core: more detailed metrics for block processing

[holiman]

This adds some more metrics, so we can get a better picture of where time is spent during block import

[holiman]

rebased to get rid of lint error due to proof.go

[karalabe]

Instead of measuring and updating, perhaps it would be cleaner to update on the fly?

		start := time.Now()
		receipts, logs, usedGas, err := bc.processor.Process(block, state, bc.vmConfig)
		blockExecutionTimer.Update(time.Since(start))

[karalabe]

Instead of measuring and updating, perhaps it would be cleaner to update on the fly?

		start = time.Now()
		status, err := bc.WriteBlockWithState(block, receipts, state)
		blockWriteTimer.Update(time.Since(start))

[holiman]

I kind of disagree. The three variables will always cover the entire time spent, the next one continuing from where the last one left off. Your methods leaves gaps, and we're not certain that we don't miss something

[karalabe]

The only thing it doesn't cover is the timer updates. But my solution does cover bad blocks and failures too, whereas yours only counts successful blocks.

[karalabe]

Maybe its just the t0, t1, etc that bother me. Perhaps if they had more meaningful names it would be fine. This feels like some left in debug code., not production.

[holiman]

Ah, ok. You want to measure N identifiable and specific things, whereas my approach "we have this chunk of code and want to divvy it up in N places". So for me t0, t1.. makes sense, but I can see why you dislike it....

[holiman]

squashed, rebased