khushalsagar@chromium.org chrome-gpu-metric-alerts@google.com

Time spent completing all work for a compositor rasterization task. This includes the time in the renderer process for sending GL or paint commands to the GPU process and the time for flushing these commands to the driver. WARNING: This metric is only reported for clients with high resolution clocks, which means this will be biased to a portion of the population on windows. {RasterTaskTypeGpu}

vmpstr@chromium.org thiabaud@google.com

Gpu image decode usage statistics. Images are decoded and locked prior to upload; this indicates how that decode is used during tile management. Warning: this histogram was expired from 2021-08-09 to 2022-05-20; data may be missing.

sashamcintosh@chromium.org vmpstr@chromium.org thiabaud@google.com

The current number of bytes locked by the GpuImageDecodeCache compared to the maximum allowed number of bytes to lock. This value is recorded every time the tile manager assigns GPU memory to tiles. Warning: this histogram was expired from 2021-08-09 to 2022-05-20; data may be missing.

vmpstr@chromium.org thiabaud@google.com

Indication whether the first lock of an image decode was wasted (image was not used). Images are decoded and locked prior to raster; this indicates whether the decode was used or not during the first lock. Warning: this histogram was expired from 2021-06-06 to 2022-05-20; data may be missing. {OutOfRaster}

khushalsagar@chromium.org vmpstr@chromium.org thiabaud@google.com

Gpu image upload usage statistics. Images are uploaded and locked prior to use; this indicates how that upload is used during tile management. Warning: this histogram was expired from 2021-06-06 to 2022-05-20; data may be missing.

vmpstr@chromium.org thiabaud@google.com

Indication whether the first lock of an image upload was wasted (image was not used). Images are uploaded and locked prior to raster; this indicates whether the upload was used or not during the first lock. Warning: this histogram was expired from M77 to 2022-05-20, and 2022-12-31 to 2023-01-09, and 2023-12-31 to 2024-01-30; data may be missing.

jonross@chromium.org chrome-gpu-metric-alerts@chromium.org

Whether gpu rasterization is enabled (checked once after the page is painted for the first time).

cblume@chromium.org vmpstr@chromium.org sashamcintosh@chromium.org

This metric records the duration of an image decode for the raster path in the compositor. It is recorded every time we decode an image. It is suffixed by the type of rasterization we're in (either Gpu or Software). The rasterization suffix is optionally prefixed by the image type (WebP, Jpeg, etc.) or OutOfRaster. Warning: Before version M78 this metric may include reports from clients with low-resolution clocks (i.e. on Windows, ref. |TimeTicks::IsHighResolution()|). Such reports will cause this metric to have an abnormal distribution.

flackr@chromium.org pdr@chromium.org input-dev@chromium.org

Ideally we'd always scroll on the impl thread, but there are a variety of situations where we need to scroll on main. We should try to drive these down. For every gesture, we record whether the scroll occurred on the impl thread ("Not scrolling on main (no reason)"), or the main thread ("Scrolling on main (any reason)") along with the reasons. There can be multiple main-thread reasons for each scroll, so the percentage of a reason shown in the UMA page is not the percentage of gesture scrolls with the reason in all gesture scrolls. The actual percentage of gesture scrolls with a reason would be count(the reason) / (count("Not scrolling on main (no reason)" + count("Scrolling on main (any reason)")).

flackr@chromium.org pdr@chromium.org input-dev@chromium.org

Ideally we'd always scroll on the impl thread, but there are a variety of situations where we need to scroll on main. We should try to drive these down. For every wheel tick, we record whether the scroll occurred on the impl thread ("Not scrolling on main (no reason)"), or the main thread ("Scrolling on main (any reason)") along with the reasons. There can be multiple main-thread reasons for each wheel tick, so the percentage of a reason shown in the UMA page is not the percentage of wheel scrolls with the reason in all wheel scrolls. The actual percentage of wheel scrolls with a reason would be count(the reason) / (count("Not scrolling on main (no reason)" + count("Scrolling on main (any reason)")).

khushalsagar@chromium.org chrome-gpu-metric-alerts@google.com

Time spent completing all work for a compositor rasterization task. This includes the time in the renderer process for sending GL or paint commands to the GPU process and the time for flushing these commands to the driver. WARNING: This metric is only reported for clients with high resolution clocks, which means this will be biased to a portion of the population on windows. {RasterTaskTypeGpu}

bokan@chromium.org input-dev@chromium.org

For each scroll gesture (i.e. at GestureScrollBegin), records whether the scroll is: handled unblocked on the compositor thread, handled on the compositor thread but blocked on main thread event listeners, or being handled on the main thread. Broken down by scroll device. {ScrollSourceDevice}