The objective is to build the fastest replacement for RTL Power.
Because the RTL Power function relies on the FPU control word beeing default
so can all functions.
Validation and Benchmark Tool PowerBV121.zip
Tables of results
Target | Function | Author | Speed up over RTL |
P4 Northwood | PowerDKCIA32Ext_5 | Dennis Christensen | 1,47 |
P4 Prescott | PowerDKCIA32Ext_5 | Dennis Christensen | 1,24 |
Pentium M | PowerJOH | John O'Harrow | 1,39 |
P3 | PowerDKCIA32_2 | Dennis Christensen | 1,42 |
Athlon XP | PowerDKCIA32Ext_5 | Dennis Christensen | 1,51 |
Opteron | PowerDKCIA32_3 | Dennis Christensen | 1,43 |
RTL Suggestion | PowerJOH | John O'Harrow | 1,39 |
Pascal | PowerDKCPas4 | Dennis Christensen | 1,13 |
Blended | PowerDKCIA32Ext_5 | Dennis Christensen | 1,41 |
Target | Function | Author | Speed up over RTL |
P4 Northwood | PowerDKCSSE2_2 | Dennis Christensen | 1,38 |
P4 Prescott | PowerDKCSSE2_1 | Dennis Christensen | 1,14 |
Pentium M | PowerJOH | John O'Harrow | 1,40 |
P3 | PowerDKCIA32_2 | Dennis Christensen | 1,40 |
Athlon XP | PowerDKCIA32Ext_4 | Dennis Christensen | 1,70 |
Opteron | PowerDKCSSE2_2 | Dennis Christensen | 1,43 |
RTL Suggestion | PowerDKCIA32_2 | Dennis Christensen | 1,58 |
Pascal | PowerDKCPas4 | Dennis Christensen | 1,36 |
Blended | PowerDKCIA32Ext_5 | Dennis Christensen | 1,38 |
Target | Function | Author | Speed up over RTL |
P4 Northwood | PowerDKCSSE_2 | Dennis Christensen | 1,69 |
P4 Prescott | PowerDKCSSE_2 | Dennis Christensen | 1,37 |
Pentium M | PowerJOH | John O'Harrow | 1,61 |
P3 | PowerDKCSSE_2 | Dennis Christensen | 1,55 |
Athlon XP | PowerDKCIA32Ext_5 | Dennis Christensen | 1,88 |
Opteron | PowerDKCSSE_2 | Dennis Christensen | 1,74 |
RTL Suggestion | PowerJOH | John O'Harrow | 1,55 |
Pascal | PowerDKCPas4 | Dennis Christensen | 1,22 |
Blended | PowerDKCIA32Ext_5 | Dennis Christensen | 1,57 |