topic Re: SPT3.1 1024-128-16range don't work! in Other NXP Products

SPT3.1 1024-128-16range don't work!

Shi_test — Wed, 22 Mar 2023 10:04:02 GMT

S32R45:

The s32ds project is "RSDK_offline_example".

SPT:"range_1024smp_128crp_16ch.pspt"

I try "./xxx.elf" in S32R45 linux,but It seems to be stuck with "pdma" command in "range_1024smp_128crp_16ch.pspt".

This pic have pdma.

This pic not have pdma.

Has anyone tested it successfully?

Re: SPT3.1 1024-128-16range don't work!

petervlna — Thu, 23 Mar 2023 08:28:01 GMT

Hello,

I have contacted RADAR team. They will post here ASAP.

Best regards,
Peter

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Fri, 24 Mar 2023 05:37:00 GMT

I already know which parameter is the problem.

There is PDMA command：

pdma.ind .signext .16cmplx .opram2sysram .async .notrace .nocoherent, CHANNELS_NO*OUTPUT_SAMPLES_PER_CH, RSDK_SPT_CUBE_BASE_ADDR, WR_16, 0x0, 0x1, (8*(CHIRPS_PER_FRAME - 1)), 0x8

If I replace WE_16 with WR_20.He's going to succeed.

But i still don't know why.

There is Reference Manual pic about WR_16-31:

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Fri, 24 Mar 2023 07:16:20 GMT

And then we analyze it again and we get WR_5，if we change WR_5，it will succeed，but the result still wrong.

Re: SPT3.1 1024-128-16range don't work!

petervlna — Fri, 24 Mar 2023 08:55:29 GMT

Hello,

So little information has been disclosed for analysis:

What RSDK version are you using?
What hardware are you using? If are you using RSDK pre-build demo, I can confirm it works well on the S32R45 EVB. Based on the snapshots you provided, I don't think the PDMA instruction will the stuck program, if there is an error with PDMA, the program won't stop there and will throw out an error message via an interrupt.

Best regards,

Peter

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Fri, 24 Mar 2023 09:23:08 GMT

1.SDK：Rev.1. 1 .0 S32DS:3.4 and 3.5
2.S32R45 Evaluation Board
Yes, From the phenomenon, the error is WR_5.In 16 channels SPT，WR_5 is Work Registers for OUTPUT buffer.
So,Could WR_5 possibly lead to this outcome?(may be WR_5_CTRL_REG)
I will try SDK1.0.0.
Thank you.

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Fri, 24 Mar 2023 09:58:43 GMT

I tried to confirm that something was wrong with PDMA:

pdma.ind .signext .16cmplx .opram2sysram .sync .notrace .nocoherent, CHANNELS_NO*OUTPUT_SAMPLES_PER_CH, RSDK_SPT_CUBE_BASE_ADDR, WR_16, 0x0, 0x1, (8*(CHIRPS_PER_FRAME - 1)), 0x8

If PDMA is not stuck,terminal will have "times:1",but there is "0".

If i change ".sync" to "async"，terminal will have "times 1".So i'm sure PDMA is not stuck.

And then i change "WR_16" to "WR_11",PDMA will go on.

So i check "WR_16"：

add .noshift .mod24 WR_5, #(OR_2_0_0), WR_16

use the same method I find that "WR_5" is the source

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Tue, 28 Mar 2023 02:19:54 GMT

SDK1.0.0 is no better than SDK1.1.0.
I can't find the reason.

Re: SPT3.1 1024-128-16range don't work!

Shi_test — Wed, 29 Mar 2023 06:42:18 GMT

I solved the problem.

The reason is I need to match S32DS input with SPT input.