2.2. Performance Guide

2.2.1. RT Kernel Performance Guide

Read This First

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

Name Description
AM65x EVM AM65x Evaluation Module rev 1.0 with ARM running at 800MHz, DDR4-2400 (1600 MT/S), TMDX654GPEVM

Table: Evaluation Modules

About This Manual

This document provides performance data for each of the device drivers which are part of the Process SDK Linux package. This document should be used in conjunction with release notes and user guides provided with the Process 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 http://e2e.ti.com/ or http://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 http://lmbench.sourceforge.net/whatis_lmbench.html and http://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 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.

Benchmarks am654x-evm: perf
af_unix_sock_stream_latency (microsec) 60.48
af_unix_socket_stream_bandwidth (MBs) 1107.59
bw_file_rd-io-1mb (MB/s) 975.61
bw_file_rd-o2c-1mb (MB/s) 502.34
bw_mem-bcopy-16mb (MB/s) 878.4
bw_mem-bcopy-1mb (MB/s) 1051.89
bw_mem-bcopy-2mb (MB/s) 880.67
bw_mem-bcopy-4mb (MB/s) 874.41
bw_mem-bcopy-8mb (MB/s) 865.99
bw_mem-bzero-16mb (MB/s) 1648.8
bw_mem-bzero-1mb (MB/s) 2734.88 (min 1051.89, max 4417.86)
bw_mem-bzero-2mb (MB/s) 1605.06 (min 880.67, max 2329.45)
bw_mem-bzero-4mb (MB/s) 1286.46 (min 874.41, max 1698.51)
bw_mem-bzero-8mb (MB/s) 1254.52 (min 865.99, max 1643.05)
bw_mem-cp-16mb (MB/s) 590.17
bw_mem-cp-1mb (MB/s) 2522.16 (min 675.11, max 4369.20)
bw_mem-cp-2mb (MB/s) 1460.09 (min 591.89, max 2328.29)
bw_mem-cp-4mb (MB/s) 1143.89 (min 586.85, max 1700.92)
bw_mem-cp-8mb (MB/s) 1117.19 (min 578.91, max 1655.46)
bw_mem-fcp-16mb (MB/s) 822.24
bw_mem-fcp-1mb (MB/s) 2690.44 (min 963.02, max 4417.86)
bw_mem-fcp-2mb (MB/s) 1578.75 (min 828.04, max 2329.45)
bw_mem-fcp-4mb (MB/s) 1258.84 (min 819.17, max 1698.51)
bw_mem-fcp-8mb (MB/s) 1232.33 (min 821.61, max 1643.05)
bw_mem-frd-16mb (MB/s) 1275.51
bw_mem-frd-1mb (MB/s) 1271.68 (min 963.02, max 1580.33)
bw_mem-frd-2mb (MB/s) 1120.11 (min 828.04, max 1412.18)
bw_mem-frd-4mb (MB/s) 1054.96 (min 819.17, max 1290.74)
bw_mem-frd-8mb (MB/s) 1049.58 (min 821.61, max 1277.55)
bw_mem-fwr-16mb (MB/s) 1646.6
bw_mem-fwr-1mb (MB/s) 2974.77 (min 1580.33, max 4369.20)
bw_mem-fwr-2mb (MB/s) 1870.24 (min 1412.18, max 2328.29)
bw_mem-fwr-4mb (MB/s) 1495.83 (min 1290.74, max 1700.92)
bw_mem-fwr-8mb (MB/s) 1466.51 (min 1277.55, max 1655.46)
bw_mem-rd-16mb (MB/s) 1297.54
bw_mem-rd-1mb (MB/s) 3319.10 (min 3250.27, max 3387.92)
bw_mem-rd-2mb (MB/s) 1174.07 (min 910.33, max 1437.81)
bw_mem-rd-4mb (MB/s) 1032.56 (min 753.86, max 1311.26)
bw_mem-rd-8mb (MB/s) 1021.93 (min 741.77, max 1302.08)
bw_mem-rdwr-16mb (MB/s) 730.43
bw_mem-rdwr-1mb (MB/s) 1861.95 (min 675.11, max 3048.78)
bw_mem-rdwr-2mb (MB/s) 740.19 (min 591.89, max 888.49)
bw_mem-rdwr-4mb (MB/s) 661.28 (min 586.85, max 735.70)
bw_mem-rdwr-8mb (MB/s) 653.29 (min 578.91, max 727.67)
bw_mem-wr-16mb (MB/s) 742.94
bw_mem-wr-1mb (MB/s) 3149.53 (min 3048.78, max 3250.27)
bw_mem-wr-2mb (MB/s) 899.41 (min 888.49, max 910.33)
bw_mem-wr-4mb (MB/s) 744.78 (min 735.70, max 753.86)
bw_mem-wr-8mb (MB/s) 734.72 (min 727.67, max 741.77)
bw_mmap_rd-mo-1mb (MB/s) 2662.3
bw_mmap_rd-o2c-1mb (MB/s) 529.94
bw_pipe (MB/s) 318.65
bw_unix (MB/s) 1107.59
lat_connect (us) 143.08
lat_ctx-2-128k (us) 8.38
lat_ctx-2-256k (us) 6.13
lat_ctx-4-128k (us) 7.94
lat_ctx-4-256k (us) 3.82
lat_fs-0k (num_files) 136
lat_fs-10k (num_files) 61
lat_fs-1k (num_files) 109
lat_fs-4k (num_files) 107
lat_mem_rd-stride128-sz1000k (ns) 24.99
lat_mem_rd-stride128-sz125k (ns) 9.88
lat_mem_rd-stride128-sz250k (ns) 10.21
lat_mem_rd-stride128-sz31k (ns) 6.39
lat_mem_rd-stride128-sz50 (ns) 3.76
lat_mem_rd-stride128-sz500k (ns) 11.82
lat_mem_rd-stride128-sz62k (ns) 9.16
lat_mmap-1m (us) 81
lat_ops-double-add (ns) 0.91
lat_ops-double-mul (ns) 5.04
lat_ops-float-add (ns) 0.91
lat_ops-float-mul (ns) 5.03
lat_ops-int-add (ns) 1.26
lat_ops-int-bit (ns) 0.84
lat_ops-int-div (ns) 7.52
lat_ops-int-mod (ns) 7.98
lat_ops-int-mul (ns) 3.81
lat_ops-int64-add (ns) 1.27
lat_ops-int64-bit (ns) 0.84
lat_ops-int64-div (ns) 11.96
lat_ops-int64-mod (ns) 9.23
lat_pagefault (us) 1.93
lat_pipe (us) 33.24
lat_proc-exec (us) 1703.25
lat_proc-fork (us) 1468.25
lat_proc-proccall (us) 0.01
lat_select (us) 55.76
lat_sem (us) 8.18
lat_sig-catch (us) 9.92
lat_sig-install (us) 1.02
lat_sig-prot (us) 0.69
lat_syscall-fstat (us) 2.44
lat_syscall-null (us) 0.5
lat_syscall-open (us) 317.94
lat_syscall-read (us) 0.98
lat_syscall-stat (us) 6.83
lat_syscall-write (us) 0.85
lat_tcp (us) 1.05
lat_unix (us) 60.48
latency_for_0.50_mb_block_size (nanosec) 11.82
latency_for_1.00_mb_block_size (nanosec) 12.49 (min 0.00, max 24.99)
pipe_bandwidth (MBs) 318.65
pipe_latency (microsec) 33.24
procedure_call (microsec) 0.01
select_on_200_tcp_fds (microsec) 55.76
semaphore_latency (microsec) 8.18
signal_handler_latency (microsec) 1.02
signal_handler_overhead (microsec) 9.92
tcp_ip_connection_cost_to_localhost (microsec) 143.08
tcp_latency_using_localhost (microsec) 1.05

Table: LM Bench Metrics Dhrystone

Dhrystone is a core only benchmark that runs from warm L1 caches in all modern processors. It scales linearly with clock speed. For standard ARM cores the DMIPS/MHz score will be identical with the same compiler and flags.

Benchmarks am654x-evm: perf
cpu_clock (MHz) 800.00
dhrystone_per_mhz (DMIPS/MHz) 2.96
dhrystone_per_second (DhrystoneP) 4166666.80

Table: Dhrystone Benchmark Whetstone

Benchmarks am654x-evm: perf
whetstone (MIPS) 3333.30

Table: Whetstone Benchmark Linpack

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

Benchmarks am654x-evm: perf
linpack (Kflops) 326564.00

Table: Linpack Benchmark NBench

NBench which stands for Native Benchmark is used to measure macro benchmarks for commonly used operations such as sorting and analysis algorithms. More information about NBench at https://en.wikipedia.org/wiki/NBench and https://nbench.io/articles/index.html

Benchmarks am654x-evm: perf
assignment (Iterations) 7.82
fourier (Iterations) 12991.00
fp_emulation (Iterations) 61.24
huffman (Iterations) 670.09
idea (Iterations) 1962.70
lu_decomposition (Iterations) 313.09
neural_net (Iterations) 4.48
numeric_sort (Iterations) 283.49
string_sort (Iterations) 94.73

Table: NBench Benchmarks 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.

Benchmarks am654x-evm: perf  
add (MB/s) 2506.20 1625.80
copy (MB/s) 3755.90 1783.80
scale (MB/s) 3313.10 1782.20
triad (MB/s) 2283.10 1521.80

Table: Stream 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.

Benchmarks am654x-evm: perf
4m-check (workloads/) 356.68
4m-check-reassembly (workloads/) 69.2
4m-check-reassembly-tcp (workloads/) 35.77
4m-check-reassembly-tcp-cmykw2-rotatew2 (workloads/) 9.53
4m-check-reassembly-tcp-x264w2 (workloads/) 1.13
4m-cmykw2 (workloads/) 138.03
4m-cmykw2-rotatew2 (workloads/) 9.69
4m-reassembly (workloads/) 48.08
4m-rotatew2 (workloads/) 13.07
4m-tcp-mixed (workloads/) 70.48
4m-x264w2 (workloads/) 1.16
empty-wld (workloads/) 1
idct-4m (workloads/) 10.97
idct-4mw1 (workloads/) 10.99
ippktcheck-4m (workloads/) 356.63
ippktcheck-4mw1 (workloads/) 355.82
ipres-4m (workloads/) 62.89
ipres-4mw1 (workloads/) 62.47
md5-4m (workloads/) 15.04
md5-4mw1 (workloads/) 15.07
rgbcmyk-4m (workloads/) 36.54
rgbcmyk-4mw1 (workloads/) 36.42
rotate-4ms1 (workloads/) 14.51
rotate-4ms1w1 (workloads/) 14.32
rotate-4ms64 (workloads/) 14.52
rotate-4ms64w1 (workloads/) 14.53
x264-4mq (workloads/) 0.33
x264-4mqw1 (workloads/) 0.33

Table: MultiBench Boot-time Measurement Boot media: MMCSD

Boot Configuration am654x-evm: boot time (sec)
Kernel boot time test when bootloader, kernel and sdk-rootfs are in mmc-sd 31.58 (min 31.14, max 32.10)
Kernel boot time test when init is /bin/sh and bootloader, kernel and sdk-rootfs are in mmc-sd 5.88

Table: Boot time MMC/SD ALSA SoC Audio Driver

  1. Access type - RW_INTERLEAVED
  2. Channels - 2
  3. Format - S16_LE
  4. Period size - 64
Sampling Rate (Hz) am654x-evm: Throughput (bits/sec) am654x-evm: CPU Load (%)
8000 255999.00 0.45
11025 352798.00 0.61
16000 511997.00 0.49
22050 705596.00 0.86
24000 705596.00 0.80
32000 1023993.00 0.62
44100 1411190.00 1.18
48000 1535989.00 0.73
88200 2822376.00 1.91
96000 3071973.00 6.12

Table: Audio Capture Graphics SGX/RGX Driver Glmark2

Run Glmark2 and capture performance reported (Score). All display outputs (HDMI, Displayport and/or LCD) are connected when running these tests

Benchmark am654x-evm: Score
Glmark2-Wayland 192.00

Table: Glmark2 Sensor Capture

Performance and Benchmarks not available in this release. MMC/SD 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. AM654x-EVM

Buffer size (bytes) am654x-evm: Write EXT4 Throughput (Mbytes/sec) am654x-evm: Write EXT4 CPU Load (%) am654x-evm: Read EXT4 Throughput (Mbytes/sec) am654x-evm: Read EXT4 CPU Load (%)
1m 28.60 1.37 85.40 1.84
4m 28.60 1.18 85.40 1.66
4k 3.22 3.81 12.40 9.55
256k 24.90 1.57 83.70 2.51

The performance numbers were captured using the following:

  • SanDisk 8GB MicroSDHC Class 10 Memory Card
  • Partition was mounted with async option UART Driver

Performance and Benchmarks not available in this release. I2C Driver

Performance and Benchmarks not available in this release. EDMA Driver

Performance and Benchmarks not available in this release. Touchscreen Driver

Performance and Benchmarks not available in this release. CRYPTO Driver

Performance and Benchmarks not available in this release. IPSec Software Performance

Algorithm am654x-evm: Throughput (Mbps) am654x-evm: Packets/Sec am654x-evm: CPU Load
3des 52.40 4.00 59.49
aes256 91.30 8.00 36.35 DCAN Driver

Performance and Benchmarks not available in this release.