From: peter.noreply@tin.it   
      
   On Wed, 16 Jul 2025 15:39:26 GMT   
   anton@mips.complang.tuwien.ac.at (Anton Ertl) wrote:   
      
   > anton@mips.complang.tuwien.ac.at (Anton Ertl) writes:   
   > >I did not do any accuracy measurements, but I did performance   
   > >measurements on a Ryzen 5800X:   
   > >   
   > >                                cycles:u   
   > >  gforth-fast     iforth           lxf          SwiftForth     VFX   
   > > 3_057_979_501  6_482_017_334  6_087_130_593  6_021_777_424  6_034_560_441   
   NAI   
   > > 6_601_284_920  6_452_716_125  7_001_806_497  6_606_674_147  6_713_703_069   
   UNR   
   > > 3_787_327_724  2_949_273_264  1_641_710_689  7_437_654_901  1_298_257_315   
   REC   
   > > 9_150_679_812 14_634_786_781                                                
   SR    
   >    
                                                                                    
   > >                                cycles:   
                                            
   >    
   > This second table is about instructions:u   
   >    
   > >  gforth-fast     iforth           lxf          SwiftForth       
   VFX               
   > >13_113_842_702  6_264_132_870  9_011_308_923 11_011_828_048  8_072_637_768   
   NAI   
   > > 6_802_702_884  2_553_418_501  4_238_099_417 11_277_658_203  3_244_590_981   
   UNR   
   > > 9_370_432_755  4_489_562_792  4_955_679_285 12_283_918_226  3_915_367_813   
   REC   
   > >51_113_853_111 29_264_267_850                                                
   SR    
   >    
   > - anton   
      
   I have run this test now on my Ryzen 9950X for lxf, lxf64 ans a snapshot of   
   gforth   
      
   Here are the results   
      
   Ryzen 9950X   
      
           lxf64   
        5,010,566,495     NAI cycles:u   
        2,011,359,782     UNR cycles:u   
          646,926,001     REC cycles:u   
        3,589,863,082     SR  cycles:u   
      
           lxf64        
        7,019,247,519     NAI instructions:u          
        4,128,689,843     UNR instructions:u           
        4,643,499,656     REC instructions:u    
       25,019,182,759     SR  instructions:u    
      
      
           gforth-fast 20250219   
        2,048,316,578      NAI cycles:u   
        7,157,520,448      UNR cycles:u   
        3,589,638,677      REC cycles:u   
       17,199,889,916      SR  cycles:u   
      
           gforth-fast 20250219   
       13,107,999,739      NAI instructions:u    
        6,789,041,049      UNR instructions:u   
        9,348,969,966      REC instructions:u    
       50,108,032,223      SR  instructions:u    
      
           lxf   
        6,005,617,374      NAI cycles:u   
        6,004,157,635      UNR cycles:u   
        1,303,627,835      REC cycles:u   
        9,187,422,499      SR  cycles:u   
      
           lxf   
        9,010,888,196      NAI instructions:u   
        4,237,679,129      UNR instructions:u    
        4,955,258,040      REC instructions:u    
       26,018,680,499      SR  instructions:u   
      
   Doing the milliseconds timing gives   
      
   lxf64 native code   
   timer-reset ' naive-sum bench .elapsed            889 ms elapsed ok   
   timer-reset ' unrolled-sum bench .elapsed         360 ms elapsed ok   
   timer-reset ' recursive-sum bench .elapsed        114 ms elapsed ok   
   timer-reset ' shift-reduce-sum bench .elapsed     647 ms elapsed ok   
      
   lxf64 token code   
   timer-reset ' naive-sum bench .elapsed          2´284 ms elapsed ok   
   timer-reset ' unrolled-sum bench .elapsed       2´723 ms elapsed ok   
   timer-reset ' recursive-sum bench .elapsed      3´474 ms elapsed ok   
   timer-reset ' shift-reduce-sum bench .elapsed   6´842 ms elapsed ok   
      
   lxf   
   timer-reset ' naive-sum bench .elapsed           1073 milli-seconds ok   
   timer-reset ' unrolled-sum bench .elapsed        1103 milli-seconds ok   
   timer-reset ' recursive-sum bench .elapsed        234 milli-seconds ok   
   timer-reset ' shift-reduce-sum bench .elapsed    1632 milli-seconds ok   
      
   It is interesting to note how the Best algorithm" change depending   
   on the underlying system implementation.   
   lxf uses the x87 builtin fp stack, lxf64 uses sse4 and a large fp stack    
      
   Thanks for these tests, they uncovered a problem with the lxf64 code   
   generator. It could only handle 114 immediate values in a basic block!   
   Both sum128 and nsum128 compiles gigantic functions of over 2k compile code.   
      
   Best Regards   
   Peter   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)