Turn on NVPR 4x MSAA by default when supported in DM and nanobench.

dm/DMGpuGMTask.cpp

Index: dm/DMGpuGMTask.cpp
diff --git a/dm/DMGpuGMTask.cpp b/dm/DMGpuGMTask.cpp
index fb170a26b9796ac2404e7d7a8a9d1d2a0fdc1aae..2890483f55455e022ac3a29702adb290e4462b93 100644
--- a/dm/DMGpuGMTask.cpp
+++ b/dm/DMGpuGMTask.cpp
@@ -22,6 +22,8 @@ GpuGMTask::GpuGMTask(const char* config,
     , fSampleCount(sampleCount)
     {}
 
+static bool gAlreadyWarned[GrContextFactory::kGLContextTypeCnt][kGrGLStandardCnt];
+
 void GpuGMTask::draw(GrContextFactory* grFactory) {
     SkImageInfo info = SkImageInfo::Make(SkScalarCeilToInt(fGM->width()),
                                          SkScalarCeilToInt(fGM->height()),
@@ -30,7 +32,11 @@ void GpuGMTask::draw(GrContextFactory* grFactory) {
     SkAutoTUnref<SkSurface> surface(NewGpuSurface(grFactory, fContextType, fGpuAPI, info,
                                                   fSampleCount));
     if (!surface) {
-        this->fail("Could not create context for the config and the api.");
+        if (!gAlreadyWarned[fContextType][fGpuAPI]) {
+            SkDebugf("FYI: couldn't create GPU context, type %d API %d.  Will skip.\n",
+                     fContextType, fGpuAPI);
+            gAlreadyWarned[fContextType][fGpuAPI] = true;
+        }
         return;
     }
     SkCanvas* canvas = surface->getCanvas();