beagleboard, Cortex-A8 500MHz, 1280x1024-16@57 framebuffer --- Running benchmarks (average case/perfect alignment case) --- very small data test: memcpy_neon_old : (3 bytes copy) = 72.2 MB/s / 75.5 MB/s memcpy_neon_a : (3 bytes copy) = 75.7 MB/s / 77.6 MB/s memcpy_neon_u : (3 bytes copy) = 74.8 MB/s / 87.1 MB/s memcpy_arm : (3 bytes copy) = 66.6 MB/s / 67.7 MB/s memcpy_neon_old : (4 bytes copy) = 74.4 MB/s / 76.1 MB/s memcpy_neon_a : (4 bytes copy) = 86.8 MB/s / 88.8 MB/s memcpy_neon_u : (4 bytes copy) = 80.9 MB/s / 116.7 MB/s memcpy_arm : (4 bytes copy) = 50.8 MB/s / 55.7 MB/s memcpy_neon_old : (5 bytes copy) = 93.3 MB/s / 95.0 MB/s memcpy_neon_a : (5 bytes copy) = 108.8 MB/s / 110.5 MB/s memcpy_neon_u : (5 bytes copy) = 100.4 MB/s / 144.4 MB/s memcpy_arm : (5 bytes copy) = 56.8 MB/s / 69.5 MB/s memcpy_neon_old : (7 bytes copy) = 105.1 MB/s / 107.0 MB/s memcpy_neon_a : (7 bytes copy) = 150.9 MB/s / 155.7 MB/s memcpy_neon_u : (7 bytes copy) = 128.2 MB/s / 204.1 MB/s memcpy_arm : (7 bytes copy) = 69.1 MB/s / 97.5 MB/s memcpy_neon_old : (8 bytes copy) = 100.7 MB/s / 102.3 MB/s memcpy_neon_a : (8 bytes copy) = 128.5 MB/s / 131.8 MB/s memcpy_neon_u : (8 bytes copy) = 127.4 MB/s / 231.6 MB/s memcpy_arm : (8 bytes copy) = 78.5 MB/s / 100.8 MB/s memcpy_neon_old : (11 bytes copy) = 119.5 MB/s / 121.0 MB/s memcpy_neon_a : (11 bytes copy) = 176.6 MB/s / 181.4 MB/s memcpy_neon_u : (11 bytes copy) = 164.1 MB/s / 320.4 MB/s memcpy_arm : (11 bytes copy) = 93.8 MB/s / 139.1 MB/s memcpy_neon_old : (12 bytes copy) = 114.3 MB/s / 115.5 MB/s memcpy_neon_a : (12 bytes copy) = 155.4 MB/s / 159.3 MB/s memcpy_neon_u : (12 bytes copy) = 161.4 MB/s / 349.6 MB/s memcpy_arm : (12 bytes copy) = 104.7 MB/s / 136.3 MB/s memcpy_neon_old : (15 bytes copy) = 127.5 MB/s / 128.4 MB/s memcpy_neon_a : (15 bytes copy) = 194.8 MB/s / 198.3 MB/s memcpy_neon_u : (15 bytes copy) = 195.0 MB/s / 437.5 MB/s memcpy_arm : (15 bytes copy) = 117.5 MB/s / 170.7 MB/s memcpy_neon_old : (16 bytes copy) = 119.4 MB/s / 207.8 MB/s memcpy_neon_a : (16 bytes copy) = 127.9 MB/s / 323.1 MB/s memcpy_neon_u : (16 bytes copy) = 154.7 MB/s / 338.3 MB/s memcpy_arm : (16 bytes copy) = 124.4 MB/s / 165.8 MB/s memcpy_neon_old : (24 bytes copy) = 179.2 MB/s / 321.4 MB/s memcpy_neon_a : (24 bytes copy) = 184.5 MB/s / 329.7 MB/s memcpy_neon_u : (24 bytes copy) = 214.8 MB/s / 507.9 MB/s memcpy_arm : (24 bytes copy) = 158.6 MB/s / 211.4 MB/s memcpy_neon_old : (31 bytes copy) = 230.8 MB/s / 394.3 MB/s memcpy_neon_a : (31 bytes copy) = 248.7 MB/s / 357.4 MB/s memcpy_neon_u : (31 bytes copy) = 291.5 MB/s / 655.7 MB/s memcpy_arm : (31 bytes copy) = 182.6 MB/s / 254.1 MB/s L1 cached data: memcpy_neon_old : (4096 bytes copy) = 1734.7 MB/s / 1838.1 MB/s memcpy_neon_a : (4096 bytes copy) = 1957.0 MB/s / 2099.6 MB/s memcpy_neon_u : (4096 bytes copy) = 1994.8 MB/s / 2063.5 MB/s memcpy_arm : (4096 bytes copy) = 781.7 MB/s / 1314.5 MB/s memcpy_neon_old : (6144 bytes copy) = 1776.2 MB/s / 1844.4 MB/s memcpy_neon_a : (6144 bytes copy) = 2011.8 MB/s / 2098.6 MB/s memcpy_neon_u : (6144 bytes copy) = 2038.1 MB/s / 2081.3 MB/s memcpy_arm : (6144 bytes copy) = 784.8 MB/s / 1305.8 MB/s L2 cached data: memcpy_neon_old : (65536 bytes copy) = 626.1 MB/s / 695.8 MB/s memcpy_neon_a : (65536 bytes copy) = 621.7 MB/s / 704.5 MB/s memcpy_neon_u : (65536 bytes copy) = 621.7 MB/s / 704.5 MB/s memcpy_arm : (65536 bytes copy) = 537.6 MB/s / 506.1 MB/s memcpy_neon_old : (98304 bytes copy) = 551.5 MB/s / 586.1 MB/s memcpy_neon_a : (98304 bytes copy) = 550.6 MB/s / 589.4 MB/s memcpy_neon_u : (98304 bytes copy) = 550.6 MB/s / 589.3 MB/s memcpy_arm : (98304 bytes copy) = 429.0 MB/s / 414.6 MB/s SDRAM: memcpy_neon_old : (2097152 bytes copy) = 268.5 MB/s / 272.9 MB/s memcpy_neon_a : (2097152 bytes copy) = 268.1 MB/s / 271.0 MB/s memcpy_neon_u : (2097152 bytes copy) = 268.2 MB/s / 271.0 MB/s memcpy_arm : (2097152 bytes copy) = 170.3 MB/s / 181.4 MB/s