[VL] Improving performance by using xsimd::memcpy

[zuochunwei]

What changes were proposed in this pull request?

(Please fill in changes proposed in this fix)

(Fixes: #ISSUE-ID)

How was this patch tested?

(Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests)

(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)

[github-actions]

Thanks for opening a pull request!

Could you open an issue for this pull request on Github Issues?

https://github.com/oap-project/gluten/issues

Then could you also rename commit message and pull request title in the following format?

[GLUTEN-${ISSUES_ID}][COMPONENT]feat/fix: ${detailed message}

See also:

• Other pull requests

[zuochunwei]

/Benchmark Velox

[GlutenPerfBot]

===== Performance report for TPCH SF2000 with Velox backend, for reference only ====

query	log/native_3365_time.csv	log/native_master_10_09_2023_65a72b1c3_time.csv	difference	percentage
q1	42.54	42.87	0.322	100.76%
q2	24.83	24.57	-0.259	98.96%
q3	37.83	37.52	-0.307	99.19%
q4	41.84	41.91	0.071	100.17%
q5	71.52	69.35	-2.173	96.96%
q6	6.56	6.58	0.017	100.26%
q7	84.17	86.66	2.491	102.96%
q8	79.97	81.22	1.252	101.57%
q9	118.31	117.00	-1.309	98.89%
q10	48.94	46.63	-2.309	95.28%
q11	19.32	19.22	-0.100	99.48%
q12	26.20	25.03	-1.171	95.53%
q13	49.50	49.58	0.073	100.15%
q14	19.45	13.15	-6.302	67.61%
q15	28.37	29.50	1.134	104.00%
q16	15.79	16.05	0.263	101.66%
q17	122.23	122.83	0.593	100.49%
q18	165.42	164.00	-1.423	99.14%
q19	15.08	12.01	-3.068	79.65%
q20	27.45	26.66	-0.795	97.10%
q21	236.89	236.75	-0.142	99.94%
q22	15.58	15.77	0.193	101.24%
total	1297.82	1284.87	-12.949	99.00%

[zuochunwei]

/Benchmark Velox

[GlutenPerfBot]

===== Performance report for TPCH SF2000 with Velox backend, for reference only ====

query	log/native_3365_time.csv	log/native_master_10_09_2023_65a72b1c3_time.csv	difference	percentage
q1	41.64	42.87	1.222	102.93%
q2	24.63	24.57	-0.052	99.79%
q3	35.86	37.52	1.668	104.65%
q4	42.94	41.91	-1.024	97.61%
q5	70.62	69.35	-1.266	98.21%
q6	6.52	6.58	0.058	100.89%
q7	84.62	86.66	2.046	102.42%
q8	81.05	81.22	0.174	100.21%
q9	114.74	117.00	2.259	101.97%
q10	46.27	46.63	0.361	100.78%
q11	19.45	19.22	-0.235	98.79%
q12	23.43	25.03	1.598	106.82%
q13	50.01	49.58	-0.427	99.15%
q14	15.87	13.15	-2.719	82.87%
q15	27.40	29.50	2.102	107.67%
q16	15.96	16.05	0.092	100.58%
q17	123.34	122.83	-0.515	99.58%
q18	161.64	164.00	2.357	101.46%
q19	12.13	12.01	-0.116	99.05%
q20	25.59	26.66	1.066	104.16%
q21	236.34	236.75	0.408	100.17%
q22	15.23	15.77	0.540	103.54%
total	1275.27	1284.87	9.597	100.75%

[marin-ma]

Is this PR ready to merge?

[zuochunwei]

/Benchmark Velox

[zuochunwei]

Is this PR ready to merge?

Yes, I removed the WIP tag from this PR's title

[marin-ma]

Looks like the TPCH workload doesn't have much performance improvement. Is this because memcpy is not hotspot in this kind of workload?

[zuochunwei]

Looks like the TPCH workload doesn't have much performance improvement. Is this because memcpy is not hotspot in this kind of workload?

it seems to be right.
the TPCH workload test may be not enough precise.

[GlutenPerfBot]

===== Performance report for TPCH SF2000 with Velox backend, for reference only ====

query	log/native_3365_time.csv	log/native_master_10_12_2023_becfd1d13_time.csv	difference	percentage
q1	43.86	42.54	-1.321	96.99%
q2	23.10	24.58	1.478	106.40%
q3	38.67	37.68	-0.991	97.44%
q4	41.91	41.71	-0.200	99.52%
q5	70.36	69.95	-0.411	99.42%
q6	6.82	5.44	-1.378	79.79%
q7	84.13	86.47	2.344	102.79%
q8	82.09	82.02	-0.066	99.92%
q9	116.39	119.24	2.854	102.45%
q10	47.71	48.00	0.290	100.61%
q11	19.95	19.26	-0.687	96.56%
q12	27.01	24.58	-2.430	91.00%
q13	50.70	52.68	1.984	103.91%
q14	17.94	15.32	-2.620	85.39%
q15	32.22	29.53	-2.683	91.67%
q16	15.84	16.15	0.310	101.96%
q17	120.54	121.78	1.237	101.03%
q18	165.73	163.32	-2.418	98.54%
q19	12.67	12.31	-0.360	97.16%
q20	30.18	27.43	-2.741	90.92%
q21	243.49	241.93	-1.556	99.36%
q22	16.24	15.40	-0.844	94.80%
total	1307.55	1297.34	-10.210	99.22%