* [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
@ 2019-10-14 3:45 Xuelei Zhang
2019-10-15 12:05 ` Szabolcs Nagy
2019-10-16 2:14 ` Yikun Jiang
0 siblings, 2 replies; 13+ messages in thread
From: Xuelei Zhang @ 2019-10-14 3:45 UTC (permalink / raw)
To: libc-alpha
This is an optimized implementation of the memcpy and memmove on the
Huawei Kunpeng processor.
Based on the prefetch mechanism on Kunpeng arch, branch to handle 96
to 2K bytes in memcpy is written without prfm instruction. Hence,
memcpy has an optimization effect above 128 bytes, 18% improvement
for copies above 2K bytes, and 38% for larger bytes, such as 32M
bytes around.
And for memmove, there are two main changes: i) Q register is used
instead of X register. ii) dst address is aligned instead of src
address aligned to improve store operation. Hence, memmove
implementation also has improvement above 128 bytes, that about 30%
for 2k to 8M bytes, and about 50% for 32M or more.
* sysdeps/aarch64/multiarch/Makefile:
Add memcpy_kunpeng and memmove_kunpeng
* sysdeps/aarch64/multiarch/ifunc-impl-list.c:
Increment to 5, add __memcpy_kunpeng and __memmove_kunpeng
* sysdeps/aarch64/multiarch/memcpy.c:
Add IS_KUNPENG checks.
* sysdeps/aarch64/multiarch/memcpy_kunpeng.S: New file
Here's a memcpy and memmove result tested by benchmark-walk:
Function: memcpy
Variant: walk
__memcpy_thunderx __memcpy_thunderx2 __memcpy_falkor __memcpy_kunpeng __memcpy_generic
========================================================================================================================
length=128: 52.13 (-32.61%) 44.10 (-12.17%) 41.34 ( -5.15%) 39.73 ( -1.07%) 39.31
length=144: 39.83 ( -0.15%) 42.36 ( -6.50%) 40.79 ( -2.55%) 39.83 ( -0.13%) 39.78
length=129: 32.69 ( -0.12%) 34.77 ( -6.49%) 35.79 ( -9.61%) 32.85 ( -0.59%) 32.65
length=143: 42.10 ( 0.05%) 40.22 ( 4.54%) 44.56 ( -5.78%) 42.18 ( -0.12%) 42.13
length=130: 32.74 ( 0.08%) 35.03 ( -6.94%) 35.74 ( -9.08%) 32.90 ( -0.41%) 32.76
length=142: 40.85 ( -0.15%) 39.69 ( 2.70%) 43.21 ( -5.94%) 40.84 ( -0.12%) 40.79
length=131: 34.12 ( 0.00%) 35.66 ( -4.51%) 36.95 ( -8.28%) 34.24 ( -0.34%) 34.12
length=141: 40.73 ( 0.22%) 39.39 ( 3.50%) 43.07 ( -5.52%) 40.92 ( -0.24%) 40.82
length=132: 33.15 ( -0.44%) 35.32 ( -6.99%) 35.80 ( -8.46%) 33.17 ( -0.49%) 33.01
length=140: 38.54 ( -0.20%) 38.77 ( -0.80%) 40.58 ( -5.52%) 38.53 ( -0.18%) 38.46
length=133: 35.53 ( -0.17%) 36.66 ( -3.34%) 38.08 ( -7.34%) 35.64 ( -0.48%) 35.47
length=139: 39.55 ( -0.08%) 38.59 ( 2.35%) 41.65 ( -5.38%) 39.63 ( -0.26%) 39.52
length=134: 35.64 ( 0.02%) 36.83 ( -3.32%) 38.27 ( -7.35%) 35.79 ( -0.40%) 35.65
length=138: 38.26 ( 0.01%) 38.25 ( 0.05%) 40.53 ( -5.90%) 38.35 ( -0.21%) 38.27
length=135: 37.01 ( -0.05%) 37.48 ( -1.33%) 39.35 ( -6.38%) 37.06 ( -0.18%) 36.99
length=137: 38.40 ( 0.01%) 38.26 ( 0.37%) 40.59 ( -5.69%) 38.48 ( -0.20%) 38.41
length=256: 79.36 ( 0.36%) 83.17 ( -4.42%) 86.01 ( -7.99%) 80.23 ( -0.73%) 79.65
length=272: 75.03 ( -0.12%) 80.47 ( -7.38%) 79.49 ( -6.07%) 75.15 ( -0.28%) 74.94
length=257: 63.07 ( -0.23%) 72.04 (-14.49%) 69.53 (-10.50%) 63.10 ( -0.28%) 62.92
length=271: 77.57 ( 0.05%) 81.50 ( -5.01%) 80.93 ( -4.27%) 77.56 ( 0.06%) 77.61
length=258: 63.71 ( -0.16%) 72.85 (-14.53%) 70.04 (-10.11%) 63.71 ( -0.16%) 63.61
length=270: 81.77 ( -8.63%) 80.70 ( -7.20%) 79.72 ( -5.90%) 75.42 ( -0.18%) 75.28
length=259: 64.69 ( -0.29%) 73.12 (-13.36%) 71.36 (-10.63%) 64.72 ( -0.34%) 64.50
length=269: 74.63 ( -0.13%) 79.69 ( -6.93%) 79.34 ( -6.45%) 74.55 ( -0.02%) 74.53
length=260: 64.97 ( 0.19%) 74.34 (-14.20%) 71.05 ( -9.14%) 65.15 ( -0.08%) 65.10
length=268: 70.65 ( -0.21%) 78.63 (-11.54%) 77.13 ( -9.41%) 70.63 ( -0.19%) 70.50
length=261: 66.31 ( -0.12%) 74.26 (-12.12%) 73.23 (-10.56%) 66.41 ( -0.27%) 66.24
length=267: 71.53 ( -0.36%) 77.99 ( -9.42%) 77.77 ( -9.11%) 71.46 ( -0.26%) 71.28
length=262: 66.83 ( -0.20%) 74.88 (-12.26%) 73.59 (-10.33%) 66.76 ( -0.10%) 66.70
length=266: 73.68 ( -0.20%) 84.47 (-14.89%) 81.79 (-11.24%) 73.93 ( -0.55%) 73.53
length=263: 71.94 ( -6.04%) 79.07 (-16.54%) 76.58 (-12.88%) 68.53 ( -1.02%) 67.84
length=265: 69.30 ( -0.10%) 76.80 (-10.93%) 76.61 (-10.65%) 69.37 ( -0.19%) 69.23
length=512: 157.74 ( 0.08%) 159.60 ( -1.10%) 166.31 ( -5.34%) 158.12 ( -0.16%) 157.87
length=528: 151.11 ( 1.81%) 175.27 (-13.89%) 171.76 (-11.61%) 158.47 ( -2.97%) 153.90
length=513: 141.74 ( -0.41%) 157.20 (-11.36%) 151.86 ( -7.58%) 141.12 ( 0.03%) 141.16
length=527: 152.77 ( -0.10%) 162.24 ( -6.30%) 165.36 ( -8.34%) 152.66 ( -0.02%) 152.62
length=514: 141.79 ( 0.10%) 157.62 (-11.06%) 152.80 ( -7.66%) 141.94 ( -0.01%) 141.92
length=526: 151.73 ( 1.39%) 171.96 (-11.76%) 177.73 (-15.51%) 158.00 ( -2.69%) 153.86
length=515: 143.83 ( 28.74%) 158.29 ( 21.58%) 206.81 ( -2.46%) 227.73 (-12.82%) 201.84
length=525: 211.49 ( -9.31%) 202.89 ( -4.87%) 198.17 ( -2.43%) 192.59 ( 0.46%) 193.48
length=516: 188.92 ( 0.46%) 320.09 (-68.66%) 255.17 (-34.45%) 201.30 ( -6.07%) 189.79
length=524: 196.97 ( 2.82%) 208.82 ( -3.03%) 206.75 ( -2.01%) 201.97 ( 0.35%) 202.69
length=517: 197.87 ( 0.06%) 209.84 ( -5.98%) 203.56 ( -2.81%) 197.54 ( 0.23%) 198.00
length=523: 203.66 ( -1.16%) 217.97 ( -8.27%) 216.06 ( -7.32%) 216.16 ( -7.37%) 201.33
length=518: 193.27 ( 1.02%) 205.92 ( -5.46%) 201.31 ( -3.10%) 192.87 ( 1.23%) 195.26
length=522: 200.56 ( 0.83%) 211.23 ( -4.45%) 207.92 ( -2.81%) 201.78 ( 0.23%) 202.24
length=519: 201.32 ( 0.29%) 211.15 ( -4.58%) 206.97 ( -2.51%) 202.22 ( -0.15%) 201.91
length=521: 202.41 ( -0.13%) 211.92 ( -4.83%) 207.75 ( -2.77%) 202.20 ( -0.03%) 202.15
length=1024: 448.09 ( -2.81%) 448.11 ( -2.81%) 452.17 ( -3.75%) 437.73 ( -0.43%) 435.84
length=1040: 420.41 ( -8.95%) 386.29 ( -0.11%) 397.24 ( -2.94%) 388.43 ( -0.66%) 385.88
length=1025: 383.70 ( 5.26%) 403.82 ( 0.29%) 394.16 ( 2.68%) 405.78 ( -0.19%) 405.00
length=1039: 419.39 ( -1.83%) 411.37 ( 0.12%) 411.08 ( 0.19%) 412.37 ( -0.13%) 411.85
length=1026: 401.69 ( 1.39%) 409.12 ( -0.43%) 405.97 ( 0.34%) 408.59 ( -0.30%) 407.37
length=1038: 414.31 ( -0.64%) 408.25 ( 0.83%) 408.66 ( 0.73%) 409.73 ( 0.47%) 411.68
length=1027: 402.12 ( 0.01%) 406.26 ( -1.02%) 395.30 ( 1.70%) 402.26 ( -0.03%) 402.15
length=1037: 414.08 ( -0.48%) 409.45 ( 0.64%) 410.09 ( 0.49%) 413.57 ( -0.36%) 412.11
length=1028: 501.62 (-63.30%) 421.34 (-37.16%) 330.36 ( -7.55%) 309.33 ( -0.70%) 307.18
length=1036: 316.73 ( 0.25%) 316.02 ( 0.47%) 328.82 ( -3.56%) 316.26 ( 0.39%) 317.51
length=1029: 309.04 ( -0.25%) 310.88 ( -0.85%) 319.09 ( -3.51%) 307.71 ( 0.18%) 308.27
length=1035: 315.97 ( 0.01%) 313.42 ( 0.81%) 325.84 ( -3.12%) 314.30 ( 0.53%) 315.99
length=1030: 309.89 ( 0.01%) 312.01 ( -0.67%) 321.51 ( -3.74%) 309.14 ( 0.26%) 309.93
length=1034: 315.50 ( 0.07%) 314.24 ( 0.47%) 326.20 ( -3.32%) 314.40 ( 0.41%) 315.71
length=1031: 311.32 ( 0.02%) 311.65 ( -0.09%) 321.93 ( -3.39%) 310.57 ( 0.26%) 311.38
length=1033: 313.60 ( -0.17%) 311.70 ( 0.44%) 323.40 ( -3.30%) 312.62 ( 0.15%) 313.07
length=2048: 541.23 ( -8.53%) 406.86 ( 18.42%) 512.53 ( -2.77%) 500.25 ( -0.31%) 498.70
length=2064: 515.11 ( -5.65%) 388.60 ( 20.30%) 527.30 ( -8.15%) 675.08 (-38.46%) 487.54
length=2049: 590.89 (-21.64%) 446.63 ( 8.06%) 458.19 ( 5.68%) 420.47 ( 13.44%) 485.77
length=2063: 551.10 ( -9.21%) 450.41 ( 10.75%) 491.83 ( 2.54%) 443.58 ( 12.10%) 504.64
length=2050: 502.53 ( -5.31%) 429.91 ( 9.91%) 449.30 ( 5.85%) 416.95 ( 12.63%) 477.21
length=2062: 534.68 ( -5.60%) 439.32 ( 13.23%) 483.60 ( 4.49%) 438.03 ( 13.49%) 506.34
length=2051: 507.00 ( -5.23%) 420.22 ( 12.78%) 458.06 ( 4.93%) 413.85 ( 14.11%) 481.82
length=2061: 529.21 ( -6.13%) 444.95 ( 10.77%) 485.45 ( 2.65%) 437.72 ( 12.22%) 498.64
length=2052: 493.09 ( -4.55%) 419.95 ( 10.96%) 438.86 ( 6.95%) 411.16 ( 12.83%) 471.65
length=2060: 518.78 ( -6.11%) 429.52 ( 12.15%) 467.37 ( 4.40%) 426.55 ( 12.75%) 488.90
length=2053: 515.00 ( -5.52%) 422.37 ( 13.46%) 465.47 ( 4.63%) 416.66 ( 14.63%) 488.08
length=2059: 524.42 ( 0.87%) 441.28 ( 16.59%) 517.29 ( 2.22%) 496.89 ( 6.08%) 529.03
length=2054: 547.42 (-11.75%) 438.11 ( 10.57%) 459.71 ( 6.16%) 422.28 ( 13.80%) 489.88
length=2058: 531.32 ( -6.94%) 435.25 ( 12.40%) 473.51 ( 4.70%) 426.63 ( 14.13%) 496.86
length=2055: 521.61 ( -6.10%) 432.22 ( 12.08%) 468.23 ( 4.76%) 423.05 ( 13.95%) 491.63
length=2057: 520.85 ( 2.95%) 832.58 (-55.13%) 592.29 (-10.36%) 459.92 ( 14.30%) 536.68
length=4096: 975.17 ( -2.54%) 655.45 ( 31.08%) 917.03 ( 3.57%) 762.73 ( 19.80%) 950.98
length=4112: 964.94 ( -3.63%) 667.76 ( 28.29%) 876.17 ( 5.91%) 692.94 ( 25.58%) 931.17
length=4097: 948.37 ( -1.21%) 682.39 ( 27.17%) 892.41 ( 4.76%) 677.27 ( 27.72%) 937.00
length=4111: 978.69 ( -1.92%) 716.71 ( 25.37%) 906.23 ( 5.63%) 712.06 ( 25.85%) 960.29
length=4098: 949.38 ( -2.14%) 679.25 ( 26.93%) 888.81 ( 4.38%) 673.31 ( 27.56%) 929.52
length=4110: 970.83 ( -1.66%) 708.96 ( 25.76%) 902.78 ( 5.46%) 704.57 ( 26.22%) 954.97
length=4099: 949.87 ( -1.71%) 679.67 ( 27.22%) 892.95 ( 4.38%) 675.80 ( 27.63%) 933.87
length=4109: 967.45 ( -1.49%) 705.64 ( 25.97%) 903.67 ( 5.20%) 702.48 ( 26.31%) 953.23
length=4100: 945.82 ( 1.14%) 675.72 ( 29.37%) 886.98 ( 7.29%) 674.15 ( 29.53%) 956.68
length=4108: 1141.85 ( -4.79%) 854.97 ( 21.54%) 930.53 ( 14.61%) 838.30 ( 23.07%) 1089.68
length=4101: 1128.48 ( -8.13%) 837.73 ( 19.73%) 919.94 ( 11.85%) 792.95 ( 24.02%) 1043.65
length=4107: 1059.05 ( -4.74%) 783.50 ( 22.51%) 911.18 ( 9.89%) 761.32 ( 24.71%) 1011.15
length=4102: 1010.60 ( -3.77%) 733.65 ( 24.67%) 898.25 ( 7.76%) 719.14 ( 26.16%) 973.87
length=4106: 991.69 (-20.82%) 811.75 ( 1.10%) 988.04 (-20.38%) 727.25 ( 11.40%) 820.79
length=4103: 871.04 ( -0.39%) 713.14 ( 17.81%) 721.12 ( 16.89%) 810.37 ( 6.60%) 867.64
length=4105: 911.67 (-14.04%) 719.09 ( 10.05%) 702.96 ( 12.06%) 710.58 ( 11.11%) 799.40
length=8192: 1360.42 ( -1.14%) 1118.26 ( 16.86%) 1157.03 ( 13.98%) 1326.21 ( 1.40%) 1345.02
length=8208: 1400.28 ( -8.73%) 1137.66 ( 11.66%) 1046.63 ( 18.73%) 1161.27 ( 9.83%) 1287.83
length=8193: 1398.21 ( -8.00%) 1145.11 ( 11.55%) 1091.21 ( 15.71%) 1157.04 ( 10.63%) 1294.66
length=8207: 1436.45 ( -7.68%) 1154.69 ( 13.44%) 1147.60 ( 13.97%) 1168.64 ( 12.40%) 1334.00
length=8194: 1370.49 ( -8.63%) 1118.03 ( 11.38%) 1067.82 ( 15.36%) 1126.42 ( 10.72%) 1261.64
length=8206: 1396.56 ( -1.25%) 1143.20 ( 17.12%) 1127.25 ( 18.28%) 1154.28 ( 16.32%) 1379.33
length=8195: 1496.53 (-15.44%) 1207.21 ( 6.88%) 1109.14 ( 14.44%) 1152.25 ( 11.12%) 1296.37
length=8205: 1418.96 ( -7.85%) 1148.54 ( 12.70%) 1132.73 ( 13.90%) 1153.58 ( 12.32%) 1315.63
length=8196: 1361.38 ( -6.84%) 1114.30 ( 12.55%) 1071.10 ( 15.94%) 1131.68 ( 11.19%) 1274.26
length=8204: 1369.29 ( -5.22%) 1126.63 ( 13.43%) 1122.16 ( 13.77%) 1141.87 ( 12.26%) 1301.39
length=8197: 1381.82 ( 4.95%) 1210.91 ( 16.71%) 1136.11 ( 21.85%) 1298.20 ( 10.70%) 1453.82
length=8203: 1563.37 (-10.38%) 1261.34 ( 10.94%) 1141.43 ( 19.41%) 1273.12 ( 10.11%) 1416.32
length=8198: 1499.47 ( -8.40%) 1233.36 ( 10.84%) 1097.78 ( 20.64%) 1240.02 ( 10.36%) 1383.31
length=8202: 2913.01 (-32.85%) 1696.61 ( 22.62%) 1775.53 ( 19.02%) 1678.14 ( 23.47%) 2192.68
length=8199: 2179.48 ( 4.82%) 1556.86 ( 32.01%) 1745.45 ( 23.77%) 1550.35 ( 32.29%) 2289.81
length=8201: 2321.80 (-14.25%) 1499.12 ( 26.23%) 1746.32 ( 14.07%) 1480.27 ( 27.16%) 2032.29
length=16384: 3977.00 ( -1.49%) 2340.90 ( 40.26%) 3411.18 ( 12.95%) 2829.02 ( 27.81%) 3918.80
length=16400: 3915.98 ( -2.54%) 2598.48 ( 31.96%) 3357.71 ( 12.08%) 2574.68 ( 32.58%) 3819.13
length=16385: 3868.77 ( 5.00%) 2604.50 ( 36.04%) 3392.37 ( 16.70%) 2589.65 ( 36.41%) 4072.39
length=16399: 4324.48 ( -2.36%) 3138.76 ( 25.70%) 3461.64 ( 18.06%) 3100.32 ( 26.61%) 4224.67
length=16386: 4226.01 ( -0.31%) 3042.08 ( 27.79%) 3429.44 ( 18.60%) 3149.32 ( 25.24%) 4212.85
length=16398: 4172.67 ( -4.11%) 2868.93 ( 28.42%) 3389.04 ( 15.44%) 2839.37 ( 29.16%) 4007.93
length=16387: 4026.32 ( -1.65%) 2778.45 ( 29.86%) 3397.77 ( 14.22%) 2749.61 ( 30.58%) 3961.09
length=16397: 3950.05 ( -2.22%) 2665.81 ( 31.02%) 3388.85 ( 12.31%) 2653.39 ( 31.34%) 3864.40
length=16388: 3883.05 ( -1.41%) 2623.36 ( 31.49%) 3376.21 ( 11.83%) 2622.74 ( 31.50%) 3828.99
length=16396: 3849.62 ( -1.89%) 2558.15 ( 32.29%) 3372.83 ( 10.73%) 2546.00 ( 32.61%) 3778.22
length=16389: 3809.94 ( -1.24%) 2531.80 ( 32.72%) 3391.95 ( 9.86%) 2541.10 ( 32.47%) 3763.11
length=16395: 3856.78 (-35.31%) 3044.15 ( -6.80%) 2049.70 ( 28.09%) 2717.45 ( 4.66%) 2850.41
length=16390: 2862.92 ( -3.52%) 2563.33 ( 7.31%) 2002.69 ( 27.58%) 2615.06 ( 5.44%) 2765.45
length=16394: 2806.77 ( -0.94%) 2504.83 ( 9.92%) 2023.76 ( 27.22%) 2596.99 ( 6.60%) 2780.55
length=16391: 2784.26 ( -2.14%) 2497.16 ( 8.39%) 2008.27 ( 26.33%) 2522.85 ( 7.45%) 2725.87
length=16393: 2731.38 ( -1.98%) 2440.61 ( 8.88%) 2023.48 ( 24.45%) 2490.99 ( 7.00%) 2678.47
length=32768: 7629.73 (-46.76%) 4473.26 ( 13.95%) 3947.03 ( 24.08%) 5201.79 ( -0.06%) 5198.66
length=32784: 7666.55 (-44.44%) 4705.93 ( 11.34%) 3782.04 ( 28.75%) 4759.99 ( 10.32%) 5307.92
length=32769: 7589.72 (-42.41%) 4776.54 ( 10.37%) 3889.61 ( 27.02%) 4789.02 ( 10.14%) 5329.43
length=32783: 7502.45 (-44.25%) 4688.97 ( 9.85%) 3858.48 ( 25.81%) 4714.48 ( 9.36%) 5201.10
length=32770: 7438.46 (-43.39%) 4647.77 ( 10.41%) 3848.94 ( 25.81%) 4680.76 ( 9.77%) 5187.71
length=32782: 7225.10 (-38.35%) 4609.10 ( 11.74%) 3855.37 ( 26.17%) 4643.70 ( 11.08%) 5222.19
length=32771: 7326.40 (-42.87%) 4587.85 ( 10.53%) 3828.78 ( 25.33%) 4580.34 ( 10.68%) 5127.85
length=32781: 7261.12 (-41.38%) 4548.17 ( 11.44%) 3851.30 ( 25.01%) 4584.78 ( 10.73%) 5135.97
length=32772: 7178.11 (-41.83%) 4510.12 ( 10.89%) 3802.44 ( 24.87%) 4521.19 ( 10.67%) 5061.20
length=32780: 7186.99 (-42.34%) 4481.01 ( 11.25%) 3835.00 ( 24.00%) 4532.33 ( 10.23%) 5049.00
length=32773: 7089.60 (-38.93%) 4482.70 ( 12.15%) 3830.79 ( 24.93%) 4487.18 ( 12.07%) 5102.88
length=32779: 7076.27 (-25.65%) 4498.21 ( 20.13%) 3881.99 ( 31.07%) 5371.42 ( 4.62%) 5631.73
length=32774: 8362.27 (-48.81%) 5190.76 ( 7.63%) 3845.61 ( 31.57%) 5176.08 ( 7.89%) 5619.46
length=32778: 8186.09 (-48.41%) 5109.29 ( 7.37%) 3861.96 ( 29.98%) 5211.88 ( 5.51%) 5515.68
length=32775: 8186.51 (-49.43%) 5096.24 ( 6.98%) 3846.12 ( 29.80%) 5128.88 ( 6.38%) 5478.44
length=32777: 8038.38 (-49.89%) 5001.09 ( 6.75%) 3837.79 ( 28.44%) 5095.39 ( 4.99%) 5362.88
length=65536: 15947.00 ( 18.56%) 8793.77 ( 55.09%) 7611.29 ( 61.13%) 10576.30 ( 45.99%) 19580.90
length=65552: 16992.10 (-15.71%) 9740.98 ( 33.67%) 13282.90 ( 9.55%) 9804.27 ( 33.23%) 14684.70
length=65537: 12917.20 ( 11.25%) 9576.91 ( 34.20%) 13265.60 ( 8.86%) 9720.59 ( 33.21%) 14554.80
length=65551: 12661.70 ( 12.50%) 9486.67 ( 34.44%) 13291.90 ( 8.14%) 9505.77 ( 34.31%) 14470.30
length=65538: 12416.60 ( 13.37%) 9378.05 ( 34.57%) 13265.00 ( 7.45%) 9360.94 ( 34.69%) 14333.20
length=65550: 12263.30 ( 14.26%) 9256.75 ( 35.28%) 13292.20 ( 7.06%) 9262.92 ( 35.23%) 14302.30
length=65539: 12137.70 ( 14.34%) 9172.56 ( 35.27%) 13278.40 ( 6.29%) 9231.09 ( 34.85%) 14169.40
length=65549: 12086.40 ( 14.69%) 9060.12 ( 36.05%) 13283.30 ( 6.24%) 9139.49 ( 35.49%) 14167.90
length=65540: 11880.40 ( 15.30%) 9043.20 ( 35.53%) 13261.50 ( 5.45%) 8997.29 ( 35.85%) 14026.10
length=65548: 11793.00 ( 15.62%) 8951.62 ( 35.95%) 13248.10 ( 5.21%) 8948.26 ( 35.98%) 13976.20
length=65541: 11679.80 ( 16.33%) 8905.88 ( 36.20%) 13265.60 ( 4.97%) 8840.90 ( 36.67%) 13960.10
length=65547: 11588.60 ( 16.80%) 8830.41 ( 36.61%) 13275.00 ( 4.70%) 8842.02 ( 36.52%) 13929.20
length=65542: 11566.10 ( 16.65%) 8844.02 ( 36.26%) 13250.80 ( 4.51%) 8820.78 ( 36.43%) 13876.00
length=65546: 11519.40 ( 16.83%) 8771.78 ( 36.67%) 13263.10 ( 4.24%) 8750.37 ( 36.82%) 13850.70
length=65543: 11448.10 ( 16.90%) 8715.39 ( 36.73%) 13252.10 ( 3.80%) 8710.74 ( 36.77%) 13775.90
length=65545: 11458.40 ( 16.86%) 8659.75 ( 37.17%) 13304.10 ( 3.47%) 8745.23 ( 36.55%) 13782.50
length=131072: 22342.00 ( 18.63%) 15174.50 ( 44.73%) 26413.10 ( 3.80%) 17999.50 ( 34.44%) 27456.40
length=131088: 22635.80 ( 17.37%) 17165.30 ( 37.34%) 26469.10 ( 3.37%) 17158.00 ( 37.36%) 27392.80
length=131073: 22526.40 ( 17.65%) 17224.40 ( 37.03%) 26434.70 ( 3.36%) 17158.70 ( 37.27%) 27354.30
length=131087: 22412.20 ( 18.01%) 17047.10 ( 37.64%) 26462.80 ( 3.20%) 17173.50 ( 37.18%) 27336.80
length=131074: 22424.20 ( 22.24%) 17079.40 ( 40.77%) 26465.00 ( 8.22%) 17125.30 ( 40.61%) 28835.90
length=131086: 40078.80 (-74.05%) 22007.40 ( 4.43%) 14988.40 ( 34.91%) 22279.10 ( 3.25%) 23027.40
length=131075: 34681.10 (-48.45%) 21243.40 ( 9.07%) 14929.10 ( 36.10%) 21932.30 ( 6.12%) 23361.40
length=131085: 34251.60 (-50.09%) 20757.70 ( 9.04%) 14877.00 ( 34.81%) 21243.60 ( 6.91%) 22820.70
length=131076: 33435.40 (-45.78%) 20673.60 ( 9.87%) 14887.50 ( 35.09%) 21088.60 ( 8.06%) 22936.30
length=131084: 33341.70 (-46.30%) 20422.30 ( 10.39%) 14848.60 ( 34.85%) 20764.90 ( 8.89%) 22790.00
length=131077: 32742.20 (-46.86%) 20348.80 ( 8.73%) 14849.80 ( 33.39%) 20570.90 ( 7.73%) 22294.30
length=131083: 32624.50 (-49.50%) 20223.30 ( 7.33%) 14842.70 ( 31.99%) 20304.80 ( 6.96%) 21822.70
length=131078: 32029.80 (-46.15%) 19946.80 ( 8.99%) 14868.50 ( 32.16%) 20330.20 ( 7.24%) 21916.10
length=131082: 31929.50 (-40.20%) 19687.50 ( 13.56%) 14793.50 ( 35.04%) 19917.50 ( 12.55%) 22774.70
length=131079: 35902.80 (-56.14%) 22017.30 ( 4.25%) 14910.50 ( 35.16%) 21891.00 ( 4.80%) 22994.70
length=131081: 34660.40 (-49.87%) 21417.10 ( 7.39%) 14950.40 ( 35.35%) 21634.10 ( 6.45%) 23126.40
length=262144: 68983.20 (-63.12%) 37265.40 ( 11.88%) 29540.60 ( 30.15%) 45874.50 ( -8.47%) 42290.80
length=262160: 67704.90 (-46.70%) 41761.70 ( 9.51%) 29389.90 ( 36.32%) 42672.00 ( 7.54%) 46152.60
length=262145: 66706.40 (-45.31%) 41534.50 ( 9.52%) 29608.80 ( 35.50%) 42510.20 ( 7.40%) 45906.60
length=262159: 81856.50 (-47.75%) 47537.70 ( 14.20%) 52798.40 ( 4.70%) 34805.20 ( 37.18%) 55402.40
length=262146: 45977.20 ( 16.41%) 35069.60 ( 36.24%) 52833.60 ( 3.94%) 34810.60 ( 36.71%) 55003.00
length=262158: 45257.20 ( 17.66%) 34440.50 ( 37.34%) 52825.00 ( 3.89%) 34518.30 ( 37.20%) 54964.10
length=262147: 45483.90 ( 17.10%) 34609.80 ( 36.92%) 52782.50 ( 3.80%) 34546.30 ( 37.04%) 54867.90
length=262157: 44816.80 ( 17.92%) 33984.20 ( 37.76%) 52817.60 ( 3.27%) 33916.20 ( 37.89%) 54602.50
length=262148: 44812.70 ( 17.79%) 34119.00 ( 37.41%) 52749.20 ( 3.23%) 34044.30 ( 37.55%) 54511.40
length=262156: 44447.20 ( 18.29%) 33973.60 ( 37.55%) 52809.00 ( 2.92%) 34057.40 ( 37.39%) 54399.30
length=262149: 44727.80 ( 17.76%) 33856.10 ( 37.75%) 52813.00 ( 2.90%) 34053.80 ( 37.39%) 54389.40
length=262155: 44715.50 ( 17.71%) 33772.90 ( 37.84%) 52864.80 ( 2.71%) 33807.10 ( 37.78%) 54336.10
length=262150: 44408.80 ( 18.31%) 33939.80 ( 37.57%) 52801.70 ( 2.87%) 33706.60 ( 37.99%) 54360.50
length=262154: 44178.20 ( 18.66%) 33739.90 ( 37.88%) 52816.00 ( 2.76%) 33990.90 ( 37.42%) 54313.40
length=262151: 45741.50 ( 33.54%) 44669.50 ( 35.10%) 53335.80 ( 22.51%) 48333.40 ( 29.78%) 68830.50
length=262153: 67710.50 ( 1.21%) 48394.80 ( 29.39%) 53387.10 ( 22.11%) 49728.80 ( 27.45%) 68540.90
length=524288: 129746.00 ( 2.00%) 77745.60 ( 41.87%) 105844.00 ( 20.00%) 92711.10 ( 30.68%) 133748.00
length=524304: 124324.00 ( 4.00%) 88915.30 ( 32.06%) 105948.00 ( 19.00%) 88982.70 ( 32.00%) 130866.00
length=524289: 119251.00 ( 7.00%) 87273.60 ( 32.35%) 106032.00 ( 17.00%) 80155.50 ( 37.87%) 129012.00
length=524303: 116137.00 ( 10.00%) 84275.30 ( 35.36%) 105867.00 ( 18.00%) 84498.10 ( 35.19%) 130376.00
length=524290: 138105.00 (-54.68%) 121209.00 (-35.76%) 63522.80 ( 28.85%) 87831.70 ( 1.63%) 89283.90
length=524302: 134105.00 (-49.00%) 83334.10 ( 7.90%) 59091.90 ( 34.69%) 83906.90 ( 7.26%) 90480.00
length=524291: 134058.00 (-53.29%) 81877.70 ( 6.37%) 58743.30 ( 32.83%) 82353.00 ( 5.83%) 87451.10
length=524301: 130047.00 (-43.81%) 80786.40 ( 10.67%) 58859.10 ( 34.91%) 82549.40 ( 8.72%) 90431.10
length=524292: 129063.00 (-51.97%) 81236.70 ( 4.35%) 58822.70 ( 30.74%) 82568.60 ( 2.78%) 84927.80
length=524300: 128313.00 (-45.40%) 79296.20 ( 10.15%) 58851.90 ( 33.31%) 81571.60 ( 7.57%) 88249.20
length=524293: 127328.00 (-44.00%) 78289.50 ( 12.02%) 58544.50 ( 34.21%) 80412.80 ( 9.63%) 88982.00
length=524299: 135595.00 (-54.83%) 81338.00 ( 7.12%) 58862.70 ( 32.79%) 80834.80 ( 7.70%) 87575.80
length=524294: 126207.00 (-49.25%) 78027.50 ( 7.72%) 58830.80 ( 30.43%) 79161.60 ( 6.38%) 84558.60
length=524298: 125189.00 (-47.64%) 76918.40 ( 9.29%) 58718.60 ( 30.75%) 77995.00 ( 8.02%) 84793.90
length=524295: 126253.00 (-47.00%) 75750.00 ( 11.00%) 58712.80 ( 31.72%) 76864.20 ( 10.61%) 85987.00
length=524297: 123030.00 (-33.76%) 75724.10 ( 17.67%) 58611.20 ( 36.28%) 82349.50 ( 10.47%) 91975.30
length=1048576: 273291.00 (-61.00%) 146604.00 ( 13.00%) 117511.00 ( 30.00%) 178378.00 ( -5.00%) 169898.00
length=1048592: 266742.00 ( 7.00%) 164992.00 ( 42.00%) 117873.00 ( 58.00%) 238306.00 ( 17.00%) 287356.00
length=1048577: 316370.00 (-51.00%) 202710.00 ( 3.00%) 120862.00 ( 42.00%) 173791.00 ( 17.00%) 210478.00
length=1048591: 310302.00 (-56.00%) 187292.00 ( 6.00%) 120591.00 ( 39.00%) 189947.00 ( 4.00%) 199304.00
length=1048578: 305883.00 ( 0.00%) 182002.00 ( 40.00%) 120063.00 ( 60.00%) 292063.00 ( 4.00%) 306638.00
length=1048590: 270399.00 ( -3.00%) 187458.00 ( 28.00%) 212570.00 ( 19.00%) 190418.00 ( 27.00%) 263049.00
length=1048579: 252772.00 ( 1.00%) 183513.00 ( 28.00%) 212058.00 ( 17.00%) 178106.00 ( 30.00%) 257625.00
length=1048589: 241352.00 ( 11.00%) 175087.00 ( 35.00%) 212020.00 ( 22.00%) 197121.00 ( 27.00%) 273018.00
length=1048580: 272863.00 ( -1.00%) 191251.00 ( 29.00%) 212480.00 ( 21.00%) 197999.00 ( 27.00%) 271294.00
length=1048588: 271516.00 ( -2.00%) 191700.00 ( 28.00%) 212554.00 ( 20.00%) 192703.00 ( 27.00%) 266482.00
length=1048581: 254257.00 ( 1.00%) 181722.00 ( 29.00%) 212162.00 ( 17.00%) 179509.00 ( 30.00%) 258385.00
length=1048587: 238944.00 ( 5.00%) 172779.00 ( 31.00%) 212316.00 ( 16.00%) 172404.00 ( 31.00%) 253201.00
length=1048582: 229499.00 ( 12.00%) 167014.00 ( 36.00%) 212926.00 ( 19.00%) 186145.00 ( 29.00%) 263337.00
length=1048586: 246753.00 ( 2.00%) 176579.00 ( 30.00%) 211904.00 ( 16.00%) 174490.00 ( 31.00%) 254016.00
length=1048583: 230403.00 ( 7.00%) 167953.00 ( 32.00%) 212186.00 ( 14.00%) 167205.00 ( 32.00%) 247924.00
length=1048585: 219784.00 ( 15.00%) 160818.00 ( 37.00%) 212837.00 ( 17.00%) 180530.00 ( 30.00%) 258848.00
length=2097152: 468514.00 (-35.00%) 303214.00 ( 12.00%) 263596.00 ( 24.00%) 375436.00 ( -9.00%) 347145.00
length=2097168: 559784.00 (-45.00%) 348186.00 ( 10.00%) 232985.00 ( 40.00%) 357944.00 ( 7.00%) 388335.00
length=2097153: 561868.00 (-42.00%) 358270.00 ( 9.00%) 232862.00 ( 41.00%) 250309.00 ( 36.00%) 395784.00
length=2097167: 541513.00 (-43.00%) 343408.00 ( 9.00%) 232726.00 ( 38.00%) 345078.00 ( 9.00%) 379829.00
length=2097154: 537004.00 (-35.00%) 358866.00 ( 10.00%) 232268.00 ( 41.00%) 317332.00 ( 20.00%) 400260.00
length=2097166: 540001.00 (-47.00%) 339144.00 ( 7.00%) 232486.00 ( 36.00%) 346319.00 ( 5.00%) 367642.00
length=2097155: 533560.00 (-44.00%) 341569.00 ( 8.00%) 232288.00 ( 37.00%) 289282.00 ( 22.00%) 372314.00
length=2097165: 527971.00 (-44.00%) 330433.00 ( 10.00%) 232416.00 ( 36.00%) 327653.00 ( 10.00%) 367777.00
length=2097156: 521450.00 (-43.00%) 320001.00 ( 12.00%) 232338.00 ( 36.00%) 307222.00 ( 15.00%) 364791.00
length=2097164: 518852.00 (-40.00%) 320562.00 ( 13.00%) 232273.00 ( 37.00%) 313861.00 ( 15.00%) 372114.00
length=2097157: 508795.00 (-38.00%) 315605.00 ( 14.00%) 232124.00 ( 37.00%) 319851.00 ( 13.00%) 370642.00
length=2097163: 510197.00 (-48.00%) 320996.00 ( 7.00%) 232101.00 ( 32.00%) 301246.00 ( 12.00%) 345614.00
length=2097158: 501512.00 (-43.00%) 316638.00 ( 10.00%) 233321.00 ( 33.00%) 318923.00 ( 9.00%) 351839.00
length=2097162: 497261.00 (-45.00%) 310550.00 ( 9.00%) 232054.00 ( 32.00%) 318091.00 ( 7.00%) 344919.00
length=2097159: 489207.00 (-40.00%) 309784.00 ( 11.00%) 232894.00 ( 33.00%) 317947.00 ( 9.00%) 349449.00
length=2097161: 492754.00 (-33.00%) 304773.00 ( 18.00%) 232418.00 ( 37.00%) 318688.00 ( 14.00%) 372033.00
length=4194304: 1129420.00 (-64.03%) 605284.00 ( 12.00%) 464642.00 ( 32.00%) 747822.00 ( -9.00%) 688544.00
length=4194320: 1281930.00 (-55.59%) 1261300.00 (-53.08%) 476589.00 ( 42.00%) 853061.00 ( -4.00%) 823925.00
length=4194305: 1326790.00 (-52.66%) 838598.00 ( 3.00%) 472888.00 ( 45.00%) 553020.00 ( 36.00%) 869124.00
length=4194319: 1296550.00 (-58.93%) 789772.00 ( 3.00%) 472919.00 ( 42.00%) 818050.00 ( -1.00%) 815809.00
length=4194306: 1294940.00 (-29.17%) 816410.00 ( 18.56%) 779710.00 ( 22.22%) 721422.00 ( 28.04%) 1002480.00
length=4194318: 938395.00 ( 5.00%) 688034.00 ( 30.00%) 841896.00 ( 15.00%) 684446.00 ( 31.00%) 993648.00
length=4194307: 909038.00 ( 7.00%) 685291.00 ( 30.00%) 842381.00 ( 14.00%) 587871.00 ( 40.00%) 983304.00
length=4194317: 884795.00 ( 8.00%) 665151.00 ( 31.00%) 841174.00 ( 13.00%) 656075.00 ( 32.00%) 969540.00
length=4194308: 866239.00 ( 10.00%) 664831.00 ( 31.00%) 841445.00 ( 12.00%) 588519.00 ( 39.00%) 964982.00
length=4194316: 846171.00 ( 11.00%) 642814.00 ( 32.00%) 841546.00 ( 11.00%) 605698.00 ( 36.00%) 954491.00
length=4194309: 832370.00 ( 12.00%) 640740.00 ( 32.00%) 842070.00 ( 11.00%) 549245.00 ( 41.00%) 946635.00
length=4194315: 817958.00 ( 12.00%) 626052.00 ( 33.00%) 842538.00 ( 10.00%) 590811.00 ( 37.00%) 939969.00
length=4194310: 811324.00 ( 13.00%) 626351.00 ( 32.00%) 841162.00 ( 9.00%) 539605.00 ( 42.00%) 934106.00
length=4194314: 797979.00 ( 14.00%) 603701.00 ( 34.00%) 840765.00 ( 9.00%) 604139.00 ( 34.00%) 928328.00
length=4194311: 784441.00 ( 14.00%) 592495.00 ( 35.00%) 840848.00 ( 8.00%) 556625.00 ( 39.00%) 920742.00
length=4194313: 778590.00 ( 14.00%) 587148.00 ( 35.00%) 839972.00 ( 8.00%) 572760.00 ( 37.00%) 915135.00
length=8388608: 1517920.00 ( 16.70%) 1000050.00 ( 45.12%) 1679180.00 ( 7.86%) 1211230.00 ( 33.53%) 1822340.00
length=8388624: 1519240.00 ( 15.78%) 1154860.00 ( 35.98%) 1683330.00 ( 6.69%) 1086690.00 ( 39.76%) 1804000.00
length=8388609: 1513510.00 ( 16.12%) 1177920.00 ( 34.72%) 1679360.00 ( 6.93%) 920625.00 ( 48.98%) 1804410.00
length=8388623: 1492580.00 ( 16.39%) 1133610.00 ( 36.50%) 1680400.00 ( 5.87%) 1071080.00 ( 40.00%) 1785120.00
length=8388610: 1476470.00 ( 17.21%) 1158330.00 ( 35.05%) 1681210.00 ( 5.73%) 916785.00 ( 48.60%) 1783480.00
length=8388622: 1457610.00 ( 17.37%) 1121600.00 ( 36.41%) 1678160.00 ( 4.86%) 1055440.00 ( 40.17%) 1763920.00
length=8388611: 1459160.00 ( 17.22%) 1142800.00 ( 35.17%) 1682120.00 ( 4.57%) 909018.00 ( 48.43%) 1762670.00
length=8388621: 1444190.00 ( 17.56%) 1098580.00 ( 37.29%) 1679030.00 ( 4.16%) 1042870.00 ( 40.47%) 1751900.00
length=8388612: 1429770.00 ( 17.84%) 1123960.00 ( 35.41%) 1679020.00 ( 3.52%) 897165.00 ( 48.45%) 1740210.00
length=8388620: 1422220.00 ( 18.03%) 1112960.00 ( 35.85%) 1680420.00 ( 3.14%) 954310.00 ( 45.00%) 1734960.00
length=8388613: 1417650.00 ( 18.21%) 1118770.00 ( 35.45%) 1679660.00 ( 3.09%) 957055.00 ( 44.78%) 1733260.00
length=8388619: 1399940.00 ( 40.39%) 1099700.00 ( 53.18%) 1677720.00 ( 28.56%) 946458.00 ( 59.70%) 2348560.00
length=8388614: 2274360.00 (-44.04%) 1508220.00 ( 4.48%) 928390.00 ( 41.20%) 1359810.00 ( 13.88%) 1578940.00
length=8388618: 2254930.00 (-39.89%) 1455850.00 ( 9.68%) 929692.00 ( 42.32%) 1319760.00 ( 18.12%) 1611880.00
length=8388615: 2257440.00 (-42.11%) 1447300.00 ( 8.89%) 926848.00 ( 41.65%) 1314240.00 ( 17.27%) 1588510.00
length=8388617: 2199110.00 (-39.96%) 1403680.00 ( 10.66%) 931030.00 ( 40.74%) 1289840.00 ( 17.91%) 1571200.00
length=16777216: 4337910.00 (-58.69%) 2382780.00 ( 12.84%) 1873900.00 ( 31.45%) 2924180.00 ( -6.97%) 2733660.00
length=16777232: 4683590.00 (-55.87%) 2975620.00 ( 0.97%) 1871120.00 ( 37.73%) 2682460.00 ( 10.73%) 3004840.00
length=16777217: 4750420.00 (-64.47%) 2942950.00 ( -1.89%) 1876280.00 ( 35.04%) 2019480.00 ( 30.08%) 2888260.00
length=16777231: 4533790.00 (-48.86%) 2921600.00 ( 4.07%) 1869750.00 ( 38.61%) 2614300.00 ( 14.16%) 3045600.00
length=16777218: 4531380.00 (-66.23%) 2867600.00 ( -5.19%) 1867880.00 ( 31.48%) 1994060.00 ( 26.85%) 2726020.00
length=16777230: 4460720.00 (-56.95%) 2847500.00 ( -0.19%) 1872530.00 ( 34.12%) 2572140.00 ( 9.50%) 2842120.00
length=16777219: 4395120.00 (-23.95%) 4268600.00 (-20.38%) 3389990.00 ( 4.40%) 1838920.00 ( 48.14%) 3546020.00
length=16777229: 2978000.00 ( 16.03%) 2341740.00 ( 33.97%) 3383900.00 ( 4.59%) 2151250.00 ( 39.34%) 3546520.00
length=16777220: 2939700.00 ( 16.66%) 2296690.00 ( 34.89%) 3384610.00 ( 4.05%) 1829960.00 ( 48.12%) 3527480.00
length=16777228: 2939380.00 ( 16.30%) 2309940.00 ( 34.23%) 3384080.00 ( 3.64%) 2125050.00 ( 39.49%) 3511970.00
length=16777221: 2909430.00 ( 16.95%) 2262640.00 ( 35.41%) 3387000.00 ( 3.32%) 1815960.00 ( 48.16%) 3503130.00
length=16777227: 2912920.00 ( 16.56%) 2292680.00 ( 34.33%) 3384680.00 ( 3.05%) 2097100.00 ( 39.93%) 3491040.00
length=16777222: 2878740.00 ( 17.31%) 2256460.00 ( 35.18%) 3385420.00 ( 2.76%) 1794210.00 ( 48.46%) 3481360.00
length=16777226: 2868980.00 ( 17.23%) 2269060.00 ( 34.54%) 3386120.00 ( 2.31%) 2077560.00 ( 40.06%) 3466280.00
length=16777223: 2857940.00 ( 17.81%) 2234260.00 ( 35.75%) 3385550.00 ( 2.64%) 1798180.00 ( 48.29%) 3477200.00
length=16777225: 2838140.00 ( 18.22%) 2256520.00 ( 34.98%) 3382380.00 ( 2.53%) 2078000.00 ( 40.12%) 3470300.00
length=33554432: 5428610.00 ( 36.24%) 3799340.00 ( 55.38%) 6766930.00 ( 20.52%) 5725590.00 ( 32.75%) 8514030.00
length=33554448: 8189030.00 ( 5.60%) 6022460.00 ( 30.57%) 6794290.00 ( 21.68%) 4487290.00 ( 48.27%) 8674640.00
length=33554433: 8505870.00 ( -1.70%) 6014980.00 ( 28.08%) 6775380.00 ( 18.99%) 4314850.00 ( 48.41%) 8363320.00
length=33554447: 7925710.00 ( 3.01%) 5746450.00 ( 29.68%) 6781540.00 ( 17.01%) 4212420.00 ( 48.45%) 8171630.00
length=33554434: 7534840.00 ( 6.09%) 5588090.00 ( 30.35%) 6775320.00 ( 15.55%) 4103770.00 ( 48.85%) 8023140.00
length=33554446: 7290170.00 ( 7.79%) 5456520.00 ( 30.98%) 6772170.00 ( 14.34%) 4040940.00 ( 48.89%) 7906270.00
length=33554435: 7078220.00 ( 19.38%) 5331420.00 ( 39.28%) 6772490.00 ( 22.86%) 4414460.00 ( 49.72%) 8779630.00
length=33554445: 8344840.00 ( -1.54%) 5927920.00 ( 27.87%) 6761630.00 ( 17.72%) 4248730.00 ( 48.30%) 8218160.00
length=33554436: 7595690.00 ( 4.67%) 5571460.00 ( 30.07%) 6782650.00 ( 14.87%) 4096580.00 ( 48.58%) 7967430.00
length=33554444: 7131220.00 ( 8.40%) 5356570.00 ( 31.20%) 6789750.00 ( 12.79%) 3983870.00 ( 48.83%) 7785220.00
length=33554437: 6817620.00 ( 10.79%) 5187320.00 ( 32.12%) 6778120.00 ( 11.31%) 3899660.00 ( 48.97%) 7642430.00
length=33554443: 6623990.00 ( 12.00%) 5064300.00 ( 32.72%) 6770280.00 ( 10.06%) 3846720.00 ( 48.90%) 7527300.00
length=33554438: 6438530.00 ( 13.60%) 4944790.00 ( 33.65%) 6769790.00 ( 9.16%) 3814180.00 ( 48.82%) 7452300.00
length=33554442: 7453860.00 ( 9.05%) 5937090.00 ( 27.55%) 6785960.00 ( 17.20%) 4224870.00 ( 48.45%) 8195170.00
length=33554439: 7500310.00 ( 5.76%) 5532480.00 ( 30.49%) 6772850.00 ( 14.90%) 4076390.00 ( 48.78%) 7958900.00
length=33554441: 7076920.00 ( 8.65%) 5322170.00 ( 31.30%) 6779540.00 ( 12.49%) 3968590.00 ( 48.78%) 7747400.00
Function: memmove
Variant: walk
__memmove_thunderx __memmove_falkor __memmove_kunpeng __memmove_generic
========================================================================================================================
length=128: 51.37 (-49.76%) 38.06 (-10.97%) 34.90 ( -1.74%) 34.30
length=144: 35.44 ( -3.25%) 36.02 ( -4.93%) 35.05 ( -2.10%) 34.33
length=129: 35.11 ( 0.09%) 37.99 ( -8.10%) 35.35 ( -0.61%) 35.14
length=143: 46.36 ( -0.17%) 42.60 ( 7.96%) 46.39 ( -0.24%) 46.28
length=130: 35.43 ( -0.16%) 37.52 ( -6.04%) 35.60 ( -0.62%) 35.38
length=142: 44.96 ( -0.16%) 41.82 ( 6.82%) 44.97 ( -0.19%) 44.89
length=131: 36.97 ( -0.12%) 38.27 ( -3.66%) 37.18 ( -0.70%) 36.92
length=141: 44.86 ( -0.09%) 41.79 ( 6.77%) 44.97 ( -0.33%) 44.83
length=132: 36.07 ( -0.09%) 37.69 ( -4.57%) 36.09 ( -0.15%) 36.04
length=140: 42.30 ( 0.17%) 40.77 ( 3.79%) 42.38 ( -0.02%) 42.37
length=133: 38.67 ( 0.01%) 39.01 ( -0.88%) 38.78 ( -0.28%) 38.67
length=139: 43.51 ( -0.01%) 41.27 ( 5.15%) 43.70 ( -0.44%) 43.50
length=134: 38.89 ( -0.01%) 39.02 ( -0.34%) 39.08 ( -0.49%) 38.89
length=138: 41.93 ( -0.12%) 40.45 ( 3.43%) 42.06 ( -0.41%) 41.88
length=135: 40.43 ( 0.11%) 39.83 ( 1.59%) 40.71 ( -0.57%) 40.48
length=137: 41.88 ( 0.03%) 40.44 ( 3.46%) 42.02 ( -0.32%) 41.89
length=256: 61.62 ( 11.97%) 72.77 ( -3.96%) 64.73 ( 7.53%) 70.00
length=272: 82.67 (-10.25%) 79.58 ( -6.13%) 78.07 ( -4.12%) 74.98
length=257: 110.00 (-33.10%) 78.46 ( 5.07%) 82.75 ( -0.12%) 82.65
length=271: 101.87 ( 0.23%) 82.67 ( 19.03%) 102.11 ( -0.01%) 102.10
length=258: 84.84 ( 0.30%) 76.20 ( 10.45%) 85.46 ( -0.43%) 85.10
length=270: 101.60 ( -0.14%) 82.52 ( 18.67%) 102.54 ( -1.06%) 101.46
length=259: 84.64 ( 0.81%) 76.52 ( 10.33%) 85.15 ( 0.21%) 85.33
length=269: 98.75 ( -0.66%) 81.08 ( 17.35%) 98.37 ( -0.27%) 98.11
length=260: 93.07 (-12.26%) 192.05 (-131.65%) 146.38 (-76.56%) 82.91
length=268: 88.50 ( -0.98%) 81.97 ( 6.47%) 87.90 ( -0.29%) 87.64
length=261: 79.76 ( -0.26%) 78.24 ( 1.64%) 79.03 ( 0.65%) 79.55
length=267: 86.23 ( -0.63%) 82.52 ( 3.69%) 85.79 ( -0.12%) 85.69
length=262: 81.69 ( -0.17%) 79.30 ( 2.76%) 81.10 ( 0.56%) 81.55
length=266: 84.99 ( 0.06%) 80.03 ( 5.89%) 84.41 ( 0.75%) 85.04
length=263: 81.83 ( -0.07%) 78.91 ( 3.50%) 81.09 ( 0.84%) 81.78
length=265: 84.92 ( -0.04%) 80.98 ( 4.60%) 84.10 ( 0.92%) 84.88
length=512: 93.63 (-37.27%) 75.94 (-11.33%) 66.20 ( 2.94%) 68.21
length=528: 73.27 ( 1.08%) 80.07 ( -8.11%) 74.84 ( -1.05%) 74.06
length=513: 154.89 ( 1.39%) 134.90 ( 14.11%) 156.24 ( 0.53%) 157.06
length=527: 166.72 ( -0.06%) 146.68 ( 11.97%) 165.36 ( 0.76%) 166.62
length=514: 159.44 ( 0.01%) 141.12 ( 11.50%) 158.39 ( 0.67%) 159.45
length=526: 167.66 ( -0.36%) 145.50 ( 12.90%) 165.94 ( 0.67%) 167.05
length=515: 159.24 ( -0.45%) 142.13 ( 10.34%) 157.64 ( 0.56%) 158.53
length=525: 164.99 ( 0.32%) 147.65 ( 10.79%) 163.86 ( 1.00%) 165.52
length=516: 163.06 ( 0.08%) 144.00 ( 11.76%) 161.98 ( 0.75%) 163.20
length=524: 168.84 ( -0.21%) 147.31 ( 12.57%) 167.38 ( 0.65%) 168.48
length=517: 160.28 ( -0.20%) 144.91 ( 9.41%) 158.44 ( 0.95%) 159.96
length=523: 164.47 ( -0.42%) 147.01 ( 10.24%) 162.57 ( 0.74%) 163.79
length=518: 161.48 ( 0.12%) 145.37 ( 10.08%) 160.42 ( 0.77%) 161.67
length=522: 164.38 ( -0.19%) 146.50 ( 10.70%) 163.17 ( 0.55%) 164.06
length=519: 161.35 ( -0.14%) 146.10 ( 9.32%) 159.59 ( 0.95%) 161.12
length=521: 271.65 ( -8.19%) 169.51 ( 32.49%) 250.02 ( 0.42%) 251.07
length=1024: 272.12 ( -0.41%) 297.19 ( -9.67%) 267.80 ( 1.18%) 270.99
length=1040: 236.44 ( 0.21%) 272.31 (-14.93%) 236.76 ( 0.07%) 236.93
length=1025: 451.94 ( -6.66%) 332.34 ( 21.57%) 447.85 ( -5.69%) 423.72
length=1039: 458.52 ( 2.37%) 329.55 ( 29.83%) 474.65 ( -1.07%) 469.64
length=1026: 524.59 ( -8.55%) 351.12 ( 27.34%) 485.88 ( -0.54%) 483.25
length=1038: 527.96 (-24.13%) 340.45 ( 19.95%) 457.44 ( -7.55%) 425.31
length=1027: 430.33 ( -0.37%) 322.11 ( 24.87%) 446.14 ( -4.06%) 428.74
length=1037: 455.58 ( -1.43%) 327.19 ( 27.16%) 469.53 ( -4.53%) 449.16
length=1028: 380.04 (-12.39%) 318.08 ( 5.93%) 381.02 (-12.68%) 338.14
length=1036: 373.06 ( -4.19%) 324.85 ( 9.28%) 402.36 (-12.37%) 358.06
length=1029: 402.50 ( 6.29%) 324.05 ( 24.56%) 448.45 ( -4.41%) 429.52
length=1035: 448.77 ( -0.89%) 327.21 ( 26.43%) 465.53 ( -4.66%) 444.79
length=1030: 420.31 ( -4.70%) 321.31 ( 19.96%) 429.95 ( -7.10%) 401.45
length=1034: 419.44 ( -1.73%) 322.84 ( 21.70%) 440.57 ( -6.86%) 412.30
length=1031: 428.93 ( 1.75%) 324.41 ( 25.69%) 455.25 ( -4.28%) 436.56
length=1033: 446.69 ( -0.85%) 325.36 ( 26.54%) 461.42 ( -4.17%) 442.94
length=2048: 509.21 ( 4.03%) 573.49 ( -8.08%) 525.59 ( 0.94%) 530.59
length=2064: 456.26 ( 0.53%) 533.25 (-16.25%) 395.50 ( 13.78%) 458.70
length=2049: 653.95 (-12.47%) 660.78 (-13.64%) 578.13 ( 0.57%) 581.46
length=2063: 667.06 ( -6.63%) 643.44 ( -2.86%) 557.57 ( 10.87%) 625.57
length=2050: 609.38 ( -5.46%) 640.02 (-10.76%) 566.68 ( 1.93%) 577.85
length=2062: 657.74 ( 13.50%) 645.01 ( 15.17%) 744.05 ( 2.14%) 760.35
length=2051: 554.96 ( -8.56%) 540.92 ( -5.82%) 482.66 ( 5.58%) 511.19
length=2061: 552.20 ( -3.16%) 535.06 ( 0.04%) 460.81 ( 13.91%) 535.28
length=2052: 497.77 ( -1.91%) 530.70 ( -8.65%) 469.79 ( 3.82%) 488.46
length=2060: 519.63 ( -1.65%) 533.80 ( -4.43%) 464.77 ( 9.08%) 511.17
length=2053: 517.69 ( -1.42%) 528.83 ( -3.60%) 469.05 ( 8.11%) 510.45
length=2059: 540.18 ( -2.07%) 531.32 ( -0.40%) 462.64 ( 12.58%) 529.21
length=2054: 515.57 ( -2.35%) 528.72 ( -4.96%) 466.24 ( 7.44%) 503.72
length=2058: 528.80 ( -2.04%) 529.37 ( -2.15%) 463.54 ( 10.55%) 518.24
length=2055: 525.71 ( -1.67%) 529.52 ( -2.40%) 465.04 ( 10.07%) 517.10
length=2057: 536.72 ( -2.28%) 531.45 ( -1.28%) 465.10 ( 11.37%) 524.74
length=4096: 881.94 ( -8.04%) 974.05 (-19.33%) 540.77 ( 33.75%) 816.28
length=4112: 731.36 ( 0.09%) 844.20 (-15.33%) 544.58 ( 25.60%) 731.99
length=4097: 890.17 ( 1.48%) 1052.36 (-16.47%) 769.43 ( 14.84%) 903.55
length=4111: 952.60 ( -0.31%) 1046.27 (-10.18%) 784.13 ( 17.43%) 949.62
length=4098: 912.19 ( -1.33%) 1046.22 (-16.22%) 766.61 ( 14.84%) 900.18
length=4110: 943.07 ( 0.13%) 1047.51 (-10.93%) 780.25 ( 17.37%) 944.31
length=4099: 923.69 ( -1.17%) 1047.81 (-14.76%) 775.22 ( 15.09%) 913.03
length=4109: 953.12 ( -0.89%) 1048.29 (-10.96%) 784.46 ( 16.97%) 944.75
length=4100: 909.17 ( -0.98%) 1055.13 (-17.19%) 766.73 ( 14.84%) 900.36
length=4108: 930.55 ( -0.25%) 1057.37 (-13.91%) 777.48 ( 16.24%) 928.22
length=4101: 929.05 ( -1.19%) 1047.62 (-14.10%) 778.69 ( 15.19%) 918.13
length=4107: 947.47 ( -0.97%) 1047.27 (-11.61%) 784.13 ( 16.43%) 938.34
length=4102: 930.43 ( -1.69%) 1045.75 (-14.30%) 775.53 ( 15.24%) 914.95
length=4106: 935.99 ( -0.75%) 1043.12 (-12.28%) 776.88 ( 16.38%) 929.07
length=4103: 933.96 ( -1.22%) 1047.82 (-13.56%) 777.32 ( 15.75%) 922.66
length=4105: 939.62 ( -0.89%) 1047.56 (-12.48%) 779.74 ( 16.28%) 931.35
length=8192: 1693.15 ( -4.61%) 1909.64 (-17.99%) 1033.45 ( 36.15%) 1618.47
length=8208: 1445.04 ( 0.10%) 1670.54 (-15.49%) 1020.78 ( 29.43%) 1446.48
length=8193: 1699.34 ( 2.32%) 2093.24 (-20.32%) 1388.96 ( 20.16%) 1739.74
length=8207: 1757.65 ( 1.94%) 2099.04 (-17.11%) 1451.55 ( 19.01%) 1792.35
length=8194: 1725.64 ( 0.23%) 2091.20 (-20.90%) 1388.15 ( 19.75%) 1729.69
length=8206: 1749.03 ( 1.30%) 2094.01 (-18.17%) 1442.46 ( 18.60%) 1772.03
length=8195: 1729.41 ( 4.45%) 2135.97 (-18.01%) 1465.03 ( 19.06%) 1810.01
length=8205: 1848.83 ( -1.58%) 2119.61 (-16.45%) 1486.65 ( 18.32%) 1820.13
length=8196: 1761.66 ( -1.45%) 2123.73 (-22.30%) 1398.56 ( 19.46%) 1736.52
length=8204: 1740.93 ( 0.85%) 2119.55 (-20.71%) 1434.39 ( 18.31%) 1755.93
length=8197: 1740.92 ( 0.40%) 2095.89 (-19.91%) 1408.64 ( 19.41%) 1747.85
length=8203: 1758.23 ( 1.00%) 2088.07 (-17.57%) 1437.80 ( 19.05%) 1776.07
length=8198: 1735.79 ( 0.49%) 2083.58 (-19.45%) 1405.62 ( 19.41%) 1744.27
length=8202: 1738.60 ( 1.34%) 2085.26 (-18.33%) 1426.53 ( 19.05%) 1762.24
length=8199: 1741.35 ( 0.40%) 2082.95 (-19.14%) 1417.40 ( 18.93%) 1748.38
length=8201: 1744.66 ( 0.97%) 2083.24 (-18.24%) 1425.20 ( 19.11%) 1761.82
length=16384: 3282.60 ( -2.66%) 3762.91 (-17.68%) 2011.40 ( 37.09%) 3197.49
length=16400: 2856.66 ( -0.16%) 3312.29 (-16.13%) 1962.10 ( 31.21%) 2852.18
length=16385: 3308.95 ( 2.91%) 4188.05 (-22.88%) 2636.16 ( 22.65%) 3408.14
length=16399: 3381.43 ( 1.28%) 4214.49 (-23.04%) 2813.89 ( 17.85%) 3425.36
length=16386: 3368.70 ( 1.74%) 4214.17 (-22.92%) 2630.54 ( 23.27%) 3428.33
length=16398: 3385.22 ( 1.20%) 4215.71 (-23.03%) 2764.28 ( 19.33%) 3426.49
length=16387: 3376.74 ( 1.95%) 4223.17 (-22.62%) 2653.92 ( 22.94%) 3444.03
length=16397: 3398.36 ( 1.19%) 4217.23 (-22.61%) 2761.99 ( 19.70%) 3439.45
length=16388: 3360.08 ( 1.87%) 4257.15 (-24.33%) 2628.56 ( 23.23%) 3424.07
length=16396: 3368.00 ( 2.31%) 4251.41 (-23.32%) 2723.26 ( 21.01%) 3447.55
length=16389: 3444.62 ( 0.73%) 4240.82 (-22.22%) 2688.86 ( 22.51%) 3469.86
length=16395: 3428.00 ( 0.64%) 4223.60 (-22.42%) 2752.34 ( 20.23%) 3450.14
length=16390: 3395.98 ( 0.76%) 4219.51 (-23.30%) 2672.76 ( 21.90%) 3422.12
length=16394: 3403.62 ( 0.64%) 4227.16 (-23.40%) 2720.07 ( 20.60%) 3425.58
length=16391: 3394.65 ( 0.98%) 4224.08 (-23.22%) 2693.65 ( 21.42%) 3428.09
length=16393: 3394.15 ( 1.01%) 4252.00 (-24.01%) 2729.86 ( 20.38%) 3428.64
length=32768: 6063.16 ( 7.10%) 7795.20 (-19.44%) 4016.20 ( 38.46%) 6526.50
length=32784: 5107.78 ( 12.14%) 6870.37 (-18.18%) 3918.07 ( 32.60%) 5813.44
length=32769: 5924.29 ( 13.57%) 8522.11 (-24.32%) 5127.22 ( 25.20%) 6854.83
length=32783: 6007.26 ( 12.29%) 8520.83 (-24.41%) 5475.73 ( 20.05%) 6849.26
length=32770: 5949.95 ( 13.43%) 8539.76 (-24.25%) 5110.44 ( 25.65%) 6873.32
length=32782: 5911.80 ( 14.04%) 8541.18 (-24.19%) 5403.79 ( 21.43%) 6877.58
length=32771: 5967.58 ( 13.37%) 8534.59 (-23.90%) 5165.10 ( 25.02%) 6888.23
length=32781: 5981.02 ( 13.24%) 8535.53 (-23.81%) 5432.46 ( 21.20%) 6894.14
length=32772: 5973.10 ( 13.14%) 8612.85 (-25.24%) 5126.35 ( 25.46%) 6876.87
length=32780: 5949.92 ( 13.44%) 8603.34 (-25.16%) 5333.63 ( 22.41%) 6873.80
length=32773: 5963.11 ( 33.86%) 8570.04 ( 4.94%) 5479.35 ( 39.22%) 9015.68
length=32779: 6775.19 ( 7.55%) 8675.92 (-18.39%) 5796.76 ( 20.90%) 7328.10
length=32774: 6562.39 ( 8.63%) 8669.82 (-20.71%) 5490.60 ( 23.55%) 7182.36
length=32778: 6351.84 ( 9.37%) 8563.24 (-22.19%) 5490.85 ( 21.65%) 7008.35
length=32775: 6084.76 ( 12.33%) 8549.78 (-23.19%) 5400.42 ( 22.19%) 6940.18
length=32777: 5989.39 ( 13.46%) 8540.92 (-23.40%) 5422.72 ( 21.65%) 6921.13
length=65536: 12035.90 ( 8.82%) 15837.90 (-19.98%) 8092.52 ( 38.69%) 13200.30
length=65552: 10312.70 ( 12.62%) 13970.90 (-18.38%) 7807.01 ( 33.85%) 11801.60
length=65537: 11783.80 ( 14.06%) 17010.90 (-24.06%) 10315.90 ( 24.77%) 13712.00
length=65551: 11820.10 ( 14.74%) 17053.40 (-23.01%) 11015.40 ( 20.54%) 13862.90
length=65538: 11737.60 ( 15.03%) 17144.20 (-24.11%) 10314.80 ( 25.33%) 13814.10
length=65550: 11773.50 ( 15.01%) 17179.70 (-24.01%) 10916.80 ( 21.20%) 13853.00
length=65539: 11742.10 ( 14.87%) 17207.50 (-24.75%) 10423.80 ( 24.43%) 13793.30
length=65549: 11756.50 ( 15.22%) 17141.20 (-23.61%) 10935.10 ( 21.14%) 13867.00
length=65540: 11742.00 ( 15.24%) 17301.20 (-24.89%) 10316.30 ( 25.53%) 13853.60
length=65548: 11709.00 ( 15.76%) 17304.90 (-24.50%) 10709.40 ( 22.95%) 13899.20
length=65541: 11714.20 ( 15.48%) 17203.60 (-24.13%) 10526.20 ( 24.05%) 13859.20
length=65547: 11691.60 ( 15.75%) 17205.70 (-23.98%) 10854.20 ( 21.78%) 13877.30
length=65542: 11715.40 ( 15.49%) 17149.90 (-23.71%) 10512.50 ( 24.17%) 13862.50
length=65546: 11718.40 ( 15.47%) 17214.30 (-24.18%) 10714.00 ( 22.71%) 13862.30
length=65543: 11740.50 ( 15.51%) 17196.10 (-23.75%) 10669.50 ( 23.22%) 13895.60
length=65545: 11736.50 ( 15.36%) 17181.40 (-23.91%) 10713.60 ( 22.73%) 13865.80
length=131072: 23477.90 ( 12.02%) 31979.40 (-19.84%) 16053.00 ( 39.84%) 26685.90
length=131088: 20288.80 ( 14.78%) 28394.50 (-19.26%) 15554.60 ( 34.67%) 23808.20
length=131073: 23065.20 ( 16.54%) 34261.60 (-23.98%) 20525.10 ( 25.73%) 27635.50
length=131087: 23368.80 ( 15.64%) 34419.10 (-24.25%) 22045.40 ( 20.41%) 27700.40
length=131074: 23329.00 ( 15.41%) 34330.80 (-24.48%) 20452.30 ( 25.84%) 27579.30
length=131086: 23289.60 ( 15.96%) 34338.60 (-23.90%) 21776.70 ( 21.42%) 27713.80
length=131075: 23393.90 ( 15.30%) 34414.20 (-24.60%) 20758.60 ( 24.84%) 27619.80
length=131085: 23497.00 ( 14.88%) 34368.40 (-24.51%) 21825.30 ( 20.93%) 27603.70
length=131076: 23355.50 ( 15.32%) 34618.90 (-25.51%) 20442.70 ( 25.88%) 27581.60
length=131084: 23302.90 ( 15.53%) 34596.20 (-25.41%) 21367.70 ( 22.54%) 27586.20
length=131077: 23294.20 ( 15.80%) 34378.90 (-24.27%) 20934.90 ( 24.33%) 27664.80
length=131083: 23252.30 ( 16.14%) 34303.60 (-23.72%) 21519.60 ( 22.39%) 27726.20
length=131078: 23541.40 ( 15.10%) 34488.70 (-24.37%) 20906.40 ( 24.61%) 27729.70
length=131082: 23233.00 ( 16.01%) 34376.90 (-24.27%) 21297.70 ( 23.01%) 27662.90
length=131079: 23640.00 ( 14.76%) 34384.70 (-23.98%) 21113.30 ( 23.87%) 27733.90
length=131081: 23294.10 ( 20.18%) 34314.20 (-17.59%) 22388.50 ( 23.28%) 29182.30
length=262144: 53968.50 ( 5.74%) 68583.00 (-19.78%) 36214.30 ( 36.75%) 57256.60
length=262160: 42725.60 ( 13.62%) 58633.60 (-18.55%) 34563.30 ( 30.12%) 49460.90
length=262145: 49340.90 ( 13.12%) 69482.90 (-22.34%) 42630.90 ( 24.94%) 56793.90
length=262159: 49072.80 ( 13.06%) 69202.80 (-22.60%) 44677.80 ( 20.85%) 56444.70
length=262146: 48050.20 ( 13.60%) 69100.20 (-24.26%) 41265.90 ( 25.80%) 55611.40
length=262158: 47355.30 ( 14.80%) 69020.10 (-24.18%) 43630.40 ( 21.50%) 55581.20
length=262147: 46992.90 ( 14.87%) 68914.80 (-24.85%) 41279.10 ( 25.22%) 55198.90
length=262157: 46511.20 ( 15.87%) 68783.80 (-24.42%) 43559.80 ( 21.21%) 55284.70
length=262148: 46358.80 ( 15.55%) 69367.50 (-26.36%) 40682.10 ( 25.89%) 54896.50
length=262156: 46715.20 ( 15.58%) 69434.10 (-25.47%) 42471.40 ( 23.25%) 55339.80
length=262149: 46967.60 ( 14.86%) 68611.30 (-24.37%) 41739.00 ( 24.34%) 55167.60
length=262155: 46597.90 ( 15.46%) 68977.30 (-25.14%) 43168.80 ( 21.68%) 55119.20
length=262150: 46568.50 ( 15.73%) 68783.00 (-24.47%) 41573.80 ( 24.77%) 55262.40
length=262154: 46608.80 ( 15.63%) 68973.70 (-24.85%) 42562.30 ( 22.96%) 55246.10
length=262151: 46122.80 ( 16.61%) 68919.90 (-24.61%) 42162.30 ( 23.77%) 55310.60
length=262153: 46312.00 ( 16.31%) 68993.40 (-24.68%) 42357.10 ( 23.46%) 55336.80
length=524288: 93954.10 ( 11.66%) 127651.00 (-21.00%) 64050.20 ( 39.78%) 106359.00
length=524304: 81301.60 ( 14.63%) 113156.00 (-18.82%) 61764.80 ( 35.14%) 95230.80
length=524289: 90748.10 ( 17.78%) 137172.00 (-25.00%) 82059.70 ( 25.65%) 110372.00
length=524303: 92724.10 ( 16.14%) 137006.00 (-24.00%) 87656.20 ( 20.72%) 110565.00
length=524290: 91348.40 ( 16.74%) 137148.00 (-25.00%) 81992.70 ( 25.27%) 109720.00
length=524302: 92311.20 ( 16.16%) 137114.00 (-25.00%) 86483.60 ( 21.45%) 110107.00
length=524291: 93077.30 ( 16.05%) 137631.00 (-25.00%) 82565.30 ( 25.53%) 110870.00
length=524301: 93835.50 ( 14.70%) 137259.00 (-25.00%) 86335.30 ( 21.51%) 110001.00
length=524292: 91692.20 ( 16.78%) 138335.00 (-26.00%) 81453.10 ( 26.07%) 110174.00
length=524300: 92474.50 ( 16.05%) 138115.00 (-26.00%) 84655.50 ( 23.14%) 110149.00
length=524293: 93196.20 ( 15.43%) 136978.00 (-25.00%) 83221.40 ( 24.48%) 110200.00
length=524299: 93586.70 ( 15.47%) 136751.00 (-24.00%) 85860.90 ( 22.45%) 110719.00
length=524294: 92551.70 ( 16.38%) 137318.00 (-25.00%) 82913.00 ( 25.00%) 110685.00
length=524298: 92245.80 ( 16.20%) 137114.00 (-25.00%) 84564.80 ( 23.18%) 110076.00
length=524295: 93268.80 ( 15.34%) 137287.00 (-25.00%) 83766.40 ( 23.97%) 110171.00
length=524297: 92812.50 ( 15.50%) 136723.00 (-25.00%) 84459.40 ( 23.10%) 109837.00
length=1048576: 189055.00 ( 11.00%) 255632.00 (-20.00%) 127520.00 ( 40.00%) 214272.00
length=1048592: 163877.00 ( 14.00%) 227786.00 (-19.00%) 125765.00 ( 34.00%) 192024.00
length=1048577: 196083.00 ( 14.00%) 276655.00 (-21.00%) 173274.00 ( 24.00%) 230236.00
length=1048591: 186074.00 ( 16.00%) 277172.00 (-25.00%) 178324.00 ( 19.00%) 222188.00
length=1048578: 186906.00 ( 21.00%) 280199.00 (-19.00%) 175363.00 ( 26.00%) 237219.00
length=1048590: 208326.00 ( 11.00%) 280487.00 (-20.00%) 185080.00 ( 21.00%) 235400.00
length=1048579: 201976.00 ( 11.00%) 282193.00 (-24.00%) 173491.00 ( 24.00%) 228632.00
length=1048589: 196144.00 ( 13.00%) 279221.00 (-24.00%) 181174.00 ( 20.00%) 226852.00
length=1048580: 189535.00 ( 14.00%) 282644.00 (-28.00%) 166542.00 ( 24.00%) 222000.00
length=1048588: 188098.00 ( 15.00%) 280967.00 (-26.00%) 172497.00 ( 22.00%) 223014.00
length=1048581: 189548.00 ( 15.00%) 279455.00 (-25.00%) 171365.00 ( 23.00%) 224151.00
length=1048587: 186756.00 ( 16.00%) 277845.00 (-25.00%) 175841.00 ( 21.00%) 223729.00
length=1048582: 184490.00 ( 16.00%) 280701.00 (-27.00%) 168527.00 ( 24.00%) 221829.00
length=1048586: 187660.00 ( 15.00%) 278802.00 (-26.00%) 172051.00 ( 22.00%) 222710.00
length=1048583: 189992.00 ( 15.00%) 279054.00 (-25.00%) 170535.00 ( 23.00%) 223912.00
length=1048585: 190160.00 ( 14.00%) 278805.00 (-26.00%) 172724.00 ( 22.00%) 222572.00
length=2097152: 376478.00 ( 11.00%) 511187.00 (-21.00%) 254444.00 ( 40.00%) 425699.00
length=2097168: 323464.00 ( 14.00%) 453921.00 (-20.00%) 250295.00 ( 34.00%) 379489.00
length=2097153: 410148.00 ( 13.00%) 551739.00 (-16.00%) 360928.00 ( 24.00%) 476018.00
length=2097167: 367602.00 ( 17.00%) 559306.00 (-26.00%) 357094.00 ( 19.00%) 444567.00
length=2097154: 391584.00 ( 14.00%) 555734.00 (-21.00%) 345333.00 ( 24.00%) 460365.00
length=2097166: 368121.00 ( 17.00%) 562643.00 (-27.00%) 357188.00 ( 19.00%) 446051.00
length=2097155: 371663.00 ( 16.00%) 557379.00 (-25.00%) 338519.00 ( 24.00%) 446731.00
length=2097165: 366809.00 ( 20.00%) 562591.00 (-23.00%) 360409.00 ( 21.00%) 459892.00
length=2097156: 393411.00 ( 14.00%) 572898.00 (-25.00%) 344439.00 ( 25.00%) 460712.00
length=2097164: 389985.00 ( 15.00%) 573819.00 (-25.00%) 354104.00 ( 23.00%) 461854.00
length=2097157: 387254.00 ( 15.00%) 564170.00 (-24.00%) 343420.00 ( 24.00%) 455872.00
length=2097163: 363951.00 ( 17.00%) 565508.00 (-28.00%) 359229.00 ( 18.00%) 441900.00
length=2097158: 373182.00 ( 16.00%) 559072.00 (-25.00%) 339402.00 ( 24.00%) 447429.00
length=2097162: 376629.00 ( 17.00%) 563218.00 (-24.00%) 348134.00 ( 23.00%) 455264.00
length=2097159: 363108.00 ( 17.00%) 560458.00 (-28.00%) 338364.00 ( 22.00%) 438209.00
length=2097161: 375238.00 ( 16.00%) 563131.00 (-27.00%) 341900.00 ( 23.00%) 446726.00
length=4194304: 749314.00 ( 12.00%) 1041320.00 (-21.34%) 508135.00 ( 40.00%) 858176.00
length=4194320: 643940.00 ( 16.00%) 918396.00 (-20.00%) 501105.00 ( 35.00%) 771141.00
length=4194305: 804616.00 ( 15.00%) 1119910.00 (-17.40%) 713385.00 ( 25.00%) 953948.00
length=4194319: 774539.00 ( 17.00%) 1134600.00 (-21.11%) 709754.00 ( 24.00%) 936814.00
length=4194306: 819070.00 ( 14.00%) 1128910.00 (-17.81%) 719244.00 ( 24.00%) 958285.00
length=4194318: 723455.00 ( 18.00%) 1135260.00 (-27.46%) 710518.00 ( 20.00%) 890714.00
length=4194307: 837508.00 ( 12.00%) 1131850.00 (-18.23%) 722016.00 ( 24.00%) 957321.00
length=4194317: 750670.00 ( 17.00%) 1135660.00 (-25.05%) 710216.00 ( 21.00%) 908166.00
length=4194308: 779111.00 ( 16.00%) 1139110.00 (-22.07%) 684378.00 ( 26.00%) 933169.00
length=4194316: 814919.00 ( 12.00%) 1168830.00 (-25.88%) 727064.00 ( 21.00%) 928514.00
length=4194309: 823692.00 ( 13.00%) 1136610.00 (-19.95%) 704266.00 ( 25.00%) 947549.00
length=4194315: 764915.00 ( 14.00%) 1142620.00 (-27.94%) 717230.00 ( 19.00%) 893064.00
length=4194310: 766340.00 ( 16.00%) 1133600.00 (-24.13%) 671266.00 ( 26.00%) 913245.00
length=4194314: 758784.00 ( 16.00%) 1138640.00 (-25.83%) 667595.00 ( 26.00%) 904895.00
length=4194311: 721838.00 ( 17.00%) 1136080.00 (-30.34%) 671049.00 ( 23.00%) 871641.00
length=4194313: 739005.00 ( 17.00%) 1135990.00 (-27.01%) 679886.00 ( 23.00%) 894418.00
length=8388608: 1492280.00 ( 13.22%) 2080460.00 (-20.98%) 1021480.00 ( 40.60%) 1719620.00
length=8388624: 1257050.00 ( 18.06%) 1855470.00 (-20.95%) 958960.00 ( 37.49%) 1534100.00
length=8388609: 1619780.00 ( 14.59%) 2239890.00 (-18.11%) 1418620.00 ( 25.20%) 1896460.00
length=8388623: 1543680.00 ( 17.09%) 2245380.00 (-20.60%) 1412280.00 ( 24.15%) 1861910.00
length=8388610: 1626290.00 ( 14.44%) 2247150.00 (-18.23%) 1420220.00 ( 25.28%) 1900740.00
length=8388622: 1541570.00 ( 17.41%) 2259510.00 (-21.06%) 1413580.00 ( 24.27%) 1866510.00
length=8388611: 1621980.00 ( 14.91%) 2251140.00 (-18.10%) 1423660.00 ( 25.31%) 1906140.00
length=8388621: 1521500.00 ( 18.41%) 2259640.00 (-21.17%) 1410280.00 ( 24.37%) 1864800.00
length=8388612: 1624520.00 ( 14.77%) 2264740.00 (-18.82%) 1425720.00 ( 25.20%) 1906020.00
length=8388620: 1447160.00 ( 18.46%) 2284810.00 (-28.74%) 1414230.00 ( 20.31%) 1774740.00
length=8388613: 1618460.00 ( 15.34%) 2244840.00 (-17.43%) 1432300.00 ( 25.08%) 1911660.00
length=8388619: 1449900.00 ( 18.83%) 2244000.00 (-25.62%) 1427800.00 ( 20.07%) 1786330.00
length=8388614: 1675980.00 ( 13.36%) 2256060.00 (-16.62%) 1436500.00 ( 25.74%) 1934520.00
length=8388618: 1468570.00 ( 17.28%) 2245000.00 (-26.45%) 1278590.00 ( 27.98%) 1775350.00
length=8388615: 1547400.00 ( 16.39%) 2263280.00 (-22.29%) 1373370.00 ( 25.80%) 1850820.00
length=8388617: 1548250.00 ( 15.74%) 2245720.00 (-22.21%) 1414830.00 ( 23.00%) 1837550.00
length=16777216: 3000940.00 ( 12.88%) 4187300.00 (-21.56%) 2039470.00 ( 40.79%) 3444540.00
length=16777232: 2641300.00 ( 16.09%) 3652870.00 (-16.05%) 1823980.00 ( 42.05%) 3147710.00
length=16777217: 3192100.00 ( 16.29%) 4455240.00 (-16.84%) 2838200.00 ( 25.57%) 3813140.00
length=16777231: 3245480.00 ( 15.48%) 4484010.00 (-16.78%) 2848240.00 ( 25.82%) 3839850.00
length=16777218: 3257810.00 ( 14.56%) 4452740.00 (-16.78%) 2829940.00 ( 25.78%) 3812960.00
length=16777230: 3222920.00 ( 15.75%) 4476790.00 (-17.03%) 2851140.00 ( 25.47%) 3825360.00
length=16777219: 3210610.00 ( 15.61%) 4477900.00 (-17.70%) 2829920.00 ( 25.62%) 3804480.00
length=16777229: 3209900.00 ( 16.23%) 4493540.00 (-17.27%) 2853500.00 ( 25.53%) 3831880.00
length=16777220: 3244780.00 ( 14.60%) 4488820.00 (-18.14%) 2827340.00 ( 25.59%) 3799720.00
length=16777228: 3245340.00 ( 15.63%) 4524140.00 (-17.61%) 2861600.00 ( 25.61%) 3846620.00
length=16777221: 3278220.00 ( 13.98%) 4482320.00 (-17.61%) 2835000.00 ( 25.61%) 3811080.00
length=16777227: 3240360.00 ( 15.43%) 4490800.00 (-17.21%) 2856230.00 ( 25.45%) 3831450.00
length=16777222: 3267210.00 ( 26.25%) 4494270.00 ( -1.45%) 2857180.00 ( 35.51%) 4430130.00
length=16777226: 3808300.00 ( 6.40%) 4557000.00 (-12.00%) 2968720.00 ( 27.03%) 4068580.00
length=16777223: 3620960.00 ( 8.70%) 4540540.00 (-14.49%) 2930680.00 ( 26.10%) 3965800.00
length=16777225: 3488200.00 ( 10.88%) 4509180.00 (-15.21%) 2869670.00 ( 26.68%) 3913840.00
length=33554432: 6280650.00 ( 12.31%) 8688930.00 (-21.32%) 4289930.00 ( 40.10%) 7162220.00
length=33554448: 6296310.00 ( 11.40%) 6905540.00 ( 2.82%) 4111960.00 ( 42.14%) 7106250.00
length=33554433: 6679210.00 ( 14.72%) 8847130.00 (-12.96%) 5851070.00 ( 25.29%) 7831840.00
length=33554447: 6665510.00 ( 14.16%) 8862040.00 (-14.13%) 5826230.00 ( 24.97%) 7765140.00
length=33554434: 6585410.00 ( 15.18%) 8862270.00 (-14.14%) 5790490.00 ( 25.42%) 7764050.00
length=33554446: 6569800.00 ( 15.66%) 8850790.00 (-13.62%) 5800580.00 ( 25.54%) 7790060.00
length=33554435: 6582510.00 ( 15.15%) 8849150.00 (-14.06%) 5804800.00 ( 25.18%) 7758130.00
length=33554445: 6587410.00 ( 15.55%) 8821360.00 (-13.09%) 5798640.00 ( 25.66%) 7800540.00
length=33554436: 6625050.00 ( 14.60%) 8875050.00 (-14.40%) 5790940.00 ( 25.35%) 7757910.00
length=33554444: 6528030.00 ( 16.13%) 8883860.00 (-14.14%) 5799910.00 ( 25.48%) 7783500.00
length=33554437: 6547760.00 ( 15.66%) 8842220.00 (-13.90%) 5808500.00 ( 25.18%) 7763300.00
length=33554443: 6597490.00 ( 15.31%) 8825810.00 (-13.30%) 5799360.00 ( 25.55%) 7790020.00
length=33554438: 6566200.00 ( 15.87%) 8830900.00 (-13.14%) 5792320.00 ( 25.79%) 7805180.00
length=33554442: 6599360.00 ( 14.73%) 8845460.00 (-14.29%) 5830230.00 ( 24.67%) 7739690.00
length=33554439: 6638320.00 ( 14.56%) 8831880.00 (-13.67%) 5810000.00 ( 25.22%) 7769450.00
length=33554441: 6622430.00 ( 15.30%) 8856990.00 (-13.28%) 5808460.00 ( 25.71%) 7818650.00
length=128: 66.27 (-36.08%) 53.19 ( -9.23%) 45.22 ( 7.14%) 48.70
length=144: 48.00 ( -0.97%) 48.57 ( -2.16%) 44.77 ( 5.82%) 47.54
length=129: 44.08 ( -7.72%) 42.10 ( -2.89%) 38.21 ( 6.62%) 40.92
length=143: 52.58 ( -0.61%) 45.84 ( 12.29%) 48.18 ( 7.80%) 52.26
length=130: 41.20 ( -0.50%) 40.80 ( 0.50%) 38.03 ( 7.23%) 41.00
length=142: 51.04 ( -0.58%) 45.10 ( 11.12%) 46.64 ( 8.08%) 50.74
length=131: 42.94 ( -0.43%) 41.47 ( 3.00%) 39.53 ( 7.53%) 42.75
length=141: 51.07 ( -0.56%) 45.31 ( 10.77%) 46.56 ( 8.32%) 50.78
length=132: 41.22 ( -0.47%) 41.01 ( 0.03%) 38.03 ( 7.29%) 41.02
length=140: 47.91 ( -0.18%) 43.92 ( 8.15%) 43.91 ( 8.18%) 47.82
length=133: 47.25 ( -6.16%) 43.96 ( 1.23%) 41.28 ( 7.26%) 44.51
length=139: 49.60 ( -0.34%) 44.47 ( 10.03%) 45.28 ( 8.38%) 49.43
length=134: 44.74 ( -0.64%) 42.16 ( 5.16%) 40.94 ( 7.92%) 44.46
length=138: 48.15 ( -0.55%) 43.70 ( 8.76%) 43.81 ( 8.52%) 47.89
length=135: 46.50 ( -0.32%) 42.78 ( 7.71%) 42.45 ( 8.44%) 46.36
length=137: 48.22 ( -0.53%) 43.63 ( 9.05%) 43.81 ( 8.66%) 47.97
length=256: 80.65 (-17.14%) 78.30 (-13.73%) 65.95 ( 4.21%) 68.85
length=272: 65.72 ( 2.00%) 73.41 ( -9.47%) 64.38 ( 4.00%) 67.06
length=257: 73.95 ( 3.12%) 78.75 ( -3.17%) 74.63 ( 2.23%) 76.33
length=271: 89.63 ( 0.04%) 84.94 ( 5.27%) 86.55 ( 3.47%) 89.66
length=258: 76.43 ( -0.42%) 78.76 ( -3.49%) 74.70 ( 1.84%) 76.10
length=270: 88.04 ( -0.27%) 83.97 ( 4.37%) 85.13 ( 3.05%) 87.80
length=259: 77.93 ( -0.39%) 79.58 ( -2.51%) 76.11 ( 1.96%) 77.63
length=269: 87.63 ( -0.31%) 83.95 ( 3.90%) 84.71 ( 3.02%) 87.36
length=260: 77.04 ( -0.48%) 79.08 ( -3.15%) 75.18 ( 1.94%) 76.67
length=268: 84.46 ( -0.22%) 82.44 ( 2.18%) 82.09 ( 2.59%) 84.28
length=261: 79.66 ( -0.05%) 80.59 ( -1.21%) 77.64 ( 2.50%) 79.62
length=267: 85.39 ( -0.16%) 83.01 ( 2.62%) 82.81 ( 2.85%) 85.25
length=262: 79.78 ( -0.10%) 80.26 ( -0.70%) 77.89 ( 2.28%) 79.70
length=266: 83.67 ( -0.21%) 82.11 ( 1.66%) 81.43 ( 2.48%) 83.50
length=263: 81.45 ( -0.18%) 81.61 ( -0.38%) 79.47 ( 2.26%) 81.30
length=265: 83.84 ( -0.67%) 82.45 ( 1.00%) 81.13 ( 2.58%) 83.28
length=512: 152.83 (-18.98%) 157.46 (-22.58%) 125.56 ( 2.25%) 128.45
length=528: 121.63 ( -0.21%) 137.09 (-12.94%) 118.77 ( 2.15%) 121.38
length=513: 146.12 ( 2.89%) 158.51 ( -5.35%) 153.03 ( -1.70%) 150.47
length=527: 158.00 ( -0.22%) 160.44 ( -1.77%) 155.19 ( 1.57%) 157.65
length=514: 149.18 ( 0.05%) 154.75 ( -3.68%) 146.54 ( 1.82%) 149.25
length=526: 158.04 ( -0.12%) 159.86 ( -1.27%) 155.56 ( 1.45%) 157.85
length=515: 149.95 ( -0.87%) 155.20 ( -4.41%) 146.59 ( 1.38%) 148.65
length=525: 156.30 ( -0.12%) 158.74 ( -1.68%) 153.84 ( 1.45%) 156.11
length=516: 151.82 ( 0.81%) 155.90 ( -1.85%) 149.22 ( 2.51%) 153.06
length=524: 158.80 ( -0.26%) 158.87 ( -0.31%) 154.73 ( 2.31%) 158.39
length=517: 151.81 ( -1.31%) 156.11 ( -4.18%) 147.92 ( 1.28%) 149.84
length=523: 154.53 ( -0.12%) 158.44 ( -2.64%) 152.12 ( 1.45%) 154.35
length=518: 151.57 ( -0.02%) 156.10 ( -3.01%) 149.00 ( 1.67%) 151.54
length=522: 154.78 ( 0.25%) 164.84 ( -6.23%) 153.96 ( 0.78%) 155.17
length=519: 152.23 ( -0.23%) 156.51 ( -3.05%) 149.14 ( 1.80%) 151.88
length=521: 153.49 ( -0.46%) 157.22 ( -2.90%) 150.71 ( 1.36%) 152.79
length=1024: 300.02 (-21.76%) 311.62 (-26.47%) 244.04 ( 0.96%) 246.41
length=1040: 228.82 ( 0.15%) 267.20 (-16.60%) 228.05 ( 0.49%) 229.17
length=1025: 301.38 ( 1.38%) 311.68 ( -1.99%) 299.25 ( 2.08%) 305.60
length=1039: 326.57 ( -0.99%) 316.49 ( 2.13%) 318.47 ( 1.52%) 323.37
length=1026: 308.05 ( -0.64%) 309.72 ( -1.18%) 301.78 ( 1.41%) 306.10
length=1038: 325.21 ( -0.24%) 314.27 ( 3.13%) 316.98 ( 2.30%) 324.43
length=1027: 309.16 ( -0.33%) 312.08 ( -1.27%) 302.74 ( 1.76%) 308.15
length=1037: 323.15 ( -0.75%) 315.46 ( 1.65%) 315.89 ( 1.52%) 320.76
length=1028: 309.12 ( -1.41%) 307.63 ( -0.92%) 304.28 ( 0.18%) 304.83
length=1036: 319.39 ( -1.76%) 310.65 ( 1.03%) 312.03 ( 0.59%) 313.88
length=1029: 311.70 ( -0.80%) 312.82 ( -1.16%) 304.84 ( 1.42%) 309.22
length=1035: 318.63 ( 4.84%) 314.50 ( 6.08%) 312.65 ( 6.63%) 334.84
length=1030: 323.14 ( -3.78%) 317.45 ( -1.96%) 306.51 ( 1.56%) 311.36
length=1034: 320.22 ( -1.42%) 312.46 ( 1.03%) 312.04 ( 1.17%) 315.72
length=1031: 313.92 ( -0.74%) 313.76 ( -0.69%) 306.80 ( 1.54%) 311.61
length=1033: 317.15 ( -0.97%) 313.84 ( 0.08%) 310.31 ( 1.21%) 314.11
length=2048: 583.91 (-22.36%) 619.40 (-29.80%) 475.65 ( 0.33%) 477.21
length=2064: 441.37 ( 0.41%) 529.05 (-19.37%) 346.21 ( 21.88%) 443.19
length=2049: 557.70 ( 5.57%) 598.78 ( -1.39%) 477.83 ( 19.09%) 590.60
length=2063: 616.61 ( -1.14%) 605.57 ( 0.67%) 474.43 ( 22.18%) 609.65
length=2050: 599.67 ( 0.11%) 600.66 ( -0.06%) 480.02 ( 20.04%) 600.32
length=2062: 615.52 ( -0.38%) 605.63 ( 1.23%) 486.73 ( 20.62%) 613.18
length=2051: 598.48 ( -0.97%) 599.18 ( -1.09%) 471.74 ( 20.41%) 592.74
length=2061: 614.90 ( -2.06%) 605.01 ( -0.42%) 475.26 ( 21.12%) 602.49
length=2052: 591.65 ( 0.98%) 596.87 ( 0.11%) 475.49 ( 20.42%) 597.52
length=2060: 601.77 ( -0.77%) 606.16 ( -1.50%) 482.20 ( 19.26%) 597.19
length=2053: 598.25 ( -0.49%) 605.99 ( -1.79%) 468.73 ( 21.26%) 595.31
length=2059: 609.89 ( -1.69%) 606.33 ( -1.10%) 474.83 ( 20.83%) 599.76
length=2054: 600.35 ( -1.35%) 603.25 ( -1.84%) 471.24 ( 20.45%) 592.37
length=2058: 602.73 ( -1.52%) 602.87 ( -1.54%) 472.11 ( 20.48%) 593.70
length=2055: 601.95 ( -0.54%) 611.06 ( -2.06%) 470.20 ( 21.46%) 598.70
length=2057: 610.13 ( -1.59%) 613.92 ( -2.22%) 470.61 ( 21.64%) 600.59
length=4096: 1115.88 (-18.50%) 1198.22 (-27.25%) 615.80 ( 34.60%) 941.64
length=4112: 867.86 ( -0.05%) 1047.58 (-20.77%) 626.42 ( 27.79%) 867.45
length=4097: 1041.49 ( 7.56%) 1176.94 ( -4.46%) 824.96 ( 26.78%) 1126.72
length=4111: 1137.29 ( 0.20%) 1186.28 ( -4.10%) 834.53 ( 26.77%) 1139.56
length=4098: 1122.45 ( -0.18%) 1171.37 ( -4.55%) 828.69 ( 26.04%) 1120.43
length=4110: 1125.74 ( 1.13%) 1189.22 ( -4.44%) 831.24 ( 27.00%) 1138.65
length=4099: 1125.48 ( 0.06%) 1170.83 ( -3.97%) 831.53 ( 26.16%) 1126.14
length=4109: 1133.93 ( 6.43%) 1267.96 ( -4.63%) 873.15 ( 27.95%) 1211.86
length=4100: 1176.48 ( -2.88%) 1231.29 ( -7.68%) 842.10 ( 26.36%) 1143.51
length=4108: 1130.34 ( 0.39%) 1193.59 ( -5.18%) 830.01 ( 26.86%) 1134.82
length=4101: 1120.91 ( 1.04%) 1182.03 ( -4.36%) 829.82 ( 26.74%) 1132.68
length=4107: 1137.13 ( 0.32%) 1179.60 ( -3.40%) 826.49 ( 27.55%) 1140.81
length=4102: 1128.99 ( -0.00%) 1175.44 ( -4.12%) 827.50 ( 26.70%) 1128.96
length=4106: 1128.90 ( 0.84%) 1177.22 ( -3.40%) 824.91 ( 27.54%) 1138.51
length=4103: 1136.80 ( -0.07%) 1180.73 ( -3.94%) 827.38 ( 27.16%) 1135.96
length=4105: 1140.69 ( 0.63%) 1183.87 ( -3.14%) 829.11 ( 27.77%) 1147.88
length=8192: 2135.74 (-12.41%) 2368.70 (-24.67%) 1185.59 ( 37.60%) 1899.97
length=8208: 1744.90 ( 0.11%) 2122.77 (-21.52%) 1204.95 ( 31.02%) 1746.84
length=8193: 2011.12 ( 6.94%) 2321.43 ( -7.42%) 1571.89 ( 27.27%) 2161.16
length=8207: 2144.63 ( 2.90%) 2337.80 ( -5.84%) 1593.75 ( 27.84%) 2208.74
length=8194: 2117.12 ( 2.14%) 2321.88 ( -7.32%) 1571.67 ( 27.35%) 2163.44
length=8206: 2125.81 ( 2.98%) 2319.91 ( -5.87%) 1609.70 ( 26.54%) 2191.20
length=8195: 2120.04 ( 2.80%) 2319.93 ( -6.36%) 1567.57 ( 28.13%) 2181.21
length=8205: 2134.25 ( 2.94%) 2319.78 ( -5.50%) 1583.60 ( 27.98%) 2198.89
length=8196: 2104.65 ( 2.27%) 2312.04 ( -7.36%) 1576.79 ( 26.78%) 2153.45
length=8204: 2099.97 ( 3.49%) 2312.66 ( -6.28%) 1602.45 ( 26.36%) 2175.94
length=8197: 2105.41 ( 3.45%) 2312.77 ( -6.06%) 1568.83 ( 28.05%) 2180.55
length=8203: 2133.16 ( 3.40%) 2318.83 ( -5.01%) 1589.13 ( 28.04%) 2208.24
length=8198: 2123.02 ( 2.21%) 2313.01 ( -6.54%) 1588.55 ( 26.83%) 2171.10
length=8202: 2122.60 ( 2.45%) 2311.70 ( -6.25%) 1594.16 ( 26.73%) 2175.82
length=8199: 2125.08 ( 2.89%) 2311.98 ( -5.65%) 1574.55 ( 28.05%) 2188.34
length=8201: 2128.49 ( 3.06%) 2307.06 ( -5.07%) 1582.46 ( 27.93%) 2195.65
length=16384: 4036.38 ( -6.67%) 4624.36 (-22.21%) 2300.30 ( 39.21%) 3783.95
length=16400: 3477.43 ( 0.84%) 4240.38 (-20.91%) 2352.72 ( 32.91%) 3507.04
length=16385: 3956.05 ( 6.20%) 4627.32 ( -9.72%) 3060.36 ( 27.44%) 4217.55
length=16399: 4091.51 ( 4.04%) 4607.30 ( -8.06%) 3138.83 ( 26.38%) 4263.81
length=16386: 4125.76 ( 1.98%) 4584.94 ( -8.93%) 3043.21 ( 27.70%) 4208.89
length=16398: 4077.08 ( 13.93%) 4829.99 ( -1.97%) 3328.65 ( 29.73%) 4736.76
length=16387: 4635.86 ( -3.78%) 4969.70 (-11.25%) 3142.77 ( 29.65%) 4467.08
length=16397: 4344.94 ( 0.13%) 4731.51 ( -8.75%) 3158.14 ( 27.41%) 4350.68
length=16388: 4201.90 ( 1.67%) 4684.62 ( -9.63%) 3068.76 ( 28.19%) 4273.19
length=16396: 4096.45 ( 3.73%) 4634.13 ( -8.91%) 3104.00 ( 27.05%) 4255.19
length=16389: 4093.97 ( 3.39%) 4616.48 ( -8.94%) 3083.18 ( 27.24%) 4237.55
length=16395: 4101.94 ( 3.90%) 4601.39 ( -7.80%) 3119.61 ( 26.92%) 4268.49
length=16390: 4103.63 ( 2.94%) 4595.93 ( -8.71%) 3071.57 ( 27.35%) 4227.74
length=16394: 4080.00 ( 4.00%) 4597.61 ( -7.87%) 3083.94 ( 27.64%) 4262.00
length=16391: 4091.93 ( 3.14%) 4595.08 ( -8.77%) 3097.15 ( 26.69%) 4224.61
length=16393: 4107.73 ( 3.10%) 4605.84 ( -8.66%) 3104.48 ( 26.76%) 4238.94
length=32768: 6650.02 ( 10.65%) 9174.17 (-23.27%) 4526.37 ( 39.18%) 7442.25
length=32784: 5435.11 ( 20.71%) 8372.00 (-22.13%) 4493.79 ( 34.45%) 6855.03
length=32769: 6501.52 ( 21.50%) 9147.42 (-10.45%) 6006.67 ( 27.47%) 8281.81
length=32783: 6575.55 ( 21.63%) 9187.63 ( -9.51%) 6210.61 ( 25.98%) 8390.03
length=32770: 6587.31 ( 20.71%) 9140.36 (-10.02%) 6027.14 ( 27.45%) 8307.74
length=32782: 6555.64 ( 21.67%) 9160.63 ( -9.46%) 6172.61 ( 26.25%) 8369.14
length=32771: 6586.04 ( 20.83%) 9158.43 (-10.09%) 6047.49 ( 27.30%) 8318.96
length=32781: 6579.04 ( 21.29%) 9154.36 ( -9.52%) 6173.93 ( 26.14%) 8358.52
length=32772: 6558.14 ( 20.96%) 9184.54 (-10.69%) 5994.79 ( 27.75%) 8297.18
length=32780: 6549.35 ( 21.41%) 9175.79 (-10.11%) 6115.82 ( 26.61%) 8333.21
length=32773: 6562.04 ( 21.29%) 9150.94 ( -9.77%) 6056.51 ( 27.35%) 8336.61
length=32779: 6598.91 ( 21.22%) 9163.61 ( -9.39%) 6149.70 ( 26.59%) 8376.67
length=32774: 6606.88 ( 20.94%) 9154.47 ( -9.55%) 6061.78 ( 27.46%) 8356.63
length=32778: 6587.32 ( 21.18%) 9150.47 ( -9.50%) 6137.37 ( 26.56%) 8356.96
length=32775: 6599.20 ( 21.00%) 9164.48 ( -9.71%) 6078.86 ( 27.23%) 8353.57
length=32777: 6573.49 ( 22.03%) 9148.41 ( -8.51%) 6126.53 ( 27.33%) 8431.05
length=65536: 13116.30 ( 14.52%) 18690.50 (-21.81%) 9236.41 ( 39.80%) 15343.80
length=65552: 11055.00 ( 20.98%) 17103.20 (-22.25%) 9201.68 ( 34.23%) 13990.40
length=65537: 13182.90 ( 20.79%) 18410.00 (-10.62%) 11968.50 ( 28.08%) 16642.30
length=65551: 13059.50 ( 21.81%) 18433.80 (-10.37%) 12358.20 ( 26.01%) 16701.40
length=65538: 12993.90 ( 21.56%) 18293.50 (-10.44%) 11908.50 ( 28.11%) 16564.90
length=65550: 12908.00 ( 32.41%) 18395.00 ( 3.68%) 13036.60 ( 31.74%) 19097.60
length=65539: 14774.70 ( 16.78%) 20012.30 (-12.72%) 12330.00 ( 30.55%) 17753.60
length=65549: 14078.20 ( 17.44%) 18994.00 (-12.00%) 12412.70 ( 27.21%) 17053.00
length=65540: 13352.40 ( 19.41%) 18538.60 (-11.89%) 11991.00 ( 27.63%) 16568.50
length=65548: 13008.80 ( 21.89%) 18425.60 (-10.64%) 12222.60 ( 26.61%) 16653.70
length=65541: 12942.00 ( 21.92%) 18303.20 (-10.42%) 12096.60 ( 27.02%) 16576.30
length=65547: 12975.40 ( 22.12%) 18323.30 ( -9.98%) 12240.80 ( 26.53%) 16661.10
length=65542: 12962.50 ( 21.66%) 18314.00 (-10.69%) 12042.10 ( 27.22%) 16545.50
length=65546: 12885.00 ( 22.56%) 18307.20 (-10.03%) 12195.20 ( 26.71%) 16638.60
length=65543: 12948.30 ( 21.90%) 18275.10 (-10.23%) 12100.80 ( 27.01%) 16578.60
length=65545: 12876.50 ( 22.42%) 18286.00 (-10.17%) 12181.10 ( 26.61%) 16597.80
length=131072: 25610.70 ( 14.01%) 36773.20 (-23.47%) 17885.90 ( 39.94%) 29781.90
length=131088: 21460.30 ( 21.46%) 33569.90 (-22.86%) 17790.40 ( 34.89%) 27322.80
length=131073: 25507.30 ( 22.69%) 36413.90 (-10.37%) 23704.50 ( 28.15%) 32993.30
length=131087: 25646.20 ( 22.42%) 36600.20 (-10.72%) 24618.50 ( 25.53%) 33057.60
length=131074: 25843.40 ( 21.90%) 36537.60 (-10.42%) 23805.20 ( 28.06%) 33090.70
length=131086: 25673.70 ( 22.38%) 36454.90 (-10.22%) 24512.80 ( 25.89%) 33074.60
length=131075: 25772.00 ( 21.65%) 36520.30 (-11.02%) 23925.60 ( 27.26%) 32893.90
length=131085: 25598.00 ( 22.00%) 36531.20 (-10.51%) 24549.60 ( 25.74%) 33057.00
length=131076: 25581.20 ( 22.32%) 36706.80 (-11.46%) 23776.20 ( 27.81%) 32933.50
length=131084: 25613.50 ( 22.53%) 36593.60 (-10.68%) 24274.80 ( 26.58%) 33063.60
length=131077: 25547.00 ( 22.54%) 36487.40 (-10.63%) 23998.30 ( 27.24%) 32981.10
length=131083: 25781.10 ( 21.95%) 36550.90 (-10.66%) 24429.60 ( 26.04%) 33029.50
length=131078: 25622.70 ( 22.40%) 36646.80 (-10.99%) 23974.20 ( 27.39%) 33017.70
length=131082: 25575.00 ( 22.55%) 36434.30 (-10.34%) 24382.40 ( 26.16%) 33020.90
length=131079: 25678.40 ( 22.17%) 36579.40 (-10.86%) 24143.90 ( 26.82%) 32994.60
length=131081: 25655.40 ( 22.32%) 36579.00 (-10.75%) 24231.90 ( 26.63%) 33027.10
length=262144: 51207.30 ( 13.98%) 72938.90 (-22.52%) 35915.20 ( 39.67%) 59531.20
length=262160: 43210.80 ( 21.95%) 67374.80 (-21.70%) 36468.00 ( 34.13%) 55360.10
length=262145: 51309.20 ( 21.80%) 72876.70 (-11.07%) 47713.80 ( 27.28%) 65613.20
length=262159: 51120.50 ( 22.17%) 73412.20 (-11.77%) 49354.00 ( 24.86%) 65680.90
length=262146: 51379.50 ( 21.63%) 72626.60 (-10.78%) 47646.20 ( 27.32%) 65558.40
length=262158: 50847.00 ( 22.56%) 73104.20 (-11.33%) 48932.70 ( 25.48%) 65661.70
length=262147: 50823.80 ( 31.35%) 72398.00 ( 2.21%) 49522.50 ( 33.11%) 74035.60
length=262157: 58526.20 ( 17.16%) 79162.70 (-12.05%) 50510.90 ( 28.51%) 70651.20
length=262148: 55606.30 ( 17.33%) 75822.00 (-12.73%) 47852.00 ( 28.86%) 67261.70
length=262156: 52492.70 ( 20.58%) 73700.10 (-11.51%) 48757.30 ( 26.23%) 66091.70
length=262149: 51340.90 ( 21.69%) 73301.80 (-11.81%) 47835.80 ( 27.03%) 65559.40
length=262155: 50855.50 ( 22.46%) 72623.10 (-10.72%) 48875.10 ( 25.48%) 65589.10
length=262150: 50648.90 ( 22.63%) 73185.50 (-11.79%) 47862.00 ( 26.89%) 65464.50
length=262154: 50788.70 ( 22.35%) 72727.20 (-11.18%) 48716.30 ( 25.52%) 65411.20
length=262151: 50842.50 ( 22.94%) 72702.90 (-10.19%) 48134.10 ( 27.05%) 65980.40
length=262153: 51335.80 ( 22.71%) 73220.00 (-10.23%) 48999.40 ( 26.23%) 66423.70
length=524288: 103239.00 ( 14.00%) 147724.00 (-22.00%) 73463.90 ( 39.41%) 121244.00
length=524304: 86967.20 ( 21.43%) 135091.00 (-23.00%) 73281.90 ( 33.80%) 110690.00
length=524289: 100735.00 ( 22.00%) 145991.00 (-13.00%) 94982.00 ( 27.00%) 130286.00
length=524303: 101670.00 ( 22.00%) 145160.00 (-11.00%) 98092.30 ( 25.10%) 130972.00
length=524290: 100919.00 ( 22.00%) 145627.00 (-13.00%) 95608.30 ( 26.45%) 129993.00
length=524302: 101060.00 ( 24.00%) 144909.00 ( -8.00%) 98590.60 ( 26.58%) 134279.00
length=524291: 104574.00 ( 20.00%) 147491.00 (-13.00%) 96004.20 ( 26.94%) 131402.00
length=524301: 102936.00 ( 21.00%) 145594.00 (-12.00%) 97421.70 ( 25.48%) 130728.00
length=524292: 102010.00 ( 22.00%) 146099.00 (-12.00%) 95049.50 ( 27.48%) 131068.00
length=524300: 101196.00 ( 22.00%) 145732.00 (-12.00%) 96657.50 ( 26.02%) 130654.00
length=524293: 101524.00 ( 22.00%) 144887.00 (-12.00%) 95421.70 ( 26.73%) 130231.00
length=524299: 101655.00 ( 22.00%) 145021.00 (-11.00%) 97079.50 ( 25.74%) 130729.00
length=524294: 101612.00 ( 22.00%) 145707.00 (-12.00%) 95804.10 ( 26.84%) 130955.00
length=524298: 101258.00 ( 22.00%) 145163.00 (-12.00%) 96505.30 ( 25.90%) 130244.00
length=524295: 101785.00 ( 21.00%) 144656.00 (-11.00%) 96120.90 ( 26.33%) 130468.00
length=524297: 101165.00 ( 22.00%) 145102.00 (-12.00%) 96348.90 ( 26.26%) 130662.00
length=1048576: 202572.00 ( 16.00%) 290419.00 (-20.00%) 148252.00 ( 38.00%) 243006.00
length=1048592: 171822.00 ( 22.00%) 266806.00 (-20.00%) 148570.00 ( 33.00%) 223142.00
length=1048577: 212137.00 ( 21.00%) 300943.00 (-12.00%) 194730.00 ( 27.00%) 269788.00
length=1048591: 207704.00 ( 21.00%) 297847.00 (-13.00%) 200116.00 ( 24.00%) 264511.00
length=1048578: 202312.00 ( 22.00%) 303038.00 (-17.00%) 188977.00 ( 27.00%) 259438.00
length=1048590: 202955.00 ( 21.00%) 295984.00 (-15.00%) 197438.00 ( 23.00%) 259635.00
length=1048579: 207280.00 ( 22.00%) 350937.00 (-32.00%) 193093.00 ( 27.00%) 267365.00
length=1048589: 202854.00 ( 32.00%) 296888.00 ( 0.00%) 204539.00 ( 31.00%) 299166.00
length=1048580: 237047.00 ( 16.00%) 337698.00 (-20.00%) 197028.00 ( 30.00%) 282857.00
length=1048588: 225748.00 ( 16.00%) 313068.00 (-16.00%) 198400.00 ( 26.00%) 269910.00
length=1048581: 214414.00 ( 19.00%) 310873.00 (-17.00%) 196363.00 ( 26.00%) 267442.00
length=1048587: 211633.00 ( 21.00%) 301078.00 (-13.00%) 199356.00 ( 25.00%) 268166.00
length=1048582: 205749.00 ( 20.00%) 303532.00 (-18.00%) 192994.00 ( 25.00%) 259009.00
length=1048586: 202573.00 ( 22.00%) 297378.00 (-15.00%) 194532.00 ( 25.00%) 259912.00
length=1048583: 205648.00 ( 21.00%) 300663.00 (-15.00%) 194682.00 ( 26.00%) 263389.00
length=1048585: 206218.00 ( 22.00%) 298019.00 (-13.00%) 196484.00 ( 25.00%) 264461.00
length=2097152: 407366.00 ( 17.00%) 585856.00 (-19.00%) 312590.00 ( 36.00%) 493615.00
length=2097168: 343420.00 ( 23.00%) 541487.00 (-21.00%) 308636.00 ( 31.00%) 448923.00
length=2097153: 449567.00 ( 20.00%) 600551.00 ( -7.00%) 407804.00 ( 27.00%) 565234.00
length=2097167: 416661.00 ( 22.00%) 601751.00 (-13.00%) 405342.00 ( 24.00%) 535177.00
length=2097154: 423269.00 ( 21.00%) 601716.00 (-13.00%) 394774.00 ( 26.00%) 536904.00
length=2097166: 412193.00 ( 22.00%) 604206.00 (-15.00%) 403032.00 ( 23.00%) 529746.00
length=2097155: 433871.00 ( 28.00%) 658755.00 ( -9.00%) 415241.00 ( 31.00%) 604566.00
length=2097165: 477727.00 ( 13.00%) 645184.00 (-17.00%) 415729.00 ( 25.00%) 555378.00
length=2097156: 424596.00 ( 18.00%) 622339.00 (-19.00%) 387786.00 ( 25.00%) 523053.00
length=2097164: 413266.00 ( 20.00%) 613514.00 (-18.00%) 396072.00 ( 24.00%) 521488.00
length=2097157: 428079.00 ( 20.00%) 605611.00 (-13.00%) 395794.00 ( 26.00%) 540399.00
length=2097163: 414076.00 ( 21.00%) 606449.00 (-15.00%) 406751.00 ( 23.00%) 530783.00
length=2097158: 422626.00 ( 21.00%) 606239.00 (-13.00%) 394724.00 ( 26.00%) 539184.00
length=2097162: 414003.00 ( 21.00%) 606936.00 (-16.00%) 397854.00 ( 24.00%) 527535.00
length=2097159: 421371.00 ( 21.00%) 604340.00 (-13.00%) 403091.00 ( 25.00%) 537462.00
length=2097161: 419372.00 ( 21.00%) 606651.00 (-15.00%) 395754.00 ( 25.00%) 531477.00
length=4194304: 816732.00 ( 17.00%) 1151150.00 (-15.75%) 666119.00 ( 33.00%) 994486.00
length=4194320: 685094.00 ( 23.00%) 1094500.00 (-22.70%) 653414.00 ( 26.00%) 891994.00
length=4194305: 890606.00 ( 19.25%) 1212020.00 ( -9.90%) 845619.00 ( 23.32%) 1102860.00
length=4194319: 847158.00 ( 20.79%) 1214160.00 (-13.53%) 844476.00 ( 21.04%) 1069480.00
length=4194306: 894996.00 ( 18.83%) 1206220.00 ( -9.40%) 847558.00 ( 23.13%) 1102590.00
length=4194318: 845055.00 ( 21.12%) 1208550.00 (-12.81%) 842018.00 ( 21.41%) 1071360.00
length=4194307: 887760.00 ( 19.74%) 1202890.00 ( -8.74%) 849276.00 ( 23.22%) 1106170.00
length=4194317: 859988.00 ( 29.33%) 1203050.00 ( 1.14%) 845961.00 ( 30.49%) 1216950.00
length=4194308: 1011340.00 ( 10.80%) 1338450.00 (-18.06%) 846686.00 ( 25.32%) 1133750.00
length=4194316: 920300.00 ( 14.63%) 1266440.00 (-17.48%) 839655.00 ( 22.11%) 1078020.00
length=4194309: 871112.00 ( 19.16%) 1215470.00 (-12.80%) 822021.00 ( 23.71%) 1077550.00
length=4194315: 852191.00 ( 19.54%) 1209040.00 (-14.15%) 843084.00 ( 20.40%) 1059180.00
length=4194310: 858644.00 ( 20.53%) 1201290.00 (-11.19%) 838059.00 ( 22.43%) 1080410.00
length=4194314: 849162.00 ( 20.39%) 1203370.00 (-12.82%) 827862.00 ( 22.38%) 1066620.00
length=4194311: 826529.00 ( 20.72%) 1204600.00 (-15.54%) 816926.00 ( 21.65%) 1042600.00
length=4194313: 843536.00 ( 18.93%) 1205540.00 (-15.87%) 824721.00 ( 20.73%) 1040450.00
length=8388608: 1712130.00 ( 17.61%) 2348700.00 (-13.02%) 1504370.00 ( 27.61%) 2078120.00
length=8388624: 1463720.00 ( 18.65%) 2182060.00 (-21.27%) 1430860.00 ( 20.48%) 1799340.00
length=8388609: 1835900.00 ( 17.78%) 2465890.00 (-10.44%) 1860840.00 ( 16.66%) 2232780.00
length=8388623: 1790540.00 ( 18.92%) 2481250.00 (-12.36%) 1865400.00 ( 15.53%) 2208250.00
length=8388610: 1834980.00 ( 17.99%) 2465940.00 (-10.21%) 1857780.00 ( 16.97%) 2237510.00
length=8388622: 1796030.00 ( 18.75%) 2482060.00 (-12.29%) 1867180.00 ( 15.53%) 2210460.00
length=8388611: 1840330.00 ( 17.87%) 2467840.00 (-10.14%) 1865340.00 ( 16.75%) 2240710.00
length=8388621: 1779400.00 ( 19.70%) 2472360.00 (-11.58%) 1864390.00 ( 15.86%) 2215810.00
length=8388612: 1869430.00 ( 18.11%) 2477290.00 ( -8.51%) 1860690.00 ( 18.49%) 2282910.00
length=8388620: 1806170.00 ( 20.04%) 2477720.00 ( -9.68%) 1862080.00 ( 17.57%) 2258950.00
length=8388613: 1872760.00 ( 18.08%) 2471000.00 ( -8.08%) 1867200.00 ( 18.33%) 2286180.00
length=8388619: 1813000.00 ( 19.72%) 2453430.00 ( -8.64%) 1863090.00 ( 17.50%) 2258410.00
length=8388614: 1869810.00 ( 18.29%) 2462170.00 ( -7.59%) 1866980.00 ( 18.41%) 2288370.00
length=8388618: 1805660.00 ( 19.45%) 2453160.00 ( -9.44%) 1858540.00 ( 17.09%) 2241530.00
length=8388615: 1760060.00 ( 19.24%) 2464380.00 (-13.08%) 1816390.00 ( 16.65%) 2179260.00
length=8388617: 1755540.00 ( 19.61%) 2465170.00 (-12.88%) 1856190.00 ( 15.00%) 2183840.00
length=16777216: 3705450.00 ( 13.62%) 4695780.00 ( -9.47%) 3610200.00 ( 15.84%) 4289500.00
length=16777232: 3770280.00 ( 11.07%) 4073680.00 ( 3.92%) 3507300.00 ( 17.28%) 4239830.00
length=16777217: 3903360.00 ( 12.04%) 4814660.00 ( -8.50%) 4276020.00 ( 3.64%) 4437650.00
length=16777231: 4113460.00 ( 9.59%) 4800360.00 ( -5.51%) 4262900.00 ( 6.30%) 4549540.00
length=16777218: 3912370.00 ( 11.22%) 4791520.00 ( -8.73%) 4260870.00 ( 3.31%) 4406780.00
length=16777230: 4074500.00 ( 10.27%) 4787480.00 ( -5.43%) 4254260.00 ( 6.31%) 4540840.00
length=16777219: 4001410.00 ( 13.90%) 5383920.00 (-15.84%) 4547010.00 ( 2.16%) 4647560.00
length=16777229: 4382360.00 ( 4.22%) 5002400.00 ( -9.33%) 4305790.00 ( 5.90%) 4575540.00
length=16777220: 3960940.00 ( 10.21%) 4834930.00 ( -9.60%) 4281000.00 ( 2.96%) 4411560.00
length=16777228: 4103180.00 ( 9.85%) 4804750.00 ( -5.57%) 4260000.00 ( 6.40%) 4551380.00
length=16777221: 3914360.00 ( 11.05%) 4795150.00 ( -8.96%) 4267440.00 ( 3.03%) 4400770.00
length=16777227: 4080360.00 ( 10.03%) 4787980.00 ( -5.57%) 4256320.00 ( 6.15%) 4535230.00
length=16777222: 4076880.00 ( 10.71%) 4786940.00 ( -4.84%) 4258830.00 ( 6.73%) 4566050.00
length=16777226: 4114630.00 ( 10.23%) 4857360.00 ( -5.97%) 4288970.00 ( 6.43%) 4583620.00
length=16777223: 4117120.00 ( 9.57%) 4824220.00 ( -5.96%) 4266900.00 ( 6.29%) 4553060.00
length=16777225: 4106500.00 ( 9.77%) 4809070.00 ( -5.66%) 4263520.00 ( 6.32%) 4551280.00
length=33554432: 5555570.00 ( 35.88%) 10902300.00 (-25.83%) 4425760.00 ( 48.92%) 8664050.00
length=33554448: 6200440.00 ( 29.46%) 11033900.00 (-25.52%) 4440250.00 ( 49.49%) 8790260.00
length=33554433: 6818220.00 ( 29.67%) 11103900.00 (-14.54%) 4705860.00 ( 51.46%) 9694040.00
length=33554447: 6858530.00 ( 29.19%) 11155600.00 (-15.17%) 4707070.00 ( 51.40%) 9685830.00
length=33554434: 6846210.00 ( 29.03%) 11144400.00 (-15.52%) 4712100.00 ( 51.15%) 9647000.00
length=33554446: 6866510.00 ( 29.12%) 11153000.00 (-15.14%) 4711300.00 ( 51.36%) 9686850.00
length=33554435: 6849040.00 ( 29.03%) 11157400.00 (-15.61%) 4705190.00 ( 51.24%) 9650550.00
length=33554445: 6859370.00 ( 29.04%) 11164500.00 (-15.49%) 4710210.00 ( 51.27%) 9666720.00
length=33554436: 6849270.00 ( 29.30%) 11174800.00 (-15.35%) 4711050.00 ( 51.37%) 9687610.00
length=33554444: 6848890.00 ( 29.37%) 11143400.00 (-14.92%) 4711510.00 ( 51.41%) 9696840.00
length=33554437: 6849790.00 ( 29.27%) 11161800.00 (-15.25%) 4711400.00 ( 51.35%) 9684640.00
length=33554443: 6842850.00 ( 29.40%) 11146400.00 (-15.00%) 4706950.00 ( 51.44%) 9692370.00
length=33554438: 6860830.00 ( 29.06%) 11159300.00 (-15.38%) 4709130.00 ( 51.31%) 9671760.00
length=33554442: 6867660.00 ( 28.99%) 11158900.00 (-15.39%) 4710000.00 ( 51.30%) 9670920.00
length=33554439: 6857250.00 ( 29.34%) 11171600.00 (-15.11%) 4725570.00 ( 51.31%) 9705220.00
length=33554441: 6848530.00 ( 29.23%) 11153000.00 (-15.25%) 4714980.00 ( 51.28%) 9677440.00
---
sysdeps/aarch64/multiarch/Makefile | 3 +-
sysdeps/aarch64/multiarch/ifunc-impl-list.c | 4 +-
sysdeps/aarch64/multiarch/memcpy.c | 5 +-
sysdeps/aarch64/multiarch/memcpy_kunpeng.S | 579 ++++++++++++++++++++++++++++
4 files changed, 588 insertions(+), 3 deletions(-)
create mode 100644 sysdeps/aarch64/multiarch/memcpy_kunpeng.S
diff --git a/sysdeps/aarch64/multiarch/Makefile b/sysdeps/aarch64/multiarch/Makefile
index 4150b89a90..aa1f612e9e 100644
--- a/sysdeps/aarch64/multiarch/Makefile
+++ b/sysdeps/aarch64/multiarch/Makefile
@@ -1,6 +1,7 @@
ifeq ($(subdir),string)
sysdep_routines += memcpy_generic memcpy_thunderx memcpy_thunderx2 \
- memcpy_falkor memmove_falkor \
+ memcpy_falkor memcpy_kunpeng \
+ memmove_falkor memmove_kunpeng \
memset_generic memset_falkor memset_emag \
memchr_generic memchr_nosimd \
strlen_generic strlen_asimd
diff --git a/sysdeps/aarch64/multiarch/ifunc-impl-list.c b/sysdeps/aarch64/multiarch/ifunc-impl-list.c
index be13b916e5..dbbe19096a 100644
--- a/sysdeps/aarch64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/aarch64/multiarch/ifunc-impl-list.c
@@ -25,7 +25,7 @@
#include <stdio.h>
/* Maximum number of IFUNC implementations. */
-#define MAX_IFUNC 4
+#define MAX_IFUNC 5
size_t
__libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
@@ -42,11 +42,13 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_thunderx)
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_thunderx2)
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_falkor)
+ IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_kunpeng)
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_generic))
IFUNC_IMPL (i, name, memmove,
IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_thunderx)
IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_thunderx2)
IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_falkor)
+ IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_kunpeng)
IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_generic))
IFUNC_IMPL (i, name, memset,
/* Enable this on non-falkor processors too so that other cores
diff --git a/sysdeps/aarch64/multiarch/memcpy.c b/sysdeps/aarch64/multiarch/memcpy.c
index 13796f987f..b5929e2718 100644
--- a/sysdeps/aarch64/multiarch/memcpy.c
+++ b/sysdeps/aarch64/multiarch/memcpy.c
@@ -32,9 +32,12 @@ extern __typeof (__redirect_memcpy) __memcpy_generic attribute_hidden;
extern __typeof (__redirect_memcpy) __memcpy_thunderx attribute_hidden;
extern __typeof (__redirect_memcpy) __memcpy_thunderx2 attribute_hidden;
extern __typeof (__redirect_memcpy) __memcpy_falkor attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_kunpeng attribute_hidden;
libc_ifunc (__libc_memcpy,
- (IS_THUNDERX (midr)
+ IS_KUNPENG(midr)
+ ?__memcpy_kunpeng
+ : (IS_THUNDERX (midr)
? __memcpy_thunderx
: (IS_FALKOR (midr) || IS_PHECDA (midr) || IS_ARES (midr)
? __memcpy_falkor
diff --git a/sysdeps/aarch64/multiarch/memcpy_kunpeng.S b/sysdeps/aarch64/multiarch/memcpy_kunpeng.S
new file mode 100644
index 0000000000..1d326d3805
--- /dev/null
+++ b/sysdeps/aarch64/multiarch/memcpy_kunpeng.S
@@ -0,0 +1,579 @@
+/* Optimized memcpy and memmove for Huawei Kunpeng processor.
+ Copyright (C) 2018-2019 Free Software Foundation, Inc.
+
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+/* Assumptions:
+ *
+ * ARMv8-a, AArch64, unaligned accesses.
+ *
+ */
+
+#define dstin x0
+#define src x1
+#define count x2
+#define dst x3
+#define srcend x4
+#define dstend x5
+#define tmp2 x6
+#define tmp3 x7
+#define tmp3w w7
+#define A_l x6
+#define A_lw w6
+#define A_h x7
+#define A_hw w7
+#define B_l x8
+#define B_lw w8
+#define B_h x9
+#define C_l x10
+#define C_h x11
+#define D_l x12
+#define D_h x13
+#define E_l src
+#define E_h count
+#define F_l srcend
+#define F_h dst
+#define G_l count
+#define G_h dst
+#define tmp1 x14
+
+#define A_q q0
+#define B_q q1
+#define C_q q2
+#define D_q q3
+#define E_q q4
+#define F_q q5
+#define G_q q6
+#define H_q q7
+#define I_q q16
+#define J_q q17
+
+#define A_v v0
+#define B_v v1
+#define C_v v2
+#define D_v v3
+#define E_v v4
+#define F_v v5
+#define G_v v6
+#define H_v v7
+#define I_v v16
+#define J_v v17
+
+#ifndef MEMMOVE
+# define MEMMOVE memmove
+#endif
+#ifndef MEMCPY
+# define MEMCPY memcpy
+#endif
+
+#if IS_IN (libc)
+
+#undef MEMCPY
+#define MEMCPY __memcpy_kunpeng
+#undef MEMMOVE
+#define MEMMOVE __memmove_kunpeng
+
+
+/* Overlapping large forward memmoves use a loop that copies backwards.
+ Otherwise memcpy is used. Small moves branch to memcopy16 directly.
+ The longer memcpy cases fall through to the memcpy head.
+*/
+
+ENTRY_ALIGN (MEMMOVE, 6)
+
+ DELOUSE (0)
+ DELOUSE (1)
+ DELOUSE (2)
+
+ sub tmp1, dstin, src
+ cmp count, 512
+ ccmp tmp1, count, 2, hi
+ b.lo L(move_long)
+ cmp count, 96
+ ccmp tmp1, count, 2, hi
+ b.lo L(move_middle)
+
+END (MEMMOVE)
+libc_hidden_builtin_def (MEMMOVE)
+
+
+/* Copies are split into 4 main cases: small copies of up to 16 bytes,
+ medium copies of 17..96 bytes which are fully unrolled. Long copies
+ of 97..2048 align dst address without prefetching. Large copies
+ of more than 2048 bytes align the destination and use load-and-merge
+ approach in the case src and dst addresses are unaligned not evenly,
+ so that, actual loads and stores are always aligned.
+ Large copies use the loops processing 64 bytes per iteration for
+ unaligned case and 128 bytes per iteration for aligned ones.
+*/
+
+#define MEMCPY_PREFETCH_LDR 640
+
+ .p2align 4
+ENTRY (MEMCPY)
+
+ DELOUSE (0)
+ DELOUSE (1)
+ DELOUSE (2)
+
+ add srcend, src, count
+ cmp count, 16
+ b.ls L(memcopy16)
+ add dstend, dstin, count
+ cmp count, 96
+ b.hi L(memcopy_long)
+
+ /* Medium copies: 17..96 bytes. */
+ ldr A_q, [src], #16
+ and tmp1, src, 15
+ ldr E_q, [srcend, -16]
+ cmp count, 64
+ b.gt L(memcpy_copy96)
+ cmp count, 48
+ b.le L(bytes_17_to_48)
+ /* 49..64 bytes */
+ ldp B_q, C_q, [src]
+ str E_q, [dstend, -16]
+ stp A_q, B_q, [dstin]
+ str C_q, [dstin, 32]
+ ret
+
+L(bytes_17_to_48):
+ /* 17..48 bytes*/
+ cmp count, 32
+ b.gt L(bytes_32_to_48)
+ /* 17..32 bytes*/
+ str A_q, [dstin]
+ str E_q, [dstend, -16]
+ ret
+
+L(bytes_32_to_48):
+ /* 32..48 */
+ ldr B_q, [src]
+ str A_q, [dstin]
+ str E_q, [dstend, -16]
+ str B_q, [dstin, 16]
+ ret
+
+ .p2align 4
+ /* Small copies: 0..16 bytes. */
+L(memcopy16):
+ cmp count, 8
+ b.lo L(bytes_0_to_8)
+ ldr A_l, [src]
+ ldr A_h, [srcend, -8]
+ add dstend, dstin, count
+ str A_l, [dstin]
+ str A_h, [dstend, -8]
+ ret
+ .p2align 4
+
+L(bytes_0_to_8):
+ tbz count, 2, L(bytes_0_to_3)
+ ldr A_lw, [src]
+ ldr A_hw, [srcend, -4]
+ add dstend, dstin, count
+ str A_lw, [dstin]
+ str A_hw, [dstend, -4]
+ ret
+
+ /* Copy 0..3 bytes. Use a branchless sequence that copies the same
+ byte 3 times if count==1, or the 2nd byte twice if count==2. */
+L(bytes_0_to_3):
+ cbz count, 1f
+ lsr tmp1, count, 1
+ ldrb A_lw, [src]
+ ldrb A_hw, [srcend, -1]
+ add dstend, dstin, count
+ ldrb B_lw, [src, tmp1]
+ strb B_lw, [dstin, tmp1]
+ strb A_hw, [dstend, -1]
+ strb A_lw, [dstin]
+1:
+ ret
+
+ .p2align 4
+
+L(memcpy_copy96):
+ /* Copying 65..96 bytes. A_q (first 16 bytes) and
+ E_q(last 16 bytes) are already loaded. The size
+ is large enough to benefit from aligned loads */
+ bic src, src, 15
+ ldp B_q, C_q, [src]
+ /* Loaded 64 bytes, second 16-bytes chunk can be
+ overlapping with the first chunk by tmp1 bytes.
+ Stored 16 bytes. */
+ sub dst, dstin, tmp1
+ add count, count, tmp1
+ /* The range of count being [65..96] becomes [65..111]
+ after tmp [0..15] gets added to it,
+ count now is <bytes-left-to-load>+48 */
+ cmp count, 80
+ b.gt L(copy96_medium)
+ ldr D_q, [src, 32]
+ stp B_q, C_q, [dst, 16]
+ str D_q, [dst, 48]
+ str A_q, [dstin]
+ str E_q, [dstend, -16]
+ ret
+
+ .p2align 4
+L(copy96_medium):
+ ldp D_q, G_q, [src, 32]
+ cmp count, 96
+ b.gt L(copy96_large)
+ stp B_q, C_q, [dst, 16]
+ stp D_q, G_q, [dst, 48]
+ str A_q, [dstin]
+ str E_q, [dstend, -16]
+ ret
+
+L(copy96_large):
+ ldr F_q, [src, 64]
+ str B_q, [dst, 16]
+ stp C_q, D_q, [dst, 32]
+ stp G_q, F_q, [dst, 64]
+ str A_q, [dstin]
+ str E_q, [dstend, -16]
+ ret
+
+ .p2align 4
+L(memcopy_long):
+ cmp count, 2048
+ b.ls L(copy2048_large)
+ ldr A_q, [src], #16
+ and tmp1, src, 15
+ bic src, src, 15
+ ldp B_q, C_q, [src], #32
+ sub dst, dstin, tmp1
+ add count, count, tmp1
+ add dst, dst, 16
+ and tmp1, dst, 15
+ ldp D_q, E_q, [src], #32
+ str A_q, [dstin]
+
+ /* Already loaded 64+16 bytes. Check if at
+ least 64 more bytes left */
+ subs count, count, 64+64+16
+ b.lt L(loop128_exit0)
+ cmp count, MEMCPY_PREFETCH_LDR + 64 + 32
+ b.lt L(loop128)
+ cbnz tmp1, L(dst_unaligned)
+ sub count, count, MEMCPY_PREFETCH_LDR + 64 + 32
+
+ .p2align 4
+
+L(loop128_prefetch):
+ prfm pldl1strm, [src, MEMCPY_PREFETCH_LDR]
+ ldp F_q, G_q, [src], #32
+ stp B_q, C_q, [dst], #32
+ ldp H_q, I_q, [src], #32
+ prfm pldl1strm, [src, MEMCPY_PREFETCH_LDR]
+ ldp B_q, C_q, [src], #32
+ stp D_q, E_q, [dst], #32
+ ldp D_q, E_q, [src], #32
+ stp F_q, G_q, [dst], #32
+ stp H_q, I_q, [dst], #32
+ subs count, count, 128
+ b.ge L(loop128_prefetch)
+
+ add count, count, MEMCPY_PREFETCH_LDR + 64 + 32
+ .p2align 4
+L(loop128):
+ ldp F_q, G_q, [src], #32
+ ldp H_q, I_q, [src], #32
+ stp B_q, C_q, [dst], #32
+ stp D_q, E_q, [dst], #32
+ subs count, count, 64
+ b.lt L(loop128_exit1)
+ ldp B_q, C_q, [src], #32
+ ldp D_q, E_q, [src], #32
+ stp F_q, G_q, [dst], #32
+ stp H_q, I_q, [dst], #32
+ subs count, count, 64
+ b.ge L(loop128)
+L(loop128_exit0):
+ ldp F_q, G_q, [srcend, -64]
+ ldp H_q, I_q, [srcend, -32]
+ stp B_q, C_q, [dst], #32
+ stp D_q, E_q, [dst]
+ stp F_q, G_q, [dstend, -64]
+ stp H_q, I_q, [dstend, -32]
+ ret
+L(loop128_exit1):
+ ldp B_q, C_q, [srcend, -64]
+ ldp D_q, E_q, [srcend, -32]
+ stp F_q, G_q, [dst], #32
+ stp H_q, I_q, [dst]
+ stp B_q, C_q, [dstend, -64]
+ stp D_q, E_q, [dstend, -32]
+ ret
+
+ /* long copies: 96..2048 bytes */
+L(copy2048_large):
+ and tmp1, dstin, 15
+ bic dst, dstin, 15
+ ldp D_l, D_h, [src]
+ sub src, src, tmp1
+ add count, count, tmp1 /* Count is now 16 too large. */
+ ldp A_l, A_h, [src, 16]
+ stp D_l, D_h, [dstin]
+ ldp B_l, B_h, [src, 32]
+ ldp C_l, C_h, [src, 48]
+ ldp D_l, D_h, [src, 64]!
+ subs count, count, 128 + 16 /* Test and readjust count. */
+ b.ls L(last64)
+
+L(loop64):
+ stp A_l, A_h, [dst, 16]
+ ldp A_l, A_h, [src, 16]
+ stp B_l, B_h, [dst, 32]
+ ldp B_l, B_h, [src, 32]
+ stp C_l, C_h, [dst, 48]
+ ldp C_l, C_h, [src, 48]
+ stp D_l, D_h, [dst, 64]
+ ldp D_l, D_h, [src, 64]
+ add dst, dst, 64
+ add src, src, 64
+ subs count, count, 64
+ b.hi L(loop64)
+
+ /* Write the last full set of 64 bytes. The remainder is at most 64
+ bytes, so it is safe to always copy 64 bytes from the end even if
+ there is just 1 byte left. */
+L(last64):
+ ldp E_l, E_h, [srcend, -64]
+ stp A_l, A_h, [dst, 16]
+ ldp A_l, A_h, [srcend, -48]
+ stp B_l, B_h, [dst, 32]
+ ldp B_l, B_h, [srcend, -32]
+ stp C_l, C_h, [dst, 48]
+ ldp C_l, C_h, [srcend, -16]
+ stp D_l, D_h, [dst, 64]
+ stp E_l, E_h, [dstend, -64]
+ stp A_l, A_h, [dstend, -48]
+ stp B_l, B_h, [dstend, -32]
+ stp C_l, C_h, [dstend, -16]
+ ret
+
+
+L(dst_unaligned_tail):
+ ldp C_q, D_q, [srcend, -64]
+ ldp E_q, F_q, [srcend, -32]
+ stp A_q, B_q, [dst], #32
+ stp H_q, I_q, [dst], #16
+ str G_q, [dst, tmp1]
+ stp C_q, D_q, [dstend, -64]
+ stp E_q, F_q, [dstend, -32]
+ ret
+
+L(dst_unaligned):
+ /* For the unaligned store case the code loads two
+ aligned chunks and then merges them using ext
+ instruction. This can be up to 30% faster than
+ the the simple unaligned store access.
+
+ Current state: tmp1 = dst % 16; C_q, D_q, E_q
+ contains data yet to be stored. src and dst points
+ to next-to-be-processed data. A_q, B_q contains
+ data already stored before, count = bytes left to
+ be load decremented by 64.
+
+ The control is passed here if at least 64 bytes left
+ to be loaded. The code does two aligned loads and then
+ extracts (16-tmp1) bytes from the first register and
+ tmp1 bytes from the next register forming the value
+ for the aligned store.
+
+ As ext instruction can only have it's index encoded
+ as immediate. 15 code chunks process each possible
+ index value. Computed goto is used to reach the
+ required code. */
+
+ /* Store the 16 bytes to dst and align dst for further
+ operations, several bytes will be stored at this
+ address once more */
+
+ ldp F_q, G_q, [src], #32
+ stp B_q, C_q, [dst], #32
+ bic dst, dst, 15
+ sub count, count, 32
+ adrp tmp2, L(ext_table)
+ add tmp2, tmp2, :lo12:L(ext_table)
+ add tmp2, tmp2, tmp1, LSL #2
+ ldr tmp3w, [tmp2]
+ add tmp2, tmp2, tmp3w, SXTW
+ br tmp2
+
+ .p2align 4
+ /* to make the loop in each chunk 16-bytes aligned */
+ nop
+#define EXT_CHUNK(shft) \
+L(ext_size_ ## shft):;\
+ ext A_v.16b, C_v.16b, D_v.16b, 16-shft;\
+ ext B_v.16b, D_v.16b, E_v.16b, 16-shft;\
+ ext H_v.16b, E_v.16b, F_v.16b, 16-shft;\
+1:;\
+ stp A_q, B_q, [dst], #32;\
+ prfm pldl1strm, [src, MEMCPY_PREFETCH_LDR];\
+ ldp C_q, D_q, [src], #32;\
+ ext I_v.16b, F_v.16b, G_v.16b, 16-shft;\
+ stp H_q, I_q, [dst], #32;\
+ ext A_v.16b, G_v.16b, C_v.16b, 16-shft;\
+ ext B_v.16b, C_v.16b, D_v.16b, 16-shft;\
+ ldp F_q, G_q, [src], #32;\
+ ext H_v.16b, D_v.16b, F_v.16b, 16-shft;\
+ subs count, count, 64;\
+ b.ge 1b;\
+2:;\
+ ext I_v.16b, F_v.16b, G_v.16b, 16-shft;\
+ b L(dst_unaligned_tail);
+
+EXT_CHUNK(1)
+EXT_CHUNK(2)
+EXT_CHUNK(3)
+EXT_CHUNK(4)
+EXT_CHUNK(5)
+EXT_CHUNK(6)
+EXT_CHUNK(7)
+EXT_CHUNK(8)
+EXT_CHUNK(9)
+EXT_CHUNK(10)
+EXT_CHUNK(11)
+EXT_CHUNK(12)
+EXT_CHUNK(13)
+EXT_CHUNK(14)
+EXT_CHUNK(15)
+
+ /* long move: more than 512 bytes align the dstend */
+ .p2align 4
+L(move_long):
+1:
+ add srcend, src, count
+ add dstend, dstin, count
+
+ and tmp1, dstend, 15
+ ldr D_q, [srcend, -16]
+ sub srcend, srcend, tmp1
+ sub count, count, tmp1
+ ldp A_q, B_q, [srcend, -32]
+ str D_q, [dstend, -16]
+ ldp C_q, D_q, [srcend, -64]!
+ sub dstend, dstend, tmp1
+ subs count, count, 128
+ b.ls 2f
+
+ .p2align 4
+1:
+ subs count, count, 64
+ stp A_q, B_q, [dstend, -32]
+ ldp A_q, B_q, [srcend, -32]
+ stp C_q, D_q, [dstend, -64]!
+ ldp C_q, D_q, [srcend, -64]!
+ b.hi 1b
+
+ /* Write the last full set of 64 bytes. The remainder is at most 64
+ bytes, so it is safe to always copy 64 bytes from the start even if
+ there is just 1 byte left. */
+2:
+ ldp E_q, F_q, [src, 32]
+ ldp G_q, H_q, [src]
+ stp A_q, B_q, [dstend, -32]
+ stp C_q, D_q, [dstend, -64]
+ stp E_q, F_q, [dstin, 32]
+ stp G_q, H_q, [dstin]
+3: ret
+
+ /* midlle move: 96..512 bytes */
+ .p2align 4
+L(move_middle):
+ cbz tmp1, 3f
+ add srcend, src, count
+ prfm PLDL1STRM, [srcend, -64]
+ add dstend, dstin, count
+ and tmp1, dstend, 15
+ ldr D_q, [srcend, -16]
+ sub srcend, srcend, tmp1
+ sub count, count, tmp1
+ ldr A_q, [srcend, -16]
+ str D_q, [dstend, -16]
+ ldr B_q, [srcend, -32]
+ ldr C_q, [srcend, -48]
+ ldr D_q, [srcend, -64]!
+ sub dstend, dstend, tmp1
+ subs count, count, 128
+ b.ls 2f
+
+1:
+ str A_q, [dstend, -16]
+ ldr A_q, [srcend, -16]
+ str B_q, [dstend, -32]
+ ldr B_q, [srcend, -32]
+ str C_q, [dstend, -48]
+ ldr C_q, [srcend, -48]
+ str D_q, [dstend, -64]!
+ ldr D_q, [srcend, -64]!
+ subs count, count, 64
+ b.hi 1b
+
+ /* Write the last full set of 64 bytes. The remainder is at most 64
+ bytes, so it is safe to always copy 64 bytes from the start even if
+ there is just 1 byte left. */
+2:
+ ldr G_q, [src, 48]
+ str A_q, [dstend, -16]
+ ldr A_q, [src, 32]
+ str B_q, [dstend, -32]
+ ldr B_q, [src, 16]
+ str C_q, [dstend, -48]
+ ldr C_q, [src]
+ str D_q, [dstend, -64]
+ str G_q, [dstin, 48]
+ str A_q, [dstin, 32]
+ str B_q, [dstin, 16]
+ str C_q, [dstin]
+3: ret
+
+
+END (MEMCPY)
+ .section .rodata
+ .p2align 4
+
+L(ext_table):
+ /* The first entry is for the alignment of 0 and is never
+ actually used (could be any value). */
+ .word 0
+ .word L(ext_size_1) -.
+ .word L(ext_size_2) -.
+ .word L(ext_size_3) -.
+ .word L(ext_size_4) -.
+ .word L(ext_size_5) -.
+ .word L(ext_size_6) -.
+ .word L(ext_size_7) -.
+ .word L(ext_size_8) -.
+ .word L(ext_size_9) -.
+ .word L(ext_size_10) -.
+ .word L(ext_size_11) -.
+ .word L(ext_size_12) -.
+ .word L(ext_size_13) -.
+ .word L(ext_size_14) -.
+ .word L(ext_size_15) -.
+
+libc_hidden_builtin_def (MEMCPY)
+#endif
--
2.14.1.windows.1
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-14 3:45 [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor Xuelei Zhang
@ 2019-10-15 12:05 ` Szabolcs Nagy
2019-10-16 2:14 ` Yikun Jiang
1 sibling, 0 replies; 13+ messages in thread
From: Szabolcs Nagy @ 2019-10-15 12:05 UTC (permalink / raw)
To: Xuelei Zhang, libc-alpha; +Cc: nd
On 14/10/2019 04:44, Xuelei Zhang wrote:
> This is an optimized implementation of the memcpy and memmove on the
> Huawei Kunpeng processor.
>
> Based on the prefetch mechanism on Kunpeng arch, branch to handle 96
> to 2K bytes in memcpy is written without prfm instruction. Hence,
> memcpy has an optimization effect above 128 bytes, 18% improvement
> for copies above 2K bytes, and 38% for larger bytes, such as 32M
> bytes around.
>
> And for memmove, there are two main changes: i) Q register is used
> instead of X register. ii) dst address is aligned instead of src
> address aligned to improve store operation. Hence, memmove
> implementation also has improvement above 128 bytes, that about 30%
> for 2k to 8M bytes, and about 50% for 32M or more.
i'd like to work on a generic memcpy that's acceptable
instead of minor variations of memcpy per uarch, i'll
have to take a look why this one is different from all
the others.
it would be nice to see the memcpy-random benchmarks too.
stopping prefetch at 2k is surprising.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-14 3:45 [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor Xuelei Zhang
2019-10-15 12:05 ` Szabolcs Nagy
@ 2019-10-16 2:14 ` Yikun Jiang
1 sibling, 0 replies; 13+ messages in thread
From: Yikun Jiang @ 2019-10-16 2:14 UTC (permalink / raw)
To: Xuelei Zhang; +Cc: libc-alpha
> diff --git a/sysdeps/aarch64/multiarch/Makefile b/sysdeps/aarch64/multiarch/Makefile
> index 4150b89a90..aa1f612e9e 100644
> --- a/sysdeps/aarch64/multiarch/Makefile
> +++ b/sysdeps/aarch64/multiarch/Makefile
> @@ -1,6 +1,7 @@
> ifeq ($(subdir),string)
> sysdep_routines += memcpy_generic memcpy_thunderx memcpy_thunderx2 \
> - memcpy_falkor memmove_falkor \
> + memcpy_falkor memcpy_kunpeng \
> + memmove_falkor memmove_kunpeng \
> memset_generic memset_falkor memset_emag \
> memchr_generic memchr_nosimd \
> strlen_generic strlen_asimd
It seems we should remove the "memmove_kunpeng", and only add
the "memcpy_kunpeng" in here.
> +ENTRY_ALIGN (MEMMOVE, 6)
> +
> + DELOUSE (0)
> + DELOUSE (1)
> + DELOUSE (2)
> +
> + sub tmp1, dstin, src
> + cmp count, 512
> + ccmp tmp1, count, 2, hi
> + b.lo L(move_long)
> + cmp count, 96
> + ccmp tmp1, count, 2, hi
> + b.lo L(move_middle)
> +
> +END (MEMMOVE)
nit: remember to remove the trailing space.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
@ 2019-11-19 13:34 Zhangxuelei (Derek)
0 siblings, 0 replies; 13+ messages in thread
From: Zhangxuelei (Derek) @ 2019-11-19 13:34 UTC (permalink / raw)
To: Wilco Dijkstra, Szabolcs Nagy, libc-alpha, yikunkero, jiangyikun; +Cc: nd
Hi Wilco,
> So is there a way to force write streaming, for example by aligning
> the source rather than the destination or use particular instructions?
Difference walk direactions of dst make the align offset different, further resulted in the different performance between memcpy-walk and memmove-walk.
In new patch, we use dst aligned rather than src aligned to solve this problem, and now both memcpy-walk and memmove-walk performe well as before with dst_unaligned code removed.
> In order to select the right memmove implementation, multiarch/
> memmove.c needs similar changes as multiarch/memcpy.c.
>
> Also since the memmove entry sequence does both check for medium
> and large cases, the full overlap check should be done in both.
As reminded, full overlap check is done now in large cases, and memmove.c is also added to new patch.
In addition, is there any reviews for the latest memset_kunpeng patch as below:
https://sourceware.org/ml/libc-alpha/2019-11/msg00044.html
Cheers,
Xuelei
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
[not found] ` <8DC571DDDE171B4094D3D33E9685917BD9F88D@DGGEMI529-MBX.china.huawei.com>
@ 2019-10-31 14:04 ` Wilco Dijkstra
0 siblings, 0 replies; 13+ messages in thread
From: Wilco Dijkstra @ 2019-10-31 14:04 UTC (permalink / raw)
To: Zhangxuelei (Derek), Szabolcs Nagy, libc-alpha, yikunkero, jiangyikun; +Cc: nd
Hi Derek,
>> There shouldn't be any performance difference between the two cases.
> With the help of our professors, the reason of different performance in unaligned
> case has been found. It is because the streaming write feature of Kunpeng processor
> is not triggered in unaligned case, that means the data needs to be read first and
> then written, time-consumingly.
So is there a way to force write streaming, for example by aligning the source rather
than the destination or use particular instructions?
> And the implementation of v1 patch happens to avoid this problem, which seems a
> better choice for Kunpeng processor at now.
I don't believe it always helps - there is still a large factor between good and bad cases,
like these results from the v1 memcpy:
length=1048578: 305883.00 ( 0.00%) 182002.00 ( 40.00%) 120063.00 ( 60.00%) 292063.00 ( 4.00%) 306638.00
Here the Falkor variant is 2.4 times faster...
Wilco
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-17 12:50 Zhangxuelei (Derek)
@ 2019-10-29 15:34 ` Wilco Dijkstra
[not found] ` <8DC571DDDE171B4094D3D33E9685917BD9F88D@DGGEMI529-MBX.china.huawei.com>
0 siblings, 1 reply; 13+ messages in thread
From: Wilco Dijkstra @ 2019-10-29 15:34 UTC (permalink / raw)
To: Zhangxuelei (Derek), Szabolcs Nagy, libc-alpha, yikunkero, jiangyikun; +Cc: nd
Hi Derek,
>> Well these results show a very significant 4% win for Falkor memcpy! It seems strange to only optimize
>> for large sizes when the vast majority of copies in real code are very small (note the distribution of the
>> sizes and alignment for the random benchmark come from SPEC).
>
> Sure, we agree the falkor memcpy has 4% win on small size. However, at the beginning we start to Kunpeng
> optimized the memcpy, one of the most important case is database case, which really need more improvement
> on large size.
Being fast on small copies is not mutually exclusive with being fast on large copies - there are different
code paths for these cases. I'd find it hard to believe databases do huge memcopies, that would be stupid!
> And what confusing us now is that, we removed dst_unaligned code in memcpy according to the previous comments,
> which did not affect performance after testing in memcpy cases. But in the case when uses memmove function and
> enters the memcpy part, unaligned cases is significantly slower than aligned case according to the results of the first
> half part of memmove-walk as shown in the bottom. So do you think we should still remove dst_unaligned code?
Well it seems to me the issue is related to prefetching/caching. memcpy-walk walks both src and dst backwards,
memmove-walk is pretty much identical for the non-overlap case but it does a forward walk on dst and a backward
one on src. There shouldn't be any performance difference between the two cases. On most microarchitectures I see
no difference between these walks, memmove-walk and memcpy-walk basically have identical performance both for
aligned and unaligned cases (alignment doesn't matter for large copies at all).
> We analyse the reason is of more judgement in the begin of memmove and may weak processor ability to handle
> this case, and so dst_unaligned make difference.
The extra instructions at the start couldn't possibly make a difference apart from slowing down the small memmoves.
Wilco
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-18 15:29 ` Wilco Dijkstra
@ 2019-10-18 15:47 ` Siddhesh Poyarekar
0 siblings, 0 replies; 13+ messages in thread
From: Siddhesh Poyarekar @ 2019-10-18 15:47 UTC (permalink / raw)
To: Wilco Dijkstra, Zhangxuelei (Derek),
Szabolcs Nagy, libc-alpha, yikunkero
Cc: nd
On 18/10/19 11:29 am, Wilco Dijkstra wrote:
> Hi Siddhesh,
>
>> The benchtests/scripts/compare_strings.py script dumps a graph png IIRC.
>
> True, but have you seen the output? Try comparing the graphs it generates
> with these: https://www.sourceware.org/ml/libc-alpha/2019-10/msg00563.html
That does look quite neat. As a follow-up to that then, I suggest
either dropping the -g flag from compare_strings.py or putting out a
deprecation warning that plot_strings.py is a better option for
generating graphs and that the -g flag might go away in future.
Siddhesh
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-16 16:25 ` Siddhesh Poyarekar
@ 2019-10-18 15:29 ` Wilco Dijkstra
2019-10-18 15:47 ` Siddhesh Poyarekar
0 siblings, 1 reply; 13+ messages in thread
From: Wilco Dijkstra @ 2019-10-18 15:29 UTC (permalink / raw)
To: Siddhesh Poyarekar, Zhangxuelei (Derek),
Szabolcs Nagy, libc-alpha, yikunkero
Cc: nd
Hi Siddhesh,
> The benchtests/scripts/compare_strings.py script dumps a graph png IIRC.
True, but have you seen the output? Try comparing the graphs it generates
with these: https://www.sourceware.org/ml/libc-alpha/2019-10/msg00563.html
Cheers,
Wilco
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
@ 2019-10-17 12:50 Zhangxuelei (Derek)
2019-10-29 15:34 ` Wilco Dijkstra
0 siblings, 1 reply; 13+ messages in thread
From: Zhangxuelei (Derek) @ 2019-10-17 12:50 UTC (permalink / raw)
To: Wilco Dijkstra, Szabolcs Nagy, libc-alpha, yikunkero, jiangyikun; +Cc: nd
> Well it looks the dst_unaligned code (which deals with a specific issue on ThunderX2) is completely
> unnecessary on Kunpeng since the unaligned cases in eg. Falkor and generic aren't slower than the
> aligned cases. So I'd suggest to remove this code - it's adds a lot of code, thus making memcpy
> unnecessarily large.
Yes, thanks for the reminder, we will remove the dst_unaligned code in next patch, note that we have tested this version and the results are same as before removed .
> Well these results show a very significant 4% win for Falkor memcpy! It seems strange to only optimize
> for large sizes when the vast majority of copies in real code are very small (note the distribution of the
> sizes and alignment for the random benchmark come from SPEC).
Sure, we agree the falkor memcpy has 4% win on small size. However, at the beginning we start to Kunpeng optimized the memcpy, one of the most important case is database case, which really need more improvement on large size.
So we use memcpy-bench-large to help us to choose the baseline, the result seems that ThunderX2 better in Kunpeng env (see below large size benchset). And it really confusing us why falkor acts so differently between bench-walk and bench-large.
And we also know the small/medium size is very important case for us, so we do more optimizations on medium size in branch L(copy2048_large), and the gap existing in small size is acceptable on our opinion, at least better than generic.
Function: memcpy
Variant: large
__memcpy_thunderx __memcpy_thunderx2 __memcpy_falkor __memcpy_kunpeng __memcpy_generic
========================================================================================================================
length=65543, align1=0, align2=0: 4238.12 (-101.40%) 2295.00 ( -9.06%) 2156.25 ( -2.46%) 2301.25 ( -9.36%) 2104.38
length=65551, align1=0, align2=3: 3101.88 ( -2.97%) 2283.75 ( 24.19%) 3562.50 (-18.26%) 3332.50 (-10.62%) 3012.50
length=65567, align1=3, align2=0: 2899.38 ( 18.38%) 2285.62 ( 35.66%) 3800.00 ( -6.97%) 3333.75 ( 6.16%) 3552.50
length=65599, align1=3, align2=5: 2866.88 ( 2.09%) 2258.75 ( 22.86%) 3333.75 (-13.85%) 3316.25 (-13.26%) 2928.12
length=131079, align1=0, align2=0: 6905.00 (-43.54%) 5153.75 ( -7.13%) 5848.75 (-21.58%) 5185.62 ( -7.80%) 4810.62
length=131087, align1=0, align2=3: 7232.50 ( -1.16%) 5138.75 ( 28.12%) 7966.88 (-11.43%) 7117.50 ( 0.45%) 7149.38
length=131103, align1=3, align2=0: 6889.38 ( -0.69%) 5145.00 ( 24.81%) 7562.50 (-10.52%) 7094.38 ( -3.68%) 6842.50
length=131135, align1=3, align2=5: 7096.88 ( 0.38%) 5116.88 ( 28.17%) 7501.88 ( -5.31%) 6989.38 ( 1.89%) 7123.75
length=262151, align1=0, align2=0: 15736.90 (-37.76%) 11581.90 ( -1.39%) 13088.10 (-14.58%) 11993.80 ( -5.00%) 11423.10
length=262159, align1=0, align2=3: 16581.20 ( -1.58%) 11655.00 ( 28.60%) 24495.60 (-50.07%) 16983.10 ( -4.05%) 16322.50
length=262175, align1=3, align2=0: 14940.60 ( -1.26%) 11769.40 ( 20.23%) 22577.50 (-53.02%) 16925.00 (-15.00%) 14755.00
length=262207, align1=3, align2=5: 16167.50 ( 0.64%) 11700.60 ( 28.09%) 23373.10 (-43.64%) 16959.40 ( -4.23%) 16271.90
length=524295, align1=0, align2=0: 59790.60 (-11.75%) 37941.90 ( 29.09%) 55081.90 ( -2.95%) 38127.50 ( 28.74%) 53505.00
length=524303, align1=0, align2=3: 58881.90 ( 11.26%) 38826.90 ( 41.48%) 82475.60 (-24.30%) 45186.20 ( 31.90%) 66351.20
length=524319, align1=3, align2=0: 52656.90 ( 13.39%) 36772.50 ( 39.52%) 79804.40 (-31.26%) 44925.60 ( 26.11%) 60800.60
length=524351, align1=3, align2=5: 60072.50 ( 9.98%) 38996.90 ( 41.56%) 80605.60 (-20.79%) 44833.80 ( 32.82%) 66731.90
length=1048583, align1=0, align2=0: 140744.00 ( 0.00%) 92834.40 ( 34.15%) 156891.00 (-12.00%) 92700.60 ( 34.25%) 140981.00
length=1048591, align1=0, align2=3: 138818.00 ( 14.00%) 93137.50 ( 42.85%) 181426.00 (-12.00%) 99160.00 ( 39.00%) 162969.00
length=1048607, align1=3, align2=0: 123231.00 ( 19.00%) 85615.00 ( 44.00%) 180895.00 (-19.00%) 99104.40 ( 35.31%) 153204.00
length=1048639, align1=3, align2=5: 139292.00 ( 14.00%) 93651.20 ( 42.53%) 181196.00 (-12.00%) 99200.00 ( 39.00%) 162948.00
length=2097159, align1=0, align2=0: 305711.00 ( -3.00%) 194030.00 ( 34.00%) 341827.00 (-15.00%) 193635.00 ( 35.00%) 298108.00
length=2097167, align1=0, align2=3: 284683.00 ( 14.00%) 194907.00 ( 41.00%) 369128.00 (-11.00%) 199646.00 ( 40.00%) 332992.00
length=2097183, align1=3, align2=0: 252849.00 ( 20.00%) 175724.00 ( 44.00%) 368446.00 (-17.00%) 199992.00 ( 36.00%) 316792.00
length=2097215, align1=3, align2=5: 284985.00 ( 14.00%) 194299.00 ( 41.00%) 370709.00 (-11.00%) 199978.00 ( 40.00%) 334178.00
length=4194311, align1=0, align2=0: 569834.00 ( 3.00%) 374032.00 ( 36.00%) 678767.00 (-16.00%) 372642.00 ( 36.00%) 588650.00
length=4194319, align1=0, align2=3: 576691.00 ( 20.00%) 378198.00 ( 47.00%) 872069.00 (-21.00%) 479148.00 ( 33.00%) 723708.00
length=4194335, align1=3, align2=0: 512557.00 ( 25.00%) 342394.00 ( 50.00%) 863184.00 (-26.00%) 480661.00 ( 30.00%) 690116.00
length=4194367, align1=3, align2=5: 576482.00 ( 19.00%) 380856.00 ( 46.00%) 860296.00 (-21.00%) 479336.00 ( 32.00%) 712510.00
length=8388615, align1=0, align2=0: 1324960.00 ( -4.74%) 805671.00 ( 36.31%) 1444260.00 (-14.17%) 784482.00 ( 37.99%) 1265040.00
length=8388623, align1=0, align2=3: 1217610.00 ( 18.78%) 771131.00 ( 48.56%) 1894040.00 (-26.35%) 1143710.00 ( 23.71%) 1499080.00
length=8388639, align1=3, align2=0: 1147310.00 ( 28.52%) 856928.00 ( 46.61%) 1974130.00 (-22.99%) 1281390.00 ( 20.17%) 1605130.00
length=8388671, align1=3, align2=5: 1124090.00 ( 27.46%) 846521.00 ( 45.37%) 1928810.00 (-24.47%) 1248630.00 ( 19.43%) 1549660.00
length=16777223, align1=0, align2=0: 4243330.00 ( -8.50%) 2821000.00 ( 27.87%) 3382290.00 ( 13.51%) 3635000.00 ( 7.05%) 3910780.00
length=16777231, align1=0, align2=3: 3503780.00 ( 19.16%) 2214300.00 ( 48.91%) 4848350.00 (-11.86%) 3703890.00 ( 14.55%) 4334450.00
length=16777247, align1=3, align2=0: 2411200.00 ( 44.45%) 1467160.00 ( 66.20%) 5724860.00 (-31.89%) 4081640.00 ( 5.97%) 4340700.00
length=16777279, align1=3, align2=5: 4414480.00 ( 12.29%) 2689160.00 ( 46.57%) 6632000.00 (-31.77%) 3824970.00 ( 24.00%) 5032940.00
length=33554439, align1=0, align2=0: 10333400.00 ( 5.39%) 7028430.00 ( 35.65%) 9881280.00 ( 9.53%) 7024200.00 ( 35.69%) 10921700.00
length=33554447, align1=0, align2=3: 11757800.00 ( 6.63%) 8676040.00 ( 31.10%) 15423100.00 (-22.47%) 8309720.00 ( 34.01%) 12592900.00
length=33554463, align1=3, align2=0: 6959020.00 ( 34.78%) 5700270.00 ( 46.58%) 15931700.00 (-49.32%) 8625250.00 ( 19.16%) 10669700.00
length=33554495, align1=3, align2=5: 11190000.00 ( 12.07%) 8721580.00 ( 31.47%) 15759900.00 (-23.84%) 8557940.00 ( 32.75%) 12726300.00
> This has the effect of slowing down all small memmoves and no-overlap memmoves (ie. 99% of calls).
> Is there a reason to special case 96-512? I don't see an obvious difference between the cases, there is
> one extra prefetch but outside the loop. Even if it helps somehow, why not do the test for >512 in the
> move_long code? That removes 4 instructions (3 and a NOP) from the memmove fallthrough path.
+ .p2align 4
+ 1:
+ subs count, count, 64
+ stp A_q, B_q, [dstend, -32]
+ ldp A_q, B_q, [srcend, -32]
+ stp C_q, D_q, [dstend, -64]!
+ ldp C_q, D_q, [srcend, -64]!
+ b.hi 1b
The branch prediction algorithm of Kunpeng chip requires that the jump judgment and the jump address cannot be in the same 32 instruction cacheline. The original loop body (see ablove) will cause the prediction failure leading to performance degradation, especially below 512 bytes. So the ldr/str instruction replaces the ldp/stp instruction to lengthen the loop body to satisfy the branch prediction algorithm, in case 96-512.
> Btw do you have any plans to post other string functions that you can discuss here? If so, would these
> add more ifuncs or improve the generic versions?
Yes, memcmp, strlen, strnlen, strcpy, memrchr will be included, we will summited the patch and test results as soon as possible.
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-16 16:20 ` Wilco Dijkstra
@ 2019-10-16 16:25 ` Siddhesh Poyarekar
2019-10-18 15:29 ` Wilco Dijkstra
0 siblings, 1 reply; 13+ messages in thread
From: Siddhesh Poyarekar @ 2019-10-16 16:25 UTC (permalink / raw)
To: Wilco Dijkstra, Zhangxuelei (Derek),
Szabolcs Nagy, libc-alpha, yikunkero
Cc: nd
On 16/10/19 12:20 pm, Wilco Dijkstra wrote:
>
> It doesn't help the existing benchmarks don't report an average across all the inputs for each ifunc...
> A colleague is working on a script to visualise benchmark results in a graph which should make these
> comparisons much easier.
>
The benchtests/scripts/compare_strings.py script dumps a graph png IIRC.
Siddhesh
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
2019-10-16 13:53 Zhangxuelei (Derek)
@ 2019-10-16 16:20 ` Wilco Dijkstra
2019-10-16 16:25 ` Siddhesh Poyarekar
0 siblings, 1 reply; 13+ messages in thread
From: Wilco Dijkstra @ 2019-10-16 16:20 UTC (permalink / raw)
To: Zhangxuelei (Derek), Szabolcs Nagy, libc-alpha, yikunkero; +Cc: nd
Hi Derek,
> We do vary based on ThunderX2 for its good performance for large copies, which is needed by us firstly.
But it's not clear whether that is a win over the Falkor version according to memcpy-walk output. Consider eg.
__memcpy_thunderx __memcpy_thunderx2 __memcpy_falkor __memcpy_kunpeng __memcpy_generic
length=32768: 7629.73 (-46.76%) 4473.26 ( 13.95%) 3947.03 ( 24.08%) 5201.79 ( -0.06%) 5198.66
length=32784: 7666.55 (-44.44%) 4705.93 ( 11.34%) 3782.04 ( 28.75%) 4759.99 ( 10.32%) 5307.92
length=32769: 7589.72 (-42.41%) 4776.54 ( 10.37%) 3889.61 ( 27.02%) 4789.02 ( 10.14%) 5329.43
length=32783: 7502.45 (-44.25%) 4688.97 ( 9.85%) 3858.48 ( 25.81%) 4714.48 ( 9.36%) 5201.10
length=32770: 7438.46 (-43.39%) 4647.77 ( 10.41%) 3848.94 ( 25.81%) 4680.76 ( 9.77%) 5187.71
length=32782: 7225.10 (-38.35%) 4609.10 ( 11.74%) 3855.37 ( 26.17%) 4643.70 ( 11.08%) 5222.19
length=32771: 7326.40 (-42.87%) 4587.85 ( 10.53%) 3828.78 ( 25.33%) 4580.34 ( 10.68%) 5127.85
length=32781: 7261.12 (-41.38%) 4548.17 ( 11.44%) 3851.30 ( 25.01%) 4584.78 ( 10.73%) 5135.97
length=32772: 7178.11 (-41.83%) 4510.12 ( 10.89%) 3802.44 ( 24.87%) 4521.19 ( 10.67%) 5061.20
length=32780: 7186.99 (-42.34%) 4481.01 ( 11.25%) 3835.00 ( 24.00%) 4532.33 ( 10.23%) 5049.00
length=32773: 7089.60 (-38.93%) 4482.70 ( 12.15%) 3830.79 ( 24.93%) 4487.18 ( 12.07%) 5102.88
length=32779: 7076.27 (-25.65%) 4498.21 ( 20.13%) 3881.99 ( 31.07%) 5371.42 ( 4.62%) 5631.73
length=32774: 8362.27 (-48.81%) 5190.76 ( 7.63%) 3845.61 ( 31.57%) 5176.08 ( 7.89%) 5619.46
length=32778: 8186.09 (-48.41%) 5109.29 ( 7.37%) 3861.96 ( 29.98%) 5211.88 ( 5.51%) 5515.68
length=32775: 8186.51 (-49.43%) 5096.24 ( 6.98%) 3846.12 ( 29.80%) 5128.88 ( 6.38%) 5478.44
length=32777: 8038.38 (-49.89%) 5001.09 ( 6.75%) 3837.79 ( 28.44%) 5095.39 ( 4.99%) 5362.88
Here the Falkor variant is 20-30% faster...
It doesn't help the existing benchmarks don't report an average across all the inputs for each ifunc...
A colleague is working on a script to visualise benchmark results in a graph which should make these
comparisons much easier.
> And we do find the detect of ThunderX2 version for 96 to 2M bytes copy, at least when running on the
> Kunpeng arch, even the falkor version is not much better.
Well it looks the dst_unaligned code (which deals with a specific issue on ThunderX2) is completely
unnecessary on Kunpeng since the unaligned cases in eg. Falkor and generic aren't slower than the
aligned cases. So I'd suggest to remove this code - it's adds a lot of code, thus making memcpy
unnecessarily large.
> Therefore, branch was written and we used generic copy, that 64 bytes loop, dst aligned, without prefetch,
> and it works. We also have simply tried Q register replacing X register in this branch, but it didn't make more sense.
Yes, using Q register copy is best on modern micro-architectures (hence the idea to do this even in the
the generic version).
> And here is the result of memcpy-random benchmarks :
__memcpy_thunderx __memcpy_thunderx2 __memcpy_falkor __memcpy_kunpeng __memcpy_generic
max-size=4096: 32558.90 ( -2.08%) 31987.80 ( -0.29%) 30474.30 ( 4.46%) 31666.30 ( 0.72%) 31896.60
max-size=8192: 31796.80 ( -1.18%) 31423.90 ( 0.01%) 29974.40 ( 4.62%) 30917.90 ( 1.62%) 31427.40
max-size=16384: 33122.80 ( -1.05%) 32058.30 ( 2.20%) 30470.40 ( 7.05%) 31727.90 ( 3.21%) 32779.90
max-size=32768: 32530.10 ( -1.22%) 31912.80 ( 0.71%) 29960.80 ( 6.78%) 31567.60 ( 1.78%) 32139.40
max-size=65536: 33373.60 ( -0.40%) 32476.30 ( 2.30%) 30957.70 ( 6.87%) 32137.00 ( 3.32%) 33240.10
Well these results show a very significant 4% win for Falkor memcpy! It seems strange to only optimize
for large sizes when the vast majority of copies in real code are very small (note the distribution of the
sizes and alignment for the random benchmark come from SPEC).
+ENTRY_ALIGN (MEMMOVE, 6)
...
+ sub tmp1, dstin, src
+ cmp count, 512
+ ccmp tmp1, count, 2, hi
+ b.lo L(move_long)
+ cmp count, 96
+ ccmp tmp1, count, 2, hi
+ b.lo L(move_middle)
This has the effect of slowing down all small memmoves and no-overlap memmoves (ie. 99% of calls).
Is there a reason to special case 96-512? I don't see an obvious difference between the cases, there is
one extra prefetch but outside the loop. Even if it helps somehow, why not do the test for >512 in the
move_long code? That removes 4 instructions (3 and a NOP) from the memmove fallthrough path.
Btw do you have any plans to post other string functions that you can discuss here? If so, would these
add more ifuncs or improve the generic versions?
Cheers,
Wilco
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
@ 2019-10-16 13:53 Zhangxuelei (Derek)
2019-10-16 16:20 ` Wilco Dijkstra
0 siblings, 1 reply; 13+ messages in thread
From: Zhangxuelei (Derek) @ 2019-10-16 13:53 UTC (permalink / raw)
To: Szabolcs Nagy, libc-alpha, yikunkero, Wilco.Dijkstra; +Cc: nd
> i'd like to work on a generic memcpy that's acceptable instead of
> minor variations of memcpy per uarch, i'll have to take a look why
> this one is different from all the others.
>
> it would be nice to see the memcpy-random benchmarks too.
>
> stopping prefetch at 2k is surprising.
Thanks for all the reviews @Sz @Wilco @kero.
We do vary based on ThunderX2 for its good performance for large copies, which is needed by us firstly. And we do find the detect of ThunderX2 version for 96 to 2M bytes copy, at least when running on the Kunpeng arch, even the falkor version is not much better. Therefore, branch was written and we used generic copy, that 64 bytes loop, dst aligned, without prefetch, and it works. We also have simply tried Q register replacing X register in this branch, but it didn't make more sense.
And here is the result of memcpy-random benchmarks :
Function: memcpy
Variant: random
__memcpy_thunderx __memcpy_thunderx2 __memcpy_falkor __memcpy_kunpeng __memcpy_generic
========================================================================================================================
max-size=4096: 32558.90 ( -2.08%) 31987.80 ( -0.29%) 30474.30 ( 4.46%) 31666.30 ( 0.72%) 31896.60
max-size=8192: 31796.80 ( -1.18%) 31423.90 ( 0.01%) 29974.40 ( 4.62%) 30917.90 ( 1.62%) 31427.40
max-size=16384: 33122.80 ( -1.05%) 32058.30 ( 2.20%) 30470.40 ( 7.05%) 31727.90 ( 3.21%) 32779.90
max-size=32768: 32530.10 ( -1.22%) 31912.80 ( 0.71%) 29960.80 ( 6.78%) 31567.60 ( 1.78%) 32139.40
max-size=65536: 33373.60 ( -0.40%) 32476.30 ( 2.30%) 30957.70 ( 6.87%) 32137.00 ( 3.32%) 33240.10
> It seems we should remove the "memmove_kunpeng", and only
> add the "memcpy_kunpeng" in here.
Indeed, redundant wirte of 'memmove_kunpeng' in multiarch/Makefile will be removed in patch v2.
regards,
Xuelei
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor
@ 2019-10-15 13:26 Wilco Dijkstra
0 siblings, 0 replies; 13+ messages in thread
From: Wilco Dijkstra @ 2019-10-15 13:26 UTC (permalink / raw)
To: 'GNU C Library', zhangxuelei4, Szabolcs Nagy; +Cc: nd
Hi,
> Sz
> i'd like to work on a generic memcpy that's acceptable
> instead of minor variations of memcpy per uarch, i'll
> have to take a look why this one is different from all
> the others.
Yes it seems the key thing we need is a generic Q-register memcpy.
> it would be nice to see the memcpy-random benchmarks too.
>
> stopping prefetch at 2k is surprising.
Also it's odd it's based on the ThunderX2 variant rather than the Falkor
one, particularly since for large copies misaligned accesses are cheap.
Briefly looking at the data for memcpy, it seems the Falkor results are typically
faster for large copy sizes, eg. from 512KB to 4MBytes.
On the other hand, the memmove results for Kunpeng look genuinely faster than
the existing implementations - and that is without prefetching or special code to
handle unaligned cases. So that suggests to me these don't help much, and all
we need is code that does Q-register copies/moves (clearly using LDP/STP as that
is where the memmove seems to win).
Wilco
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2019-11-19 13:34 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-10-14 3:45 [PATCH 2/2] aarch64: Optimized memcpy and memmove for Kunpeng processor Xuelei Zhang
2019-10-15 12:05 ` Szabolcs Nagy
2019-10-16 2:14 ` Yikun Jiang
2019-10-15 13:26 Wilco Dijkstra
2019-10-16 13:53 Zhangxuelei (Derek)
2019-10-16 16:20 ` Wilco Dijkstra
2019-10-16 16:25 ` Siddhesh Poyarekar
2019-10-18 15:29 ` Wilco Dijkstra
2019-10-18 15:47 ` Siddhesh Poyarekar
2019-10-17 12:50 Zhangxuelei (Derek)
2019-10-29 15:34 ` Wilco Dijkstra
[not found] ` <8DC571DDDE171B4094D3D33E9685917BD9F88D@DGGEMI529-MBX.china.huawei.com>
2019-10-31 14:04 ` Wilco Dijkstra
2019-11-19 13:34 Zhangxuelei (Derek)
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).