-
Notifications
You must be signed in to change notification settings - Fork 96
Mele A2000 (Allwinner A10)
ssvb edited this page Mar 15, 2013
·
5 revisions
Processor : ARMv7 Processor rev 2 (v7l) BogoMIPS : 1001.88 Features : swp half thumb fastmult vfp edsp neon vfpv3 CPU implementer : 0x41 CPU architecture: 7 CPU variant : 0x3 CPU part : 0xc08 CPU revision : 2 Hardware : sun4i Revision : 0000 Serial : 0000000000000000
2x16-bit DDR3 is clocked at 480MHz (up from the default 360MHz)
tinymembench v0.2 (simple benchmark for memory throughput and latency)
==========================================================================
== Memory bandwidth tests ==
== ==
== Note 1: 1MB = 1000000 bytes ==
== Note 2: Results for 'copy' tests show how many bytes can be ==
== copied per second (adding together read and writen ==
== bytes would have provided twice higher numbers) ==
== Note 3: 2-pass copy means that we are using a small temporary buffer ==
== to first fetch data into it, and only then write it to the ==
== destination (source -> L1 cache, L1 cache -> destination) ==
== Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
== brackets ==
==========================================================================
C copy backwards : 322.9 MB/s (1.3%)
C copy : 320.9 MB/s (1.2%)
C copy prefetched (32 bytes step) : 730.2 MB/s
C copy prefetched (64 bytes step) : 730.4 MB/s
C 2-pass copy : 291.5 MB/s (1.1%)
C 2-pass copy prefetched (32 bytes step) : 560.8 MB/s (1.6%)
C 2-pass copy prefetched (64 bytes step) : 560.9 MB/s (1.7%)
C fill : 1461.8 MB/s (0.4%)
---
standard memcpy : 570.6 MB/s
standard memset : 1460.8 MB/s
---
NEON read : 1326.3 MB/s
NEON read prefetched (32 bytes step) : 1055.6 MB/s
NEON read prefetched (64 bytes step) : 1059.0 MB/s (0.2%)
NEON copy : 862.3 MB/s
NEON copy prefetched (32 bytes step) : 917.8 MB/s
NEON copy prefetched (64 bytes step) : 920.6 MB/s (0.1%)
NEON unrolled copy : 862.6 MB/s
NEON unrolled copy prefetched (32 bytes step) : 807.9 MB/s
NEON unrolled copy prefetched (64 bytes step) : 808.4 MB/s (0.1%)
NEON copy backwards : 865.2 MB/s
NEON copy backwards prefetched (32 bytes step) : 918.1 MB/s
NEON copy backwards prefetched (64 bytes step) : 918.9 MB/s (0.1%)
NEON 2-pass copy : 621.0 MB/s
NEON 2-pass copy prefetched (32 bytes step) : 622.2 MB/s
NEON 2-pass copy prefetched (64 bytes step) : 639.1 MB/s
NEON unrolled 2-pass copy : 637.9 MB/s
NEON unrolled 2-pass copy prefetched (32 bytes step) : 590.8 MB/s
NEON unrolled 2-pass copy prefetched (64 bytes step) : 600.2 MB/s
NEON fill : 1466.8 MB/s
NEON fill backwards : 1466.8 MB/s
ARM fill (STRD) : 1462.0 MB/s
ARM fill (STM with 8 registers) : 1463.2 MB/s
ARM fill (STM with 4 registers) : 1462.2 MB/s
ARM copy prefetched (incr pld) : 774.0 MB/s
ARM copy prefetched (wrap pld) : 775.7 MB/s
ARM 2-pass copy prefetched (incr pld) : 567.4 MB/s
ARM 2-pass copy prefetched (wrap pld) : 543.2 MB/s
==========================================================================
== Memory latency test ==
== ==
== Average time is measured for random memory accesses in the buffers ==
== of different sizes. The larger is the buffer, the more significant ==
== are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
== accesses. For extremely large buffer sizes we are expecting to see ==
== page table walk with total 3 requests to SDRAM for almost every ==
== memory access (though 64MiB is not large enough to experience this ==
== effect to its fullest). ==
== ==
== Note 1: All the numbers are representing extra time, which needs to ==
== be added to L1 cache latency. The cycle timings for L1 cache ==
== latency can be usually found in the processor documentation. ==
== Note 1: Dual random read means that we are simultaneously performing ==
== two independent memory accesses at a time. In the case if ==
== the memory subsystem can't handle multiple outstanding ==
== requests, dual random read has the same timings as two ==
== single reads performed one after another. ==
==========================================================================
block size : read access time (single random read / dual random read)
2 : 0.0 ns / 0.0 ns
4 : 0.0 ns / 0.0 ns
8 : 0.0 ns / 0.0 ns
16 : 0.0 ns / 0.0 ns
32 : 0.0 ns / 0.0 ns
64 : 0.0 ns / 0.0 ns
128 : 0.0 ns / 0.0 ns
256 : 0.0 ns / 0.0 ns
512 : 0.0 ns / 0.0 ns
1024 : 0.0 ns / 0.0 ns
2048 : 0.0 ns / 0.0 ns
4096 : 0.0 ns / 0.0 ns
8192 : 0.0 ns / 0.0 ns
16384 : 0.0 ns / 0.0 ns
32768 : 0.0 ns / 0.0 ns
65536 : 5.0 ns / 9.7 ns
131072 : 7.5 ns / 14.7 ns
262144 : 36.5 ns / 70.6 ns
524288 : 106.1 ns / 213.7 ns
1048576 : 144.4 ns / 292.4 ns
2097152 : 164.1 ns / 332.8 ns
4194304 : 174.8 ns / 354.5 ns
8388608 : 181.8 ns / 369.0 ns
16777216 : 188.8 ns / 385.4 ns
33554432 : 200.1 ns / 418.0 ns
67108864 : 221.8 ns / 471.5 ns
Driving a 1920x1080-32@60Hz HDMI monitor
tinymembench v0.2 (simple benchmark for memory throughput and latency)
==========================================================================
== Memory bandwidth tests ==
== ==
== Note 1: 1MB = 1000000 bytes ==
== Note 2: Results for 'copy' tests show how many bytes can be ==
== copied per second (adding together read and writen ==
== bytes would have provided twice higher numbers) ==
== Note 3: 2-pass copy means that we are using a small temporary buffer ==
== to first fetch data into it, and only then write it to the ==
== destination (source -> L1 cache, L1 cache -> destination) ==
== Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
== brackets ==
==========================================================================
C copy backwards : 317.4 MB/s
C copy : 316.4 MB/s
C copy prefetched (32 bytes step) : 525.9 MB/s
C copy prefetched (64 bytes step) : 526.1 MB/s
C 2-pass copy : 267.2 MB/s (1.2%)
C 2-pass copy prefetched (32 bytes step) : 516.8 MB/s
C 2-pass copy prefetched (64 bytes step) : 517.0 MB/s
C fill : 555.6 MB/s
---
standard memcpy : 518.0 MB/s (2.1%)
standard memset : 555.8 MB/s
---
NEON read : 1224.1 MB/s
NEON read prefetched (32 bytes step) : 985.5 MB/s
NEON read prefetched (64 bytes step) : 985.5 MB/s (3.0%)
NEON copy : 530.4 MB/s
NEON copy prefetched (32 bytes step) : 532.7 MB/s (1.6%)
NEON copy prefetched (64 bytes step) : 533.2 MB/s (1.6%)
NEON unrolled copy : 530.8 MB/s (0.1%)
NEON unrolled copy prefetched (32 bytes step) : 528.8 MB/s (1.6%)
NEON unrolled copy prefetched (64 bytes step) : 528.7 MB/s
NEON copy backwards : 603.2 MB/s
NEON copy backwards prefetched (32 bytes step) : 625.0 MB/s
NEON copy backwards prefetched (64 bytes step) : 625.1 MB/s
NEON 2-pass copy : 520.2 MB/s
NEON 2-pass copy prefetched (32 bytes step) : 520.3 MB/s
NEON 2-pass copy prefetched (64 bytes step) : 521.0 MB/s
NEON unrolled 2-pass copy : 521.2 MB/s
NEON unrolled 2-pass copy prefetched (32 bytes step) : 519.0 MB/s
NEON unrolled 2-pass copy prefetched (64 bytes step) : 519.0 MB/s
NEON fill : 556.5 MB/s
NEON fill backwards : 1040.1 MB/s
ARM fill (STRD) : 556.5 MB/s
ARM fill (STM with 8 registers) : 556.5 MB/s
ARM fill (STM with 4 registers) : 556.5 MB/s
ARM copy prefetched (incr pld) : 528.0 MB/s
ARM copy prefetched (wrap pld) : 526.2 MB/s
ARM 2-pass copy prefetched (incr pld) : 516.2 MB/s
ARM 2-pass copy prefetched (wrap pld) : 490.5 MB/s
==========================================================================
== Memory latency test ==
== ==
== Average time is measured for random memory accesses in the buffers ==
== of different sizes. The larger is the buffer, the more significant ==
== are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
== accesses. For extremely large buffer sizes we are expecting to see ==
== page table walk with total 3 requests to SDRAM for almost every ==
== memory access (though 64MiB is not large enough to experience this ==
== effect to its fullest). ==
== ==
== Note 1: All the numbers are representing extra time, which needs to ==
== be added to L1 cache latency. The cycle timings for L1 cache ==
== latency can be usually found in the processor documentation. ==
== Note 1: Dual random read means that we are simultaneously performing ==
== two independent memory accesses at a time. In the case if ==
== the memory subsystem can't handle multiple outstanding ==
== requests, dual random read has the same timings as two ==
== single reads performed one after another. ==
==========================================================================
block size : read access time (single random read / dual random read)
2 : 0.0 ns / 0.0 ns
4 : 0.0 ns / 0.0 ns
8 : 0.0 ns / 0.0 ns
16 : 0.0 ns / 0.0 ns
32 : 0.0 ns / 0.0 ns
64 : 0.0 ns / 0.0 ns
128 : 0.0 ns / 0.0 ns
256 : 0.0 ns / 0.0 ns
512 : 0.0 ns / 0.0 ns
1024 : 0.0 ns / 0.0 ns
2048 : 0.0 ns / 0.0 ns
4096 : 0.0 ns / 0.0 ns
8192 : 0.0 ns / 0.0 ns
16384 : 0.0 ns / 0.0 ns
32768 : 0.0 ns / 0.0 ns
65536 : 5.0 ns / 9.7 ns
131072 : 7.5 ns / 14.8 ns
262144 : 36.9 ns / 72.1 ns
524288 : 108.5 ns / 218.3 ns
1048576 : 147.4 ns / 298.8 ns
2097152 : 167.8 ns / 340.0 ns
4194304 : 178.6 ns / 362.2 ns
8388608 : 185.6 ns / 376.3 ns
16777216 : 191.8 ns / 388.9 ns
33554432 : 201.7 ns / 411.9 ns
67108864 : 221.1 ns / 463.4 ns
Driving two 1920x1080-32@60Hz monitors (HDMI and VGA) at the same time
tinymembench v0.2.9 (simple benchmark for memory throughput and latency)
==========================================================================
== Memory bandwidth tests ==
== ==
== Note 1: 1MB = 1000000 bytes ==
== Note 2: Results for 'copy' tests show how many bytes can be ==
== copied per second (adding together read and writen ==
== bytes would have provided twice higher numbers) ==
== Note 3: 2-pass copy means that we are using a small temporary buffer ==
== to first fetch data into it, and only then write it to the ==
== destination (source -> L1 cache, L1 cache -> destination) ==
== Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
== brackets ==
==========================================================================
C copy backwards : 209.6 MB/s
C copy : 217.8 MB/s
C copy prefetched (32 bytes step) : 296.2 MB/s (0.5%)
C copy prefetched (64 bytes step) : 296.1 MB/s
C 2-pass copy : 167.3 MB/s
C 2-pass copy prefetched (32 bytes step) : 247.4 MB/s
C 2-pass copy prefetched (64 bytes step) : 247.3 MB/s
C fill : 519.0 MB/s
---
standard memcpy : 315.7 MB/s
standard memset : 519.0 MB/s
---
NEON read : 814.3 MB/s (1.2%)
NEON read prefetched (32 bytes step) : 589.4 MB/s
NEON read prefetched (64 bytes step) : 590.4 MB/s
NEON copy : 351.6 MB/s
NEON copy prefetched (32 bytes step) : 351.2 MB/s
NEON copy prefetched (64 bytes step) : 349.7 MB/s
NEON unrolled copy : 358.1 MB/s
NEON unrolled copy prefetched (32 bytes step) : 313.4 MB/s
NEON unrolled copy prefetched (64 bytes step) : 348.1 MB/s
NEON copy backwards : 340.4 MB/s (0.5%)
NEON copy backwards prefetched (32 bytes step) : 341.3 MB/s
NEON copy backwards prefetched (64 bytes step) : 341.3 MB/s
NEON 2-pass copy : 304.6 MB/s
NEON 2-pass copy prefetched (32 bytes step) : 301.0 MB/s
NEON 2-pass copy prefetched (64 bytes step) : 303.8 MB/s
NEON unrolled 2-pass copy : 305.0 MB/s
NEON unrolled 2-pass copy prefetched (32 bytes step) : 268.4 MB/s (0.4%)
NEON unrolled 2-pass copy prefetched (64 bytes step) : 296.1 MB/s
NEON fill : 519.1 MB/s
NEON fill backwards : 495.5 MB/s
ARM fill (STRD) : 519.1 MB/s
ARM fill (STM with 8 registers) : 519.0 MB/s
ARM fill (STM with 4 registers) : 519.1 MB/s
ARM copy prefetched (incr pld) : 347.6 MB/s (0.5%)
ARM copy prefetched (wrap pld) : 317.4 MB/s
ARM 2-pass copy prefetched (incr pld) : 281.2 MB/s
ARM 2-pass copy prefetched (wrap pld) : 268.8 MB/s
==========================================================================
== Memory latency test ==
== ==
== Average time is measured for random memory accesses in the buffers ==
== of different sizes. The larger is the buffer, the more significant ==
== are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
== accesses. For extremely large buffer sizes we are expecting to see ==
== page table walk with total 3 requests to SDRAM for almost every ==
== memory access (though 64MiB is not large enough to experience this ==
== effect to its fullest). ==
== ==
== Note 1: All the numbers are representing extra time, which needs to ==
== be added to L1 cache latency. The cycle timings for L1 cache ==
== latency can be usually found in the processor documentation. ==
== Note 2: Dual random read means that we are simultaneously performing ==
== two independent memory accesses at a time. In the case if ==
== the memory subsystem can't handle multiple outstanding ==
== requests, dual random read has the same timings as two ==
== single reads performed one after another. ==
==========================================================================
block size : read access time (single random read / dual random read)
2 : 0.0 ns / 0.0 ns
4 : 0.0 ns / 0.0 ns
8 : 0.0 ns / 0.0 ns
16 : 0.0 ns / 0.0 ns
32 : 0.0 ns / 0.0 ns
64 : 0.0 ns / 0.0 ns
128 : 0.0 ns / 0.0 ns
256 : 0.0 ns / 0.0 ns
512 : 0.0 ns / 0.0 ns
1024 : 0.0 ns / 0.0 ns
2048 : 0.0 ns / 0.0 ns
4096 : 0.0 ns / 0.0 ns
8192 : 0.0 ns / 0.0 ns
16384 : 0.0 ns / 0.0 ns
32768 : 0.0 ns / 0.0 ns
65536 : 5.0 ns / 9.7 ns
131072 : 7.5 ns / 14.8 ns
262144 : 37.5 ns / 74.4 ns
524288 : 118.6 ns / 241.0 ns
1048576 : 164.5 ns / 336.4 ns
2097152 : 188.2 ns / 385.8 ns
4194304 : 201.1 ns / 412.5 ns
8388608 : 209.4 ns / 429.6 ns
16777216 : 217.6 ns / 445.9 ns
33554432 : 229.5 ns / 469.9 ns
67108864 : 254.8 ns / 522.1 ns
Kernel 4.9.140-tegra #1 SMP PREEMPT Wed Mar 13 00:32:22 PDT 2019 aarch64 GNU/Linux Under xorg, no compositor active, no browser or other cpu hogs.
tinymembench v0.4.9 (simple benchmark for memory thr
==========================================================================
== Memory bandwidth tests ==
== ==
== Note 1: 1MB = 1000000 bytes ==
== Note 2: Results for 'copy' tests show how many bytes can be ==
== copied per second (adding together read and writen ==
== bytes would have provided twice higher numbers) ==
== Note 3: 2-pass copy means that we are using a small temporary buffer ==
== to first fetch data into it, and only then write it to the ==
== destination (source -> L1 cache, L1 cache -> destination) ==
== Note 4: If sample standard deviation exceeds 0.1%, it is shown in ==
== brackets ==
==========================================================================
C copy backwards : 2949.7 MB/s (3.8%)
C copy backwards (32 byte blocks) : 3011.8 MB/s
C copy backwards (64 byte blocks) : 3029.2 MB/s
C copy : 3642.2 MB/s (4.1%)
C copy prefetched (32 bytes step) : 3824.4 MB/s (0.3%)
C copy prefetched (64 bytes step) : 3825.3 MB/s (0.4%)
C 2-pass copy : 2726.2 MB/s
C 2-pass copy prefetched (32 bytes step) : 2902.6 MB/s (2.5%)
C 2-pass copy prefetched (64 bytes step) : 2928.3 MB/s (0.3%)
C fill : 8541.0 MB/s (0.2%)
C fill (shuffle within 16 byte blocks) : 8518.5 MB/s (2.1%)
C fill (shuffle within 32 byte blocks) : 8537.1 MB/s (0.1%)
C fill (shuffle within 64 byte blocks) : 8528.7 MB/s (0.2%)
---
standard memcpy : 3558.8 MB/s
standard memset : 8520.2 MB/s
---
NEON LDP/STP copy : 3633.9 MB/s (4.2%)
NEON LDP/STP copy pldl2strm (32 bytes step) : 1451.0 MB/s (0.3%)
NEON LDP/STP copy pldl2strm (64 bytes step) : 1450.9 MB/s (0.5%)
NEON LDP/STP copy pldl1keep (32 bytes step) : 3882.5 MB/s (3.9%)
NEON LDP/STP copy pldl1keep (64 bytes step) : 3884.0 MB/s (0.4%)
NEON LD1/ST1 copy : 3630.8 MB/s (0.3%)
NEON STP fill : 8537.8 MB/s
NEON STNP fill : 8544.9 MB/s (1.7%)
ARM LDP/STP copy : 3635.8 MB/s (0.3%)
ARM STP fill : 8544.8 MB/s (0.1%)
ARM STNP fill : 8549.2 MB/s (1.0%)
==========================================================================
== Framebuffer read tests. ==
== ==
== Many ARM devices use a part of the system memory as the framebuffer, ==
== typically mapped as uncached but with write-combining enabled. ==
== Writes to such framebuffers are quite fast, but reads are much ==
== slower and very sensitive to the alignment and the selection of ==
== CPU instructions which are used for accessing memory. ==
== ==
== Many x86 systems allocate the framebuffer in the GPU memory, ==
== accessible for the CPU via a relatively slow PCI-E bus. Moreover, ==
== PCI-E is asymmetric and handles reads a lot worse than writes. ==
== ==
== If uncached framebuffer reads are reasonably fast (at least 100 MB/s ==
== or preferably >300 MB/s), then using the shadow framebuffer layer ==
== is not necessary in Xorg DDX drivers, resulting in a nice overall ==
== performance improvement. For example, the xf86-video-fbturbo DDX ==
== uses this trick. ==
==========================================================================
NEON LDP/STP copy (from framebuffer) : 766.0 MB/s
NEON LDP/STP 2-pass copy (from framebuffer) : 688.8 MB/s
NEON LD1/ST1 copy (from framebuffer) : 770.6 MB/s (0.1%)
NEON LD1/ST1 2-pass copy (from framebuffer) : 681.3 MB/s (0.3%)
ARM LDP/STP copy (from framebuffer) : 766.1 MB/s
ARM LDP/STP 2-pass copy (from framebuffer) : 689.1 MB/s
==========================================================================
== Memory latency test ==
== ==
== Average time is measured for random memory accesses in the buffers ==
== of different sizes. The larger is the buffer, the more significant ==
== are relative contributions of TLB, L1/L2 cache misses and SDRAM ==
== accesses. For extremely large buffer sizes we are expecting to see ==
== page table walk with several requests to SDRAM for almost every ==
== memory access (though 64MiB is not nearly large enough to experience ==
== this effect to its fullest). ==
== ==
== Note 1: All the numbers are representing extra time, which needs to ==
== be added to L1 cache latency. The cycle timings for L1 cache ==
== latency can be usually found in the processor documentation. ==
== Note 2: Dual random read means that we are simultaneously performing ==
== two independent memory accesses at a time. In the case if ==
== the memory subsystem can't handle multiple outstanding ==
== requests, dual random read has the same timings as two ==
== single reads performed one after another. ==
==========================================================================
block size : single random read / dual random read, [MADV_NOHUGEPAGE]
1024 : 0.0 ns / 0.1 ns
2048 : 0.0 ns / 0.1 ns
4096 : 0.0 ns / 0.1 ns
8192 : 0.0 ns / 0.1 ns
16384 : 0.1 ns / 0.1 ns
32768 : 1.7 ns / 2.9 ns
65536 : 6.4 ns / 9.5 ns
131072 : 9.6 ns / 12.3 ns
262144 : 13.7 ns / 17.0 ns
524288 : 15.8 ns / 19.7 ns
1048576 : 17.3 ns / 22.1 ns
2097152 : 42.1 ns / 64.2 ns
4194304 : 98.5 ns / 138.1 ns
8388608 : 143.9 ns / 186.3 ns
16777216 : 167.2 ns / 211.2 ns
33554432 : 180.1 ns / 227.1 ns
67108864 : 200.0 ns / 260.2 ns
block size : single random read / dual random read, [MADV_HUGEPAGE]
1024 : 0.0 ns / 0.0 ns
2048 : 0.0 ns / 0.0 ns
4096 : 0.0 ns / 0.0 ns
8192 : 0.0 ns / 0.0 ns
16384 : 0.0 ns / 0.0 ns
32768 : 0.0 ns / 0.0 ns
65536 : 6.4 ns / 9.4 ns
131072 : 9.5 ns / 12.2 ns
262144 : 11.2 ns / 13.1 ns
524288 : 12.1 ns / 13.5 ns
1048576 : 12.8 ns / 13.6 ns
2097152 : 27.0 ns / 33.0 ns
4194304 : 90.6 ns / 127.8 ns
8388608 : 123.9 ns / 153.8 ns
16777216 : 139.5 ns / 161.2 ns
33554432 : 147.2 ns / 163.6 ns
67108864 : 154.0 ns / 167.6 ns