Automatic parallelization has been the holy grail of performance-based computing since the 1980's or earlier, and if auto-threading compilers had arrived we would all know about it.
The fact remains that for general-purpose code, automatic parallelization is an unsolved and exceedingly difficult problem. So difficult that PG claimed it as one of his highly ambitious startup ideas.
https://research.microsoft.com/pubs/170528/msr-tr-2012-79.pd... (this is the TR linked from the Dr. Dobbs thing.)
Automatic parallelization has been the holy grail of performance-based computing since the 1980's or earlier, and if auto-threading compilers had arrived we would all know about it.
The fact remains that for general-purpose code, automatic parallelization is an unsolved and exceedingly difficult problem. So difficult that PG claimed it as one of his highly ambitious startup ideas.