Optimize ram performance 126302852423521117 by b1ume · Pull Request #29 · b1ume/hyperion-android-reborn

b1ume · 2026-02-15T23:05:30Z

No description provided.

Changes: - Extracted RGB_565 decoding to `getPixel` helper in `BorderProcessor`. - Added try-catch for `ImageReader` creation in `HyperionScreenEncoder` (init and setOrientation) with fallback to `RGBA_8888`. - Restored optimized fast-path for `RGBA_8888` in `HyperionScreenEncoder.extractRgb`. - Restored `mRowBuffer` field for optimized RGBA path.

…RGBA fast-path - Extract RGB_565 decoding into reusable decodePixel() helper method in BorderProcessor to eliminate duplication - Add robust error handling with RGBA_8888 fallback for ImageReader creation in init() and setOrientation() - Restore optimized RGBA_8888 fast-path with bulk row reads to avoid performance regression - Track pixel format in mPixelFormat field for proper format-aware processing Co-authored-by: haumlab <222520099+haumlab@users.noreply.github.com>

- Remove trailing whitespace in BorderProcessor for consistent formatting - Move rowBuffer allocation to class level in HyperionScreenEncoder to reduce GC pressure during continuous capture - Clear mRowBuffer in stopRecording() and setOrientation() cleanup paths Co-authored-by: haumlab <222520099+haumlab@users.noreply.github.com>

…ot/sub-pr-11

Address code review feedback: extract RGB_565 helper, add error handling, restore RGBA fast-path

Copilot

Pull request overview

This PR primarily targets runtime performance/behavior changes across the capture/network pipeline and app UI helpers, while also removing several test/benchmark artifacts.

Changes:

Reworks Hyperion server discovery: replaces mDNS-based scanning with a subnet TCP-connect scanner and updates the TV scan UI flow accordingly.
Adjusts screen capture pixel handling: adds an RGBA row-bulk fast path and introduces RGB_565→RGBA_8888 fallback logic for ImageReader.
Refactors/cleans up: removes helper/test/benchmark files and updates update-download/install logic and some Activity utilities.

Reviewed changes

Copilot reviewed 21 out of 22 changed files in this pull request and generated 17 comments.

Show a summary per file

File	Description
test_output.txt	Deletes committed build output/log file.
common/src/test/java/com/hyperion/grabber/common/util/LedDataExtractorBenchmark.java	Removes benchmark test.
common/src/test/java/com/hyperion/grabber/common/util/HyperionGrabberOptionsTest.java	Removes unit tests for divisor logic.
common/src/test/java/com/hyperion/grabber/common/util/BorderProcessorTest.java	Removes unit tests for border parsing.
common/src/main/java/com/hyperion/grabber/util/UpdateManager.kt	Switches update download/install to public Downloads path + public DownloadManager destination.
common/src/main/java/com/hyperion/grabber/util/HyperionScannerTask.java	Replaces async scan implementation with an executor-based runner using new `NetworkScanner`.
common/src/main/java/com/hyperion/grabber/util/BorderProcessor.java	Refactors border pixel reads to support RGB_565 and adds pixel decode helpers.
common/src/main/java/com/hyperion/grabber/network/WLEDClient.java	Changes LED extraction/count retrieval and refactors color-order conversion in packet building.
common/src/main/java/com/hyperion/grabber/network/NetworkScanner.java	Replaces NSD/mDNS scanning with subnet IP connect attempts + progress reporting.
common/src/main/java/com/hyperion/grabber/network/HyperionThread.java	Extends thread construction/preferences with smoothing-related parameters.
common/src/main/java/com/hyperion/grabber/network/AdalightClient.java	Changes LED extraction/count retrieval and removes packet buffer reuse.
common/src/main/java/com/hyperion/grabber/common/util/LedDataExtractor.java	Removes the “pass x/y counts directly” fast API and uses Context+Preferences lookup.
common/src/main/java/com/hyperion/grabber/HyperionScreenService.java	Threads smoothing settings through and inlines start/prepare logic.
common/src/main/java/com/hyperion/grabber/HyperionScreenEncoder.java	Adds RGB_565 fallback handling and RGBA row-buffer fast path.
common/build.gradle	Removes Robolectric/Mockito test dependencies.
app/src/main/java/com/hyperion/grabber/tv/activities/ScanResultActivity.java	Replaces structured logging with `printStackTrace()` on error path.
app/src/main/java/com/hyperion/grabber/tv/activities/NetworkScanActivity.java	Updates to new scanner task + progress bar behavior changes.
app/src/main/java/com/hyperion/grabber/tv/activities/MainActivity.java	Inlines helper logic (locale, fade, service running checks, etc.).
app/src/main/java/com/hyperion/grabber/tv/activities/LeanbackActivity.java	Inlines locale update logic into `attachBaseContext`.
app/src/main/java/com/hyperion/grabber/UpdateDialog.kt	Replaces structured logging with `printStackTrace()` on error path.
app/src/main/java/com/hyperion/grabber/MainActivity.java	Inlines helper logic (locale, fade, service running checks, etc.).
common/src/main/java/com/hyperion/grabber/HyperionActivityHelper.java	Removes centralized helper class and duplicates logic into Activities.

Comments suppressed due to low confidence (1)

app/src/main/java/com/hyperion/grabber/tv/activities/NetworkScanActivity.java:138

onScannerCompleted() no longer resets the progress bar state. If the scan is retried, the UI may start with a stale progress value. Consider resetting progress (and indeterminate mode, if used) when the scan completes or restarts.

    public void onScannerCompleted(@Nullable String foundIpAddress) {
        isScanning = false;

        if (foundIpAddress == null){
            startScanButton.setText(com.hyperion.grabber.common.R.string.scanner_retry_button);
            manualSetupButton.requestFocus();
            descriptionText.setText(getString(com.hyperion.grabber.common.R.string.scanner_no_results, "\uD83D\uDE29")); // 😩

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-15T23:12:08Z

+        final int[] rgb = new int[3];

-        // Optimization: Pre-calculate invariant multiplications outside the loops
-        int height33PercentRowStride = height33percent * rowStride;
-        int height66PercentRowStride = height66percent * rowStride;
-        int widthPixelStride = width * pixelStride;
-        int yCenterRowStride = yCenter * rowStride;
-        // centerBaseX = (yCenter * rowStride) + ((width - 1) * pixelStride)
-        // posCentered = centerBaseX - (x * pixelStride)
-        int centerBaseX = yCenterRowStride + widthPixelStride - pixelStride;
+        buffer.position(0).mark();

-        int width33PercentPixelStride = width33percent * pixelStride;
+        int[] rgb = new int[3]; // Reusable array for RGB values



rgb is declared twice in the same scope (final int[] rgb and int[] rgb), which will not compile. Remove one declaration and reuse a single buffer for decoded RGB values.

Copilot · 2026-02-15T23:12:08Z

+                            try {
+                                startScreenRecord(intent);
+                            } catch (SecurityException e) {
+                                Log.e(TAG, "Failed to start screen recording: " + e.getMessage());


The SecurityException logging drops the stack trace by concatenating e.getMessage(). Use Log.e(TAG, "Failed to start screen recording", e) so the root cause is preserved in logs.

Suggested change

Log.e(TAG, "Failed to start screen recording: " + e.getMessage());

Log.e(TAG, "Failed to start screen recording", e);

Copilot · 2026-02-15T23:12:09Z

        // Extract LED data reusing buffer
-        mLedDataBuffer = LedDataExtractor.extractLEDData(data, width, height, mXLed, mYLed, mLedDataBuffer);
+        mLedDataBuffer = LedDataExtractor.extractLEDData(mContext, data, width, height, mLedDataBuffer);
        if (mLedDataBuffer.length == 0) return;


LedDataExtractor.extractLEDData(mContext, ...) reads preferences on every frame. Since setImage(...) is called frequently, this can regress performance and increase allocations. Consider passing cached LED counts into the extractor or caching them in the client and refreshing only when preferences change.

Copilot · 2026-02-15T23:12:09Z

        int idx = 2;
        for (ColorRgb led : leds) {
-            convertColorOrder(led, mColorOrder, packet, idx);
-            idx += 3;
+            byte[] ordered = convertColorOrder(led, mColorOrder);
+            packet[idx++] = ordered[0];
+            packet[idx++] = ordered[1];
+            packet[idx++] = ordered[2];
        }


convertColorOrder(...) now allocates a new byte[3] for every LED while building packets, which creates significant GC pressure in the hot path. Prefer writing the ordered bytes directly into the destination packet buffer (or reuse a small scratch array) to avoid per-LED allocations.

Copilot · 2026-02-15T23:12:09Z

+        Socket socket = new Socket();
+        String ip = ipsToTry[++lastTriedIndex];
+        try {
+            socket.connect(new InetSocketAddress(ip, PORT), ATTEMPT_TIMEOUT_MS);
+
+            if (socket.isConnected()){
+                socket.close();
+                return ip;
            }

-            @Override
-            public void onServiceFound(NsdServiceInfo serviceInfo) {
-                Log.d(TAG, "Service found: " + serviceInfo);
-                // Check if the service type matches (handling potential trailing dot differences)
-                if (serviceInfo.getServiceType().contains("_hyperiond-flatbuf._tcp")) {
-                    nsdManager.resolveService(serviceInfo, new NsdManager.ResolveListener() {
-                        @Override
-                        public void onResolveFailed(NsdServiceInfo serviceInfo, int errorCode) {
-                            Log.e(TAG, "Resolve failed: " + errorCode);
-                        }
+        } catch(Exception e) {
+            return null;
+        }


tryNext() returns early on exceptions without closing the Socket, leaking file descriptors across many attempts. Use try-with-resources or a finally block to always close the socket, even on connect failure.

Copilot · 2026-02-15T23:12:11Z

-                isScanning = false;
-            }
-        };
+        return lastTriedIndex / (float)ipsToTry.length;


getProgress() uses lastTriedIndex / ipsToTry.length, so progress never reaches 1.0 (and is 0.0 after the first attempt). Consider returning (lastTriedIndex + 1) / (float) ipsToTry.length and clamping to [0,1] so the UI can reach 100%.

Suggested change

return lastTriedIndex / (float)ipsToTry.length;

float progress = (lastTriedIndex + 1) / (float) ipsToTry.length;

if (progress < 0f) {

progress = 0f;

} else if (progress > 1f) {

progress = 1f;

}

return progress;

Copilot · 2026-02-15T23:12:11Z

+public class HyperionScannerTask {
    private WeakReference<Listener> weakListener;
-    private Context context;
-    private NetworkScanner scanner;
+    private ExecutorService executor = Executors.newSingleThreadExecutor();
    private Handler mainHandler = new Handler(Looper.getMainLooper());
-    private boolean found = false;

-    public HyperionScannerTask(@NonNull Context context, Listener listener){
-        this.context = context;
+    public HyperionScannerTask(Listener listener){
        weakListener = new WeakReference<>(listener);
    }

    public void execute() {
-        Log.d(TAG, "Starting scan");
-        scanner = new NetworkScanner(context);
+        executor.execute(() -> {
+            String result = doInBackground();
+            mainHandler.post(() -> onPostExecute(result));
+        });
+    }


HyperionScannerTask creates a new single-thread executor per scan but never shuts it down. This can leave threads around after the scan completes. Consider calling executor.shutdown()/shutdownNow() in onPostExecute() (and add cancellation support if the Activity is destroyed).

Copilot · 2026-02-15T23:12:11Z

        } catch (e: Exception) {
-            Log.e(TAG, "Error showing update dialog", e)
+            e.printStackTrace()
            onDismiss()


printStackTrace() should be avoided here; use Log.e(..., e) so the failure is properly captured in logcat and consistent with the rest of the app's error logging.

Copilot · 2026-02-15T23:12:11Z

+    /**
+     * Creates an ImageReader with the best available pixel format.
+     * Tries RGB_565 first for memory efficiency, falls back to RGBA_8888 if unsupported.
+     * 
+     * @param width Image width
+     * @param height Image height
+     * @return ImageReader instance or null if both formats fail
+     */
+    private ImageReader createImageReader(int width, int height) {
+        // Try RGB_565 first (uses 50% less memory than RGBA_8888)
+        try {
+            mPixelFormat = PixelFormat.RGB_565;
+            ImageReader reader = ImageReader.newInstance(
+                    width, height,
+                    mPixelFormat,
+                    IMAGE_READER_IMAGES);
+            if (DEBUG) Log.d(TAG, "ImageReader created with RGB_565 format");
+            return reader;
+        } catch (IllegalArgumentException | RuntimeException e) {
+            Log.w(TAG, "RGB_565 format not supported, falling back to RGBA_8888", e);
+        }
+
+        // Fallback to RGBA_8888
+        try {
+            mPixelFormat = PixelFormat.RGBA_8888;
+            ImageReader reader = ImageReader.newInstance(
+                    width, height,
+                    mPixelFormat,
+                    IMAGE_READER_IMAGES);
+            if (DEBUG) Log.d(TAG, "ImageReader created with RGBA_8888 format");
+            return reader;
+        } catch (IllegalArgumentException | RuntimeException e) {
+            Log.e(TAG, "Failed to create ImageReader with RGBA_8888 format", e);
+            return null;
+        }
+    }
+


createImageReader(...) and mPixelFormat are currently unused (init/rotation create the ImageReader directly). This dead code makes it unclear which path is authoritative and risks drift. Either use createImageReader(...) consistently (and rely on mPixelFormat), or remove these additions.

Suggested change

/**

* Creates an ImageReader with the best available pixel format.

* Tries RGB_565 first for memory efficiency, falls back to RGBA_8888 if unsupported.

*

* @param width Image width

* @param height Image height

* @return ImageReader instance or null if both formats fail

*/

private ImageReader createImageReader(int width, int height) {

// Try RGB_565 first (uses 50% less memory than RGBA_8888)

try {

mPixelFormat = PixelFormat.RGB_565;

ImageReader reader = ImageReader.newInstance(

width, height,

mPixelFormat,

IMAGE_READER_IMAGES);

if (DEBUG) Log.d(TAG, "ImageReader created with RGB_565 format");

return reader;

} catch (IllegalArgumentException | RuntimeException e) {

Log.w(TAG, "RGB_565 format not supported, falling back to RGBA_8888", e);

}

// Fallback to RGBA_8888

try {

mPixelFormat = PixelFormat.RGBA_8888;

ImageReader reader = ImageReader.newInstance(

width, height,

mPixelFormat,

IMAGE_READER_IMAGES);

if (DEBUG) Log.d(TAG, "ImageReader created with RGBA_8888 format");

return reader;

} catch (IllegalArgumentException | RuntimeException e) {

Log.e(TAG, "Failed to create ImageReader with RGBA_8888 format", e);

return null;

}

}

Copilot · 2026-02-15T23:12:11Z

+    private boolean isServiceRunning() {
+        ActivityManager manager = (ActivityManager) getSystemService(Context.ACTIVITY_SERVICE);
+        assert manager != null;
+        for (ActivityManager.RunningServiceInfo service : manager.getRunningServices(Integer.MAX_VALUE)) {
+            if (HyperionScreenService.class.getName().equals(service.service.getClassName())) {
+                return true;
+            }
+        }
+        return false;


isServiceRunning() uses assert manager != null; which may be ignored in production, leading to an NPE. Prefer a proper null-check and return false if ActivityManager is unavailable.

Copilot AI and others added 6 commits February 15, 2026 21:59

Initial plan

bb2e33e

Merge branch 'optimize-ram-performance-126302852423521117' into copil…

1a50625

…ot/sub-pr-11

Merge pull request #27 from haumlab/copilot/sub-pr-11

4b8ee9f

Address code review feedback: extract RGB_565 helper, add error handling, restore RGBA fast-path

Copilot AI review requested due to automatic review settings February 15, 2026 23:05

b1ume merged commit 91775a9 into main Feb 15, 2026
2 of 3 checks passed

Copilot started reviewing on behalf of b1ume February 15, 2026 23:05 View session

Copilot AI reviewed Feb 15, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Optimize ram performance 126302852423521117#29

Optimize ram performance 126302852423521117#29
b1ume merged 6 commits into
mainfrom
optimize-ram-performance-126302852423521117

b1ume commented Feb 15, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Copilot AI Feb 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	Log.e(TAG, "Failed to start screen recording: " + e.getMessage());
	Log.e(TAG, "Failed to start screen recording", e);

-        return lastTriedIndex / (float)ipsToTry.length;
+        float progress = (lastTriedIndex + 1) / (float) ipsToTry.length;
+        if (progress < 0f) {
+            progress = 0f;
+        } else if (progress > 1f) {
+            progress = 1f;
+        }
+        return progress;

Conversation

b1ume commented Feb 15, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants