[OpenPower-Firmware] [RFC] Drop untested graphic support from the build

[Paul Menzel]

Dear Stewart,


Am 22.08.2018 um 06:51 schrieb Stewart Smith:
> Paul Menzel <pmenzel+openpower-firmware at molgen.mpg.de> writes:
>> On 08/21/18 06:53, Stewart Smith wrote:
>>> Paul Menzel <pmenzel+openpower-firmware at molgen.mpg.de> writes:
>>>> On 08/20/18 08:54, Joel Stanley wrote:
>>>>> Back in the day Tuletta (P8) machines would use Radeon GPU drivers. No
>>>>> one has tested this in many moons. To save all systems are paying for
>>>>> the kernel driver,
>>>>
>>>> I do not understand the sentence.
>>>
>>> Tuleta is the name for the S812L, S822L etc servers that IBM released as
>>> the first POWER8 systems. Allegedly you can purchase these with some
>>> really old Radeon card, although we've never had one in the LTC to be
>>> able to check if graphics come out.
>>
>> Thank you for the explanation, but I only meant the last version.
>>
>>> To save all systems are paying for the kernel driver,
>>
>> A verb seems to have snucked in. But What is paid? The image size?
> 
> Yeah, image size and boot time.

It’d be great to have specific numbers (for (at least) one test system) 
in the commit message.

> We don't really get multiple megabytes per second loading off flash, and
> the bootkernel image is close to 16mb. Additionally, the initramfs takes
> about 2.5 seconds to decompress.

If the image is smaller now, and decompression takes so long, it might 
be worth to test if LZ4 gives better timings by having a larger image 
size but faster decompression.

Additionally, I always wondered, why the initrd (if no encryption is 
used) can’t be loaded from a hard disk (preferably SSD) or a fast network.

> I believe Raptor has an out of tree change that adds another partition
> on flash for optional firmware components such as the binary only blobs
> from the linux-firmware package. I'd love to see that go upstream.

Interesting. Fingers crossed.


Kind regardrs,

Paul