From: Michael H. <mic...@ju...> - 2001-08-20 04:50:49
|
On Sun, 19 Aug 2001, Michael Hope wrote: > > Michael I seem to have lost the z80 test case, could you provide it again > > The current fix you have (!TARGET_Z80) may cause inefficient code > > generation. > > The test case is running the dhrystone test in support/tests/dhrystone for > the z80 port. Turning on loop induction breaks it. You'll need the rrz80 > emulator from the sdcc-extra module in cvs. > > -- Michael Talking to myself, but ah well. I've stripped the dhrystone code down to the minimum that will reproduce the problem. I've added a bug about the problem (#453196) and have created a test case in the regression tests (support/regression/tests/bug-453196.c) that reproduces it. It's a funky one - the code is about ~35 lines long, and removing pretty much any line obscures the problem. -- Michael |