Skip to content

Linux_Performance_Guide.rst

Latest commit

 

History

History
762 lines (599 loc) · 29.1 KB

Linux_Performance_Guide.rst

File metadata and controls

762 lines (599 loc) · 29.1 KB

Linux 11.00 Performance Guide

Read This First

All performance numbers provided in this document are gathered using following Evaluation Modules unless otherwise specified.

Name Description
AM62Ax SK AM62Ax Starter Kit with ARM running at 1.2GHz, DDR data rate 3733 MT/S

Table: Evaluation Modules

About This Manual

This document provides performance data for each of the device drivers which are part of the Processor SDK Linux package. This document should be used in conjunction with release notes and user guides provided with the Processor SDK Linux package for information on specific issues present with drivers included in a particular release.

If You Need Assistance

For further information or to report any problems, contact https://e2e.ti.com/ or https://support.ti.com/

System Benchmarks

LMBench

LMBench is a collection of microbenchmarks of which the memory bandwidth and latency related ones are typically used to estimate processor memory system performance. More information about lmbench at https://lmbench.sourceforge.net/whatis_lmbench.html and https://lmbench.sourceforge.net/man/lmbench.8.html

Latency: lat_mem_rd-stride128-szN, where N is equal to or smaller than the cache size at given level measures the cache miss penalty. N that is at least double the size of last level cache is the latency to external memory.

Bandwidth: bw_mem_bcopy-N, where N is equal to or smaller than the cache size at a given level measures the achievable memory bandwidth from software doing a memcpy() type operation. Typical use is for external memory bandwidth calculation. The bandwidth is calculated as byte read and written counts as 1 which should be roughly half of STREAM copy result.

Execute the LMBench with the following:

cd /opt/ltp
./runltp -P j721e-idk-gw -f ddt/lmbench -s LMBENCH_L_PERF_0001
LMBench Benchmarks
Benchmarks am62axx_sk-fs: perf
af_unix_sock_stream_latency (microsec) 28.80
af_unix_socket_stream_bandwidth (MBs) 1158.39
bw_file_rd-io-1mb (MB/s) 1367.99
bw_file_rd-o2c-1mb (MB/s) 771.72
bw_mem-bcopy-16mb (MB/s) 1805.05
bw_mem-bcopy-1mb (MB/s) 1992.39
bw_mem-bcopy-2mb (MB/s) 1734.30
bw_mem-bcopy-4mb (MB/s) 1786.78
bw_mem-bcopy-8mb (MB/s) 1796.74
bw_mem-bzero-16mb (MB/s) 7947.02
bw_mem-bzero-1mb (MB/s) 4970.90 (min 1992.39, max 7949.41)
bw_mem-bzero-2mb (MB/s) 4832.55 (min 1734.30, max 7930.79)
bw_mem-bzero-4mb (MB/s) 4864.51 (min 1786.78, max 7942.24)
bw_mem-bzero-8mb (MB/s) 4867.28 (min 1796.74, max 7937.82)
bw_mem-cp-16mb (MB/s) 930.29
bw_mem-cp-1mb (MB/s) 4563.39 (min 916.59, max 8210.18)
bw_mem-cp-2mb (MB/s) 4473.88 (min 893.12, max 8054.63)
bw_mem-cp-4mb (MB/s) 4472.48 (min 943.51, max 8001.45)
bw_mem-cp-8mb (MB/s) 4465.06 (min 959.35, max 7970.77)
bw_mem-fcp-16mb (MB/s) 1724.32
bw_mem-fcp-1mb (MB/s) 4842.31 (min 1735.21, max 7949.41)
bw_mem-fcp-2mb (MB/s) 4754.04 (min 1577.29, max 7930.79)
bw_mem-fcp-4mb (MB/s) 4822.07 (min 1701.89, max 7942.24)
bw_mem-fcp-8mb (MB/s) 4749.32 (min 1560.82, max 7937.82)
bw_mem-frd-16mb (MB/s) 1904.31
bw_mem-frd-1mb (MB/s) 1886.69 (min 1735.21, max 2038.17)
bw_mem-frd-2mb (MB/s) 1663.69 (min 1577.29, max 1750.09)
bw_mem-frd-4mb (MB/s) 1642.30 (min 1582.70, max 1701.89)
bw_mem-frd-8mb (MB/s) 1637.86 (min 1560.82, max 1714.90)
bw_mem-fwr-16mb (MB/s) 7964.16
bw_mem-fwr-1mb (MB/s) 5124.18 (min 2038.17, max 8210.18)
bw_mem-fwr-2mb (MB/s) 4902.36 (min 1750.09, max 8054.63)
bw_mem-fwr-4mb (MB/s) 4792.08 (min 1582.70, max 8001.45)
bw_mem-fwr-8mb (MB/s) 4842.84 (min 1714.90, max 7970.77)
bw_mem-rd-16mb (MB/s) 1951.93
bw_mem-rd-1mb (MB/s) 1994.86 (min 1764.60, max 2225.11)
bw_mem-rd-2mb (MB/s) 1756.29 (min 1568.63, max 1943.95)
bw_mem-rd-4mb (MB/s) 1841.36 (min 1734.35, max 1948.37)
bw_mem-rd-8mb (MB/s) 1892.72 (min 1830.87, max 1954.56)
bw_mem-rdwr-16mb (MB/s) 1863.50
bw_mem-rdwr-1mb (MB/s) 1331.97 (min 916.59, max 1747.34)
bw_mem-rdwr-2mb (MB/s) 1205.75 (min 893.12, max 1518.37)
bw_mem-rdwr-4mb (MB/s) 1310.86 (min 943.51, max 1678.20)
bw_mem-rdwr-8mb (MB/s) 1371.14 (min 959.35, max 1782.93)
bw_mem-wr-16mb (MB/s) 1805.67
bw_mem-wr-1mb (MB/s) 1755.97 (min 1747.34, max 1764.60)
bw_mem-wr-2mb (MB/s) 1543.50 (min 1518.37, max 1568.63)
bw_mem-wr-4mb (MB/s) 1706.28 (min 1678.20, max 1734.35)
bw_mem-wr-8mb (MB/s) 1806.90 (min 1782.93, max 1830.87)
bw_mmap_rd-mo-1mb (MB/s) 2130.68
bw_mmap_rd-o2c-1mb (MB/s) 731.93
bw_pipe (MB/s) 791.41
bw_unix (MB/s) 1158.39
lat_connect (us) 56.82
lat_ctx-2-128k (us) 7.87
lat_ctx-2-256k (us) 8.59
lat_ctx-4-128k (us) 6.76
lat_ctx-4-256k (us) 9.67
lat_fs-0k (num_files) 249.00
lat_fs-10k (num_files) 119.00
lat_fs-1k (num_files) 153.00
lat_fs-4k (num_files) 146.00
lat_mem_rd-stride128-sz1000k (ns) 30.87
lat_mem_rd-stride128-sz125k (ns) 5.59
lat_mem_rd-stride128-sz250k (ns) 5.84
lat_mem_rd-stride128-sz31k (ns) 4.19
lat_mem_rd-stride128-sz50 (ns) 2.15
lat_mem_rd-stride128-sz500k (ns) 11.18
lat_mem_rd-stride128-sz62k (ns) 4.53
lat_mmap-1m (us) 50.00
lat_ops-double-add (ns) 2.86
lat_ops-double-div (ns) 15.74
lat_ops-double-mul (ns) 2.86
lat_ops-float-add (ns) 2.86
lat_ops-float-div (ns) 9.30
lat_ops-float-mul (ns) 2.86
lat_ops-int-add (ns) 0.72
lat_ops-int-bit (ns) 0.48
lat_ops-int-div (ns) 4.29
lat_ops-int-mod (ns) 4.53
lat_ops-int-mul (ns) 3.08
lat_ops-int64-add (ns) 0.72
lat_ops-int64-bit (ns) 0.48
lat_ops-int64-div (ns) 6.81
lat_ops-int64-mod (ns) 5.25
lat_ops-int64-mul (ns) 3.55
lat_pagefault (us) 1.06
lat_pipe (us) 25.31
lat_proc-exec (us) 693.38
lat_proc-fork (us) 585.00
lat_proc-proccall (us) 0.01
lat_select (us) 34.04
lat_sem (us) 2.90
lat_sig-catch (us) 5.55
lat_sig-install (us) 0.68
lat_sig-prot (us) 0.85
lat_syscall-fstat (us) 1.96
lat_syscall-null (us) 0.46
lat_syscall-open (us) 212.00
lat_syscall-read (us) 0.80
lat_syscall-stat (us) 4.80
lat_syscall-write (us) 0.76
lat_tcp (us) 0.91
lat_unix (us) 28.80
latency_for_0.50_mb_block_size (nanosec) 11.18
latency_for_1.00_mb_block_size (nanosec) 15.44 (min 0.00, max 30.87)
pipe_bandwidth (MBs) 791.41
pipe_latency (microsec) 25.31
procedure_call (microsec) 0.01
select_on_200_tcp_fds (microsec) 34.04
semaphore_latency (microsec) 2.90
signal_handler_latency (microsec) 0.68
signal_handler_overhead (microsec) 5.55
tcp_ip_connection_cost_to_localhost (microsec) 56.82
tcp_latency_using_localhost (microsec) 0.91

Dhrystone

Dhrystone is a core only benchmark that runs from warm L1 caches in all modern processors. It scales linearly with clock speed.

Please take note, different run may produce different slightly results. This is advised to run this test multiple times in order to get maximum performance numbers.

Execute the benchmark with the following:

runDhrystone
Dhrystone Benchmarks
Benchmarks am62axx_sk-fs: perf
cpu_clock (MHz) 1400.00
dhrystone_per_mhz (DMIPS/MHz) 2.90
dhrystone_per_second (DhrystoneP) 7142857.00

Whetstone

Whetstone is a benchmark primarily measuring floating-point arithmetic performance.

Execute the benchmark with the following:

runWhetstone
Whetstone Benchmarks
Benchmarks am62axx_sk-fs: perf
whetstone (MIPS) 5000.00

Linpack

Linpack measures peak double precision (64 bit) floating point performance in solving a dense linear system.

Linpack Benchmarks
Benchmarks am62axx_sk-fs: perf
linpack (Kflops) 577662.00

Stream

STREAM is a microbenchmark for measuring data memory system performance without any data reuse. It is designed to miss on caches and exercise data prefetcher and speculative accesses. It uses double precision floating point (64bit) but in most modern processors the memory access will be the bottleneck. The four individual scores are copy, scale as in multiply by constant, add two numbers, and triad for multiply accumulate. For bandwidth, a byte read counts as one and a byte written counts as one, resulting in a score that is double the bandwidth LMBench will show.

Execute the benchmark with the following:

stream_c
Stream Benchmarks
Benchmarks am62axx_sk-fs: perf
add (MB/s) 2638.50
copy (MB/s) 3791.10
scale (MB/s) 3538.50
triad (MB/s) 2426.20

CoreMarkPro

CoreMark®-Pro is a comprehensive, advanced processor benchmark that works with and enhances the market-proven industry-standard EEMBC CoreMark® benchmark. While CoreMark stresses the CPU pipeline, CoreMark-Pro tests the entire processor, adding comprehensive support for multicore technology, a combination of integer and floating-point workloads, and data sets for utilizing larger memory subsystems.

CoreMarkPro Benchmarks
Benchmarks am62axx_sk-fs: perf
cjpeg-rose7-preset (workloads/) 41.67
core (workloads/) 0.30
coremark-pro () 940.71
linear_alg-mid-100x100-sp (workloads/) 14.68
loops-all-mid-10k-sp (workloads/) 0.71
nnet_test (workloads/) 1.09
parser-125k (workloads/) 8.77
radix2-big-64k (workloads/) 72.61
sha-test (workloads/) 81.30
zip-test (workloads/) 22.22
CoreMarkProFourCore Benchmarks
Benchmarks am62axx_sk-fs: perf
cjpeg-rose7-preset (workloads/) 161.29
core (workloads/) 1.20
coremark-pro () 2630.14
linear_alg-mid-100x100-sp (workloads/) 56.37
loops-all-mid-10k-sp (workloads/) 2.07
nnet_test (workloads/) 3.62
parser-125k (workloads/) 9.76
radix2-big-64k (workloads/) 100.84
sha-test (workloads/) 270.27
zip-test (workloads/) 78.43

MultiBench

MultiBench™ is a suite of benchmarks that allows processor and system designers to analyze, test, and improve multicore processors. It uses three forms of concurrency: Data decomposition: multiple threads cooperating on achieving a unified goal and demonstrating a processor’s support for fine grain parallelism. Processing multiple data streams: uses common code running over multiple threads and demonstrating how well a processor scales over scalable data inputs. Multiple workload processing: shows the scalability of general-purpose processing, demonstrating concurrency over both code and data. MultiBench combines a wide variety of application-specific workloads with the EEMBC Multi-Instance-Test Harness (MITH), compatible and portable with most any multicore processors and operating systems. MITH uses a thread-based API (POSIX-compliant) to establish a common programming model that communicates with the benchmark through an abstraction layer and provides a flexible interface to allow a wide variety of thread-enabled workloads to be tested.

Multibench Benchmarks
Benchmarks am62axx_sk-fs: perf
4m-check (workloads/) 418.20
4m-check-reassembly (workloads/) 124.07
4m-check-reassembly-tcp (workloads/) 60.24
4m-check-reassembly-tcp-cmykw2-rotatew2 (workloads/) 34.25
4m-check-reassembly-tcp-x264w2 (workloads/) 1.89
4m-cmykw2 (workloads/) 248.76
4m-cmykw2-rotatew2 (workloads/) 50.76
4m-reassembly (workloads/) 83.19
4m-rotatew2 (workloads/) 56.50
4m-tcp-mixed (workloads/) 117.65
4m-x264w2 (workloads/) 1.97
idct-4m (workloads/) 19.25
idct-4mw1 (workloads/) 19.25
ippktcheck-4m (workloads/) 423.44
ippktcheck-4mw1 (workloads/) 421.30
ipres-4m (workloads/) 112.02
ipres-4mw1 (workloads/) 113.12
md5-4m (workloads/) 28.39
md5-4mw1 (workloads/) 28.29
rgbcmyk-4m (workloads/) 65.66
rgbcmyk-4mw1 (workloads/) 65.66
rotate-4ms1 (workloads/) 23.60
rotate-4ms1w1 (workloads/) 23.54
rotate-4ms64 (workloads/) 23.80
rotate-4ms64w1 (workloads/) 23.84
x264-4mq (workloads/) 0.58
x264-4mqw1 (workloads/) 0.58

Boot-time Measurement

Boot media: MMCSD

Linux boot time MMCSD
Boot Configuration am62axx_sk-fs: Boot time in seconds: avg(min,max)
Linux boot time from SD with default rootfs (20 boot cycles) 13.68 (min 13.50, max 13.97)

Boot time numbers [avg, min, max] are measured from "Starting kernel" to Linux prompt across 20 boot cycles.


Ethernet

Ethernet performance benchmarks were measured using Netperf 2.7.1 https://hewlettpackard.github.io/netperf/doc/netperf.html Test procedures were modeled after those defined in RFC-2544: https://tools.ietf.org/html/rfc2544, where the DUT is the TI device and the "tester" used was a Linux PC. To produce consistent results, it is recommended to carry out performance tests in a private network and to avoid running NFS on the same interface used in the test. In these results, CPU utilization was captured as the total percentage used across all cores on the device, while running the performance test over one external interface.

UDP Throughput (0% loss) was measured by the procedure defined in RFC-2544 section 26.1: Throughput. In this scenario, netperf options burst_size (-b) and wait_time (-w) are used to limit bandwidth during different trials of the test, with the goal of finding the highest rate at which no loss is seen. For example, to limit bandwidth to 500Mbits/sec with 1472B datagram:

burst_size = <bandwidth (bits/sec)> / 8 (bits -> bytes) / <UDP datagram size> / 100 (seconds -> 10 ms)
burst_size = 500000000 / 8 / 1472 / 100 = 425

wait_time = 10 milliseconds (minimum supported by Linux PC used for testing)

UDP Throughput (possible loss) was measured by capturing throughput and packet loss statistics when running the netperf test with no bandwidth limit (remove -b/-w options).

In order to start a netperf client on one device, the other device must have netserver running. To start netserver:

netserver [-p <port_number>] [-4 (IPv4 addressing)] [-6 (IPv6 addressing)]

Running the following shell script from the DUT will trigger netperf clients to measure bidirectional TCP performance for 60 seconds and report CPU utilization. Parameter -k is used in client commands to summarize selected statistics on their own line and -j is used to gain additional timing measurements during the test.

#!/bin/bash
for i in 1
do
   netperf -H <tester ip> -j -c -l 60 -t TCP_STREAM --
      -k DIRECTION,THROUGHPUT,MEAN_LATENCY,LOCAL_CPU_UTIL,REMOTE_CPU_UTIL,LOCAL_BYTES_SENT,REMOTE_BYTES_RECVD,LOCAL_SEND_SIZE &

   netperf -H <tester ip> -j -c -l 60 -t TCP_MAERTS --
      -k DIRECTION,THROUGHPUT,MEAN_LATENCY,LOCAL_CPU_UTIL,REMOTE_CPU_UTIL,LOCAL_BYTES_SENT,REMOTE_BYTES_RECVD,LOCAL_SEND_SIZE &
done

Running the following commands will trigger netperf clients to measure UDP burst performance for 60 seconds at various burst/datagram sizes and report CPU utilization.

  • For UDP egress tests, run netperf client from DUT and start netserver on tester.
netperf -H <tester ip> -j -c -l 60 -t UDP_STREAM -b <burst_size> -w <wait_time> -- -m <UDP datagram size>
   -k DIRECTION,THROUGHPUT,MEAN_LATENCY,LOCAL_CPU_UTIL,REMOTE_CPU_UTIL,LOCAL_BYTES_SENT,REMOTE_BYTES_RECVD,LOCAL_SEND_SIZE
  • For UDP ingress tests, run netperf client from tester and start netserver on DUT.
netperf -H <DUT ip> -j -C -l 60 -t UDP_STREAM -b <burst_size> -w <wait_time> -- -m <UDP datagram size>
   -k DIRECTION,THROUGHPUT,MEAN_LATENCY,LOCAL_CPU_UTIL,REMOTE_CPU_UTIL,LOCAL_BYTES_SENT,REMOTE_BYTES_RECVD,LOCAL_SEND_SIZE

CPSW/CPSW2g/CPSW3g Ethernet Driver

  • CPSW3g: AM62ax

TCP Bidirectional Throughput

CPSW2g TCP Bidirectional Throughput
Command Used am62axx_sk-fs: THROUGHPUT (Mbits/sec) am62axx_sk-fs: CPU Load % (LOCAL_CPU_UTIL)
netperf -H 192.168.0.1 -j -c -C -l 60 -t TCP_STREAM; netperf -H 192.168.0.1 -j -c -C -l 60 -t TCP_MAERTS 1855.56 65.98

TCP Bidirectional Throughput Interrupt Pacing

CPSW2g TCP Bidirectional Throughput Interrupt Pacing
Command Used am62axx_sk-fs: THROUGHPUT (Mbits/sec) am62axx_sk-fs: CPU Load % (LOCAL_CPU_UTIL)
netperf -H 192.168.0.1 -j -c -C -l 60 -t TCP_STREAM; netperf -H 192.168.0.1 -j -c -C -l 60 -t TCP_MAERTS 1876.44 38.25

UDP Throughput

CPSW2g UDP Egress Throughput 0 loss
Frame Size(bytes) am62axx_sk-fs: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE) am62axx_sk-fs: THROUGHPUT (Mbits/sec) am62axx_sk-fs: Packets Per Second (kPPS) am62axx_sk-fs: CPU Load % (LOCAL_CPU_UTIL)
64 18.00 15.73 109.00 40.72
128 82.00 69.74 106.00 40.24
256 210.00 177.12 105.00 40.18
1024 978.00 800.79 102.00 39.91
1518 1472.00 955.95 81.00 35.45
CPSW2g UDP Ingress Throughput 0 loss
Frame Size(bytes) am62axx_sk-fs: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE) am62axx_sk-fs: THROUGHPUT (Mbits/sec) am62axx_sk-fs: Packets Per Second (kPPS) am62axx_sk-fs: CPU Load % (LOCAL_CPU_UTIL)
64 18.00 1.02 7.00 1.11
128 82.00 5.25 8.00 3.50
256 210.00 13.10 8.00 2.10
1024 978.00 102.49 13.00 7.14
1518 1472.00 100.09 8.00 3.16
CPSW2g UDP Ingress Throughput possible loss
Frame Size(bytes) am62axx_sk-fs: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE) am62axx_sk-fs: THROUGHPUT (Mbits/sec) am62axx_sk-fs: Packets Per Second (kPPS) am62axx_sk-fs: CPU Load % (LOCAL_CPU_UTIL) am62axx_sk-fs: Packet Loss %
64 18.00 33.07 230.00 42.67 50.74
128 82.00 134.23 205.00 45.22 52.22
256 210.00 337.01 201.00 44.35 47.91
1024 978.00 898.26 115.00 41.05 0.26
1518 1472.00 931.76 79.00 40.01 2.10

EMMC Driver

Important

The performance numbers can be severely affected if the media is mounted in sync mode. Hot plug scripts in the filesystem mount removable media in sync mode to ensure data integrity. For performance sensitive applications, umount the auto-mounted filesystem and re-mount in async mode.

EMMC EXT4 FIO 1G

EMMC EXT4 FIO 1G
Buffer size (bytes) am62axx_sk-fs: Write EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Write EXT4 CPU Load (%) am62axx_sk-fs: Read EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Read EXT4 CPU Load (%)
1m 91.70 1.48 172.00 1.54
4m 97.50 1.07 171.00 1.38
4k 64.00 21.81 93.00 22.36
256k 92.20 1.68 171.00 2.20

EMMC RAW FIO 1G

EMMC RAW FIO 1G
Buffer size (bytes) am62axx_sk-fs: Write Raw Throughput (Mbytes/sec) am62axx_sk-fs: Write Raw CPU Load (%) am62axx_sk-fs: Read Raw Throughput (Mbytes/sec) am62axx_sk-fs: Read Raw CPU Load (%)
1m 91.00 1.42 174.00 1.83
4m 97.60 1.02 174.00 1.35
4k 64.10 16.11 92.70 20.72
256k 90.90 1.77 173.00 2.43

EMMC EXT4

EMMC EXT4
Buffer size (bytes) am62axx_sk-fs: Write EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Write EXT4 CPU Load (%) am62axx_sk-fs: Read EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Read EXT4 CPU Load (%)
102400 82.79 (min 80.86, max 84.20) 5.89 (min 5.33, max 7.53) 178.87 8.62
262144 79.25 (min 56.62, max 88.47) 5.91 (min 3.91, max 8.43) 180.55 9.13
524288 79.95 (min 57.79, max 88.21) 5.89 (min 3.99, max 8.38) 181.15 8.70
1048576 74.08 (min 52.01, max 88.83) 5.40 (min 3.35, max 8.03) 181.22 8.26
5242880 75.34 (min 53.85, max 87.76) 5.38 (min 3.47, max 8.02) 181.47 9.48

EMMC VFAT

EMMC VFAT
Buffer size (bytes) am62axx_sk-fs: Write VFAT Throughput (Mbytes/sec) am62axx_sk-fs: Write VFAT CPU Load (%) am62axx_sk-fs: Read VFAT Throughput (Mbytes/sec) am62axx_sk-fs: Read VFAT CPU Load (%)
102400 28.13 (min 3.67, max 34.37) 3.45 (min 2.27, max 3.95) 171.77 10.20
262144 43.39 (min 11.54, max 60.99) 5.67 (min 4.16, max 6.97) 175.90 9.70
524288 41.35 (min 3.90, max 51.60) 4.24 (min 2.41, max 4.93) 171.23 9.35
1048576 42.74 (min 3.97, max 53.26) 4.21 (min 2.26, max 4.84) 171.45 9.47
5242880 42.69 (min 4.00, max 52.77) 3.86 (min 2.30, max 4.35) 171.53 9.02

UBoot EMMC Driver

UBOOT EMMC RAW
File size (bytes in hex) am62axx_sk-fs: Write Throughput (Kbytes/sec) am62axx_sk-fs: Read Throughput (Kbytes/sec)
2000000 59362.32 168907.22
4000000 59850.23 172918.21

MMCSD

Important

The performance numbers can be severely affected if the media is mounted in sync mode. Hot plug scripts in the filesystem mount removable media in sync mode to ensure data integrity. For performance sensitive applications, umount the auto-mounted filesystem and re-mount in async mode.

MMC EXT4 FIO 1G

MMC EXT4 FIO 1G
Buffer size (bytes) am62axx_sk-fs: Write EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Write EXT4 CPU Load (%) am62axx_sk-fs: Read EXT4 Throughput (Mbytes/sec) am62axx_sk-fs: Read EXT4 CPU Load (%)
1m 42.20 1.02 87.40 1.26
4m 43.20 0.75 87.20 1.06
4k 2.84 1.67 12.90 4.27
256k 39.20 1.31 83.60 1.57

MMC RAW FIO 1G

MMC RAW FIO 1G
Buffer size (bytes) am62axx_sk-fs: Write Raw Throughput (Mbytes/sec) am62axx_sk-fs: Write Raw CPU Load (%) am62axx_sk-fs: Read Raw Throughput (Mbytes/sec) am62axx_sk-fs: Read Raw CPU Load (%)
1m 43.00 0.84 88.10 1.14
4m 43.00 0.64 88.00 0.89
4k 2.81 1.40 13.00 3.90
256k 36.80 0.94 84.20 1.38

MMC EXT4

MMC EXT4
Buffer size (bytes) am62axx_sk-fs: Write Raw Throughput (Mbytes/sec) am62axx_sk-fs: Write Raw CPU Load (%) am62axx_sk-fs: Read Raw Throughput (Mbytes/sec) am62axx_sk-fs: Read Raw CPU Load (%)
102400 10.40 (min 10.24, max 10.69) 0.83 (min 0.72, max 1.12) 10.80 0.67
262144 10.37 (min 10.24, max 10.50) 0.91 (min 0.64, max 1.48) 10.99 0.79
524288 10.47 (min 10.29, max 10.56) 0.81 (min 0.71, max 1.13) 11.05 0.66
1048576 10.38 (min 10.26, max 10.59) 0.92 (min 0.76, max 1.19) 10.79 0.77
5242880 10.79 (min 10.66, max 10.92) 0.83 (min 0.73, max 1.04) 12.03 0.75

The performance numbers were captured using the following:

  • SanDisk Max Endurance SD card (SDSQQVR-032G-GN6IA)
  • Partition was mounted with async option

UBoot MMCSD

UBOOT MMCSD FAT

UBOOT MMCSD FAT
File size (bytes in hex) am62axx_sk-fs: Write Throughput (Kbytes/sec) am62axx_sk-fs: Read Throughput (Kbytes/sec)
400000 38641.51 81920.00
800000 45765.36 87148.94
1000000 48046.92 90021.98

The performance numbers were captured using the following:

  • SanDisk Max Endurance SD card (SDSQQVR-032G-GN6IA)

USB Driver

USB Device Controller

USBDEVICE HIGHSPEED SLAVE_READ_THROUGHPUT
Number of Blocks am62axx_sk-fs: Throughput (MB/sec)
150 43.80
USBDEVICE HIGHSPEED SLAVE_WRITE_THROUGHPUT
Number of Blocks am62axx_sk-fs: Throughput (MB/sec)
150 36.10

CRYPTO Driver

OpenSSL Performance

OpenSSL Performance
Algorithm Buffer Size (in bytes) am62axx_sk-fs: throughput (KBytes/Sec)
aes-128-cbc 1024 22690.82
aes-128-cbc 16 407.87
aes-128-cbc 16384 83771.39
aes-128-cbc 256 6693.03
aes-128-cbc 64 1769.71
aes-128-cbc 8192 70991.87
aes-128-ecb 1024 23238.66
aes-128-ecb 16 415.95
aes-128-ecb 16384 87064.58
aes-128-ecb 256 6831.70
aes-128-ecb 64 1811.97
aes-128-ecb 8192 72884.22
aes-192-cbc 1024 22330.03
aes-192-cbc 16 412.24
aes-192-cbc 16384 76540.59
aes-192-cbc 256 6668.46
aes-192-cbc 64 1779.07
aes-192-cbc 8192 65191.94
aes-192-ecb 1024 22708.57
aes-192-ecb 16 415.37
aes-192-ecb 16384 78255.45
aes-192-ecb 256 6798.68
aes-192-ecb 64 1813.08
aes-192-ecb 8192 67029.67
aes-256-cbc 1024 21670.23
aes-256-cbc 16 405.98
aes-256-cbc 16384 70068.91
aes-256-cbc 256 6556.16
aes-256-cbc 64 1761.22
aes-256-cbc 8192 60661.76
aes-256-ecb 1024 22268.25
aes-256-ecb 16 415.66
aes-256-ecb 16384 71734.61
aes-256-ecb 256 6744.66
aes-256-ecb 64 1808.04
aes-256-ecb 8192 60866.56
sha256 1024 34747.73
sha256 16 571.94
sha256 16384 286829.23
sha256 256 9044.22
sha256 64 2270.38
sha256 8192 193448.62
sha512 1024 24330.24
sha512 16 566.06
sha512 16384 67687.77
sha512 256 8116.39
sha512 64 2266.01
sha512 8192 60424.19
OpenSSL CPU Load
Algorithm am62axx_sk-fs: CPU Load
aes-128-cbc 30.00
aes-128-ecb 32.00
aes-192-cbc 30.00
aes-192-ecb 31.00
aes-256-cbc 30.00
aes-256-ecb 30.00
sha256 96.00
sha512 96.00

Listed for each algorithm are the code snippets used to run each benchmark test.

time -v openssl speed -elapsed -evp aes-128-cbc

Low Power Performance

Power Performance

Deep Sleep Power Performance
Rail name Rail voltage(V) Power (mW)
vdd_core 0.85 19.16
vddr_core 0.85 2.34
soc_dvdd_3v3 3.30 4.63
soc_dvdd_1v8 1.80 2.03
vdda_1v8 1.80 11.22
vdd_lpddr4 1.10 3.59
Total   42.96
MCU Only Power Performance
Rail name Rail voltage(V) Power (mW)
vdd_core 0.85 199.73
vddr_core 0.85 3.28
soc_dvdd_3v3 3.30 10.49
soc_dvdd_1v8 1.80 1.80
vdda_1v8 1.80 19.42
vdd_lpddr4 1.10 4.41
Total   239.14

Partial I/O Data - All voltage rails were measured to be near 0V

Note

The measurements shown are from an AM62Ax rev E3. Results may vary based off of the board revision being used.

Further optimizations are possible for these low power modes. Please refer to the AM62x Power Consumption App Note (https://www.ti.com/lit/pdf/spradg1)

Resume Latency Performance

LPM Resume Latency Performance
Low Power Mode Total Resume Latency (ms)
Deep Sleep 146.45
MCU Only 110.40

The performance numbers are measured without the Linux printk logs. To remove the Linux printk logs, run the following commands in the terminal:

# Detach kernel serial console
consoles=$(find /sys/bus/platform/devices/*.serial/ -name console)
for console in ${consoles}; do
     echo -n N > ${console}
done

Note

The measurements shown are from using the default SDK with no extra optimizations.