differences between individual rasterize_and_record_micro benchmarks

[David Grogan]

I'm trying to understand why record_time is an order of magnitude smaller than record_time_painting_disabled. Doesn't record_time do everything that record_time_painting_disabled does, except actually painting?

In other words, what are the two benchmarks measuring?

The run I'm looking at is https://00e9e64baccdc024014d86f7ff81c16b15e91e07535af117b8-apidata.googleusercontent.com/download/storage/v1/b/cluster-telemetry/o/tasks%2Fchromium_perf_runs%2Fdgrogan-20170302032746%2Fhtml%2Findex.html?qk=AD5uMEtgVKjyAalamMrYAoIEPsjiT4bNYkxjGgOVsvYpmvWOHu6tcTmKVLfJgCkeE4BMOLUlzpU2-nBYFke48qbAst1Es4xFo4asnZhvGZE-F2sKldhB2T3pvl2gjs4F4dP6dMgNHqzJSBc-zsu5XhxB94_Uiv3cFqqbbg23npBldn0zDmeMRoRRXxBdz4sazfosh-Wl3N6GgjwNm_vwz1l9b19skHoTFQNN-a7SLDSLhAQxihPEUVsHg2_sKKK1_ITcA1R--Vnujorjy5JtfU47QN9PQOHZMddmnc0Od1CIa0h-nY_9XSwMPYn2G4u4UG7jFRQthKFJEzIk5uUIcsJEs5Hh8ntP-jgeu2lPU4LGQv62SLN5GmioD9vZqHW95Jq-VN5IWKFetCFbNcYk45C8rkvPMnUh_4pNCIQmR3KfoRGB4hPiQFH36-o3oQedH8UAKbmlsBan1UkyI2exT7e2RXnOX7NBBqiRBFOdBsnbyIh3hov7tS64kirdf9zsi8kRqM_gTZOByHCZ0iZgyfK5PtNwz4wscX-OUPzSXBgYjGSJrglU1gBa9A2r07wh2mD3n9cvPlWcbCuc7s1wknQ53ivk-WFPafc73PCRItCh9zjOdCXjH5kVMFrhukoOuDK6y2xrh3iB2taR9F4JTLzrZbWZcHStMTjlXVHCQ5r75CUyBOhhIqGBN2cPlrG1lo_dD44n5LcqaXIrg3GBRtXrQmtWu0PsTcfBPsU0v2jVZ7c461r30fyOPQbvnyvKn8HDM1WQ4JzDBvf670tXXOozB-nhF7-AqfBORbUV5xaLx36hWdqAd50Xy_yqut7iLABQjevX1VZSBubxgt0zLLCwv4Z5ogIzjJsQ8rssbvRPm2xwPsbywJ0

[Ned]

+Adrienne Walker +Vladimir Levin  are the experts here

--
You received this message because you are subscribed to the Google Groups "telemetry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+...@chromium.org.

[Adrienne Walker]

chrishtr is the person to ask about the various record flavors.

To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+unsubscribe@chromium.org.

[Vladimir Levin]

From looking at https://cs.chromium.org/chromium/src/third_party/WebKit/Source/platform/graphics/ContentLayerDelegate.cpp?l=73

I believe that disabling painting also disables any caching that normally happens, which is like to do more work. +chrishtr to confirm. 

Vlad

To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+unsubscribe@chromium.org.

[Chris Harrelson]

On Wed, Apr 26, 2017 at 5:45 PM, Vladimir Levin <vmp...@google.com> wrote:

From looking at https://cs.chromium.org/chromium/src/third_party/WebKit/Source/platform/graphics/ContentLayerDelegate.cpp?l=73

I believe that disabling painting also disables any caching that normally happens, which is like to do more work. +chrishtr to confirm. 

That's exactly it. record_time verifies that display list caching is working for painted content, and so always has a very small value. record_time_painting_disabled forces the paint tree walk, but doesn't actually paint anything. record_time_caching_disabled forces the paint tree walk and actually paints.

Chris

[David Grogan]

Got it, thanks for the explanation.