2.4.2. Linux Performance Guide

Read This First

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

Name	Description
AM62x EVM	AM62x Starter Kit rev E2 and E3 with ARM running at 1400 MHz, DDR4 data rate 1600 MT/S

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/

2.4.2.1. System Benchmarks

2.4.2.1.1. 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 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	am62xx-sk: perf
af_unix_sock_stream_latency (microsec)	32.39
af_unix_socket_stream_bandwidth (MBs)	1199.02
bw_file_rd-io-1mb (MB/s)	804.76
bw_file_rd-o2c-1mb (MB/s)	480.31
bw_mem-bcopy-16mb (MB/s)	768.38
bw_mem-bcopy-1mb (MB/s)	839.77
bw_mem-bcopy-2mb (MB/s)	759.30
bw_mem-bcopy-4mb (MB/s)	786.32
bw_mem-bcopy-8mb (MB/s)	816.16
bw_mem-bzero-16mb (MB/s)	1736.49
bw_mem-bzero-1mb (MB/s)	1287.94 (min 839.77, max 1736.11)
bw_mem-bzero-2mb (MB/s)	1246.96 (min 759.30, max 1734.61)
bw_mem-bzero-4mb (MB/s)	1260.97 (min 786.32, max 1735.61)
bw_mem-bzero-8mb (MB/s)	1275.86 (min 816.16, max 1735.55)
bw_mem-cp-16mb (MB/s)	441.34
bw_mem-cp-1mb (MB/s)	1172.98 (min 475.74, max 1870.21)
bw_mem-cp-2mb (MB/s)	1074.07 (min 439.90, max 1708.23)
bw_mem-cp-4mb (MB/s)	1130.11 (min 469.70, max 1790.51)
bw_mem-cp-8mb (MB/s)	1122.44 (min 470.45, max 1774.43)
bw_mem-fcp-16mb (MB/s)	687.85
bw_mem-fcp-1mb (MB/s)	1291.64 (min 847.17, max 1736.11)
bw_mem-fcp-2mb (MB/s)	1247.03 (min 759.45, max 1734.61)
bw_mem-fcp-4mb (MB/s)	1240.46 (min 745.30, max 1735.61)
bw_mem-fcp-8mb (MB/s)	1241.43 (min 747.31, max 1735.55)
bw_mem-frd-16mb (MB/s)	1138.22
bw_mem-frd-1mb (MB/s)	1045.48 (min 847.17, max 1243.78)
bw_mem-frd-2mb (MB/s)	934.56 (min 759.45, max 1109.67)
bw_mem-frd-4mb (MB/s)	936.03 (min 745.30, max 1126.76)
bw_mem-frd-8mb (MB/s)	937.51 (min 747.31, max 1127.71)
bw_mem-fwr-16mb (MB/s)	1757.28
bw_mem-fwr-1mb (MB/s)	1557.00 (min 1243.78, max 1870.21)
bw_mem-fwr-2mb (MB/s)	1408.95 (min 1109.67, max 1708.23)
bw_mem-fwr-4mb (MB/s)	1458.64 (min 1126.76, max 1790.51)
bw_mem-fwr-8mb (MB/s)	1451.07 (min 1127.71, max 1774.43)
bw_mem-rd-16mb (MB/s)	1152.24
bw_mem-rd-1mb (MB/s)	993.19 (min 697.47, max 1288.90)
bw_mem-rd-2mb (MB/s)	871.06 (min 619.58, max 1122.54)
bw_mem-rd-4mb (MB/s)	898.79 (min 691.68, max 1105.89)
bw_mem-rd-8mb (MB/s)	960.42 (min 791.37, max 1129.46)
bw_mem-rdwr-16mb (MB/s)	853.65
bw_mem-rdwr-1mb (MB/s)	580.16 (min 475.74, max 684.58)
bw_mem-rdwr-2mb (MB/s)	524.74 (min 439.90, max 609.57)
bw_mem-rdwr-4mb (MB/s)	587.65 (min 469.70, max 705.59)
bw_mem-rdwr-8mb (MB/s)	626.77 (min 470.45, max 783.09)
bw_mem-wr-16mb (MB/s)	815.79
bw_mem-wr-1mb (MB/s)	691.03 (min 684.58, max 697.47)
bw_mem-wr-2mb (MB/s)	614.58 (min 609.57, max 619.58)
bw_mem-wr-4mb (MB/s)	698.64 (min 691.68, max 705.59)
bw_mem-wr-8mb (MB/s)	787.23 (min 783.09, max 791.37)
bw_mmap_rd-mo-1mb (MB/s)	1277.37
bw_mmap_rd-o2c-1mb (MB/s)	504.03
bw_pipe (MB/s)	499.88
bw_unix (MB/s)	1199.02
lat_connect (us)	70.08
lat_ctx-2-128k (us)	4.54
lat_ctx-2-256k (us)	4.16
lat_ctx-4-128k (us)	4.28
lat_ctx-4-256k (us)	3.35
lat_fs-0k (num_files)	276.00
lat_fs-10k (num_files)	114.00
lat_fs-1k (num_files)	194.00
lat_fs-4k (num_files)	198.00
lat_mem_rd-stride128-sz1000k (ns)	52.11
lat_mem_rd-stride128-sz125k (ns)	5.53
lat_mem_rd-stride128-sz250k (ns)	5.83
lat_mem_rd-stride128-sz31k (ns)	3.65
lat_mem_rd-stride128-sz50 (ns)	2.15
lat_mem_rd-stride128-sz500k (ns)	13.10
lat_mem_rd-stride128-sz62k (ns)	5.24
lat_mmap-1m (us)	47.00
lat_ops-double-add (ns)	0.52
lat_ops-double-mul (ns)	2.86
lat_ops-float-add (ns)	0.52
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)	2.17
lat_ops-int64-add (ns)	0.72
lat_ops-int64-bit (ns)	0.48
lat_ops-int64-div (ns)	6.80
lat_ops-int64-mod (ns)	5.24
lat_pagefault (us)	1.24
lat_pipe (us)	17.74
lat_proc-exec (us)	1167.60
lat_proc-fork (us)	923.00
lat_proc-proccall (us)	0.01
lat_select (us)	35.68
lat_sem (us)	1.56
lat_sig-catch (us)	4.45
lat_sig-install (us)	0.47
lat_sig-prot (us)	0.54
lat_syscall-fstat (us)	1.15
lat_syscall-null (us)	0.29
lat_syscall-open (us)	161.65
lat_syscall-read (us)	0.57
lat_syscall-stat (us)	3.16
lat_syscall-write (us)	0.48
lat_tcp (us)	0.59
lat_unix (us)	32.39
latency_for_0.50_mb_block_size (nanosec)	13.10
latency_for_1.00_mb_block_size (nanosec)	26.06 (min 0.00, max 52.11)
pipe_bandwidth (MBs)	499.88
pipe_latency (microsec)	17.74
procedure_call (microsec)	0.01
select_on_200_tcp_fds (microsec)	35.68
semaphore_latency (microsec)	1.56
signal_handler_latency (microsec)	0.47
signal_handler_overhead (microsec)	4.45
tcp_ip_connection_cost_to_localhost (microsec)	70.08
tcp_latency_using_localhost (microsec)	0.59

Table: LM Bench Metrics

2.4.2.1.2. 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.

Execute the benchmark with the following:

runDhrystone

Benchmarks	am62xx-sk: perf
cpu_clock (MHz)	1400.00
dhrystone_per_mhz (DMIPS/MHz)	3.00
dhrystone_per_second (DhrystoneP)	7407407.50

Table: Dhrystone Benchmark

2.4.2.1.3. Whetstone

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

Execute the benchmark with the following:

runWhetstone

Benchmarks	am62xx-sk: perf
whetstone (MIPS)	5000.00

Table: Whetstone Benchmark

2.4.2.1.4. Linpack

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

Benchmarks	am62xx-sk: perf
linpack (Kflops)	581773.00

Table: Linpack Benchmark

2.4.2.1.5. 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	am62xx-sk: perf
assignment (Iterations)	13.63
fourier (Iterations)	22879.00
fp_emulation (Iterations)	107.20
huffman (Iterations)	1174.50
idea (Iterations)	3436.40
lu_decomposition (Iterations)	542.73
neural_net (Iterations)	7.85
numeric_sort (Iterations)	505.15
string_sort (Iterations)	165.88

Table: NBench Benchmarks

2.4.2.1.6. 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.

Execute the benchmark with the following:

stream_c

Benchmarks	am62xx-sk: perf
add (MB/s)	1446.10
copy (MB/s)	1640.40
scale (MB/s)	1810.50
triad (MB/s)	1487.90

Table: Stream

2.4.2.1.7. 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.

Benchmarks	am62xx-sk: perf
cjpeg-rose7-preset (workloads/)	41.67
core (workloads/)	0.30
coremark-pro ()	904.65
linear_alg-mid-100x100-sp (workloads/)	14.66
loops-all-mid-10k-sp (workloads/)	0.66
nnet_test (workloads/)	1.09
parser-125k (workloads/)	8.47
radix2-big-64k (workloads/)	59.17
sha-test (workloads/)	80.00
zip-test (workloads/)	21.74

Table: CoreMarkPro

Benchmarks	am62xx-sk: perf
cjpeg-rose7-preset (workloads/)	83.33
core (workloads/)	0.60
coremark-pro ()	1518.99
linear_alg-mid-100x100-sp (workloads/)	29.33
loops-all-mid-10k-sp (workloads/)	1.14
nnet_test (workloads/)	2.18
parser-125k (workloads/)	12.42
radix2-big-64k (workloads/)	42.58
sha-test (workloads/)	161.29
zip-test (workloads/)	39.22

Table: CoreMarkPro for Two Cores

2.4.2.1.8. 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	am62xx-sk: perf
4m-check (workloads/)	361.38
4m-check-reassembly (workloads/)	58.00
4m-check-reassembly-tcp (workloads/)	44.56
4m-check-reassembly-tcp-cmykw2-rotatew2 (workloads/)	24.70
4m-check-reassembly-tcp-x264w2 (workloads/)	1.73
4m-cmykw2 (workloads/)	200.20
4m-cmykw2-rotatew2 (workloads/)	39.84
4m-reassembly (workloads/)	53.91
4m-rotatew2 (workloads/)	45.91
4m-tcp-mixed (workloads/)	108.84
4m-x264w2 (workloads/)	1.87
empty-wld (workloads/)	1.00
idct-4m (workloads/)	18.59
idct-4mw1 (workloads/)	18.58
ippktcheck-4m (workloads/)	364.01
ippktcheck-4mw1 (workloads/)	360.33
ipres-4m (workloads/)	60.85
ipres-4mw1 (workloads/)	61.10
md5-4m (workloads/)	28.42
md5-4mw1 (workloads/)	28.21
rgbcmyk-4m (workloads/)	63.49
rgbcmyk-4mw1 (workloads/)	63.71
rotate-4ms1 (workloads/)	18.44
rotate-4ms1w1 (workloads/)	18.48
rotate-4ms64 (workloads/)	18.56
rotate-4ms64w1 (workloads/)	18.59
x264-4mq (workloads/)	0.56
x264-4mqw1 (workloads/)	0.51

Table: Multibench

2.4.2.2. Graphics SGX/RGX Driver

2.4.2.2.1. GFXBench

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

Benchmark	am62xx-sk: Score	am62xx-sk: Fps
GFXBench 5.x gl_5_high_off	11.48	0.18

Table: GFXBench

2.4.2.2.2. Glmark2

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

Benchmark	am62xx-sk: Score
Glmark2-Wayland	186.00

Table: Glmark2

2.4.2.3. 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

2.4.2.3.1. CPSW/CPSW2g/CPSW3g Ethernet Driver

TCP Bidirectional Throughput

Command Used	am62xx-sk: THROUGHPUT (Mbits/sec)	am62xx-sk: 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	1719.54	46.31

Table: CPSW TCP Bidirectional Throughput

UDP Throughput

Frame Size(bytes)	am62xx-sk: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE)	am62xx-sk: THROUGHPUT (Mbits/sec)	am62xx-sk: CPU Load % (LOCAL_CPU_UTIL)
64	18.00	13.77	37.27
128	82.00	61.93	37.11
256	210.00	157.63	37.11
1024	978.00	882.49	42.39
1518	1472.00	135.42	3.10

Table: CPSW UDP Egress Throughput

Frame Size(bytes)	am62xx-sk: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE)	am62xx-sk: THROUGHPUT (Mbits/sec)	am62xx-sk: CPU Load % (LOCAL_CPU_UTIL)
64	18.00	8.55	17.12
128	82.00	11.15	3.70
256	210.00	109.53	16.04
1024	978.00	212.03	12.93
1518	1472.00	957.04	39.57

Table: CPSW UDP Ingress Throughput (0% loss)

Frame Size(bytes)	am62xx-sk: UDP Datagram Size(bytes) (LOCAL_SEND_SIZE)	am62xx-sk: THROUGHPUT (Mbits/sec)	am62xx-sk: CPU Load % (LOCAL_CPU_UTIL)	am62xx-sk: Packet Loss %
64	18.00	25.42	39.69	36.21
128	82.00	115.68	41.30	53.69
256	210.00	281.24	44.48	53.92
1024	978.00	935.93	39.21	0.06
1518	1472.00	956.59	39.46	0.04

Table: CPSW UDP Ingress Throughput (possible loss)

2.4.2.4. 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.

2.4.2.4.1. AM62XX-SK

Buffer size (bytes)	am62xx-sk: Write EXT4 Throughput (Mbytes/sec)	am62xx-sk: Write EXT4 CPU Load (%)	am62xx-sk: Read EXT4 Throughput (Mbytes/sec)	am62xx-sk: Read EXT4 CPU Load (%)
1m	18.00	0.87	87.60	1.41
4m	18.30	0.75	80.60	1.30
4k	5.35	2.84	16.80	6.02
256k	19.00	1.17	85.30	1.72

The performance numbers were captured using the following:

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

2.4.2.5. CRYPTO Driver

2.4.2.5.1. OpenSSL Performance

Algorithm	Buffer Size (in bytes)	am62xx-sk: throughput (KBytes/Sec)
aes-128-cbc	1024	231939.41
aes-128-cbc	16	4927.82
aes-128-cbc	16384	654098.43
aes-128-cbc	256	72672.43
aes-128-cbc	64	19475.09
aes-128-cbc	8192	579556.69
aes-192-cbc	1024	219130.20
aes-192-cbc	16	4923.08
aes-192-cbc	16384	565848.75
aes-192-cbc	256	70948.35
aes-192-cbc	64	19254.91
aes-192-cbc	8192	512087.38
aes-256-cbc	1024	211012.61
aes-256-cbc	16	4903.49
aes-256-cbc	16384	515320.49
aes-256-cbc	256	70183.34
aes-256-cbc	64	19079.66
aes-256-cbc	8192	469909.50
des-cbc	1024	26105.17
des-cbc	16	5203.38
des-cbc	16384	27667.11
des-cbc	256	21917.10
des-cbc	64	13308.95
des-cbc	8192	27552.43
des3	1024	10566.66
des3	16	4052.52
des3	16384	11239.42
des3	256	8542.98
des3	64	7796.95
des3	8192	11231.23
md5	1024	51471.70
md5	16	1120.82
md5	16384	153938.60
md5	256	16328.79
md5	64	4373.87
md5	8192	135965.35
sha1	1024	62358.53
sha1	16	1086.12
sha1	16384	363550.04
sha1	256	16986.45
sha1	64	4329.37
sha1	8192	273566.38
sha224	1024	60982.27
sha224	16	1063.24
sha224	16384	371081.22
sha224	256	16469.42
sha224	64	4200.04
sha224	8192	275884.71
sha256	1024	61202.09
sha256	16	1061.38
sha256	16384	370130.94
sha256	256	16495.10
sha256	64	4202.71
sha256	8192	274909.87
sha384	1024	36339.37
sha384	16	1037.43
sha384	16384	74257.75
sha384	256	13717.85
sha384	64	4153.30
sha384	8192	69307.05
sha512	1024	36271.79
sha512	16	1031.84
sha512	16384	74290.52
sha512	256	13665.19
sha512	64	4131.11
sha512	8192	69296.13

Algorithm	am62xx-sk: CPU Load
aes-128-cbc	98.00
aes-192-cbc	98.00
aes-256-cbc	98.00
des-cbc	98.00
des3	96.00
md5	98.00
sha1	98.00
sha224	98.00
sha256	98.00
sha384	98.00
sha512	98.00

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

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

2.4.2.5.2. IPSec Software Performance

Algorithm	am62xx-sk: Throughput (Mbps)	am62xx-sk: Packets/Sec	am62xx-sk: CPU Load
3des	69.00	6.00	25.06
aes128	238.00	21.00	25.50
aes192	237.90	21.00	25.46
aes256	239.30	21.00	25.51