[EXT] [PATCH v6] app/dma-perf: introduce dma-perf application

[Anoob Joseph]

Hi Cheng,

Please see inline.

Thanks,
Anoob

> -----Original Message-----
> From: Jiang, Cheng1 <cheng1.jiang at intel.com>
> Sent: Sunday, June 18, 2023 11:05 AM
> To: Anoob Joseph <anoobj at marvell.com>
> Cc: dev at dpdk.org; Hu, Jiayu <jiayu.hu at intel.com>; Ding, Xuan
> <xuan.ding at intel.com>; thomas at monjalon.net; Richardson, Bruce
> <bruce.richardson at intel.com>; mb at smartsharesystems.com; Xia, Chenbo
> <chenbo.xia at intel.com>; Amit Prakash Shukla
> <amitprakashs at marvell.com>; Ma, WenwuX <wenwux.ma at intel.com>;
> Wang, YuanX <yuanx.wang at intel.com>; He, Xingguang
> <xingguang.he at intel.com>
> Subject: RE: [EXT] [PATCH v6] app/dma-perf: introduce dma-perf application
> 
> Hi Anoob,
> 
> I've looked into the memory footprint issue, and I didn't find any.
> So could you please help to share the config file you are using? Maybe it can
> help me to accelerate the debug process.

[Anoob] We had to increase the hugepages to get the test running. Otherwise simple memory allocations were failing. Some of structs have redundant fields which can be easily addressed.

For example please check below members in struct lcore_params,
1. scenario_id
2. lcore_id
3. dma_name
4. worker_id
5. test_secs

Also, some of the parameters in the above struct is const for a thread (like buf_size, kick_batch etc). So I was thinking may be it is better to split lcore_params into two portions and have only the dynamic part as volatile. But that is something we can take up later. I leave that to your judgement.

> 
> Thanks a lot,
> Cheng
> 
> > -----Original Message-----
> > From: Jiang, Cheng1
> > Sent: Friday, June 16, 2023 11:16 PM
> > To: Anoob Joseph <anoobj at marvell.com>
> > Cc: dev at dpdk.org; Hu, Jiayu <Jiayu.Hu at intel.com>; Ding, Xuan
> > <Xuan.Ding at intel.com>; thomas at monjalon.net; Richardson, Bruce
> > <bruce.richardson at intel.com>; mb at smartsharesystems.com; Xia, Chenbo
> > <Chenbo.Xia at intel.com>; Amit Prakash Shukla
> > <amitprakashs at marvell.com>; Ma, WenwuX <WenwuX.Ma at intel.com>;
> Wang,
> > YuanX <yuanx.wang at intel.com>; He, Xingguang
> <xingguang.he at intel.com>
> > Subject: RE: [EXT] [PATCH v6] app/dma-perf: introduce dma-perf
> > application
> >
> > Hi Anoob,
> >
> > Replies are inline.
> >
> > Thanks,
> > Cheng
> >
> > > -----Original Message-----
> > > From: Anoob Joseph <anoobj at marvell.com>
> > > Sent: Friday, June 16, 2023 6:53 PM
> > > To: Jiang, Cheng1 <cheng1.jiang at intel.com>
> > > Cc: dev at dpdk.org; Hu, Jiayu <jiayu.hu at intel.com>; Ding, Xuan
> > > <xuan.ding at intel.com>; thomas at monjalon.net; Richardson, Bruce
> > > <bruce.richardson at intel.com>; mb at smartsharesystems.com; Xia,
> Chenbo
> > > <chenbo.xia at intel.com>; Amit Prakash Shukla
> > > <amitprakashs at marvell.com>; Ma, WenwuX <wenwux.ma at intel.com>;
> > Wang,
> > > YuanX <yuanx.wang at intel.com>; He, Xingguang
> <xingguang.he at intel.com>
> > > Subject: RE: [EXT] [PATCH v6] app/dma-perf: introduce dma-perf
> > > application
> > >
> > > Hi Cheng,
> > >
> > > I think there is a bug in EAL parsing. Please check the below diff
> > > and see if we should do something similar.
> > >
> > > diff --git a/app/test-dma-perf/main.c b/app/test-dma-perf/main.c
> > > index
> > > d65655b87b..7fcaa5acf6 100644
> > > --- a/app/test-dma-perf/main.c
> > > +++ b/app/test-dma-perf/main.c
> > > @@ -432,7 +432,7 @@ append_eal_args(int argc, char **argv, const
> > > char *eal_args, char **new_argv)
> > >                         i++;
> > >                         continue;
> > >                 }
> > > -               strlcpy(new_argv[new_argc], argv[i],
> sizeof(new_argv[new_argc]));
> > > +               strlcpy(new_argv[new_argc], argv[i],
> > > + MAX_EAL_PARAM_LEN);
> > >                 new_argc++;
> > >         }
> > >
> > > Thanks,
> > > Anoob
> >
> > [Cheng] yes there is an issue in it. And I have the same fix. I'll
> > submit it later, thanks.
> >
> > >
> > > > -----Original Message-----
> > > > From: Anoob Joseph <anoobj at marvell.com>
> > > > Sent: Friday, June 16, 2023 3:19 PM
> > > > To: Jiang, Cheng1 <cheng1.jiang at intel.com>
> > > > Cc: dev at dpdk.org; Hu, Jiayu <jiayu.hu at intel.com>; Ding, Xuan
> > > > <xuan.ding at intel.com>; thomas at monjalon.net; Richardson, Bruce
> > > > <bruce.richardson at intel.com>; mb at smartsharesystems.com; Xia,
> > Chenbo
> > > > <chenbo.xia at intel.com>; Amit Prakash Shukla
> > > > <amitprakashs at marvell.com>; Ma, WenwuX
> <wenwux.ma at intel.com>;
> > > Wang,
> > > > YuanX <yuanx.wang at intel.com>; He, Xingguang
> > <xingguang.he at intel.com>
> > > > Subject: RE: [EXT] [PATCH v6] app/dma-perf: introduce dma-perf
> > > > application
> > > >
> > > > Hi Cheng,
> > > >
> > > > > [Cheng] sure, no problem. Feel free to contact me if you have
> > > > > any more questions.
> > > >
> > > > No further comments from my side. Can you submit next version?
> >
> > [Cheng] Glad to know. Yes, the next version is almost ready, I'll
> > submit it this weekend.
> >
> > > >
> > > > Top level, I think you may need to add a doc file for the new app.
> >
> > [Cheng] sure, I was thinking about this, and I'll add a doc for this, thanks.
> >
> > > >
> > > > Few issues that we are also checking in our end, 1. The app is
> > > > having significant memory footprint. Need to see where we can
> improve.
> > > > 2. EAL args passing doesn't seem to be working. Are you able to
> > > > enable specific devices by making using of '-a' arg?
> >
> > [Cheng] I'll take a look at the significant memory footprint issue later,
> thanks.
> > As for the eal args process, yes there is an issue, Dengdui's comments
> > also mentioned it, and I have fixed it in the v7 patch.
> >
> > > >
> > > > Thanks,
> > > > Anoob