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/

2.2.1.1. System Benchmarks¶

2.2.1.1.1. 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)	62.08
af_unix_socket_stream_bandwidth (MBs)	1105.78
bw_file_rd-io-1mb (MB/s)	1001.67
bw_file_rd-o2c-1mb (MB/s)	559.81
bw_mem-bcopy-16mb (MB/s)	880.91
bw_mem-bcopy-1mb (MB/s)	1054.48
bw_mem-bcopy-2mb (MB/s)	883.00
bw_mem-bcopy-4mb (MB/s)	878.16
bw_mem-bcopy-8mb (MB/s)	884.76
bw_mem-bzero-16mb (MB/s)	1645.08
bw_mem-bzero-1mb (MB/s)	2764.18 (min 1054.48, max 4473.87)
bw_mem-bzero-2mb (MB/s)	1603.91 (min 883.00, max 2324.81)
bw_mem-bzero-4mb (MB/s)	1290.75 (min 878.16, max 1703.34)
bw_mem-bzero-8mb (MB/s)	1267.04 (min 884.76, max 1649.31)
bw_mem-cp-16mb (MB/s)	584.99
bw_mem-cp-1mb (MB/s)	2584.51 (min 672.95, max 4496.07)
bw_mem-cp-2mb (MB/s)	1457.14 (min 570.78, max 2343.49)
bw_mem-cp-4mb (MB/s)	1146.70 (min 587.63, max 1705.76)
bw_mem-cp-8mb (MB/s)	1126.56 (min 599.03, max 1654.09)
bw_mem-fcp-16mb (MB/s)	822.58
bw_mem-fcp-1mb (MB/s)	2730.36 (min 986.84, max 4473.87)
bw_mem-fcp-2mb (MB/s)	1576.71 (min 828.61, max 2324.81)
bw_mem-fcp-4mb (MB/s)	1257.56 (min 811.77, max 1703.34)
bw_mem-fcp-8mb (MB/s)	1236.99 (min 824.66, max 1649.31)
bw_mem-frd-16mb (MB/s)	1278.47
bw_mem-frd-1mb (MB/s)	1273.30 (min 986.84, max 1559.76)
bw_mem-frd-2mb (MB/s)	1112.14 (min 828.61, max 1395.67)
bw_mem-frd-4mb (MB/s)	1042.32 (min 811.77, max 1272.87)
bw_mem-frd-8mb (MB/s)	1042.15 (min 824.66, max 1259.64)
bw_mem-fwr-16mb (MB/s)	1655.63
bw_mem-fwr-1mb (MB/s)	3027.92 (min 1559.76, max 4496.07)
bw_mem-fwr-2mb (MB/s)	1869.58 (min 1395.67, max 2343.49)
bw_mem-fwr-4mb (MB/s)	1489.32 (min 1272.87, max 1705.76)
bw_mem-fwr-8mb (MB/s)	1456.87 (min 1259.64, max 1654.09)
bw_mem-rd-16mb (MB/s)	1301.24
bw_mem-rd-1mb (MB/s)	3298.78 (min 3141.75, max 3455.80)
bw_mem-rd-2mb (MB/s)	1164.23 (min 912.27, max 1416.18)
bw_mem-rd-4mb (MB/s)	1033.02 (min 750.89, max 1315.14)
bw_mem-rd-8mb (MB/s)	1022.41 (min 742.74, max 1302.08)
bw_mem-rdwr-16mb (MB/s)	731.19
bw_mem-rdwr-1mb (MB/s)	1768.05 (min 672.95, max 2863.14)
bw_mem-rdwr-2mb (MB/s)	720.81 (min 570.78, max 870.83)
bw_mem-rdwr-4mb (MB/s)	663.98 (min 587.63, max 740.33)
bw_mem-rdwr-8mb (MB/s)	666.96 (min 599.03, max 734.89)
bw_mem-wr-16mb (MB/s)	742.22
bw_mem-wr-1mb (MB/s)	3002.45 (min 2863.14, max 3141.75)
bw_mem-wr-2mb (MB/s)	891.55 (min 870.83, max 912.27)
bw_mem-wr-4mb (MB/s)	745.61 (min 740.33, max 750.89)
bw_mem-wr-8mb (MB/s)	738.82 (min 734.89, max 742.74)
bw_mmap_rd-mo-1mb (MB/s)	2857.65
bw_mmap_rd-o2c-1mb (MB/s)	567.86
bw_pipe (MB/s)	298.19
bw_unix (MB/s)	1105.78
lat_connect (us)	105.25
lat_ctx-2-128k (us)	8.20
lat_ctx-2-256k (us)	6.49
lat_ctx-4-128k (us)	8.36
lat_ctx-4-256k (us)	3.72
lat_fs-0k (num_files)	129.00
lat_fs-10k (num_files)	73.00
lat_fs-1k (num_files)	89.00
lat_fs-4k (num_files)	101.00
lat_mem_rd-stride128-sz1000k (ns)	26.78
lat_mem_rd-stride128-sz125k (ns)	9.72
lat_mem_rd-stride128-sz250k (ns)	10.23
lat_mem_rd-stride128-sz31k (ns)	3.80
lat_mem_rd-stride128-sz50 (ns)	3.76
lat_mem_rd-stride128-sz500k (ns)	11.52
lat_mem_rd-stride128-sz62k (ns)	7.95
lat_mmap-1m (us)	78.00
lat_ops-double-add (ns)	0.91
lat_ops-double-mul (ns)	5.03
lat_ops-float-add (ns)	0.92
lat_ops-float-mul (ns)	5.03
lat_ops-int-add (ns)	1.26
lat_ops-int-bit (ns)	0.83
lat_ops-int-div (ns)	7.53
lat_ops-int-mod (ns)	7.96
lat_ops-int-mul (ns)	3.81
lat_ops-int64-add (ns)	1.26
lat_ops-int64-bit (ns)	0.84
lat_ops-int64-div (ns)	11.98
lat_ops-int64-mod (ns)	9.21
lat_pagefault (us)	1.91
lat_pipe (us)	33.59
lat_proc-exec (us)	1733.25
lat_proc-fork (us)	1484.25
lat_proc-proccall (us)	0.01
lat_select (us)	52.71
lat_sem (us)	8.20
lat_sig-catch (us)	9.89
lat_sig-install (us)	1.02
lat_sig-prot (us)	0.41
lat_syscall-fstat (us)	2.47
lat_syscall-null (us)	0.53
lat_syscall-open (us)	218.84
lat_syscall-read (us)	0.96
lat_syscall-stat (us)	6.53
lat_syscall-write (us)	0.81
lat_tcp (us)	1.00
lat_unix (us)	62.08
latency_for_0.50_mb_block_size (nanosec)	11.52
latency_for_1.00_mb_block_size (nanosec)	13.39 (min 0.00, max 26.78)
pipe_bandwidth (MBs)	298.19
pipe_latency (microsec)	33.59
procedure_call (microsec)	0.01
select_on_200_tcp_fds (microsec)	52.71
semaphore_latency (microsec)	8.20
signal_handler_latency (microsec)	1.02
signal_handler_overhead (microsec)	9.89
tcp_ip_connection_cost_to_localhost (microsec)	105.25
tcp_latency_using_localhost (microsec)	1.00

Table: LM Bench Metrics

2.2.1.1.2. 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

2.2.1.1.3. Whetstone¶

Benchmarks	am654x-evm: perf
whetstone (MIPS)	3333.30

Table: Whetstone Benchmark

2.2.1.1.4. Linpack¶

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

Benchmarks	am654x-evm: perf
linpack (Kflops)	331179.00

Table: Linpack Benchmark

2.2.1.1.5. 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)	7.80
fourier (Iterations)	13066.00
fp_emulation (Iterations)	61.24
huffman (Iterations)	670.69
idea (Iterations)	1962.80
lu_decomposition (Iterations)	313.08
neural_net (Iterations)	4.48
numeric_sort (Iterations)	288.25
string_sort (Iterations)	94.74

Table: NBench Benchmarks

2.2.1.1.6. 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)	1602.90
copy (MB/s)	1747.10
scale (MB/s)	1756.80
triad (MB/s)	1507.20

Table: Stream

2.2.1.2. Boot-time Measurement¶

2.2.1.2.1. Boot media: MMCSD¶

Boot Configuration	am654x-evm: boot time (sec)
Kernel boot time test when init is /bin/sh and bootloader, kernel and sdk-rootfs are in mmc-sd	5.85 (min 5.78, max 6.13)

Table: Boot time MMC/SD

2.2.1.3. ALSA SoC Audio Driver¶

Access type - RW_INTERLEAVED
Channels - 2
Format - S16_LE
Period size - 64

Sampling Rate (Hz)	am654x-evm: Throughput (bits/sec)	am654x-evm: CPU Load (%)
8000	255999.00	0.53
11025	352798.00	0.44
16000	511997.00	0.32
22050	705596.00	0.66
24000	705596.00	0.62
32000	1023993.00	0.43
44100	1411190.00	1.12
48000	1535989.00	0.55
88200	2822375.00	1.90
96000	3071973.00	5.93

Table: Audio Capture

2.2.1.4. Sensor Capture¶

Capture video frames (MMAP buffers) with v4l2c-ctl and record the reported fps

Resolution	Format	am654x-evm: Fps	am654x-evm: Sensor
176x144	uyvy	30.02	ov5640
1920x1080	uyvy	30.00	ov5640

Table: Sensor Capture

2.2.1.5. MMC/SD Driver¶

Warning

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.

2.2.1.5.1. 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	30.90	1.06	85.80	1.56
4m	29.40	0.91	85.50	1.28
4k	3.22	3.51	12.50	9.60
256k	27.90	1.27	84.00	2.02

The performance numbers were captured using the following:

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