two tiny potential bugs
Thanks. Both are good catches: The first is a bug that might result in wrong code being generated.
The second one is more subtle: I still remember, when I wrote that code, I thought that ld a, yl, is twice as expensive as exg a, yl, so I could just fall through to the exg case below without losing anything. But I only thought of code size. In terms of speed, the ld a, yl is actually faster by one cycle than the double exg a, yl.
Applied in revision #8954.
Reverted because something got broken.
I'll have another look at this after the 3.4.0 release.
I have done regression test for stm8 on both rev8954 and rev8955 locally, it seemed that rev8954 passed but rev8955 got 4 failures. So should this patch (8953 -> 8954) itself be OK ?
I applied the patch in the sdcc-stm8 branch, which I created for changes to the stm8 backend. The branch can be merged after the 3.4.0 release.
You can find the sdcc-stm8 branch using:
svn co https://svn.code.sf.net/p/sdcc/code/branches/sdcc-stm8/sdcc sdcc-stm8
Log in to post a comment.
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.