| commit | 083bc1576a0e3e214496be9d8803e4bd6ba0b45d | [log] [tgz] |
|---|---|---|
| author | Michael Ludwig <michaelludwig@google.com> | Mon Oct 01 17:15:15 2018 -0400 |
| committer | Skia Commit-Bot <skia-commit-bot@chromium.org> | Mon Oct 01 22:59:00 2018 +0000 |
| tree | 8318a257653137f5ff0bc66da0783ca49f892b61 | |
| parent | e50efdfdedf070d94f695369d0964c07b01d523b [diff] |
Update parameter generation for randomized gradient tests Previously, randomized gradients used layout parameters between 0 and 1. However, that means when the ProcessorOptimizationValidationTest renders the generated gradients, the sub-pixel gradients are stretched a lot. This may be the source of instability in the test, if we are reaching precision issues due to the small values scaled up so much when actually rendered. Here is an example of a valid 2 point conical gradient produced by the original parameter generation: https://drive.google.com/file/d/1CMc69AHcWBjLu2Ir2ty0qw29T-Vd1cOg Here is a very similar case, rendered on another device, that also has corruption: https://drive.google.com/open?id=1K2HPXCSda2bWoGwV-x_iKwu9_xLENBt- Here is the equivalent 2 point conical gradient from the first link, generated using the new parameters from this CL: https://drive.google.com/open?id=1Z_M0xJ2h-o-bPb5Il7-P5X4z703nWt5R Bug: skia: Change-Id: I89f5ef0e91fa9ff5664079936f97e90a8dbfcd19 Reviewed-on: https://skia-review.googlesource.com/158561 Reviewed-by: Brian Salomon <bsalomon@google.com> Commit-Queue: Michael Ludwig <michaelludwig@google.com>