• What is the future of Parabix compiler technology?
• What are the issues for completion of icgrep?

icgrep as a Parabix Showcase

• Demonstration that parallel techniques can outperform sequential techniques.
• Demonstration that search programs compiled from declarative specifications can outperform the best hand-written sequential programs.
• Tool to allow exploration of Parabix algorithms
  • ShowREs
  • ShowPablo
  • ShowIR
  • ShowASM
  • DumpTrace
  • TraceCounts

• Full Unicode support meeting Unicode Level 2 requirements of Unicode Technical Standard 18.

• Suitable as a replacement for grep (or ripgrep, ag, ...) for (almost) all applications.

icgrep Gap Analysis

• Colorized output
  • Need to find matches within lines
  • icgrep algorithm finds end of match only
  • run RE in reverse through reversed lines to find start
• Full lookahead assertions
  • optimize away some cases by RE transformation
  • general approach: run assertion in reverse through reversed input
• Context lines (-A, -B, -C)
  • ContextGroup kernel
• Byte offset printing (-b)
• Recursive path filtering
  • include/exclude support
  • .gitignore, .svn, others

• Unicode level 2
  • Full Unicode caseless matching
  • Canonical and compatibility equivalent matching
• Backreferences
  • Implement the known fast cases using parallel methods
  • Reject backreferences that aren't supported?
  • Generalized backreference kernel after initial scan
• Performance
  • Further accelerate the simple cases using multiple kernels/multiple cores
  • Extend the log2 bounded repetition for more cases
  • Extended star normal form, (A?{B{0,7}c*D{0,1000})* => [ABcD]*
  • Extend MatchStar for all cases handled by log2 bounded repetition
  • Local subexpression optimization
  • Better RE factoring for multiple regular expressions

Parabix Framework Development

• Individual Kernel development
  • Consolidate deletion/PEXT kernels
  • PDEP kernel review and implementation
  • Gather kernels
• Sequential scan language/compilation
• Length-sorted symbol resolution
• Unix pipe notation support
  • Auto optimizations

Parabix Application Development

• Transducers
  • CSV to JSON
  • grammar-based transducers