Skip to contentSkip to conversation

drm/mediatek: Filter modes according to hardware capability

PATCH Shawn Sung (宋孝謙) 2 participants 2 messages Dec 2, 2024 – Dec 3, 2024 1 message not in the archive Arm MediaTek

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

Hi Michał,

Quoted 5 lines · Michał Kopeć

External email : Please do not click links or open attachments until you have verified the sender or the content.

Hi Shawn, thanks for replying!

Thank you for bringing this issue to our attention.

Since every email from outside of the company is first moved to an isolation folder, I replied your mail before moving it out and noticed that there is a "[SPAM]" tag in the subject line.
I apologize for what happened and have fixed it this time.

Quoted 31 lines
Quoted 5 lines

As the reference number 8250 = 594000 / 72:
594000 is KHz of the clock rate in mode setting
72 is lines during VBP (Vertical Back Porch)
Can you share your reference number or the EDID in 3440x1440@144

with

Quoted 1 line

us?

Sure, here's the mode defined in the EDID:

DTD: 3440x1440 143.922761 Hz 0:0 217.323 kHz 799.750000 MHz (aspect undefined, no 3D stereo, preferred)
Hfront 48 Hsync 32 Hback 160 Hpol P
Vfront 3 Vsync 10 Vback 57 Vpol N

799750/57 is 14030 so definitely much higher than supported according to the code.

There's also other modes:

DTD: 3440x1440 119.999642 Hz 0:0 182.999 kHz 670.510000 MHz (aspect undefined, no 3D stereo)
Hfront 144 Hsync 32 Hback 48 Hpol P
Vfront 3 Vsync 4 Vback 78 Vpol N

DTD 2: 3440x1440 99.981604 Hz 43:18 150.972 kHz 543.500000 MHz (800 mm
x 335 mm)
Hfront 48 Hsync 32 Hback 80 Hpol P
Vfront 3 Vsync 4 Vback 63 Vpol N

Quoted 6 lines

670510/78 = 8596
543500/63 = 8626

So the 120Hz and 100Hz modes look like they might almost work :) The manufacturer neglected to add a 60Hz mode which I suspect might have worked without issue.

Yes, these two numbers seems to work as they are just less than 5% above the spec.

Quoted 26 lines
Quoted 3 lines

The reference number 8250 is established based on our hardware specifications. Numbers higher than that may function, but could

be

Quoted 1 line

unstable. For example, the monitor may seem to work properly at

first,

Quoted 1 line

but you could encounter black screens or glitching from time to

time.

Understood. I was just wondering if it might be dependent on other factors than Vbp and pixel clock.

Quoted 1 line

May I confirm if you are going to implement the patch just for

local

Quoted 2 lines

testing or for upstream? As 8250 is calculated based on the specifications and was verified through a series of procedures.

Any

Quoted 1 line

changes to it may require thorough verifications before being

accepted.

I was going to initially try to implement something for personal use but
I wanted to eventually try to upstream it. However, from what I understand,
it might not be easy or possible.

Quoted 4 lines

If I develop a patch that works reliably at least for me, would you be
open to me adding a kernel module parameter to allow unverified data rates, with clear description that it may cause unstable behavior?

As I also have a high refresh rate monitor, I can completely understand how you feel. Could you please change 8250 to 8250 * 1.05 and see if it works in your situation?

Since the official configuration usually prefer a more stable setting, we have initiated an internal discussion about this request.

Quoted 7 lines
Quoted 3 lines

Regards,
Shawn

Regards,
Michał

Regards,
Shawn

#2 lore raw

Hi Michał,

[snip]

Quoted 38 lines
Quoted 26 lines
Quoted 2 lines

The reference number 8250 is established based on our hardware specifications. Numbers higher than that may function, but could

be

Quoted 1 line

unstable. For example, the monitor may seem to work properly at

first,

Quoted 1 line

but you could encounter black screens or glitching from time to

time.

Understood. I was just wondering if it might be dependent on other factors than Vbp and pixel clock.

Quoted 1 line

May I confirm if you are going to implement the patch just for

local

Quoted 2 lines

testing or for upstream? As 8250 is calculated based on the specifications and was verified through a series of procedures.

Any

Quoted 1 line

changes to it may require thorough verifications before being

accepted.

I was going to initially try to implement something for personal use
but
I wanted to eventually try to upstream it. However, from what I understand,
it might not be easy or possible.

Quoted 5 lines

If I develop a patch that works reliably at least for me, would you be
open to me adding a kernel module parameter to allow unverified data
rates, with clear description that it may cause unstable behavior?

As I also have a high refresh rate monitor, I can completely understand
how you feel. Could you please change 8250 to 8250 * 1.05 and see if
it works in your situation?

Quoted 4 lines

Since the official configuration usually prefer a more stable setting,
we have initiated an internal discussion about this request.

Here is the reply by hardware designer after an internal discussion with us:

Due to internal regulations, we are unable to disclose the actual hardware specifications.

The current approach has been validated against the hardware limitations:
1. pixel clock rate < 594000 KHz
2. pixel clock rate / VBP < 8250

While it is possible that exceeding these limitations might still work, it could potentially go beyond the guarantees provided by the DRAM controller and Hardware Real-Time (HRT) capabilities, leading to a risk of encountering unstable underflow issues.

I think this scaling patch (without hardware guarantees) is more suitable for the specific product lines applied to Chrome, not upstream.

Regards,
Jason-JH Lin

Quoted 11 lines
Quoted 7 lines
Quoted 3 lines

Regards,
Shawn

Regards,
Michał

Regards,
Shawn