#avr Logs

May 08 2019

#avr Calendar

08:47 AM rue_mohr: --
08:54 AM cehteh: mhmpf ... 64bit math pulls in shitloads of libs while i only need little if it, increment/compare mostly
08:55 AM cehteh: eventually i should implement my own lib doing that with less code
09:37 AM splud: cehteh - you'd be far better off manually doing your own increment and rollover checks.
09:37 AM cehteh: yeah planning to do so
09:38 AM splud: breaking the 32 bit into a pair of 16 bit operations, or even series of 8's.
09:38 AM cehteh: still wondering why it pulls in so much code even with -flto
09:38 AM splud: The maths are EXTREMELY slow.
09:39 AM cehteh: i need it only for global timer, 32 bit isnt enough 64bit is too much by far (would last serveral 100000 years :D)
09:39 AM cehteh: so i prolly implement my own 5 or 6 byte based integer for that later
09:39 AM splud: suggest you do a simple program with conditional builds for an 8 bit variable, 16 bit variable, and 32 bit variable, where you increment and compare, then build each with output of intermediate assembly and see for yourself what the overhead is.
09:40 AM splud: Stuff is moving in and out of 8-bit registers
09:40 AM cehteh: nah i just add some generic uint8_t array math
09:40 AM cehteh: after all its a 8 bit cpu 16 bit wont improve significantly over that
09:41 AM cehteh: but for now i just go with uint64_t to make it work
11:25 AM gsi__ is now known as gsi_
12:54 PM davor_ is now known as davor
04:55 PM cehteh: hah -O0 and disabling lto gives readable assembler code ... but over twice the size :D
11:22 PM day__ is now known as day