am 93369c24: Add the \'hello-neon\' sample app to demonstrate cpufeatures and ARM Neon usage.

Merge commit '93369c247069b116f39c30530287809f5a190c22' into eclair-plus-aosp

* commit '93369c247069b116f39c30530287809f5a190c22':
  Add the 'hello-neon' sample app to demonstrate cpufeatures and ARM Neon usage.

ndk/apps/hello-gl2/project/jni/gl_code.cpp

@@ -42,12 +42,14 @@ static void checkGlError(const char* op) {
     }
 }
 
-static const char gVertexShader[] = "attribute vec4 vPosition;\n"
+static const char gVertexShader[] = 
+    "attribute vec4 vPosition;\n"
     "void main() {\n"
     "  gl_Position = vPosition;\n"
     "}\n";
 
-static const char gFragmentShader[] = "precision mediump float;\n"
+static const char gFragmentShader[] = 
+    "precision mediump float;\n"
     "void main() {\n"
     "  gl_FragColor = vec4(0.0, 1.0, 0.0, 1.0);\n"
     "}\n";

ndk/apps/hello-neon/Application.mk

@@ -0,0 +1,3 @@
+APP_PROJECT_PATH := $(call my-dir)/project
+APP_MODULES      := helloneon cpufeatures
+APP_ABI          := armeabi armeabi-v7a

ndk/apps/hello-neon/project/AndroidManifest.xml

@@ -0,0 +1,15 @@
+<?xml version="1.0" encoding="utf-8"?>
+<manifest xmlns:android="http://schemas.android.com/apk/res/android"
+      package="com.example.neon"
+      android:versionCode="1"
+      android:versionName="1.0">
+    <application android:label="@string/app_name">
+        <activity android:name=".HelloNeon"
+                  android:label="@string/app_name">
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN" />
+                <category android:name="android.intent.category.LAUNCHER" />
+            </intent-filter>
+        </activity>
+    </application>
+</manifest> 

ndk/apps/hello-neon/project/build.properties

@@ -0,0 +1,20 @@
+# This file is used to override default values used by the Ant build system.
+# 
+# This file must be checked in Version Control Systems, as it is
+# integral to the build system of your project.
+
+# This file is only used by the Ant script.
+
+# You can use this to override default values such as
+#  'source.dir' for the location of your java source folder and
+#  'out.dir' for the location of your output folder.
+
+# You can also use it define how the release builds are signed by declaring
+# the following properties:
+#  'key.store' for the location of your keystore and
+#  'key.alias' for the name of the key to use.
+# The password will be asked during the build when you use the 'release' target.
+
+# The name of your application package as defined in the manifest.
+# Used by the 'uninstall' rule.
+application.package=com.example.neon

ndk/apps/hello-neon/project/default.properties

@@ -0,0 +1,11 @@
+# This file is automatically generated by Android Tools.
+# Do not modify this file -- YOUR CHANGES WILL BE ERASED!
+# 
+# This file must be checked in Version Control Systems.
+# 
+# To customize properties used by the Ant build system use,
+# "build.properties", and override values to adapt the script to your
+# project structure.
+
+# Project target.
+target=android-3

ndk/apps/hello-neon/project/jni/Android.mk

@@ -0,0 +1,22 @@
+LOCAL_PATH := $(call my-dir)
+
+include $(CLEAR_VARS)
+
+LOCAL_MODULE := helloneon
+
+LOCAL_SRC_FILES := helloneon.c
+
+ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
+    LOCAL_CFLAGS := -DHAVE_NEON=1
+    LOCAL_SRC_FILES += helloneon-intrinsics.c.neon
+endif
+
+LOCAL_C_INCLUDES := sources/cpufeatures
+
+LOCAL_STATIC_LIBRARIES := cpufeatures
+
+LOCAL_LDLIBS := -llog
+
+include $(BUILD_SHARED_LIBRARY)
+
+include sources/cpufeatures/Android.mk

ndk/apps/hello-neon/project/jni/helloneon-intrinsics.c

@@ -0,0 +1,48 @@
+#include "helloneon-intrinsics.h"
+#include <arm_neon.h>
+
+/* this source file should only be compiled by Android.mk when targeting
+ * the armeabi-v7a ABI, and should be built in NEON mode
+ */
+void
+fir_filter_neon_intrinsics(short *output, const short* input, const short* kernel, int width, int kernelSize)
+{
+#if 1
+   int nn, offset = -kernelSize/2;
+
+   for (nn = 0; nn < width; nn++)
+   {
+        int mm, sum = 0;
+        int32x4_t sum_vec = vdupq_n_s32(0);
+        for(mm = 0; mm < kernelSize/4; mm++)
+        {
+            int16x4_t  kernel_vec = vld1_s16(kernel + mm*4);
+            int16x4_t  input_vec = vld1_s16(input + (nn+offset+mm*4));
+            sum_vec = vmlal_s16(sum_vec, kernel_vec, input_vec);
+        }
+
+        sum += vgetq_lane_s32(sum_vec, 0);
+        sum += vgetq_lane_s32(sum_vec, 1);
+        sum += vgetq_lane_s32(sum_vec, 2);
+        sum += vgetq_lane_s32(sum_vec, 3);
+
+        if(kernelSize & 3)
+        {
+            for(mm = kernelSize - (kernelSize & 3); mm < kernelSize; mm++)
+                sum += kernel[mm] * input[nn+offset+mm];
+        }
+
+        output[nn] = (short)((sum + 0x8000) >> 16);
+    }
+#else /* for comparison purposes only */
+    int nn, offset = -kernelSize/2;
+    for (nn = 0; nn < width; nn++) {
+        int sum = 0;
+        int mm;
+        for (mm = 0; mm < kernelSize; mm++) {
+            sum += kernel[mm]*input[nn+offset+mm];
+        }
+        output[n] = (short)((sum + 0x8000) >> 16);
+    }
+#endif
+}

ndk/apps/hello-neon/project/jni/helloneon-intrinsics.h

@@ -0,0 +1,6 @@
+#ifndef HELLONEON_INTRINSICS_H
+#define HELLONEON_INTRINSICS_H
+
+void fir_filter_neon_intrinsics(short *output, const short* input, const short* kernel, int width, int kernelSize);
+
+#endif /* HELLONEON_INTRINSICS_H */

ndk/apps/hello-neon/project/jni/helloneon.c

@@ -0,0 +1,147 @@
+#include <jni.h>
+#include <time.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <cpu-features.h>
+#include "helloneon-intrinsics.h"
+
+#define DEBUG 0
+
+#if DEBUG
+#include <android/log.h>
+#  define  D(x...)  __android_log_print(ANDROID_LOG_INFO,"helloneon",x)
+#else
+#  define  D(...)  do {} while (0)
+#endif
+
+/* return current time in milliseconds */
+static double
+now_ms(void)
+{
+    struct timespec res;
+    clock_gettime(CLOCK_REALTIME, &res);
+    return 1000.0*res.tv_sec + (double)res.tv_nsec/1e6;
+}
+
+
+/* this is a FIR filter implemented in C */
+static void
+fir_filter_c(short *output, const short* input, const short* kernel, int width, int kernelSize)
+{
+    int  offset = -kernelSize/2;
+    int  nn;
+    for (nn = 0; nn < width; nn++) {
+        int sum = 0;
+        int mm;
+        for (mm = 0; mm < kernelSize; mm++) {
+            sum += kernel[mm]*input[nn+offset+mm];
+        }
+        output[nn] = (short)((sum + 0x8000) >> 16);
+    }
+}
+
+#define  FIR_KERNEL_SIZE   32
+#define  FIR_OUTPUT_SIZE   2560
+#define  FIR_INPUT_SIZE    (FIR_OUTPUT_SIZE + FIR_KERNEL_SIZE)
+#define  FIR_ITERATIONS    600
+
+static const short  fir_kernel[FIR_KERNEL_SIZE] = { 
+    0x10, 0x20, 0x40, 0x70, 0x8c, 0xa2, 0xce, 0xf0, 0xe9, 0xce, 0xa2, 0x8c, 070, 0x40, 0x20, 0x10,
+    0x10, 0x20, 0x40, 0x70, 0x8c, 0xa2, 0xce, 0xf0, 0xe9, 0xce, 0xa2, 0x8c, 070, 0x40, 0x20, 0x10 };
+
+static short        fir_output[FIR_OUTPUT_SIZE];
+static short        fir_input_0[FIR_INPUT_SIZE];
+static const short* fir_input = fir_input_0 + (FIR_KERNEL_SIZE/2);
+static short        fir_output_expected[FIR_OUTPUT_SIZE];
+
+/* This is a trivial JNI example where we use a native method
+ * to return a new VM String. See the corresponding Java source
+ * file located at:
+ *
+ *   apps/samples/hello-neon/project/src/com/example/neon/HelloNeon.java
+ */
+jstring
+Java_com_example_neon_HelloNeon_stringFromJNI( JNIEnv* env,
+                                               jobject thiz )
+{
+    char*  str;
+    uint64_t features;
+    char buffer[512];
+    char tryNeon = 0;
+    double  t0, t1, time_c, time_neon;
+
+    /* setup FIR input - whatever */
+    {
+        int  nn;
+        for (nn = 0; nn < FIR_INPUT_SIZE; nn++) {
+            fir_input_0[nn] = (5*nn) & 255;
+        }
+        fir_filter_c(fir_output_expected, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE);
+    }
+
+    /* Benchmark small FIR filter loop - C version */
+    t0 = now_ms();
+    {
+        int  count = FIR_ITERATIONS;
+        for (; count > 0; count--) {
+            fir_filter_c(fir_output, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE);
+        }
+    }
+    t1 = now_ms();
+    time_c = t1 - t0;
+
+    asprintf(&str, "FIR Filter benchmark:\nC version          : %g ms\n", time_c);
+    strlcpy(buffer, str, sizeof buffer);
+    free(str);
+
+    strlcat(buffer, "Neon version   : ", sizeof buffer);
+
+    if (android_getCpuFamily() != ANDROID_CPU_FAMILY_ARM) {
+        strlcat(buffer, "Not an ARM CPU !\n", sizeof buffer);
+        goto EXIT;
+    }
+
+    features = android_getCpuFeatures();
+    if ((features & ANDROID_CPU_ARM_FEATURE_ARMv7) == 0) {
+        strlcat(buffer, "Not an ARMv7 CPU !\n", sizeof buffer);
+        goto EXIT;
+    }
+
+    /* HAVE_NEON is defined in Android.mk ! */
+#ifdef HAVE_NEON
+    if ((features & ANDROID_CPU_ARM_FEATURE_NEON) == 0) {
+        strlcat(buffer, "CPU doesn't support NEON !\n", sizeof buffer);
+        goto EXIT;
+    }
+
+    /* Benchmark small FIR filter loop - Neon version */
+    t0 = now_ms();
+    {
+        int  count = FIR_ITERATIONS;
+        for (; count > 0; count--) {
+            fir_filter_neon_intrinsics(fir_output, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE);
+        }
+    }
+    t1 = now_ms();
+    time_neon = t1 - t0;
+    asprintf(&str, "%g ms (x%g faster)\n", time_neon, time_c / (time_neon < 1e-6 ? 1. : time_neon));
+    strlcat(buffer, str, sizeof buffer);
+    free(str);
+
+    /* check the result, just in case */
+    {
+        int  nn, fails = 0;
+        for (nn = 0; nn < FIR_OUTPUT_SIZE; nn++) {
+            if (fir_output[nn] != fir_output_expected[nn]) {
+                if (++fails < 16)
+                    D("neon[%d] = %d expected %d", nn, fir_output[nn], fir_output_expected[nn]);
+            }
+        }
+        D("%d fails\n", fails);
+    }
+#else /* !HAVE_NEON */
+    strlcat(buffer, "Program not compiled with ARMv7 support !\n", sizeof buffer);
+#endif /* !HAVE_NEON */
+EXIT:
+    return (*env)->NewStringUTF(env, buffer);
+}

ndk/apps/hello-neon/project/res/values/strings.xml

@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <string name="app_name">HelloNeon</string>
+</resources>

ndk/apps/hello-neon/project/src/com/example/neon/HelloNeon.java

@@ -0,0 +1,37 @@
+package com.example.neon;
+
+import android.app.Activity;
+import android.os.Bundle;
+import android.widget.TextView;
+
+public class HelloNeon extends Activity
+{
+    /** Called when the activity is first created. */
+    @Override
+    public void onCreate(Bundle savedInstanceState)
+    {
+        super.onCreate(savedInstanceState);
+        /* Create a TextView and set its content.
+         * the text is retrieved by calling a native
+         * function.
+         */
+        TextView  tv = new TextView(this);
+        tv.setText( stringFromJNI() );
+        setContentView(tv);
+    }
+
+    /* A native method that is implemented by the
+     * 'helloneon' native library, which is packaged
+     * with this application.
+     */
+    public native String  stringFromJNI();
+
+    /* this is used to load the 'helloneon' library on application
+     * startup. The library has already been unpacked into
+     * /data/data/com.example.neon/lib/libhelloneon.so at
+     * installation time by the package manager.
+     */
+    static {
+        System.loadLibrary("helloneon");
+    }
+}

ndk/build/tools/cleanup-apps.sh

@@ -0,0 +1,12 @@
+#!/bin/sh
+#
+# This is used to cleanup the project directories before making a commit or
+# a clean release. This will get rid of auto-generated files in the
+# apps/<name>/project directories.
+#
+for projectPath in `find apps/*/project` ; do
+    rm -rf $projectPath/bin
+    rm -rf $projectPath/gen
+    rm -f  $projectPath/build.xml
+    rm -f  $projectPath/local.properties
+done

ndk/docs/CPU-ARM-NEON.TXT

@@ -116,3 +116,12 @@ that has the ANDROID_CPU_ARM_FEATURE_NEON flag set, as  in:
     }
 
     ...
+
+Sample code:
+------------
+
+Look at the source code for the "hello-neon" sample in this NDK for an example
+on how to use the 'cpufeatures' library and Neon intrinsics at the same time.
+
+This implements a tiny benchmark for a FIR filter loop using a C version, and
+a NEON-optimized one for devices that support it.