rk3229 tv box

#Linux armbian 4.4.194-rk322x #3 SMP Thu May 6 22:00:38 UTC 2021 armv7l armv7l armv7l GNU/Linux tinymembench v0.4.9 (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 : 332.7 MB/s (0.3%) C copy backwards (32 byte blocks) : 846.1 MB/s (0.5%) C copy backwards (64 byte blocks) : 841.9 MB/s (0.5%) C copy : 841.9 MB/s (1.1%) C copy prefetched (32 bytes step) : 936.8 MB/s (0.5%) C copy prefetched (64 bytes step) : 966.8 MB/s (0.6%) C 2-pass copy : 911.6 MB/s (0.4%) C 2-pass copy prefetched (32 bytes step) : 917.1 MB/s (0.3%) C 2-pass copy prefetched (64 bytes step) : 932.0 MB/s (5.1%) C fill : 2249.5 MB/s C fill (shuffle within 16 byte blocks) : 2252.1 MB/s C fill (shuffle within 32 byte blocks) : 517.5 MB/s (1.0%) C fill (shuffle within 64 byte blocks) : 516.8 MB/s (0.5%)

standard memcpy : 699.9 MB/s (0.4%) standard memset : 2250.3 MB/s

NEON read : 1722.6 MB/s (0.6%) NEON read prefetched (32 bytes step) : 1848.3 MB/s (3.7%) NEON read prefetched (64 bytes step) : 1844.2 MB/s (0.4%) NEON read 2 data streams : 523.8 MB/s NEON read 2 data streams prefetched (32 bytes step) : 964.9 MB/s (0.2%) NEON read 2 data streams prefetched (64 bytes step) : 1042.8 MB/s NEON copy : 843.0 MB/s (0.6%) NEON copy prefetched (32 bytes step) : 874.9 MB/s (0.6%) NEON copy prefetched (64 bytes step) : 928.1 MB/s (1.1%) NEON unrolled copy : 840.2 MB/s (0.4%) NEON unrolled copy prefetched (32 bytes step) : 884.5 MB/s (0.5%) NEON unrolled copy prefetched (64 bytes step) : 921.7 MB/s (3.5%) NEON copy backwards : 853.3 MB/s (1.6%) NEON copy backwards prefetched (32 bytes step) : 852.3 MB/s (0.7%) NEON copy backwards prefetched (64 bytes step) : 916.2 MB/s (0.8%) NEON 2-pass copy : 931.9 MB/s (0.5%) NEON 2-pass copy prefetched (32 bytes step) : 957.3 MB/s (1.2%) NEON 2-pass copy prefetched (64 bytes step) : 978.5 MB/s (0.6%) NEON unrolled 2-pass copy : 861.9 MB/s (0.7%) NEON unrolled 2-pass copy prefetched (32 bytes step) : 810.9 MB/s (0.3%) NEON unrolled 2-pass copy prefetched (64 bytes step) : 851.1 MB/s (0.2%) NEON fill : 2252.1 MB/s (3.1%) NEON fill backwards : 2250.8 MB/s (0.1%) VFP copy : 842.6 MB/s (0.5%) VFP 2-pass copy : 817.6 MB/s (0.3%) ARM fill (STRD) : 2251.4 MB/s (0.3%) ARM fill (STM with 8 registers) : 2250.5 MB/s ARM fill (STM with 4 registers) : 2253.1 MB/s (0.6%) ARM copy prefetched (incr pld) : 911.2 MB/s (1.2%) ARM copy prefetched (wrap pld) : 859.5 MB/s (4.0%) ARM 2-pass copy prefetched (incr pld) : 931.6 MB/s (0.5%) ARM 2-pass copy prefetched (wrap pld) : 901.9 MB/s (0.4%)

========================================================================== == 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 read (from framebuffer) : 79.2 MB/s (0.1%) NEON copy (from framebuffer) : 76.6 MB/s (0.1%) NEON 2-pass copy (from framebuffer) : 76.2 MB/s (0.2%) NEON unrolled copy (from framebuffer) : 73.5 MB/s NEON 2-pass unrolled copy (from framebuffer) : 75.2 MB/s VFP copy (from framebuffer) : 341.6 MB/s (0.1%) VFP 2-pass copy (from framebuffer) : 380.0 MB/s ARM copy (from framebuffer) : 267.5 MB/s (0.8%) ARM 2-pass copy (from framebuffer) : 245.0 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 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 : 6.1 ns / 10.2 ns 131072 : 9.4 ns / 14.1 ns 262144 : 13.1 ns / 19.1 ns 524288 : 74.0 ns / 117.3 ns 1048576 : 108.2 ns / 154.0 ns 2097152 : 130.8 ns / 172.6 ns 4194304 : 142.5 ns / 181.0 ns 8388608 : 150.9 ns / 188.3 ns 16777216 : 159.0 ns / 198.4 ns 33554432 : 171.4 ns / 219.7 ns 67108864 : 193.1 ns / 262.7 ns