Comment 1 for bug 490792

In previous discussions with Adobe we decided that Thumb1 would not be a good target for a JIT. The code cache is supposed to contain code that is inherently hot, so trading off compactness for performance for hot code seems silly. Hence we removed Thumb1. We also discussed that Thumb2 most likely would actually be an improvement, so we want a Thumb2 target. I am not sure how to do the configuration. We seem to be drifting towards runtime configuration (vlad just added VFP detection). If the backend is small enough we might tolerate having a Thumb2 code generation mode in parallel to regular ARM. Any comments? vlad? ed?