ESP32 Support for VSPI and HSPI

[alx-uta]

I started using this library in a personal project and I realised that the ESP32 SPI part is missing.

After checking the other library (MCP_ADC) I was able to use the same code style and format.

[RobTillaart]

Thanks for the PR, I will look into the details asap. Might take a few days.

[RobTillaart]

@alx-uta Looks good, I assume you have tested both HSPI and VSPI with ESP32?

Do you have time to run the performance example and add some results to a (new) performance_0.2.4.md ? you can use the performance_0.2.2.md file as a template. Would be appreciated!

I am working on another PR and as this one is affecting less files I will merge this one first.

[alx-uta]

@RobTillaart that's perfect thanks for checking it.

Yes, I tested using MCP.selectVSPI(); and MCP.selectHSPI();.

I'll have a look at the performance file and I'll run the test.

[alx-uta]

@RobTillaart I used the test file. What I'll need to run for SW SPI?

VAL1: 0 VAL8: 67 VAL16: 17219

Action	SW SPI	HW 10 MHz	notes
TEST digitalWrite(0, value)	------	27.56
TEST digitalWrite(pin, value)	------	24.56
TEST digitalRead(pin)	------	18.06
TEST write8(port, mask)	------	19.50
TEST read8(port)	------	21.00
TEST write16(mask)	------	18.00
TEST read16()	------	18.50

[RobTillaart]

To use SW SPI you must change the parameters of the constructor call.in the performance sketch.

// MCP23S17 MCP(10, 12, 11, 13);

You can (re)use the HSPI or VSPI pins.

Expectations are a performance quite close to the HW SPI on an ESP32..

[RobTillaart]

Thanks for taking time to test!

[alx-uta]

@RobTillaart I just added the performance results. Sorry, I didn't realise yesterday that the SW was the Software SPI.

[RobTillaart]

Thanks for the update, appreciated! Merged into master branch, Will merge into the other PR asap.

[alx-uta]

Np. Thanks for this great library :)

[RobTillaart]

Did you notice that on ESP32 the software SPI is not far behind the hardware SPI?

[alx-uta]

@RobTillaart yes, the data was quite nice. Is there a specific part of the code that helped with that? Or, that's because of the esp32?

[RobTillaart]

It's how the ESP32 implemented their digitalWrite() et al. On AVR (UNO) you need to find the right register and then the right bit. The ESP32, it looks like directly mapped upon the IO pins, and of course the higher clock speed compensates a lot.

When looking on a scope you can see that software SPI is slower and has a nicer (more square) signal, IN the hardware SPI you see smaller pulses and thus more the RC effect (curve of the rising edge)

[alx-uta]

Ah, I see. That's quite nice then.

I need to get a scope if I want to continue playing with this.