I'm currently workink on a speedup of my TFT lib. I use the DMA to transfer the pixel to the spi. It work - cls is 6 times faster - but on my final tests I came across a problem :
I use malloc to get a buffer for the pixel data of a char. The font only store bit data which has to be repaced by 16 bit colors. It work for small fonts, but if I use a big font the malloc stop. For 42 * 35 pixel I need 1470 words. It work for some chars, but then the malloc stops.
I check all malloc / free and it looks good.
If I look on the addresses I get from malloc, I see that I get a different address for the last working char ...
There has to be a bug in Mbed malloc / free.
I will try to export the code to the offline compiler to check.
Peter
I'm currently workink on a speedup of my TFT lib. I use the DMA to transfer the pixel to the spi. It work - cls is 6 times faster - but on my final tests I came across a problem :
I use malloc to get a buffer for the pixel data of a char. The font only store bit data which has to be repaced by 16 bit colors. It work for small fonts, but if I use a big font the malloc stop. For 42 * 35 pixel I need 1470 words. It work for some chars, but then the malloc stops. I check all malloc / free and it looks good. If I look on the addresses I get from malloc, I see that I get a different address for the last working char ... There has to be a bug in Mbed malloc / free.
I will try to export the code to the offline compiler to check.
Peter