some definition changes to save some bytes

[night-ghost]

-int analogRead(uint8_t pin)
+unsigned int analogRead(uint8_t pin)

because it's really unsigned

-void attachInterrupt(uint8_t interruptNum, void (_userFunc)(void), int mode) {
+void attachInterrupt(uint8_t interruptNum, void (_userFunc)(void), uint8_t mode) {

because mode has only low values

-int digitalRead(uint8_t pin)
+uint8_t digitalRead(uint8_t pin)

because it returns only 0 or 1

[NicoHood]

doesnt analogwrite() also take an int, while only 0-255 are accepted anyways?

I'd also adapt the changes suggested, if it does not break anything.

[matthijskooijman]

@night-ghost, you say this saves some bytes, you mean the code compiles smaller with these changes?

The proposed changes look good to me, perhaps a pull request along with some statistics about sketch sizes with and without the PR applied would be useful?

[night-ghost]

Yes, transferring of uint8_t takes less commands on 8-bit processor than int, and unneeded sign cause using signed arithmetic in expressions like "4000/analogRead()" which can not be used anywhere else.

[night-ghost]

analogwrite() on other platforms that has hardware DAC uses 10-12bit int

[night-ghost]

"some statistics" - size difference depends on frequency of calls. attachInterrupt saves 2 words on each call, digitalRead one word on each call. Where project size becomes larger than the memory size you'll count each byte

[matthijskooijman]

Thanks, sounds like good savings. Care to prepare a pull-request?

[vbextreme]

If a function returns a value wastes the same time both a int8_t as a int16_t.
With a int8_t the return value will be put into r24 and r25 reset .
With int16_t will be put into r24 and r25 if necessary.
No waste of memory.
To call a function instead you can save a reg , but the memory remains unchanged.

[night-ghost]

if function used as

uint8_t b=digitalRead()

then you right. But when used as

if(digitalRead()==HIGH)

then check compiled as

if(digitalRead(pushButton) == HIGH)
208: lds r24, 0x0106
20c: std Y+1, r22 ; 0x01
20e: call 0x60a ; 0x60a
212: ldd r22, Y+1 ; 0x01
214: subi r24, 0x01 ; 1
216: sbci r25, 0 ; 0

You see unneeded sbci?

[vbextreme]

digitalRead ret 8 bit

 1d0:   81 e0           ldi r24, 0x01   ; 1
 1d2:   09 f4           brne    .+2         ; 0x1d6 <lowdigitalRead+0x44>
 1d4:   80 e0           ldi r24, 0x00   ; 0
 1d6:   df 91           pop r29
 1d8:   cf 91           pop r28
 1da:   08 95           ret

digitalRead ret 16 bit

17e:    81 e0           ldi r24, 0x01   ; 1
 180:   90 e0           ldi r25, 0x00   ; 0
 182:   21 f4           brne    .+8         ; 0x18c <digitalRead+0x4c>
 184:   80 e0           ldi r24, 0x00   ; 0
 186:   02 c0           rjmp    .+4         ; 0x18c <digitalRead+0x4c>
 188:   80 e0           ldi r24, 0x00   ; 0
 18a:   90 e0           ldi r25, 0x00   ; 0
 18c:   df 91           pop r29
 18e:   cf 91           pop r28
 190:   08 95           ret

actually I was wrong.

[tico-tico]

I was going to file the same bug about digitalRead() but found this one. So, guys, can any of you create the pull request? Thanks.

[tico-tico]

Ok, here's my attempt arduino/Arduino#4625 .
I haven't touched interrupts since there is some work on them in the pull section.

[aentinger]

@night-ghost a similiar proposition was made in arduino/Arduino#4625 and refused based on the argument of breaking the Arduino API signator (see arduino/Arduino#4625 (comment) ) therefore I am closing this issue.