JSON parser integration

[karthikeyann]

trafficstars

Description

Integrates the experimental full gpu json parser. Replaces existing cpu json parser(+ gpu tokenizer)

All test passes and no cuda memory errors.

Benchmark Results

nested_json_gpu_parser

[0] Quadro GV100

string_size	Samples	CPU Time	Noise	GPU Time	Noise	Elem/s
2^20 = 1048576	1976x	7.572 ms	7.25%	7.565 ms	7.25%	138.608M
2^21 = 2097152	256x	9.088 ms	15.72%	9.082 ms	15.72%	230.924M
2^22 = 4194304	880x	11.419 ms	14.07%	11.413 ms	14.07%	367.518M
2^23 = 8388608	80x	17.619 ms	9.65%	17.612 ms	9.65%	476.293M
2^24 = 16777216	500x	30.022 ms	5.33%	30.015 ms	5.33%	558.957M
2^25 = 33554432	286x	52.556 ms	4.52%	52.549 ms	4.52%	638.533M
2^26 = 67108864	11x	100.289 ms	0.40%	100.282 ms	0.40%	669.200M
2^27 = 134217728	77x	195.659 ms	1.45%	195.654 ms	1.45%	685.994M
2^28 = 268435456	39x	387.014 ms	0.58%	387.013 ms	0.58%	693.609M
2^29 = 536870912	11x	773.444 ms	0.38%	773.448 ms	0.38%	694.126M
2^30 = 1073741824	10x	1.554 s	0.29%	1.554 s	0.29%	691.001M

Checklist

• [x] I am familiar with the Contributing Guidelines.
• [ ] New or existing tests cover these changes.
• [ ] The documentation is up to date with these changes.

[codecov[bot]]

Codecov Report

:exclamation: No coverage uploaded for pull request base (branch-22.10@466a90d). Click here to learn what that means. Patch has no changes to coverable lines.

:exclamation: Current head 87f5575 differs from pull request most recent head 9486c20. Consider uploading reports for the commit 9486c20 to get more accurate results

Additional details and impacted files

@@               Coverage Diff               @@
##             branch-22.10   #11717   +/-   ##
===============================================
  Coverage                ?   87.51%           
===============================================
  Files                   ?      133           
  Lines                   ?    21803           
  Branches                ?        0           
===============================================
  Hits                    ?    19080           
  Misses                  ?     2723           
  Partials                ?        0           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.

[karthikeyann]

Benchmark Results

Dated 19-Sept-2022

nested_json_gpu_parser

[0] Quadro GV100

string_size	Samples	CPU Time	Noise	GPU Time	Noise	Elem/s
2^20 = 1048576	1976x	7.572 ms	7.25%	7.565 ms	7.25%	138.608M
2^21 = 2097152	256x	9.088 ms	15.72%	9.082 ms	15.72%	230.924M
2^22 = 4194304	880x	11.419 ms	14.07%	11.413 ms	14.07%	367.518M
2^23 = 8388608	80x	17.619 ms	9.65%	17.612 ms	9.65%	476.293M
2^24 = 16777216	500x	30.022 ms	5.33%	30.015 ms	5.33%	558.957M
2^25 = 33554432	286x	52.556 ms	4.52%	52.549 ms	4.52%	638.533M
2^26 = 67108864	11x	100.289 ms	0.40%	100.282 ms	0.40%	669.200M
2^27 = 134217728	77x	195.659 ms	1.45%	195.654 ms	1.45%	685.994M
2^28 = 268435456	39x	387.014 ms	0.58%	387.013 ms	0.58%	693.609M
2^29 = 536870912	11x	773.444 ms	0.38%	773.448 ms	0.38%	694.126M
2^30 = 1073741824	10x	1.554 s	0.29%	1.554 s	0.29%	691.001M

[karthikeyann]

@karthikeyann NC_VAL to NC_STR in token to tree conversion itself c89a982 this effectively remove NC_VAL in tree nodes completely context: duration types may have mixed literal and string types. JSON_TEST JsonReaderParamTest Durations failed because it ignored the literals.

(Note: Idea, it could be done without loosing NC_VAL by hashing NC_VAL and NC_STR to same node_type in hashing in tree traversal code).

[karthikeyann]

Benchmark Results

Dated 21-Sept-2022

nested_json_gpu_parser

[0] Quadro GV100

string_size	Samples	CPU Time	Noise	GPU Time	Noise	Elem/s
2^20 = 1048576	2501x	5.975 ms	7.89%	5.968 ms	7.88%	175.689M
2^21 = 2097152	672x	6.677 ms	4.13%	6.670 ms	4.13%	314.402M
2^22 = 4194304	592x	8.323 ms	4.69%	8.316 ms	4.69%	504.346M
2^23 = 8388608	1008x	12.778 ms	13.18%	12.771 ms	13.18%	656.833M
2^24 = 16777216	750x	19.987 ms	5.99%	19.980 ms	5.99%	839.719M
2^25 = 33554432	80x	37.238 ms	4.29%	37.230 ms	4.29%	901.272M
2^26 = 67108864	220x	68.162 ms	1.98%	68.154 ms	1.98%	984.658M
2^27 = 134217728	80x	131.745 ms	1.09%	131.738 ms	1.09%	1.019G
2^28 = 268435456	11x	260.182 ms	0.35%	260.175 ms	0.35%	1.032G
2^29 = 536870912	11x	520.952 ms	0.19%	520.946 ms	0.19%	1.031G
2^30 = 1073741824	11x	1.057 s	0.27%	1.057 s	0.27%	1.016G

[karthikeyann]

Benchmark Results

Dated 27-Sept-2022

nested_json_gpu_parser

[0] Quadro GV100

string_size	Samples	CPU Time	Noise	GPU Time	Noise	Elem/s	bytes_per_second	peak_memory_usage
2^20 = 1048576	2192x	5.784 ms	11.10%	5.777 ms	11.10%	181.518M	181518461	17.149 MiB
2^21 = 2097152	960x	6.158 ms	5.10%	6.150 ms	5.09%	340.980M	340979979	34.106 MiB
2^22 = 4194304	192x	7.798 ms	15.82%	7.791 ms	15.82%	538.363M	538362570	68.020 MiB
2^23 = 8388608	848x	10.036 ms	4.47%	10.029 ms	4.47%	836.432M	836432150	135.848 MiB
2^24 = 16777216	80x	16.555 ms	9.49%	16.547 ms	9.49%	1.014G	1013889148	271.504 MiB
2^25 = 33554432	552x	27.182 ms	7.52%	27.175 ms	7.52%	1.235G	1234774323	542.817 MiB
2^26 = 67108864	80x	48.934 ms	3.19%	48.927 ms	3.19%	1.372G	1371621284	1.060 GiB
2^27 = 134217728	166x	90.397 ms	1.64%	90.390 ms	1.64%	1.485G	1484872039	2.120 GiB
2^28 = 268435456	87x	172.521 ms	0.99%	172.514 ms	0.99%	1.556G	1556019625	4.239 GiB
2^29 = 536870912	43x	351.175 ms	0.77%	351.170 ms	0.77%	1.529G	1528807741	8.479 GiB
2^30 = 1073741824	11x	667.835 ms	0.18%	667.831 ms	0.18%	1.608G	1607803624	16.957 GiB