686 */
 /* #undef STRESSAPPTEST_CPU_I686 */
-
-/* Defined if the target CPU is MIPS */
-/* #undef STRESSAPPTEST_CPU_MIPS */
-
 /* Defined if the target CPU is PowerPC */
 /* #undef STRESSAPPTEST_CPU_PPC */
-
 /* Defined if the target CPU is x86_64 */
 /* #undef STRESSAPPTEST_CPU_X86_64 */
-
 /* Defined if the target OS is BSD based */
 /* #undef STRESSAPPTEST_OS_BSD */
-
 /* Defined if the target OS is OSX */
 /* #undef STRESSAPPTEST_OS_DARWIN */
-
 /* Defined if the target OS is Linux */
 #define STRESSAPPTEST_OS_LINUX /**/
-
 /* Timestamp when ./configure was executed */
 #ifndef STRESSAPPTEST_TIMESTAMP
 #define STRESSAPPTEST_TIMESTAMP "Android version"
 #endif
-
 /* Define to 1 if you can safely include both <sys/time.h> and <time.h>. */
 #define TIME_WITH_SYS_TIME 1
-
 /* Version number of package */
-#define VERSION "1.0.4_autoconf"
-
+#define VERSION "1.0.9_autoconf"
 /* Define to empty if `const' does not conform to ANSI C. */
 /* #undef const */
-
 /* Define to `__inline__' or `__inline' if that's what the C compiler
    calls it, or to nothing if 'inline' is not supported under any name.  */
 #ifndef __cplusplus
 /* #undef inline */
 #endif
-
 /* Define to `int' if <sys/types.h> does not define. */
 /* #undef pid_t */
-
 /* Define to the equivalent of the C99 'restrict' keyword, or to
    nothing if this is not supported.  Do not define if restrict is
    supported directly.  */
@@ -230,17 +160,13 @@
 # define _Restrict
 # define __restrict__
 #endif
-
 /* Define to `unsigned int' if <sys/types.h> does not define. */
 /* #undef size_t */
-
 /* Define to `int' if <sys/types.h> does not define. */
 /* #undef ssize_t */
-
 /* Define to the type of an unsigned integer type of width exactly 16 bits if
    such a type exists and the standard includes do not define it. */
 /* #undef uint16_t */
-
 /* Define to empty if the keyword `volatile' does not work. Warning: valid
    code using `volatile' can become incorrect without. Disable with care. */
 /* #undef volatile */
diff --git a/src/worker.cc b/src/worker.cc
index 7f201ec..745a816 100644
--- a/src/worker.cc
+++ b/src/worker.cc
@@ -1525,8 +1525,10 @@ bool CopyThread::Work() {
   bool result = true;
   int64 loops = 0;
 
-  logprintf(9, "Log: Starting copy thread %d: cpu %s, mem %x\n",
-            thread_num_, cpuset_format(&cpu_mask_).c_str(), tag_);
+  logprintf(9, "Log: Starting copy thread %d: cpu %s, "
+            "mem %x, warm: %d, has_vector: %d\n",
+            thread_num_, cpuset_format(&cpu_mask_).c_str(), tag_,
+            sat_->warm(), os_->has_vector());
 
   while (IsReadyToRun()) {
     // Pop the needed pages.