Changes

Page history
New tests for Haswell authored by Thomas.Roehl's avatar Thomas.Roehl
Show whitespace changes
Inline Side-by-side
AccuracyHaswell.md
View page @ 49de1e43
......@@ -14,12 +14,12 @@ Not all performance groups can be tested for accuracy. We don't have a test appl
| Group | Tests |
|-------|-------|
| BRANCH | load, triad, copy, store, stream |
| INSTRUCTIONS | load, daxpy, stream, ddot, triad, copy, store |
| INST_RETIRED_ANY | load, daxpy, stream, ddot, triad, copy, store |
| DATA | triad, copy, stream, store |
| FLOPS_AVX | triad_avx, stream_avx, daxpy_avx, ddot_avx |
| FLOPS_AVX | ddot_avx, triad_avx, striad_avx, daxpy_avx |
| L2 | load, triad, copy, stream, store |
| L3 | load, triad, copy, stream, store |
| UOPS | ddot_uops, stream_uops |
| UOPS_RETIRED_ANY | triad, daxpy, ddot, stream |
## Information about the pictures
......@@ -41,8 +41,8 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 11.00 | 10.95 | -0.43% | 11.00 | 10.95 | -0.43% | 11.00 | 10.95 | 0.34% |
| 1MB | 11.00 | 11.00 | -0.01% | 11.00 | 11.00 | -0.01% | 11.00 | 11.00 | 0.00% |
| 12kB | 11.00 | 10.95 | 0.43% | 11.00 | 10.95 | 0.43% | 11.00 | 10.95 | 0.43% |
| 1MB | 11.00 | 11.00 | 0.01% | 11.00 | 11.00 | 0.01% | 11.00 | 11.00 | 0.01% |
| 4MB | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% |
| 1GB | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% |
......@@ -61,9 +61,9 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 19.00 | 18.69 | -1.65% | 19.00 | 18.69 | -1.65% | 19.00 | 18.69 | 1.32% |
| 1MB | 19.00 | 19.00 | -0.02% | 19.00 | 19.00 | -0.02% | 19.00 | 19.00 | 0.02% |
| 4MB | 19.00 | 19.00 | -0.01% | 19.00 | 19.00 | -0.01% | 19.00 | 19.00 | 0.00% |
| 12kB | 19.00 | 18.69 | 1.65% | 19.00 | 18.69 | 1.65% | 19.00 | 18.69 | 1.65% |
| 1MB | 19.00 | 19.00 | 0.02% | 19.00 | 19.00 | 0.02% | 19.00 | 19.00 | 0.02% |
| 4MB | 19.00 | 19.00 | 0.01% | 19.00 | 19.00 | 0.01% | 19.00 | 19.00 | 0.01% |
| 1GB | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% |
......@@ -81,7 +81,7 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 11.00 | 10.96 | -0.38% | 11.00 | 10.96 | -0.38% | 11.00 | 10.96 | 0.31% |
| 12kB | 11.00 | 10.96 | 0.38% | 11.00 | 10.96 | 0.38% | 11.00 | 10.96 | 0.38% |
| 1MB | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% |
| 4MB | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% |
| 1GB | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% | 11.00 | 11.00 | 0.00% |
......@@ -101,7 +101,7 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 7.00 | 7.02 | 0.26% | 7.00 | 7.02 | 0.26% | 7.00 | 7.02 | 0.21% |
| 12kB | 7.00 | 7.02 | 0.26% | 7.00 | 7.02 | 0.26% | 7.00 | 7.02 | 0.26% |
| 1MB | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% |
| 4MB | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% |
| 1GB | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% | 7.00 | 7.00 | 0.00% |
......@@ -121,152 +121,152 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 19.00 | 18.77 | -1.19% | 19.00 | 18.77 | -1.19% | 19.00 | 18.77 | 0.95% |
| 1MB | 19.00 | 19.00 | -0.01% | 19.00 | 19.00 | -0.01% | 19.00 | 19.00 | 0.01% |
| 12kB | 19.00 | 18.77 | 1.19% | 19.00 | 18.77 | 1.19% | 19.00 | 18.77 | 1.19% |
| 1MB | 19.00 | 19.00 | 0.01% | 19.00 | 19.00 | 0.01% | 19.00 | 19.00 | 0.01% |
| 4MB | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% |
| 1GB | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% | 19.00 | 19.00 | 0.00% |
### Verification of Group INSTRUCTIONS
### Verification of Group INST_RETIRED_ANY
#### Verification of Group INSTRUCTIONS with Test load
#### Verification of Group INST_RETIRED_ANY with Test load
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_load](./images/accuracy/haswell/INSTRUCTIONS_load.jpg)
![haswell_INST_RETIRED_ANY_load](./images/accuracy/haswell/INST_RETIRED_ANY_load.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 2669128434.00 | 2669134000.00 | 0.00% | 2593941718.00 | 2593948000.00 | 0.00% | 2639053747.60 | 2639059600.00 | 0.00% |
| 426801kB | 2640830956.00 | 2640836000.00 | 0.00% | 2640830956.00 | 2640836000.00 | 0.00% | 2655502239.00 | 2655507200.00 | 0.00% |
| 634877kB | 2618867608.00 | 2618872000.00 | 0.00% | 2400628642.00 | 2400633000.00 | 0.00% | 2531572021.60 | 2531576400.00 | 0.00% |
| 842953kB | 2607885736.00 | 2607890000.00 | 0.00% | 2173238116.00 | 2173243000.00 | 0.00% | 2520956212.00 | 2520960600.00 | 0.00% |
| 223231kB | 2685747696.00 | 2685754000.00 | 0.00% | 2800851168.00 | 2800858000.00 | 0.00% | 2739462649.60 | 2739469200.00 | 0.00% |
| 430488kB | 2737634641.00 | 2737641000.00 | 0.00% | 2737634641.00 | 2737641000.00 | 0.00% | 2737634641.00 | 2737641000.00 | 0.00% |
| 637745kB | 2740310391.00 | 2740316000.00 | 0.00% | 2740310391.00 | 2740316000.00 | 0.00% | 2740310391.00 | 2740316000.00 | 0.00% |
| 845002kB | 2759459620.00 | 2759465000.00 | 0.00% | 2759459620.00 | 2759465000.00 | 0.00% | 2759459620.00 | 2759465000.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test daxpy
#### Verification of Group INST_RETIRED_ANY with Test daxpy
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_daxpy](./images/accuracy/haswell/INSTRUCTIONS_daxpy.jpg)
![haswell_INST_RETIRED_ANY_daxpy](./images/accuracy/haswell/INST_RETIRED_ANY_daxpy.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 3701236991.00 | 3701243000.00 | 0.00% | 3571369027.00 | 3571375000.00 | 0.00% | 3649289805.40 | 3649295800.00 | 0.00% |
| 426801kB | 3547782997.00 | 3547788000.00 | 0.00% | 3294369927.00 | 3294375000.00 | 0.00% | 3522441690.00 | 3522446800.00 | 0.00% |
| 634877kB | 3581103050.00 | 3581108000.00 | 0.00% | 3581103050.00 | 3581108000.00 | 0.00% | 3581103050.00 | 3581108000.00 | 0.00% |
| 842953kB | 3503523257.00 | 3503528000.00 | 0.00% | 3503523257.00 | 3503528000.00 | 0.00% | 3503523257.00 | 3503528000.00 | 0.00% |
| 223231kB | 3644943297.00 | 3644950000.00 | 0.00% | 3711214993.00 | 3711222000.00 | 0.00% | 3697960653.80 | 3697967600.00 | 0.00% |
| 430488kB | 3706232642.00 | 3706238000.00 | 0.00% | 3578431517.00 | 3578437000.00 | 0.00% | 3680672417.00 | 3680677800.00 | 0.00% |
| 637745kB | 3597280182.00 | 3597286000.00 | 0.00% | 3597280182.00 | 3597286000.00 | 0.00% | 3597280182.00 | 3597286000.00 | 0.00% |
| 845002kB | 3512039513.00 | 3512045000.00 | 0.00% | 3512039513.00 | 3512045000.00 | 0.00% | 3512039513.00 | 3512045000.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test stream
#### Verification of Group INST_RETIRED_ANY with Test stream
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_stream](./images/accuracy/haswell/INSTRUCTIONS_stream.jpg)
![haswell_INST_RETIRED_ANY_stream](./images/accuracy/haswell/INST_RETIRED_ANY_stream.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 2770516177.00 | 2770522000.00 | 0.00% | 2770516177.00 | 2770522000.00 | 0.00% | 2770516177.00 | 2770522000.00 | 0.00% |
| 426801kB | 2703072561.00 | 2703078000.00 | 0.00% | 2703072561.00 | 2703078000.00 | 0.00% | 2703072561.00 | 2703078000.00 | 0.00% |
| 634877kB | 2638707515.00 | 2638712000.00 | 0.00% | 2513054777.00 | 2513060000.00 | 0.00% | 2588446419.80 | 2588451200.00 | 0.00% |
| 842953kB | 2669351057.00 | 2669356000.00 | 0.00% | 2335682177.00 | 2335687000.00 | 0.00% | 2602617281.00 | 2602622200.00 | 0.00% |
| 223231kB | 2827591569.00 | 2827598000.00 | 0.00% | 2827591569.00 | 2827598000.00 | 0.00% | 2827591569.00 | 2827598000.00 | 0.00% |
| 430488kB | 2726424017.00 | 2726430000.00 | 0.00% | 2811624767.00 | 2811631000.00 | 0.00% | 2777544467.00 | 2777550600.00 | 0.00% |
| 637745kB | 2776848003.00 | 2776854000.00 | 0.00% | 2776848003.00 | 2776854000.00 | 0.00% | 2776848003.00 | 2776854000.00 | 0.00% |
| 845002kB | 2675839633.00 | 2675845000.00 | 0.00% | 2675839633.00 | 2675845000.00 | 0.00% | 2675839633.00 | 2675845000.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test ddot
#### Verification of Group INST_RETIRED_ANY with Test ddot
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_ddot](./images/accuracy/haswell/INSTRUCTIONS_ddot.jpg)
![haswell_INST_RETIRED_ANY_ddot](./images/accuracy/haswell/INST_RETIRED_ANY_ddot.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 3434665907.00 | 3434672000.00 | 0.00% | 3485929577.00 | 3485936000.00 | 0.00% | 3475676843.00 | 3475683200.00 | 0.00% |
| 426801kB | 3401070167.00 | 3401075000.00 | 0.00% | 3401070167.00 | 3401075000.00 | 0.00% | 3401070167.00 | 3401075000.00 | 0.00% |
| 634877kB | 3422383802.00 | 3422389000.00 | 0.00% | 3422383802.00 | 3422389000.00 | 0.00% | 3422383802.00 | 3422389000.00 | 0.00% |
| 842953kB | 3358640717.00 | 3358645000.00 | 0.00% | 3358640717.00 | 3358645000.00 | 0.00% | 3358640717.00 | 3358645000.00 | 0.00% |
| 223231kB | 3505423937.00 | 3505431000.00 | 0.00% | 3557743697.00 | 3557751000.00 | 0.00% | 3547279745.00 | 3547287000.00 | 0.00% |
| 430488kB | 3531346892.00 | 3531353000.00 | 0.00% | 3531346892.00 | 3531353000.00 | 0.00% | 3531346892.00 | 3531353000.00 | 0.00% |
| 637745kB | 3587315417.00 | 3587321000.00 | 0.00% | 3437843942.00 | 3437850000.00 | 0.00% | 3527526827.00 | 3527532600.00 | 0.00% |
| 845002kB | 3366804797.00 | 3366810000.00 | 0.00% | 3564852137.00 | 3564858000.00 | 0.00% | 3406414265.00 | 3406419600.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test triad
#### Verification of Group INST_RETIRED_ANY with Test triad
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_triad](./images/accuracy/haswell/INSTRUCTIONS_triad.jpg)
![haswell_INST_RETIRED_ANY_triad](./images/accuracy/haswell/INST_RETIRED_ANY_triad.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 2110354431.00 | 2110360000.00 | 0.00% | 2077887440.00 | 2077893000.00 | 0.00% | 2090874236.40 | 2090879800.00 | 0.00% |
| 426801kB | 2090657530.00 | 2090663000.00 | 0.00% | 2090657530.00 | 2090663000.00 | 0.00% | 2065316226.80 | 2065321600.00 | 0.00% |
| 634877kB | 1979030440.00 | 1979035000.00 | 0.00% | 2073269984.00 | 2073275000.00 | 0.00% | 2035574166.40 | 2035579000.00 | 0.00% |
| 842953kB | 2127139126.00 | 2127144000.00 | 0.00% | 1876887466.00 | 1876892000.00 | 0.00% | 2077088794.00 | 2077093600.00 | 0.00% |
| 223231kB | 2120694288.00 | 2120701000.00 | 0.00% | 2153830136.00 | 2153839000.00 | 0.00% | 2147202966.40 | 2147210200.00 | 0.00% |
| 430488kB | 2108718265.00 | 2108724000.00 | 0.00% | 2172618818.00 | 2172625000.00 | 0.00% | 2121498375.60 | 2121504200.00 | 0.00% |
| 637745kB | 2082635692.00 | 2082641000.00 | 0.00% | 2082635692.00 | 2082641000.00 | 0.00% | 2082635692.00 | 2082641000.00 | 0.00% |
| 845002kB | 2006879728.00 | 2006885000.00 | 0.00% | 2006879728.00 | 2006885000.00 | 0.00% | 2006879728.00 | 2006885000.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test copy
#### Verification of Group INST_RETIRED_ANY with Test copy
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_copy](./images/accuracy/haswell/INSTRUCTIONS_copy.jpg)
![haswell_INST_RETIRED_ANY_copy](./images/accuracy/haswell/INST_RETIRED_ANY_copy.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 2218008138.00 | 2218014000.00 | 0.00% | 2218008138.00 | 2218014000.00 | 0.00% | 2202970794.80 | 2202976600.00 | 0.00% |
| 426801kB | 2200692466.00 | 2200697000.00 | 0.00% | 2200692466.00 | 2200697000.00 | 0.00% | 2200692466.00 | 2200697000.00 | 0.00% |
| 634877kB | 2182389676.00 | 2182394000.00 | 0.00% | 2182389676.00 | 2182394000.00 | 0.00% | 2182389676.00 | 2182394000.00 | 0.00% |
| 842953kB | 2028355576.00 | 2028360000.00 | 0.00% | 2173238116.00 | 2173243000.00 | 0.00% | 2144261608.00 | 2144266400.00 | 0.00% |
| 223231kB | 2225333808.00 | 2225340000.00 | 0.00% | 2225333808.00 | 2225340000.00 | 0.00% | 2225333808.00 | 2225340000.00 | 0.00% |
| 430488kB | 2145713641.00 | 2145719000.00 | 0.00% | 2219703766.00 | 2219710000.00 | 0.00% | 2160511666.00 | 2160517200.00 | 0.00% |
| 637745kB | 2192248316.00 | 2192254000.00 | 0.00% | 2192248316.00 | 2192254000.00 | 0.00% | 2192248316.00 | 2192254000.00 | 0.00% |
| 845002kB | 2178520756.00 | 2178526000.00 | 0.00% | 2033286040.00 | 2033291000.00 | 0.00% | 2120426869.60 | 2120432000.00 | 0.00% |
#### Verification of Group INSTRUCTIONS with Test store
#### Verification of Group INST_RETIRED_ANY with Test store
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_INSTRUCTIONS_store](./images/accuracy/haswell/INSTRUCTIONS_store.jpg)
![haswell_INST_RETIRED_ANY_store](./images/accuracy/haswell/INST_RETIRED_ANY_store.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 2153074160.00 | 2153080000.00 | 0.00% | 2153074160.00 | 2153080000.00 | 0.00% | 2153074160.00 | 2153080000.00 | 0.00% |
| 426801kB | 2053979794.00 | 2053985000.00 | 0.00% | 2053979794.00 | 2053985000.00 | 0.00% | 2072652337.40 | 2072657400.00 | 0.00% |
| 634877kB | 1944310808.00 | 1944315000.00 | 0.00% | 2083190150.00 | 2083195000.00 | 0.00% | 2055414281.60 | 2055419000.00 | 0.00% |
| 842953kB | 2028355657.00 | 2028360000.00 | 0.00% | 1843959690.00 | 1843964000.00 | 0.00% | 1991476463.60 | 1991480800.00 | 0.00% |
| 223231kB | 2148598164.00 | 2148604000.00 | 0.00% | 2099766388.00 | 2099773000.00 | 0.00% | 2138831808.80 | 2138837800.00 | 0.00% |
| 430488kB | 2071723520.00 | 2071729000.00 | 0.00% | 2071723520.00 | 2071729000.00 | 0.00% | 2071723520.00 | 2071729000.00 | 0.00% |
| 637745kB | 2092600775.00 | 2092606000.00 | 0.00% | 2092600775.00 | 2092606000.00 | 0.00% | 2092600775.00 | 2092606000.00 | 0.00% |
| 845002kB | 2033286044.00 | 2033291000.00 | 0.00% | 2033286044.00 | 2033291000.00 | 0.00% | 2033286044.00 | 2033291000.00 | 0.00% |
### Verification of Group DATA
......@@ -285,10 +285,10 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 3.00 | 2.98 | -0.70% | 3.00 | 2.98 | -0.70% | 3.00 | 2.98 | 0.56% |
| 1MB | 3.00 | 3.00 | -0.01% | 3.00 | 3.00 | -0.01% | 3.00 | 3.00 | 0.01% |
| 4MB | 3.00 | 3.00 | 0.00% | 3.00 | 3.00 | -0.02% | 3.00 | 3.00 | 0.01% |
| 1GB | 3.00 | 3.00 | 0.00% | 3.00 | 3.00 | -0.01% | 3.00 | 3.00 | 0.01% |
| 12kB | 3.00 | 2.98 | 0.70% | 3.00 | 2.98 | 0.70% | 3.00 | 2.98 | 0.70% |
| 1MB | 3.00 | 3.00 | 0.01% | 3.00 | 3.00 | 0.01% | 3.00 | 3.00 | 0.01% |
| 4MB | 3.00 | 3.00 | 0.00% | 3.00 | 3.00 | 0.02% | 3.00 | 3.00 | 0.01% |
| 1GB | 3.00 | 3.00 | 0.00% | 3.00 | 3.00 | 0.04% | 3.00 | 3.00 | 0.01% |
#### Verification of Group DATA with Test copy
......@@ -305,10 +305,10 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 1.00 | 1.01 | 0.53% | 1.00 | 1.01 | 0.53% | 1.00 | 1.01 | 0.42% |
| 1MB | 1.00 | 1.00 | 0.01% | 1.00 | 1.00 | -0.03% | 1.00 | 1.00 | 0.01% |
| 4MB | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | -0.01% | 1.00 | 1.00 | 0.00% |
| 1GB | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | -0.01% | 1.00 | 1.00 | 0.00% |
| 12kB | 1.00 | 1.01 | 0.53% | 1.00 | 1.01 | 0.53% | 1.00 | 1.01 | 0.53% |
| 1MB | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | 0.01% | 1.00 | 1.00 | 0.00% |
| 4MB | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | 0.05% | 1.00 | 1.00 | 0.02% |
| 1GB | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | 0.00% | 1.00 | 1.00 | 0.00% |
#### Verification of Group DATA with Test stream
......@@ -325,10 +325,10 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 12kB | 2.00 | 2.00 | -0.10% | 2.00 | 2.00 | -0.10% | 2.00 | 2.00 | 0.08% |
| 1MB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | -0.02% | 2.00 | 2.00 | 0.01% |
| 4MB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | -0.03% | 2.00 | 2.00 | 0.01% |
| 1GB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | -0.01% | 2.00 | 2.00 | 0.01% |
| 12kB | 2.00 | 2.00 | 0.10% | 2.00 | 2.00 | 0.10% | 2.00 | 2.00 | 0.10% |
| 1MB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | 0.01% | 2.00 | 2.00 | 0.01% |
| 4MB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | 0.02% | 2.00 | 2.00 | 0.01% |
| 1GB | 2.00 | 2.00 | 0.00% | 2.00 | 2.00 | 0.01% | 2.00 | 2.00 | 0.00% |
#### Verification of Group DATA with Test store
......@@ -353,7 +353,7 @@ Each data size is tested 5 times, hence the first 5 entries on the x-axis corres
### Verification of Group FLOPS_AVX
#### Verification of Group FLOPS_AVX with Test triad_avx
#### Verification of Group FLOPS_AVX with Test ddot_avx
Benchmark performed with data sizes: 24kB, 128kB, 2MB, 1GB.
......@@ -362,18 +362,18 @@ Each data size is tested 10 times, hence the first 10 entries on the x-axis corr
![haswell_FLOPS_AVX_triad_avx](./images/accuracy/haswell/FLOPS_AVX_triad_avx.jpg)
![haswell_FLOPS_AVX_ddot_avx](./images/accuracy/haswell/FLOPS_AVX_ddot_avx.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 24kB | 12122.00 | 12124.39 | 0.02% | 12093.00 | 12096.26 | 0.03% | 12059.10 | 12061.82 | 0.02% |
| 128kB | 4366.00 | 4366.53 | 0.01% | 4409.00 | 4410.37 | 0.03% | 4393.30 | 4394.30 | 0.02% |
| 2MB | 2542.00 | 2542.52 | 0.02% | 2539.00 | 2540.24 | 0.05% | 2541.40 | 2542.28 | 0.03% |
| 1GB | 951.00 | 951.12 | 0.01% | 945.00 | 946.04 | 0.11% | 947.30 | 947.99 | 0.07% |
| 24kB | 9745.00 | 9759.77 | 0.15% | 9731.00 | 9747.19 | 0.17% | 9740.60 | 9755.90 | 0.16% |
| 128kB | 9276.00 | 9279.53 | 0.04% | 9292.00 | 9296.52 | 0.05% | 9175.60 | 9179.47 | 0.04% |
| 2MB | 5789.00 | 5789.92 | 0.02% | 5729.00 | 5730.72 | 0.03% | 5759.60 | 5760.82 | 0.02% |
| 1GB | 2109.00 | 2109.35 | 0.02% | 2000.00 | 2001.31 | 0.07% | 2139.70 | 2140.41 | 0.03% |
#### Verification of Group FLOPS_AVX with Test stream_avx
#### Verification of Group FLOPS_AVX with Test triad_avx
Benchmark performed with data sizes: 24kB, 128kB, 2MB, 1GB.
......@@ -382,18 +382,18 @@ Each data size is tested 10 times, hence the first 10 entries on the x-axis corr
![haswell_FLOPS_AVX_stream_avx](./images/accuracy/haswell/FLOPS_AVX_stream_avx.jpg)
![haswell_FLOPS_AVX_triad_avx](./images/accuracy/haswell/FLOPS_AVX_triad_avx.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 24kB | 17166.00 | 17169.52 | 0.02% | 17105.00 | 17109.67 | 0.03% | 17103.70 | 17107.88 | 0.02% |
| 128kB | 5464.00 | 5464.66 | 0.01% | 5508.00 | 5509.45 | 0.03% | 5471.80 | 5472.82 | 0.02% |
| 2MB | 3176.00 | 3176.42 | 0.01% | 3176.00 | 3177.31 | 0.04% | 3175.10 | 3175.90 | 0.02% |
| 1GB | 1196.00 | 1196.20 | 0.02% | 1192.00 | 1193.12 | 0.09% | 1193.30 | 1193.94 | 0.05% |
| 24kB | 12106.00 | 12109.04 | 0.03% | 12108.00 | 12112.00 | 0.03% | 12084.20 | 12087.87 | 0.03% |
| 128kB | 4138.00 | 4138.76 | 0.02% | 4186.00 | 4187.49 | 0.04% | 4326.10 | 4327.27 | 0.03% |
| 2MB | 2528.00 | 2528.36 | 0.01% | 2543.00 | 2544.08 | 0.04% | 2535.90 | 2536.59 | 0.03% |
| 1GB | 962.00 | 962.20 | 0.02% | 962.00 | 963.11 | 0.12% | 960.00 | 960.68 | 0.07% |
#### Verification of Group FLOPS_AVX with Test daxpy_avx
#### Verification of Group FLOPS_AVX with Test striad_avx
Benchmark performed with data sizes: 24kB, 128kB, 2MB, 1GB.
......@@ -402,18 +402,18 @@ Each data size is tested 10 times, hence the first 10 entries on the x-axis corr
![haswell_FLOPS_AVX_daxpy_avx](./images/accuracy/haswell/FLOPS_AVX_daxpy_avx.jpg)
![haswell_FLOPS_AVX_striad_avx](./images/accuracy/haswell/FLOPS_AVX_striad_avx.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 24kB | 17383.00 | 17385.86 | 0.02% | 17203.00 | 17207.05 | 0.02% | 17355.90 | 17359.44 | 0.02% |
| 128kB | 6726.00 | 6726.94 | 0.01% | 6595.00 | 6596.53 | 0.02% | 6692.70 | 6694.00 | 0.02% |
| 2MB | 4228.00 | 4228.46 | 0.01% | 4229.00 | 4230.40 | 0.03% | 4230.80 | 4231.88 | 0.02% |
| 1GB | 1608.00 | 1608.39 | 0.02% | 1609.00 | 1610.17 | 0.07% | 1609.20 | 1609.98 | 0.04% |
| 24kB | 962.00 | 962.20 | 0.00% | 962.00 | 963.11 | 0.00% | 0.00 | 0.00 | 0.00% |
| 128kB | 962.00 | 962.20 | 0.00% | 962.00 | 963.11 | 0.00% | 0.00 | 0.00 | 0.00% |
| 2MB | 962.00 | 962.20 | 0.00% | 962.00 | 963.11 | 0.00% | 0.00 | 0.00 | 0.00% |
| 1GB | 962.00 | 962.20 | 0.00% | 962.00 | 963.11 | 0.00% | 0.00 | 0.00 | 0.00% |
#### Verification of Group FLOPS_AVX with Test ddot_avx
#### Verification of Group FLOPS_AVX with Test daxpy_avx
Benchmark performed with data sizes: 24kB, 128kB, 2MB, 1GB.
......@@ -422,15 +422,15 @@ Each data size is tested 10 times, hence the first 10 entries on the x-axis corr
![haswell_FLOPS_AVX_ddot_avx](./images/accuracy/haswell/FLOPS_AVX_ddot_avx.jpg)
![haswell_FLOPS_AVX_daxpy_avx](./images/accuracy/haswell/FLOPS_AVX_daxpy_avx.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 24kB | 9731.00 | 9745.46 | 0.15% | 9669.00 | 9684.49 | 0.16% | 9725.90 | 9740.91 | 0.14% |
| 128kB | 9300.00 | 9303.34 | 0.04% | 9249.00 | 9253.25 | 0.05% | 9282.00 | 9285.78 | 0.04% |
| 2MB | 5738.00 | 5738.82 | 0.01% | 5741.00 | 5742.66 | 0.03% | 5784.20 | 5785.43 | 0.02% |
| 1GB | 2202.00 | 2202.32 | 0.01% | 2192.00 | 2193.18 | 0.05% | 2192.50 | 2193.33 | 0.04% |
| 24kB | 17413.00 | 17416.73 | 0.02% | 17318.00 | 17322.88 | 0.03% | 17364.10 | 17368.53 | 0.03% |
| 128kB | 6695.00 | 6695.91 | 0.01% | 6678.00 | 6679.68 | 0.03% | 6680.70 | 6681.99 | 0.02% |
| 2MB | 4235.00 | 4235.59 | 0.01% | 4225.00 | 4226.39 | 0.03% | 4234.70 | 4235.67 | 0.02% |
| 1GB | 1611.00 | 1611.65 | 0.04% | 1614.00 | 1615.14 | 0.07% | 1618.30 | 1619.16 | 0.05% |
### Verification of Group L2
......@@ -438,9 +438,9 @@ Each data size is tested 10 times, hence the first 10 entries on the x-axis corr
#### Verification of Group L2 with Test load
Benchmark performed with data sizes: 1263kB, 2485kB, 3707kB, 4929kB.
Benchmark performed with data sizes: 85kB, 129kB, 173kB, 217kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 1263kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 85kB and so on.
......@@ -450,18 +450,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 1263kB | 29982.00 | 30191.52 | 0.70% | 29889.00 | 30110.24 | 0.74% | 30005.00 | 30218.57 | 0.57% |
| 2485kB | 30522.00 | 30658.55 | 0.45% | 30068.00 | 30218.23 | 0.50% | 30223.60 | 30366.41 | 0.38% |
| 3707kB | 30032.00 | 30138.36 | 0.35% | 29790.00 | 29910.56 | 0.40% | 29977.40 | 30088.09 | 0.30% |
| 4929kB | 30026.00 | 30112.42 | 0.29% | 29982.00 | 30070.73 | 0.30% | 30029.40 | 30116.80 | 0.23% |
| 85kB | 30017.00 | 30225.26 | 0.69% | 29940.00 | 30149.08 | 0.70% | 29968.00 | 30176.65 | 0.70% |
| 129kB | 30127.00 | 30268.68 | 0.47% | 29607.00 | 29795.07 | 0.64% | 30010.80 | 30163.65 | 0.51% |
| 173kB | 30062.00 | 30169.04 | 0.36% | 29978.00 | 30090.66 | 0.38% | 30048.60 | 30158.38 | 0.37% |
| 217kB | 30045.00 | 30132.34 | 0.29% | 29947.00 | 30052.79 | 0.35% | 30002.20 | 30094.70 | 0.31% |
#### Verification of Group L2 with Test triad
Benchmark performed with data sizes: 1263kB, 2485kB, 3707kB, 4929kB.
Benchmark performed with data sizes: 85kB, 129kB, 173kB, 217kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 1263kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 85kB and so on.
......@@ -471,18 +471,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 1263kB | 61258.75 | 63016.89 | 2.87% | 61191.25 | 63062.18 | 3.06% | 61612.00 | 63450.31 | 2.36% |
| 2485kB | 60535.00 | 61963.08 | 2.36% | 61886.25 | 63613.59 | 2.79% | 60761.50 | 62353.58 | 2.14% |
| 3707kB | 61212.50 | 62722.72 | 2.47% | 60452.50 | 62088.82 | 2.71% | 61368.00 | 62977.99 | 2.08% |
| 4929kB | 57853.75 | 59110.94 | 2.17% | 58141.25 | 59457.71 | 2.26% | 58107.50 | 59390.41 | 1.77% |
| 85kB | 59615.00 | 60995.19 | 2.32% | 61861.25 | 63754.02 | 3.06% | 61055.00 | 62759.23 | 2.79% |
| 129kB | 61390.00 | 62952.89 | 2.55% | 61998.75 | 63859.00 | 3.00% | 61657.25 | 63356.52 | 2.76% |
| 173kB | 58922.50 | 60322.95 | 2.38% | 61640.00 | 63331.23 | 2.74% | 60628.75 | 62191.40 | 2.58% |
| 217kB | 60598.75 | 61910.11 | 2.16% | 60471.25 | 61980.18 | 2.50% | 60244.00 | 61638.01 | 2.31% |
#### Verification of Group L2 with Test copy
Benchmark performed with data sizes: 1263kB, 2485kB, 3707kB, 4929kB.
Benchmark performed with data sizes: 85kB, 129kB, 173kB, 217kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 1263kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 85kB and so on.
......@@ -492,18 +492,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 1263kB | 70135.50 | 70506.07 | 0.53% | 70734.00 | 71115.22 | 0.54% | 70322.10 | 70697.91 | 0.43% |
| 2485kB | 71095.50 | 71372.97 | 0.39% | 71367.00 | 71648.23 | 0.39% | 69151.20 | 69414.51 | 0.31% |
| 3707kB | 68998.50 | 69195.84 | 0.29% | 72034.50 | 72248.69 | 0.30% | 70388.40 | 70595.71 | 0.24% |
| 4929kB | 67662.00 | 67802.36 | 0.21% | 68779.50 | 68926.09 | 0.21% | 68329.80 | 68474.66 | 0.17% |
| 85kB | 70909.50 | 71279.37 | 0.52% | 70821.00 | 71192.12 | 0.52% | 71042.70 | 71413.90 | 0.52% |
| 129kB | 69058.50 | 69285.61 | 0.33% | 71679.00 | 71963.22 | 0.40% | 71064.60 | 71335.96 | 0.38% |
| 173kB | 68892.00 | 69095.22 | 0.29% | 70830.00 | 71045.36 | 0.30% | 70979.40 | 71190.63 | 0.30% |
| 217kB | 66747.00 | 66883.16 | 0.20% | 70054.50 | 70208.20 | 0.22% | 68190.30 | 68336.77 | 0.21% |
#### Verification of Group L2 with Test stream
Benchmark performed with data sizes: 1263kB, 2485kB, 3707kB, 4929kB.
Benchmark performed with data sizes: 85kB, 129kB, 173kB, 217kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 1263kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 85kB and so on.
......@@ -513,18 +513,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 1263kB | 63022.42 | 63930.60 | 1.44% | 63043.76 | 64039.84 | 1.58% | 62648.30 | 63602.62 | 1.20% |
| 2485kB | 61347.80 | 62031.71 | 1.11% | 62247.78 | 63003.76 | 1.21% | 62727.23 | 63470.47 | 0.94% |
| 3707kB | 61837.12 | 62449.21 | 0.99% | 62825.10 | 63488.80 | 1.06% | 62608.83 | 63255.57 | 0.83% |
| 4929kB | 58482.54 | 58957.92 | 0.81% | 61053.14 | 61589.59 | 0.88% | 60320.89 | 60827.59 | 0.68% |
| 85kB | 63219.75 | 64180.30 | 1.52% | 63131.75 | 64155.62 | 1.62% | 63123.76 | 64112.38 | 1.57% |
| 129kB | 62913.09 | 63682.09 | 1.22% | 63425.08 | 64251.90 | 1.30% | 63347.75 | 64140.59 | 1.25% |
| 173kB | 61491.80 | 62094.36 | 0.98% | 63418.41 | 64103.10 | 1.08% | 62635.23 | 63273.94 | 1.02% |
| 217kB | 58602.53 | 59091.37 | 0.83% | 63357.08 | 63952.81 | 0.94% | 60962.21 | 61498.71 | 0.88% |
#### Verification of Group L2 with Test store
Benchmark performed with data sizes: 1263kB, 2485kB, 3707kB, 4929kB.
Benchmark performed with data sizes: 85kB, 129kB, 173kB, 217kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 1263kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 85kB and so on.
......@@ -534,10 +534,10 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 1263kB | 54404.00 | 54542.20 | 0.25% | 54772.00 | 54940.85 | 0.31% | 54694.80 | 54857.63 | 0.24% |
| 2485kB | 54990.00 | 55117.53 | 0.23% | 54870.00 | 54998.78 | 0.23% | 54751.60 | 54879.75 | 0.19% |
| 3707kB | 54620.00 | 54687.69 | 0.12% | 54878.00 | 54964.53 | 0.16% | 54814.40 | 54897.09 | 0.12% |
| 4929kB | 54606.00 | 54678.02 | 0.13% | 51196.00 | 51268.10 | 0.14% | 53521.20 | 53594.82 | 0.11% |
| 85kB | 54894.00 | 55060.25 | 0.30% | 54928.00 | 55100.67 | 0.31% | 54895.20 | 55065.69 | 0.31% |
| 129kB | 54840.00 | 54957.11 | 0.21% | 54792.00 | 54918.82 | 0.23% | 54820.00 | 54941.93 | 0.22% |
| 173kB | 54014.00 | 54031.60 | 0.03% | 54904.00 | 54985.64 | 0.15% | 54646.80 | 54714.07 | 0.12% |
| 217kB | 52796.00 | 52798.99 | 0.01% | 53082.00 | 53156.99 | 0.14% | 53202.40 | 53241.88 | 0.07% |
### Verification of Group L3
......@@ -545,9 +545,9 @@ The table contains the values after scaling the original values up using the sca
#### Verification of Group L3 with Test load
Benchmark performed with data sizes: 216473kB, 424959kB, 633445kB, 841931kB.
Benchmark performed with data sizes: 2738kB, 5144kB, 7550kB, 9956kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 216473kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 2738kB and so on.
......@@ -557,18 +557,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 216473kB | 29849.00 | 29874.82 | 0.09% | 29762.00 | 29800.20 | 0.13% | 29814.20 | 29844.18 | 0.08% |
| 424959kB | 24900.00 | 24964.61 | 0.26% | 25000.00 | 25067.64 | 0.27% | 25055.40 | 25121.63 | 0.21% |
| 633445kB | 21200.00 | 21252.83 | 0.25% | 20901.00 | 20959.90 | 0.28% | 21037.20 | 21094.01 | 0.21% |
| 841931kB | 18575.00 | 18625.81 | 0.27% | 19021.00 | 19085.54 | 0.34% | 19067.80 | 19123.58 | 0.24% |
| 2738kB | 29404.00 | 29434.23 | 0.10% | 28940.00 | 28996.57 | 0.20% | 29403.40 | 29440.33 | 0.13% |
| 5144kB | 27854.00 | 27927.26 | 0.26% | 25783.00 | 25859.28 | 0.30% | 27246.00 | 27321.15 | 0.28% |
| 7550kB | 24263.00 | 24326.36 | 0.26% | 24638.00 | 24707.56 | 0.28% | 24370.20 | 24436.19 | 0.27% |
| 9956kB | 21621.00 | 21677.32 | 0.26% | 21398.00 | 21462.88 | 0.30% | 21487.60 | 21547.40 | 0.28% |
#### Verification of Group L3 with Test triad
Benchmark performed with data sizes: 216473kB, 424959kB, 633445kB, 841931kB.
Benchmark performed with data sizes: 2738kB, 5144kB, 7550kB, 9956kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 216473kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 2738kB and so on.
......@@ -578,18 +578,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 216473kB | 53485.29 | 50079.68 | -6.37% | 54103.80 | 50643.50 | -6.40% | 53929.18 | 50485.75 | 5.11% |
| 424959kB | 45588.60 | 42681.46 | -6.38% | 44298.26 | 41393.14 | -6.56% | 44973.55 | 42057.65 | 5.17% |
| 633445kB | 31336.16 | 28951.01 | -7.61% | 31718.74 | 29241.28 | -7.81% | 31659.82 | 29221.99 | 6.17% |
| 841931kB | 24355.24 | 22171.71 | -8.97% | 24036.66 | 21771.98 | -9.42% | 24281.93 | 22048.72 | 7.35% |
| 2738kB | 53662.58 | 50250.08 | 6.36% | 53265.35 | 49861.80 | 6.39% | 53715.63 | 50289.00 | 6.38% |
| 5144kB | 53130.71 | 49816.08 | 6.24% | 52438.89 | 49153.44 | 6.27% | 52597.25 | 49308.82 | 6.25% |
| 7550kB | 43221.19 | 40386.14 | 6.56% | 42084.14 | 39278.56 | 6.67% | 43282.78 | 40427.37 | 6.60% |
| 9956kB | 34843.29 | 32308.23 | 7.28% | 32530.53 | 30090.15 | 7.50% | 34043.75 | 31517.41 | 7.42% |
#### Verification of Group L3 with Test copy
Benchmark performed with data sizes: 216473kB, 424959kB, 633445kB, 841931kB.
Benchmark performed with data sizes: 2738kB, 5144kB, 7550kB, 9956kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 216473kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 2738kB and so on.
......@@ -599,18 +599,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 216473kB | 53052.00 | 52643.91 | -0.77% | 53317.50 | 52884.67 | -0.81% | 52810.20 | 52396.08 | 0.63% |
| 424959kB | 44575.50 | 44114.58 | -1.03% | 44871.00 | 44369.52 | -1.12% | 44261.70 | 43780.72 | 0.85% |
| 633445kB | 26823.00 | 26663.11 | -0.60% | 30517.50 | 29650.14 | -2.84% | 30302.10 | 29623.23 | 1.71% |
| 841931kB | 24465.00 | 23401.80 | -4.35% | 23302.50 | 22231.72 | -4.60% | 24141.60 | 23065.88 | 3.57% |
| 2738kB | 52758.00 | 52368.79 | 0.74% | 53038.50 | 52601.42 | 0.82% | 53020.80 | 52607.12 | 0.78% |
| 5144kB | 50281.50 | 49964.82 | 0.63% | 51748.50 | 51377.63 | 0.72% | 51367.80 | 51025.23 | 0.67% |
| 7550kB | 42813.00 | 42578.89 | 0.55% | 42984.00 | 42735.88 | 0.58% | 42413.40 | 42176.07 | 0.56% |
| 9956kB | 28399.50 | 28233.12 | 0.59% | 32791.50 | 32017.53 | 2.36% | 30515.70 | 30112.54 | 1.26% |
#### Verification of Group L3 with Test stream
Benchmark performed with data sizes: 216473kB, 424959kB, 633445kB, 841931kB.
Benchmark performed with data sizes: 2738kB, 5144kB, 7550kB, 9956kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 216473kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 2738kB and so on.
......@@ -620,18 +620,18 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 216473kB | 50929.93 | 50869.62 | -0.12% | 50977.92 | 50906.66 | -0.14% | 50900.87 | 50836.44 | 0.10% |
| 424959kB | 43997.00 | 43916.35 | -0.18% | 44324.92 | 44186.80 | -0.31% | 44414.23 | 44307.24 | 0.19% |
| 633445kB | 30940.26 | 30430.21 | -1.65% | 30961.59 | 30364.38 | -1.93% | 30893.34 | 30342.43 | 1.43% |
| 841931kB | 23343.50 | 22568.57 | -3.32% | 23360.83 | 22539.85 | -3.51% | 23347.50 | 22545.19 | 2.74% |
| 2738kB | 50700.65 | 50641.70 | 0.12% | 51148.54 | 51075.05 | 0.14% | 50809.69 | 50745.89 | 0.13% |
| 5144kB | 46575.02 | 46593.72 | 0.04% | 49890.19 | 49928.31 | 0.08% | 49045.60 | 49074.95 | 0.06% |
| 7550kB | 40524.53 | 40493.64 | 0.08% | 42505.37 | 42341.35 | 0.39% | 41911.92 | 41783.27 | 0.30% |
| 9956kB | 33146.38 | 32702.68 | 1.34% | 30712.32 | 30163.74 | 1.79% | 32462.82 | 31966.37 | 1.53% |
#### Verification of Group L3 with Test store
Benchmark performed with data sizes: 216473kB, 424959kB, 633445kB, 841931kB.
Benchmark performed with data sizes: 2738kB, 5144kB, 7550kB, 9956kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 216473kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 2738kB and so on.
......@@ -641,51 +641,91 @@ The table contains the values after scaling the original values up using the sca
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 216473kB | 49448.00 | 49425.03 | -0.05% | 48412.00 | 48267.43 | -0.30% | 49028.80 | 48952.52 | 0.12% |
| 424959kB | 41018.00 | 40911.35 | -0.26% | 39620.00 | 38774.11 | -2.14% | 42063.20 | 41648.88 | 0.80% |
| 633445kB | 32838.00 | 31715.07 | -3.42% | 29584.00 | 28285.26 | -4.39% | 31995.60 | 30781.99 | 3.10% |
| 841931kB | 26528.00 | 24852.61 | -6.32% | 23622.00 | 21867.85 | -7.43% | 25873.60 | 24167.42 | 5.30% |
| 2738kB | 48922.00 | 48858.48 | 0.13% | 49218.00 | 49006.95 | 0.43% | 49137.20 | 48993.43 | 0.29% |
| 5144kB | 48028.00 | 47972.96 | 0.11% | 47688.00 | 47509.31 | 0.37% | 47530.00 | 47419.73 | 0.23% |
| 7550kB | 39904.00 | 39848.63 | 0.14% | 43046.00 | 42662.92 | 0.89% | 42693.60 | 42403.49 | 0.67% |
| 9956kB | 29090.00 | 28862.86 | 0.78% | 33750.00 | 32593.58 | 3.43% | 33204.80 | 32358.25 | 2.50% |
### Verification of Group UOPS_RETIRED_ANY
#### Verification of Group UOPS_RETIRED_ANY with Test triad
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_UOPS_RETIRED_ANY_triad](./images/accuracy/haswell/UOPS_RETIRED_ANY_triad.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 223231kB | 3400784400.00 | 3403678000.00 | 0.09% | 3400784400.00 | 3403809000.00 | 0.09% | 3390320448.00 | 3393246400.00 | 0.09% |
| 430488kB | 3329555130.00 | 3332321000.00 | 0.08% | 3329555130.00 | 3332513000.00 | 0.09% | 3329555130.00 | 3332410800.00 | 0.09% |
| 637745kB | 3288372120.00 | 3291221000.00 | 0.09% | 2989429200.00 | 2992958000.00 | 0.12% | 3198689244.00 | 3201830800.00 | 0.10% |
| 845002kB | 3168757440.00 | 3171517000.00 | 0.09% | 3168757440.00 | 3171806000.00 | 0.10% | 3168757440.00 | 3171626000.00 | 0.09% |
#### Verification of Group UOPS_RETIRED_ANY with Test daxpy
### Verification of Group UOPS
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_UOPS_RETIRED_ANY_daxpy](./images/accuracy/haswell/UOPS_RETIRED_ANY_daxpy.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 223231kB | 4987817120.00 | 4990520000.00 | 0.05% | 5078504704.00 | 5081333000.00 | 0.06% | 5005954636.80 | 5008702400.00 | 0.05% |
| 430488kB | 5071686750.00 | 5074408000.00 | 0.05% | 5071686750.00 | 5074549000.00 | 0.06% | 5071686750.00 | 5074484600.00 | 0.06% |
| 637745kB | 4922593910.00 | 4925059000.00 | 0.05% | 4922593910.00 | 4925821000.00 | 0.07% | 4922593910.00 | 4925264600.00 | 0.05% |
| 845002kB | 4805948784.00 | 4808553000.00 | 0.05% | 4805948784.00 | 4808690000.00 | 0.06% | 4805948784.00 | 4808589000.00 | 0.05% |
#### Verification of Group UOPS with Test ddot_uops
#### Verification of Group UOPS_RETIRED_ANY with Test ddot
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_UOPS_ddot_uops](./images/accuracy/haswell/UOPS_ddot_uops.jpg)
![haswell_UOPS_RETIRED_ANY_ddot](./images/accuracy/haswell/UOPS_RETIRED_ANY_ddot.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 5775706837.00 | 5779137000.00 | 0.06% | 5775706837.00 | 5779540000.00 | 0.07% | 5704621214.60 | 5708047600.00 | 0.05% |
| 426801kB | 5895188277.00 | 5897724000.00 | 0.04% | 5721800387.00 | 5725504000.00 | 0.06% | 5583090075.00 | 5586261400.00 | 0.04% |
| 634877kB | 5932131911.00 | 5934641000.00 | 0.04% | 5932131911.00 | 5935153000.00 | 0.05% | 5932131911.00 | 5934848400.00 | 0.04% |
| 842953kB | 6164093537.00 | 6166727000.00 | 0.04% | 5479194257.00 | 5481666000.00 | 0.05% | 5821643897.00 | 5824162600.00 | 0.03% |
| 223231kB | 4394859840.00 | 4397519000.00 | 0.06% | 4269292416.00 | 4272523000.00 | 0.08% | 4332076128.00 | 4334914400.00 | 0.07% |
| 430488kB | 4358691000.00 | 4361284000.00 | 0.06% | 3753317250.00 | 3756928000.00 | 0.10% | 3922821900.00 | 3926081200.00 | 0.08% |
| 637745kB | 4304778480.00 | 4307471000.00 | 0.06% | 3587315400.00 | 3590248000.00 | 0.08% | 3946046940.00 | 3948902600.00 | 0.07% |
| 845002kB | 4277822544.00 | 4280637000.00 | 0.07% | 4277822544.00 | 4280716000.00 | 0.07% | 4135228459.20 | 4137979800.00 | 0.07% |
#### Verification of Group UOPS with Test stream_uops
#### Verification of Group UOPS_RETIRED_ANY with Test stream
Benchmark performed with data sizes: 218725kB, 426801kB, 634877kB, 842953kB.
Benchmark performed with data sizes: 223231kB, 430488kB, 637745kB, 845002kB.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 218725kB and so on.
Each data size is tested 5 times, hence the first 5 entries on the x-axis correspond to the 5 runs for the first data size of 223231kB and so on.
![haswell_UOPS_stream_uops](./images/accuracy/haswell/UOPS_stream_uops.jpg)
![haswell_UOPS_RETIRED_ANY_stream](./images/accuracy/haswell/UOPS_RETIRED_ANY_stream.jpg)
| Variant | BENCH (Min) | LIKWID (Min) | Error (Min) | BENCH (Max) | LIKWID (Max) | Error (Max) | BENCH (Avg) | LIKWID (Avg) | Error (Avg) |
| ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| 218725kB | 5541032337.00 | 5543598000.00 | 0.05% | 5541032337.00 | 5543753000.00 | 0.05% | 5523716611.00 | 5526332000.00 | 0.04% |
| 426801kB | 5575087139.00 | 5577494000.00 | 0.04% | 5575087139.00 | 5577806000.00 | 0.05% | 5575087139.00 | 5577633800.00 | 0.04% |
| 634877kB | 5528720489.00 | 5531259000.00 | 0.05% | 5277415013.00 | 5280510000.00 | 0.06% | 5428198298.60 | 5430898800.00 | 0.04% |
| 842953kB | 5338702097.00 | 5340972000.00 | 0.04% | 5338702097.00 | 5341667000.00 | 0.06% | 5338702097.00 | 5341234400.00 | 0.04% |
| 223231kB | 3808877436.00 | 3811576000.00 | 0.07% | 3869335808.00 | 3872163000.00 | 0.07% | 3820969110.40 | 3823728400.00 | 0.07% |
| 430488kB | 3847486500.00 | 3850204000.00 | 0.07% | 3614305500.00 | 3617084000.00 | 0.08% | 3777532200.00 | 3780310600.00 | 0.07% |
| 637745kB | 3799897244.00 | 3802632000.00 | 0.07% | 3799897244.00 | 3802703000.00 | 0.07% | 3799897244.00 | 3802662400.00 | 0.07% |
| 845002kB | 3661675264.00 | 3664309000.00 | 0.07% | 3661675264.00 | 3664795000.00 | 0.09% | 3661675264.00 | 3664549400.00 | 0.08% |