Add benchmark fp8 script

[drisspg]

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 32/32 [02:29<00:00,  4.67s/it]
   numel  high_precision_dtype    low_precision_dtype    saturated      triton_time    pytorch_time    compiled_pytorch_time
--------  ----------------------  ---------------------  -----------  -------------  --------------  -----------------------
 2097152  torch.bfloat16          torch.float8_e4m3fn    True               39.4022         59.9545                  75.0129
 2097152  torch.bfloat16          torch.float8_e4m3fn    False              39.193          27.1596                  29.506
 2097152  torch.bfloat16          torch.float8_e5m2      True               39.1569         60.187                   79.2593
 2097152  torch.bfloat16          torch.float8_e5m2      False              38.8475         27.2665                  29.2743
 2097152  torch.float32           torch.float8_e4m3fn    True               39.8845         59.7176                  69.8467
 2097152  torch.float32           torch.float8_e4m3fn    False              38.8206         25.6547                  30.5112
 2097152  torch.float32           torch.float8_e5m2      True               39.5862         60.5091                  69.3323
 2097152  torch.float32           torch.float8_e5m2      False              39.2996         25.6836                  30.4653
 4194304  torch.bfloat16          torch.float8_e4m3fn    True               39.5744         89.2462                  89.2371
 4194304  torch.bfloat16          torch.float8_e4m3fn    False              39.4241         44.5693                  44.5856
 4194304  torch.bfloat16          torch.float8_e5m2      True               39.5765         89.2892                  89.3438
 4194304  torch.bfloat16          torch.float8_e5m2      False              39.51           45.5271                  45.5449
 4194304  torch.float32           torch.float8_e4m3fn    True               39.0992         88.103                   88.1655
 4194304  torch.float32           torch.float8_e4m3fn    False              38.7692         43.3739                  43.3968
 4194304  torch.float32           torch.float8_e5m2      True               39.4078         88.1485                  88.1937
 4194304  torch.float32           torch.float8_e5m2      False              39.0674         44.4441                  44.4622
 8388608  torch.bfloat16          torch.float8_e4m3fn    True               38.9665        201.205                  201.236
 8388608  torch.bfloat16          torch.float8_e4m3fn    False              38.6342         98.2023                  98.2161
 8388608  torch.bfloat16          torch.float8_e5m2      True               39.0337        201.23                   201.231
 8388608  torch.bfloat16          torch.float8_e5m2      False              38.9622         99.2762                  99.3148
 8388608  torch.float32           torch.float8_e4m3fn    True               39.4607        206.298                  206.345
 8388608  torch.float32           torch.float8_e4m3fn    False              38.7217         98.8632                  98.9315
 8388608  torch.float32           torch.float8_e5m2      True               38.9956        206.25                   206.253
 8388608  torch.float32           torch.float8_e5m2      False              38.9852         99.7639                  99.8383
16777216  torch.bfloat16          torch.float8_e4m3fn    True               39.6764        456.835                  456.883
16777216  torch.bfloat16          torch.float8_e4m3fn    False              38.617         217.268                  217.277
16777216  torch.bfloat16          torch.float8_e5m2      True               39.1838        456.872                  456.897
16777216  torch.bfloat16          torch.float8_e5m2      False              38.7144        218.301                  218.347
16777216  torch.float32           torch.float8_e4m3fn    True               43.0035        457.834                  457.869
16777216  torch.float32           torch.float8_e4m3fn    False              42.993         214.077                  214.139
16777216  torch.float32           torch.float8_e5m2      True               42.9996        457.954                  458.012
16777216  torch.float32           torch.float8_e5m2      False              42.9937        215.222                  215.239