u-boot-users-bounces@lists.sourceforge.net wrote:
This formula is better at avoiding integer overflow.
Signed-off-by: Gerald Van Baren vanbaren@cideas.com
This is my latest entry in the baud rate rounding dual. Since it doesn't multiply the master BRG clock but instead adds the baud rate scaled by 1/2 the clock multiplier, it should not overflow (for a master clock right at the edge of overflowing itself, it still will overflow, but that is pretty unlikely).
This compiles OK on the mpc7448hpc2. I have only tested it on a calculator. I have NOT tested it on real hardware.
I have tested it on my board and it works.
I also tested it on my board by using the clock value of the mpc7448hpc2 board, and printing the computed value which was correct.
drivers/serial/serial.c | 16 +++++----------- 1 files changed, 5 insertions(+), 11 deletions(-) ...
- /* Compute divisor value. Normally, we should simply return:
* CFG_NS16550_CLK) / MODE_X_DIV / gd->baudrate* but we need to round that value by adding 0.5 (2/4).* Rounding is especially important at high baud rates.
- /*
* Compute divisor value, rounding it properly. Rounding is* especially important at high baud rates.
Please keep the comments explaining the base formula: it is not easy to reverse-engineer an optimized formula to figure out what is the base formula.
I will resubmit a patch with your fix and my comments.
Hugo V.