Don't report array buffer allocations less than 1 Mb.

src/counters.h

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef V8_COUNTERS_H_
 #define V8_COUNTERS_H_
 
 #include "include/v8.h"
 #include "src/allocation.h"
 #include "src/base/atomic-utils.h"
@@ skipping 917 matching lines @@
   HR(debug_feature_usage, V8.DebugFeatureUsage, 1, 7, 7)                      \
   HR(incremental_marking_reason, V8.GCIncrementalMarkingReason, 0, 21, 22)    \
   HR(mark_compact_reason, V8.GCMarkCompactReason, 0, 21, 22)                  \
   HR(scavenge_reason, V8.GCScavengeReason, 0, 21, 22)                         \
   HR(young_generation_handling, V8.GCYoungGenerationHandling, 0, 2, 3)        \
   /* Asm/Wasm. */                                                             \
   HR(wasm_functions_per_asm_module, V8.WasmFunctionsPerModule.asm, 1, 100000, \
      51)                                                                      \
   HR(wasm_functions_per_wasm_module, V8.WasmFunctionsPerModule.wasm, 1,       \
      100000, 51)                                                              \
-  HR(array_buffer_big_allocations, V8.ArrayBufferBigAllocations, 0, 4096, 13) \
+  HR(array_buffer_big_allocations, V8.ArrayBufferBigAllocations, 1, 4096, 13) \

[Ilya Sherman, 2017/05/01 19:51:14]

Please rename the histogram, as you are changing the bucket layout for it. 
Otherwise, you'll get data mixed in from multiple versions, and the blended data
will look rather confusing.

[kschimpf, 2017/05/04 20:51:36]

Is this really necessary. This metric was created very recently and isn't really
being used anywhere (yet).

Otherwise, I'd be more willing to change this back to starting at zero.

[Ilya Sherman, 2017/05/04 21:21:21]

Well, I think it might actually be the same bucket layout regardless of whether
the first value is 0 or 1 -- I think the code will automatically correct a 0 to
a 1, and reserve the zero-th bucket as an underflow bucket.  I'd strongly
encourage you to double-check me on this, as I'm not familiar with what exactly
these V8 wrappers are doing under the hood.  If the actual bucket layout is
changing, then you should definitely either (a) restore the previous layout, or
(b) rename the histogram.

However, you're still blending data with rather different distributions across
multiple versions.  The best practice would be to rename the histogram.  The
next-best option would be to document the change in semantics in the
histograms.xml description... though such documentation is very easy for readers
to miss.

What is your concern w.r.t. renaming the histogram?  

[kschimpf, 2017/05/05 17:05:37]

Ok. Changing the name of the metric, and introducing it into V8.

   HR(array_buffer_new_size_failures, V8.ArrayBufferNewSizeFailures, 0, 4096, 13)
 
 #define HISTOGRAM_TIMER_LIST(HT)                                               \
   /* Garbage collection timers. */                                             \
   HT(gc_compactor, V8.GCCompactor, 10000, MILLISECOND)                         \
   HT(gc_finalize, V8.GCFinalizeMC, 10000, MILLISECOND)                         \
   HT(gc_finalize_reduce_memory, V8.GCFinalizeMCReduceMemory, 10000,            \
      MILLISECOND)                                                              \
   HT(gc_scavenger, V8.GCScavenger, 10000, MILLISECOND)                         \
   HT(gc_context, V8.GCContext, 10000,                                          \
@@ skipping 396 matching lines @@
 
   explicit Counters(Isolate* isolate);
 
   DISALLOW_IMPLICIT_CONSTRUCTORS(Counters);
 };
 
 }  // namespace internal
 }  // namespace v8
 
 #endif  // V8_COUNTERS_H_