[dts] [PATCH V2] test: add vhost/virtio pvp performance test based on qemu

Liu, Yong yong.liu at intel.com
Wed Dec 21 09:43:29 CET 2016


Hi Lei, remaining two comments.

> -----Original Message-----
> From: dts [mailto:dts-bounces at dpdk.org] On Behalf Of lei,yao
> Sent: Wednesday, December 21, 2016 4:28 PM
> To: dts at dpdk.org
> Cc: Yao, Lei A
> Subject: [dts] [PATCH V2] test: add vhost/virtio pvp performance test
> based on qemu
> 
> From: lei yao <lei.a.yao at intel.com>
> 
> The test case include 3 part:
> 1. virtio 0.95 performance test
> virtio0.95 is the default version for qemu. Vector, normal, mergeable path
> will be tested on this version
> 2. Test vector performance on virtio 1.0.
> 
> Signed-off-by: lei yao <lei.a.yao at intel.com>
> ---
>  tests/TestSuite_vhost_qemu_pvp_performance.py | 311
> ++++++++++++++++++++++++++
>  1 file changed, 311 insertions(+)
>  create mode 100644 tests/TestSuite_vhost_qemu_pvp_performance.py
> 
> diff --git a/tests/TestSuite_vhost_qemu_pvp_performance.py
> b/tests/TestSuite_vhost_qemu_pvp_performance.py
> new file mode 100644
> index 0000000..32f5c56
> --- /dev/null
> +++ b/tests/TestSuite_vhost_qemu_pvp_performance.py
> +
> +        self.number_of_ports = 1
> +        self.header_row = ["FrameSize(B)", "Throughput(Mpps)",
> "LineRate(%)", "Cycle"]
> +        self.memory_channel = 4
> +        # Specify the path of Qemu binary for different Qemu versions,
> user can modify this as necessary
> +        self.qemu_path = ["/home/lei/qemu_install/qemu_2.5/bin/qemu-
> system-x86_64",
> +                          "/home/lei/qemu_install/qemu_2.6/bin/qemu-
> system-x86_64",
> +                          "/home/lei/qemu_install/qemu_2.7/bin/qemu-
> system-x86_64"]

Please remove hard code path, its' better to use configuration file.

> +
> +    def set_up(self):
> +        #
> +        # Run before each test case.
> +        #
> +        # Launch vhost sample using default params
> +        #
> +        self.dut.send_expect("rm -rf ./vhost.out", "#")
> +        self.dut.send_expect("rm -rf ./vhost-net*", "#")
> +        self.dut.send_expect("killall -s INT vhost-switch", "#")
> +        self.dut.send_expect("killall -s INT qemu-system-x86_64", "#")
> +


Please not direct killall qemu, maybe someone are using them for other usage. You can simple trust framework stop function.
If there's still has issue in kill qemu process, let's fix it in framework.




More information about the dts mailing list