| --- Log opened Sun Apr 10 00:00:14 2016 | ||
| wallento | olofk: You probably want to only handle sections with allocate flag | 04:32 |
|---|---|---|
| wallento | but even data sections can be 11 bytes large | 04:32 |
| olofk | wallento: Yeah, the verilog backend is a bit hacky, since it doesn't care about that and just pads with 0, but that's fine with me | 05:15 |
| wallento | yes, thats makes sense | 05:17 |
| olofk | But reverse-bytes is an internal parameter used for all conversions, so it can't just pad with zero, since that could be wrong for other formats | 05:20 |
| olofk | Maybe the only option is an explicit endian switch for the verilog backend | 05:22 |
| olofk | or a verilog-specific reverse-bytes switch | 05:22 |
| wallento | how about padding first? | 06:03 |
| wallento | honestly I have no clue about the degrees of freedom | 06:03 |
| wallento | but wouldn't padding to the target width first and then reversing work with these functions? | 06:03 |
| olofk | I thought of that too, but I can't figure out how to pad every section to nearest 32-bit address | 06:21 |
| olofk | There are some padding and aligning functions in objcopy, but none of them seem to do the trick | 06:21 |
| olofk | I also tried to first convert elf to binary and then from binary to verilog with a --pad-to switch, but that didn't work either | 06:26 |
| wallento | yeah, thats probably why everyone writes their python scripts around gcc and binutils to generate .hex, .bin or .vmem | 06:40 |
| -!- p1oooop is now known as bklu | 16:43 | |
| --- Log closed Mon Apr 11 00:00:16 2016 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!