Adding new benchmark to test image decoding performance.

gyp/bench.gyp

 # GYP file to build performance testbench.
 #
 {
   'includes': [
     'apptype_console.gypi',
   ],
   'targets': [
     {
       'target_name': 'nanobench',
       'type': 'executable',
       'sources': [
+        '../bench/DecodingBench.cpp',

[scroggo, 2015/02/12 14:19:02]

Nit: I know this was not alphabetized before, but the old order had its own
consistency (bench/ was grouped together).

So long as we're looking at these sources, why are they here and not in
bench.gypi?

The result is that any gyp file which includes bench.gypi needs to also include
these files. For example, iOSShell.gypi also lists these sources. This CL does
not add DecodingBench.cpp to the iOS build, meaning it will now fail to compile
(since nanobench.cpp attempts to use it). 

[mtklein, 2015/02/12 14:41:38]

If we do move them, let's keep them together in one distinct clump in
bench.gypi.  These are the core files needed to build nanobench, as opposed to
files which provide benchmarks through registration.  These are required, the
others are optional.

[scroggo, 2015/02/12 15:00:18]

Maybe move them into their own gypi? bench_core.gypi? That seems a little odd
though; isn't this the core of bench? Maybe the real question is, why does iOS
not include this target?

I think it's because the iOS build just has one executable for nanobench and dm
(although I'm not sure why), so it just recreates this file inside iOSShell.gyp.
Maybe there's a better way to share the code?

Not to say that decision needs to hold this CL up. I'd be fine with adding
DecodingBench.cpp to iOSShell.gyp to land this, and then dealing with this
duplication in another CL.

[_cary, 2015/02/12 15:18:05]

ioSShell.gyp exists because multiple iOS applications do not play well together.
Creating a single application with an additional flag to choose whether the bot
calls nanobench or dm was the simplest way to prototype a bot-driven iOS test
environment. That said, the bots are building but not running iOS to my
knowledge so that decision could be revisited. 

This CL should maintain iOSShell.gyp by minimally adding DecodingBench.cpp to it
as well.

[msarett, 2015/02/12 18:52:56]

Done.

         '../gm/gm.cpp',
         '../bench/GMBench.cpp',
         '../bench/RecordingBench.cpp',
         '../bench/SKPBench.cpp',
         '../bench/nanobench.cpp',
       ],
       'includes': [
         'bench.gypi',
         'gmslides.gypi',
       ],
       'dependencies': [
         'flags.gyp:flags_common',
         'jsoncpp.gyp:jsoncpp',
         'skia_lib.gyp:skia_lib',
         'tools.gyp:crash_handler',
         'tools.gyp:proc_stats',
         'tools.gyp:timer',
       ],
       'conditions': [
         ['skia_android_framework', {
           'libraries': [
             '-lskia',
           ],
         }],
       ],
     },
   ],
 }