Adding this non-tile bench data will give us more confidence that any speedups we put in place are well-tested (as noted in https://codereview.appspot.com/7336043 ) Still takes about 39 seconds to run on my machine. https://codereview.appspot.com/7317048/diff/1/tools/tests/benchgraphs/Skia_Sh... File tools/tests/benchgraphs/Skia_Shuttle_Ubuntu12_ATI5770_Float_Bench_32/output-expected/graph.xhtml (right): https://codereview.appspot.com/7317048/diff/1/tools/tests/benchgraphs/Skia_Sh... tools/tests/benchgraphs/Skia_Shuttle_Ubuntu12_ATI5770_Float_Bench_32/output-expected/graph.xhtml:96: <g id="desk_amazon.skp_record_rtree__{'scalar': 'data_skp_device_bitmap_multi_4_mode_tile_256_256_timeIndividualTiles'}"> I took a *quick* look at the diff, and the before and after versions of the graph, and it all seemed kosher to me. Please take a look and see what you think: https://www.corp.google.com/~epoger/graphs/cl7317048/before.xhtml https://www.corp.google.com/~epoger/graphs/cl7317048/after.xhtml
LGTM. Both the xhtml text diff and the addition of "record_rtree" in the browser make sense.