[M3commit] CVS Update: cm3

Jay Krell jkrell at elego.de
Tue Feb 23 16:06:37 CET 2010


CVSROOT:	/usr/cvs
Changes by:	jkrell at birch.	10/02/23 16:06:37

Modified files:
	cm3/m3-sys/m3back/src/: Codex86.i3 Codex86.m3 M3BackInt.i3 
	                        M3x86.m3 

Log message:
	another not very useful function in hand.c bites the dust, at least under #ifdef
	
	use bts instruction instead of calling set_singleton
	bts's notion of how to index bits happens to line up with set_singleton (assuming test p230 is complete enough)
	bts isn't fast, but surely this is better overall due to drastically fewer instructions
	note that set_singleton test cases tend to generate very large code zeroing stuff
	
	I was not able to get coverage of several addressing modes,
	so they are left issuing errors and asserts.
	That is, among:
	bts mem, reg
	bts reg, imm8
	bts reg, reg
	possibly bts mem, imm8
	
	I was only able to get bts reg, reg to hit.
	
	Still to ponder gcc backend?
	I had hoped to get this done in the frontend but became stumped. ?
	It's just a shift and and, should be achievable portably. ?




More information about the M3commit mailing list