* [dts] [next] [PATCH V1] conf/nic_single_core_perf: take 0.00 instead of actual performance numbers
@ 2019-01-30 14:52 Lijuan Tu
2019-01-31 8:07 ` Tu, Lijuan
0 siblings, 1 reply; 2+ messages in thread
From: Lijuan Tu @ 2019-01-30 14:52 UTC (permalink / raw)
To: dts, jplsek; +Cc: Lijuan Tu
Since some NIC vendors might not public their performance numbers,
keep performance numbers private and local, it's a good option for
DTS.
Signed-off-by: Lijuan Tu <lijuan.tu@intel.com>
---
conf/nic_single_core_perf.cfg | 42 +++++++++++++++++++++++-------------------
1 file changed, 23 insertions(+), 19 deletions(-)
diff --git a/conf/nic_single_core_perf.cfg b/conf/nic_single_core_perf.cfg
index da1a050..6ecb524 100644
--- a/conf/nic_single_core_perf.cfg
+++ b/conf/nic_single_core_perf.cfg
@@ -1,39 +1,43 @@
# Throughput numbers vary in different environment.
-# Change these configuration on demand.
+# Users could change these configuration on demand:
+#
# - test_parameters defines the combination of frame size and descriptor
# numbers, and the pattern is
# {'frame size': ['descriptor number #1', 'descriptor number #2']}
+#
# - test_duration is how many seconds each combination performance will
# be recorded.
+#
# - accepted_tolerance defines the accepted tolerance between test
# results and expected numbers.
+#
# - expected_throughput is a dictionary defining expected throughput
# numbers based on NIC, and the pattern is
# {'NIC': {'frame size': {'descriptor number': 'excepted throughput'}}}
-# Default numbers of fortville_spirit and niantic is generated on this
-# environment:
-# * Intel(R) Xeon(R) Platinum 8180 CPU @ 2.50GHz
-# * Trex v3.23
+# Every user should fill it out with your actual numbers. To keep the
+# expected throughput private, dts takes 0.00 as default.
+#
# - if update_expected == Ture, and add argument "--update-expected" in
# bash command, all objects in this file will changed after the run
# for example: ./dts --update-expected
+#
+# Highlights:
+# At the begining, please change test_parameters according to your
+# requirements, then run ./dts --update-expected to get the absolute
+# results which will replace the default numbers 0.00 in this configuration.
+# So you will have your own private configuration, and could start your
+# tests as ususal.
[suite]
update_expected = True
test_parameters = {64: [512, 2048]}
test_duration = 60
accepted_tolerance = 1
-expected_throughput = {'fortville_spirit': {64: {512: 62.35, 2048: 47.89}},
- 'niantic': {64: {128: 53.435, 512: 53.699, 2048: 42.798}},
- 'fortville_25g': {64: {512: 62.35, 2048: 47.651}},
- 'ConnectX4_LX_MT4117': {'40G': {64: {128: 31.635, 256: 32.473, 512: 30.72, 2048: 26.94}},
- '25G': {64: {128: 28.178, 256: 34.581, 512: 30.528, 2048: 26.004}}},
- 'ConnectX5_MT4121': {64: {128: 42.161, 256: 56.651, 512: 47.091, 2048: 40.104}}}
-
-# legacy formate for reference
-throughput_nnt = {64: {128: 53.435, 512: 53.699, 2048: 42.798}}
-throughput_fvl25g = {64: {512: 43.777, 2048: 43.651}}
-throughput_fvl40g = {64: {512: 62.35, 2048: 47.89}}
-throughput_cx5 = {64: {128: 42.161, 256: 56.651, 512: 47.091, 2048: 40.104}}
-throughput_cx4lx25g = {64: {128: 28.178, 256: 34.581, 512: 30.528, 2048: 26.004}}
-throughput_cx4lx40g = {64: {128: 31.635, 256: 32.473, 512: 30.72, 2048: 26.94}}
+expected_throughput = {
+ 'fortville_spirit': {64: {512: 0.00, 2048: 0.00}},
+ 'niantic': {64: {128: 0.00, 512: 0.00, 2048: 0.00}},
+ 'fortville_25g': {64: {512: 0.00, 2048: 0.00}},
+ 'ConnectX4_LX_MT4117': {'40G': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048: 0.00}},
+ '25G': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048: 0.00}}},
+ 'ConnectX5_MT4121': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048: 0.00}}
+}
--
1.8.3.1
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [dts] [next] [PATCH V1] conf/nic_single_core_perf: take 0.00 instead of actual performance numbers
2019-01-30 14:52 [dts] [next] [PATCH V1] conf/nic_single_core_perf: take 0.00 instead of actual performance numbers Lijuan Tu
@ 2019-01-31 8:07 ` Tu, Lijuan
0 siblings, 0 replies; 2+ messages in thread
From: Tu, Lijuan @ 2019-01-31 8:07 UTC (permalink / raw)
To: dts, jplsek
Applied, thanks
> -----Original Message-----
> From: Tu, Lijuan
> Sent: Wednesday, January 30, 2019 10:52 PM
> To: dts@dpdk.org; jplsek@iol.unh.edu
> Cc: Tu, Lijuan <lijuan.tu@intel.com>
> Subject: [next] [PATCH V1] conf/nic_single_core_perf: take 0.00 instead of
> actual performance numbers
>
> Since some NIC vendors might not public their performance numbers, keep
> performance numbers private and local, it's a good option for DTS.
>
> Signed-off-by: Lijuan Tu <lijuan.tu@intel.com>
> ---
> conf/nic_single_core_perf.cfg | 42 +++++++++++++++++++++++-------------------
> 1 file changed, 23 insertions(+), 19 deletions(-)
>
> diff --git a/conf/nic_single_core_perf.cfg b/conf/nic_single_core_perf.cfg index
> da1a050..6ecb524 100644
> --- a/conf/nic_single_core_perf.cfg
> +++ b/conf/nic_single_core_perf.cfg
> @@ -1,39 +1,43 @@
> # Throughput numbers vary in different environment.
> -# Change these configuration on demand.
> +# Users could change these configuration on demand:
> +#
> # - test_parameters defines the combination of frame size and descriptor #
> numbers, and the pattern is
> # {'frame size': ['descriptor number #1', 'descriptor number #2']}
> +#
> # - test_duration is how many seconds each combination performance will #
> be recorded.
> +#
> # - accepted_tolerance defines the accepted tolerance between test # results
> and expected numbers.
> +#
> # - expected_throughput is a dictionary defining expected throughput #
> numbers based on NIC, and the pattern is # {'NIC': {'frame size': {'descriptor
> number': 'excepted throughput'}}} -# Default numbers of fortville_spirit and
> niantic is generated on this -# environment:
> -# * Intel(R) Xeon(R) Platinum 8180 CPU @ 2.50GHz
> -# * Trex v3.23
> +# Every user should fill it out with your actual numbers. To keep the #
> +expected throughput private, dts takes 0.00 as default.
> +#
> # - if update_expected == Ture, and add argument "--update-expected" in #
> bash command, all objects in this file will changed after the run # for
> example: ./dts --update-expected
> +#
> +# Highlights:
> +# At the begining, please change test_parameters according to your #
> +requirements, then run ./dts --update-expected to get the absolute #
> +results which will replace the default numbers 0.00 in this configuration.
> +# So you will have your own private configuration, and could start your
> +# tests as ususal.
>
> [suite]
> update_expected = True
> test_parameters = {64: [512, 2048]}
> test_duration = 60
> accepted_tolerance = 1
> -expected_throughput = {'fortville_spirit': {64: {512: 62.35, 2048: 47.89}},
> - 'niantic': {64: {128: 53.435, 512: 53.699, 2048: 42.798}},
> - 'fortville_25g': {64: {512: 62.35, 2048: 47.651}},
> - 'ConnectX4_LX_MT4117': {'40G': {64: {128: 31.635, 256: 32.473, 512: 30.72,
> 2048: 26.94}},
> - '25G': {64: {128: 28.178, 256: 34.581, 512: 30.528, 2048:
> 26.004}}},
> - 'ConnectX5_MT4121': {64: {128: 42.161, 256: 56.651, 512: 47.091, 2048:
> 40.104}}}
> -
> -# legacy formate for reference
> -throughput_nnt = {64: {128: 53.435, 512: 53.699, 2048: 42.798}} -
> throughput_fvl25g = {64: {512: 43.777, 2048: 43.651}} -throughput_fvl40g = {64:
> {512: 62.35, 2048: 47.89}}
> -throughput_cx5 = {64: {128: 42.161, 256: 56.651, 512: 47.091, 2048: 40.104}} -
> throughput_cx4lx25g = {64: {128: 28.178, 256: 34.581, 512: 30.528, 2048:
> 26.004}} -throughput_cx4lx40g = {64: {128: 31.635, 256: 32.473, 512: 30.72,
> 2048: 26.94}}
> +expected_throughput = {
> + 'fortville_spirit': {64: {512: 0.00, 2048: 0.00}},
> + 'niantic': {64: {128: 0.00, 512: 0.00, 2048: 0.00}},
> + 'fortville_25g': {64: {512: 0.00, 2048: 0.00}},
> + 'ConnectX4_LX_MT4117': {'40G': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048:
> 0.00}},
> + '25G': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048: 0.00}}},
> + 'ConnectX5_MT4121': {64: {128: 0.00, 256: 0.00, 512: 0.00, 2048:
> +0.00}} }
> --
> 1.8.3.1
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2019-01-31 8:07 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-01-30 14:52 [dts] [next] [PATCH V1] conf/nic_single_core_perf: take 0.00 instead of actual performance numbers Lijuan Tu
2019-01-31 8:07 ` Tu, Lijuan
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).