Welcome to the Lounge

13-Nov-21 11:13

I just love them. They're like little puzzle pieces you can use to solve really complicated problems.

I can't stand debugging them, but I love creating them and watching them make the compiler do really neat things in funny ways.

I don't look for excuses to use them, because they're hard to debug, but I enjoy them when they are the right tool for the job. They're incredibly versatile, and I think fun to create.

Does that make me a nerd?**

** It's okay if it does, I'm just looking for a frame of reference. Poke tongue | ;-P

Real programmers use butterflies

Greg Utas13-Nov-21 13:54

Greg Utas

13-Nov-21 13:54

The Witch wrote:
Does that make me a nerd?

Yes. And more specifically, a nerd who's into BDSM.

Full specialization is one thing, but partial specialization (even matching on function signatures when template parameters are unspecified) is another. It is write-only code because none but the author and others also into BDSM will be able to read it.

Hard to debug? If a compiler supports all of this properly, it should also provide a debugger that handles it. If not, 12 strokes of the cane are in order.

Mad | :mad:

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.

honey the codewitch13-Nov-21 14:17

13-Nov-21 14:17

Hard to debug because when you have chains of them it can be hard to know which ones are even being compiled.

Real programmers use butterflies

Greg Utas13-Nov-21 14:32

Greg Utas

13-Nov-21 14:32

I can imagine!

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.

honey the codewitch13-Nov-21 14:38

13-Nov-21 14:38

In my GFX project I allow drivers to expose different methods depending on what their hardware supports. GFX will use them when available for better performance.

You tell gfx what you want by exposing a caps "using" and filling it with what you support like this:

Supports only batching:

C++

using caps = gfx::gfx_caps<false,false,true,false,false,false,false>;

Supports everything but read ops and direct blts:

C++

using caps = gfx::gfx_caps<false,false,true,true,false,false,false>;

Each of those fields has a named argument to tell you what it is, so it's not as confusing as it looks.

If you mark a field as true, you *must* implement the methods necessary to support that feature. For example, if you mark your target as readable, you must implement the method to read a pixel from its framebuffer.

This caps template is examined through partial template specializations, and the different methods are called based on its values. It chooses the best method it can to perform an operation, and falls back when the caps doesn't support it.

This isn't really possible to do any other way, except at runtime, and that comes with lots of overhead and additional code size for unused code.

Real programmers use butterflies

Greg Utas14-Nov-21 0:58

Greg Utas

14-Nov-21 0:58

This is graphics related, so it took me a moment to realize that caps meant capabilities, not capitalization. Smile | :)

I think you want to flip the trues and falses in the second example.

I assume the non-template way would be to have an abstract Driver class define virtual functions that return false, with each driver overriding the ones that it supports and returning true?

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.

honey the codewitch14-Nov-21 5:26

14-Nov-21 5:26

No, there's actually more read operations than write. Although posting that did make me realize I turned off asynchronicity on that driver. Edit: Never mind, async is not supported under Arduino. I forgot I was looking at that and not the ESP-IDF driver.

Yes, the non-template way would require a virtual class, but then it couldn't do things like taking non-native pixel formats. Every pixel format is its own type. gfx::rgb_pixel<16> nets you a 16 bit RGB pixel, but it's a different type than gfx::rgb_pixel<24> or gfx::yuv_pixel<24>;

Real programmers use butterflies

modified 14-Nov-21 13:17pm.

englebart14-Nov-21 4:35

englebart

14-Nov-21 4:35

For readability, you could introduce a new type/enum for each parameter that converts to a boolean.

That way it would say BATCH/NOBATCH, etc

If you have a unique type for each cap, then you could not accidentally transpose parameters in the long list of booleans.

I think you could figure out a way that the compiler would enforce it all without making the runtime any larger?

My C++ is very rusty, but I have faith that you can figure it out if you think it has merit. 😊

honey the codewitch14-Nov-21 5:23

Someone answer here, please

14-Nov-21 5:23

Yeah, I could do that, but with approaching 600 downloads here, and with over 30 stars on github i'm not sure I want to make such a big breaking change. caps is used *everywhere*

Real programmers use butterflies

Nelek13-Nov-21 7:29

13-Nov-21 7:29

Random content to just check something.

I need someone to answer here with a message, please

M.D.V.

If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.

ElectronProgrammer13-Nov-21 7:39

ElectronProgrammer

13-Nov-21 7:39

Ping

Nelek13-Nov-21 11:23

13-Nov-21 11:23

Thanks

M.D.V.

Gary R. Wheeler14-Nov-21 2:37

Gary R. Wheeler

14-Nov-21 2:37

Pong[^]

Software Zen: delete this;

0x01AA13-Nov-21 7:44

0x01AA

13-Nov-21 7:44

No you can't delete your post because it has sub items. OTH you can while playing with the URL

[Edit]
Doing that will not delete the thread, it will only delete the content of the top message and show it something like 'Message Closed'

modified 13-Nov-21 13:52pm.

OriginalGriff13-Nov-21 8:39

13-Nov-21 8:39

@Nelek is a Protector - so he has additional privileges and powers. One of those is the power to delete any post, which means he can delete the whole thread (message by message) if he wants to.

"I have no idea what I did, but I'm taking full credit for it." - ThisOldTony
"Common sense is so rare these days, it should be classified as a super power" - Random T-shirt
AntiTwitter: @DalekDave is now a follower!

0x01AA13-Nov-21 8:50

0x01AA

13-Nov-21 8:50

*lol* then protector should be aware to be protectors and keep that in mind when comenting to simple members Wink | ;)

OriginalGriff13-Nov-21 8:55

13-Nov-21 8:55

It's possible he doesn't know he can: you don't get a manual with the badge! Laugh | :laugh:

0x01AA13-Nov-21 8:57

0x01AA

13-Nov-21 8:57

I'm working on the manual Laugh | :laugh:

OriginalGriff13-Nov-21 9:23

13-Nov-21 9:23

Yeah, right! I think that probably counts as "documentation" and no bugger does that these days! Laugh | :laugh:

Nelek13-Nov-21 11:21

13-Nov-21 11:21

I do know that I can.
But the protector deletion "red X" is a different button, the standard "delete" widget is greyscaled as "not allowed" as Bruno says.
I would have said before I could still "delete" my own messages, but doing it would only delete the content. Now, I can't even click it.

M.D.V.

Gary R. Wheeler14-Nov-21 2:39

Gary R. Wheeler

14-Nov-21 2:39

OriginalGriff wrote:
@Nelek is a Protector

Protector[^]?

Software Zen: delete this;

OriginalGriff14-Nov-21 2:53

14-Nov-21 2:53

Yep. Hadn't you noticed the leathery skin, swollen joints, and so forth - or did you think that was just the Spanish sun? Laugh | :laugh:

Nelek14-Nov-21 6:18