From: Mark Zhang <nvmarkzhang-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: Arto Merilainen <amerilainen-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
Cc: "thierry.reding-RM9K5IK7kjKj5M59NBduVrNAH6kLmebB@public.gmane.org"
<thierry.reding-RM9K5IK7kjKj5M59NBduVrNAH6kLmebB@public.gmane.org>,
"dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org"
<dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>,
"linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org"
<linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
Francis Hart <fhart-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>,
Terje Bergstrom
<tbergstrom-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
Subject: Re: [RFC,libdrm 1/3] tegra: Add stream library
Date: Fri, 28 Dec 2012 15:57:21 +0800 [thread overview]
Message-ID: <50DD50E1.80006@gmail.com> (raw)
In-Reply-To: <50DD4E2C.2070104-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
On 12/28/2012 03:45 PM, Arto Merilainen wrote:
> On 12/28/2012 08:47 AM, Mark Zhang wrote:
>>> +int tegra_fence_is_valid(const struct tegra_fence *fence)
>>> +{
>>> + int valid = fence ? 1 : 0;
>>> + valid = valid && fence->id != (uint32_t) -1;
>>> + valid = valid && fence->id < 32;
>>
>> Hardcode here. Assume always has 32 syncpts.
>> Change to a micro wrapped with tegra version ifdef will be better.
>>
>
> You are correct. I will fix this.
>
>>> + return valid;
>>> +}
>>> +
>> [...]
>>> +
>>> + /* Add fences */
>>> + if (num_fences) {
>>> +
>>> + tegra_stream_push(stream,
>>> + nvhost_opcode_setclass(NV_HOST1X_CLASS_ID,
>>> + host1x_uclass_wait_syncpt_r(), num_fences));
>>> +
>>> + for (; num_fences; num_fences--, fence++) {
>>> + assert(tegra_fence_is_valid(fence));
>>
>> This is useless. We already add "1 + num_fences" to num_words above. So
>> move this "assert" before adding "1 + num_fences" to num_words makes
>> sense.
>>
>
> The assertion checks the validity of a single fence - not if there is
> room in the command buffer.
>
> The goal is to prevent having invalid fences in the command stream. If
> this check were not here it would be possible to initialise a fence with
> tegra_fence_clear() and put that fence into the stream.
My idea is, if one fence is invalid, then we should not count this in
"num_words". In current code, if one fence is invalid, then this fence
will not be pushed into the command stream, and the "num_words" shows a
wrong command buffer size.
So I think we should:
- validate the fences, remove the invalid fence
- update num_words
- then you don't need to check fence here - I mean, before push a host1x
syncpt wait command into the active buffer of stream.
>
>>> +
>>> + tegra_stream_push(stream,
>>> nvhost_class_host_wait_syncpt(fence->id,
>>> + fence->value));
>>> + }
>>> + }
>>> +
>>> + if (class_id)
>>> + tegra_stream_push(stream, nvhost_opcode_setclass(class_id,
>>> 0, 0));
>>> +
>>> + return 0;
>>> +}
>>> +
>> [...]
>>> +
>>> +#endif /* TEGRA_DRMIF_H */
>>>
>
> - Arto
next prev parent reply other threads:[~2012-12-28 7:57 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-12-13 14:01 [RFC,libdrm 0/3] NVIDIA Tegra support Arto Meriläinen
2012-12-13 14:01 ` [RFC,libdrm 1/3] tegra: Add stream library Arto Meriläinen
[not found] ` <1355407268-32381-2-git-send-email-amerilainen-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2012-12-28 6:47 ` Mark Zhang
[not found] ` <50DD407B.3030306-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2012-12-28 7:45 ` Arto Merilainen
[not found] ` <50DD4E2C.2070104-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2012-12-28 7:57 ` Mark Zhang [this message]
[not found] ` <50DD50E1.80006-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2012-12-28 8:50 ` Arto Merilainen
[not found] ` <50DD5D6D.3090504-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2012-12-28 9:04 ` Mark Zhang
[not found] ` <50DD6098.60900-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2012-12-28 9:20 ` Arto Merilainen
2012-12-28 20:48 ` Thierry Reding
[not found] ` <20121228204847.GA4309-RM9K5IK7kjIyiCvfTdI0JKcOhU4Rzj621B7CTYaBSLdn68oJJulU0Q@public.gmane.org>
2012-12-31 6:22 ` Terje Bergström
[not found] ` <50E12F1A.3000803-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2013-01-02 2:44 ` Mark Zhang
[not found] ` <50E39F2A.1000405-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-01-02 6:31 ` Terje Bergström
[not found] ` <50E3D45D.5050206-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2013-01-02 6:53 ` Mark Zhang
2012-12-13 14:01 ` [RFC,libdrm 2/3] tegra: Add 2d library Arto Meriläinen
[not found] ` <1355407268-32381-3-git-send-email-amerilainen-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
2012-12-28 6:48 ` Mark Zhang
2012-12-13 14:01 ` [RFC,libdrm 3/3] tests: tegra: Add 2d tests Arto Meriläinen
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=50DD50E1.80006@gmail.com \
--to=nvmarkzhang-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
--cc=amerilainen-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org \
--cc=dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
--cc=fhart-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org \
--cc=linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=tbergstrom-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org \
--cc=thierry.reding-RM9K5IK7kjKj5M59NBduVrNAH6kLmebB@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox