Nulo/qmk_firmware
1
0
Fork 0
Code Releases Activity

Fix bit loss in cie_lightness() when doing division to resolve #15331 (#15344)

Browse source 
* Fix bit loss in cie_lightness() when doing division.

* Use the right types

* Format

Co-authored-by: zvecr <git@zvecr.com>
This commit is contained in:
Karl Shea 2021-11-29 09:07:36 -06:00 committed by GitHub
parent 23d3ebd78a
commit 84b8cdc1a4
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
quantum/backlight/backlight_avr.c
View file

@ -232,19 +232,19 @@ ISR(TIMERx_OVF_vect) {
// See http://jared.geek.nz/2013/feb/linear-led-pwm
static uint16_t cie_lightness(uint16_t v) {
if (v <= ICRx / 12) // If the value is less than or equal to ~8% of max
if (v <= (uint32_t)ICRx / 12) // If the value is less than or equal to ~8% of max
{
return v / 9; // Same as dividing by 900%
} else {
// In the next two lines values are bit-shifted. This is to avoid loosing decimals in integer math.
uint32_t y = (((uint32_t)v + ICRx / 6) << 5) / (ICRx / 6 + ICRx); // If above 8%, add ~16% of max, and normalize with (max + ~16% max)
uint32_t out = (y * y * y * ICRx) >> 15; // Cube it and undo the bit-shifting. (which is now three times as much due to the cubing)
uint32_t y = (((uint32_t)v + (uint32_t)ICRx / 6) << 5) / ((uint32_t)ICRx / 6 + ICRx); // If above 8%, add ~16% of max, and normalize with (max + ~16% max)
uint32_t out = (y * y * y * ICRx) >> 15; // Cube it and undo the bit-shifting. (which is now three times as much due to the cubing)
if (out > ICRx) // Avoid overflows
{
out = ICRx;
}
return out;
return (uint16_t)out;
}
}