2021-09-20 16:03:09

by Arnd Bergmann

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: remove debugfs #ifdef

On Mon, Sep 20, 2021 at 2:47 PM Laurent Pinchart
<[email protected]> wrote:
>
> It's only a few bytes of data in struct dma_device, but a bit more in
> .text here. Is the simplification really required in this driver ?

The intention was to not change the resulting object code in this driver,
and I still don't see where it would grow after dead-code-elimination removes
all the unused static functions. What am I missing?

Arnd


2021-09-21 01:40:29

by Laurent Pinchart

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: remove debugfs #ifdef

Hi Arnd,

On Mon, Sep 20, 2021 at 02:50:52PM +0200, Arnd Bergmann wrote:
> On Mon, Sep 20, 2021 at 2:47 PM Laurent Pinchart wrote:
> >
> > It's only a few bytes of data in struct dma_device, but a bit more in
> > .text here. Is the simplification really required in this driver ?
>
> The intention was to not change the resulting object code in this driver,
> and I still don't see where it would grow after dead-code-elimination removes
> all the unused static functions. What am I missing?

Indeed, gcc does a fairly good job there. The .text section doesn't
grow. Interestingly, there's an increase in size in the .data and
.rodata sections in the xilinx-dpdma module:

- 8 .rodata.str1.8 0000029f 0000000000000000 0000000000000000 00003660 2**3
+ 8 .rodata.str1.8 000002a7 0000000000000000 0000000000000000 00003660 2**3

- 10 .rodata 00001080 0000000000000000 0000000000000000 00003960 2**5
+ 10 .rodata 000010e0 0000000000000000 0000000000000000 00003960 2**5

- 15 .data 00001050 0000000000000000 0000000000000000 00004e40 2**5
+ 15 .data 00001090 0000000000000000 0000000000000000 00004ea0 2**5

I'm not entirely sure where it comes from, it may be related to
instrumentation caused by debugging options.

For your patch,

Reviewed-by: Laurent Pinchart <[email protected]>

--
Regards,

Laurent Pinchart