Yes, I was assuming that it wasn't supported on the StrongARM, but then I discovered that it was. So it isn't an unrecognised instruction and therefore won't cause an exception. And Sarah has concisely described the problem: that the accesses aren't presented appropriately and produce bad results, meaning that the strategy of avoiding LDRH presumably employed by Acorn won't work in a world where all the compilers want to generate it.To answer an earlier question from paulb - I suspect there actually isn't a way to trap all LDRH and patch process it - you'd need to have some kind of selective abort system to fault 'unfaulty' instructions, because at the processor level, it'll believe it's all ok.
Well now, does anyone want to make some "trivial" changes to GCC?
Statistics: Posted by paulb — Mon Feb 03, 2025 8:07 pm