—opt level
0unopt cycles
0opt cycles
1.00×speedup
Source —
Edit the buffer — the pattern matcher will say which optimizations still apply.
Counters cost model
Cycles
0
Instructions
0
Allocations
0
Cache
—
Loop iters
—
UNOPT
— c
OPT
— c
Execution trace —
Call stack
stack empty
Heap / objects
(no heap traffic this step)
Compiler IR SSA — original vs optimized
ORIGINAL · pre-opt SSA

          

          

            
AFTER · post-opt SSA

            

          

        

      


      

        

          Expression graph
          data-flow morphs as the optimization fires
        

        
        

      

    

  


  
  

    

      Why this happens
      —
    

    

      
Pick a chapter from the banner above. Each chapter peels back one layer between the C++ you write and the cycles the CPU spends.

    

  


  
  

    

      Cheat sheet
      What each -O flag enables