Request for Reviews(M): 7092905: C2: Keep track of the number of dead nodes

[Vladimir Kozlov]

Yes, your proposal looks reasonable.

Thanks,
Vladimir

Bharadwaj Yadavalli wrote:
>   As a consequence of keeping a more accurate track of live and dead 
> ideal nodes, we have the following ideal node counts to report in 
> various log messages:
> 
>    1. number of nodes created (Compile::unique()) - access time O(1)
>    2. number of nodes live as tracked via proposed code additions
>       (Compile::live_nodes()) - access time O(1)
>    3. number of nodes live computed by graph walk
>       (Compile::count_live_nodes_by_graph_walk()) - O(N) where N is the
>       number of nodes.
> 
> We have the following command line options (existing and proposed) that 
> would log one or more of the above ideal node counts:
> 
>    1. LogCompilation - existing option
>    2. PrintIdealNodeCount - proposed to print to console (tty)
>    3. VerifyIdealNodeCount - proposed to verify that tracked live nodes
>       is the same as live nodes computed by graph walk and emit a log
>       message if they are not the same.
> 
> After reading through all the review comments on my RFR, following is my 
> proposed way to incorporate the responses.
> Move all the logging functionality into TracePhase so that it augments 
> the information logged therein.
> 
>    1. LogCompilation - existing option
>           * Augment the current logging via TracePhase; currently we
>             just print out the number of nodes created (unique())
>    2. PrintIdealNodeCount - print number of nodes created, live as
>       tracked to console (tty)
>    3. VerifyIdealNodeCount - if verification fails, print total nodes
>       created, number of tracked live nodes and live nodes computed by
>       graph walk.
> 
> If this sounds reasonable, I'll send out an updated webrev.
> 
> Thanks,
> 
> Bharadwaj
>