Joerg, I am painfully aware of the consequence of avr-gcc sharing most of its codebase with a compiler primarily aimed at the "big" computers. IMHO it's a pity that the time and energy of volunteers did not went into the avr port of SDCC, which is maybe a more adequate, albeit certainly less mature, background for an mcu-specific compiler.
However, I still believe, that RFEs exist to bring everyday problems of mere users into the more competent developers' attention. Certainly an esoteric and complex problem is more attractive target for a volunteer work, but my hopes are, the volunteers want to see their effort also to bring fruit to others - who might then return ther share perhaps in some other way. To be specific, as 1-, 2- and 4-byte array initialisers are compiled in an optimal ldi/st way, and e.g. 3-bytes are not, I assume this is an optimisation performed in the avr-specific backend, to reduce the overhead when using initialised char-short-long. That's IMHO the place where a bit more work might bring benefit. The string initialiser *might* be a different issue, I see that. Jan Waclawek -----Original Message----- Note that this behaviour is only a pessimization if you've got a target machine like an AVR. For a hosted application on a "mainstream" GCC target machine, this is not an optimization at all, because setting up the data segment comes at virtually no cost there. That might explain why an enhancement request like this one would very likely not trigger much motivation on the average GCC developer. -- cheers, J"org .-.-. --... ...-- -.. . DL8DTL _______________________________________________ AVR-GCC-list mailing list AVR-GCC-list@nongnu.org http://lists.nongnu.org/mailman/listinfo/avr-gcc-list
