<html>
<head>
<style>
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
FONT-SIZE: 10pt;
FONT-FAMILY:Tahoma
}
</style>
</head>
<body class='hmmessage'>If they fit in an INT32, use an INT32.<BR>
But if they fit in an INT64, that's still probably more efficient than an otherwise "big set".<BR>
Well, of course, there's always the inline vs. non-inline size vs. speed.<BR>
And here I teeter toward hypocricy in taking advantage of two "natural" integer types.<BR>
Heck, generalize it to a list of efficient sizes: 8, 16, 32, 64, pick the smallest that fits, and in the future when otherwise there would have been LONGLONGINT, add 128 to the list.<BR>
<BR>
- Jay<BR><BR><BR>
<BLOCKQUOTE>
<HR id=EC_stopSpelling>
CC: m3devel@elegosoft.com<BR>From: hosking@cs.purdue.edu<BR>To: jayk123@hotmail.com<BR>Subject: Re: [M3devel] small set comparisons understood, now just to understand the front end code..<BR>Date: Mon, 14 Apr 2008 11:22:35 -0400<BR><BR>
<DIV><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV style="WORD-WRAP: break-word"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate"><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV>I would hesitate to implement small sets as LONGINT instead of INTEGER, since there is no guarantee that LONGINT is necessarily efficient, whereas INTEGER is intended to be the same as the natural word size of the target.</DIV>
<DIV><BR></DIV>
<DIV>I could take a look at this but not anytime soon, since I have several other things I need to work on.</DIV>
<DIV><BR></DIV>
<DIV>On Apr 14, 2008, at 10:51 AM, Jay wrote:</DIV></SPAN></SPAN></SPAN></SPAN></SPAN></SPAN></SPAN></SPAN></DIV></SPAN></DIV>
<DIV><BR class=EC_Apple-interchange-newline>
<BLOCKQUOTE><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV class=EC_hmmessage style="FONT-SIZE: 10pt; FONT-FAMILY: Tahoma"><FONT size=2>currently set <,>,<=,>= are implemented merely as<BR>integer <,>,<=,>=</FONT></DIV></SPAN></BLOCKQUOTE>
<DIV><BR></DIV>
<DIV>Yes, that seems quite wrong. I can't imagine things ever worked properly if that is how they are implemented.</DIV>
<DIV><BR></DIV>
<BLOCKQUOTE><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV class=EC_hmmessage style="FONT-SIZE: 10pt; FONT-FAMILY: Tahoma"><FONT size=2><BR> <BR>This is wrong.<BR> <BR>I believe it should be:<BR> <BR>a < b => (a & b) == a<BR>a <= b => (a & b) == a (same as <=)</FONT></DIV></SPAN></BLOCKQUOTE>
<BLOCKQUOTE><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV class=EC_hmmessage style="FONT-SIZE: 10pt; FONT-FAMILY: Tahoma"><FONT size=2>a > b => (a & b) == b<BR>a >= b => (a & b) == b (same as >)</FONT></DIV></SPAN></BLOCKQUOTE>
<DIV><BR></DIV>
<DIV>Probably should be:</DIV>
<DIV><BR></DIV>
<DIV>a <= b => (a & b) == a</DIV>
<DIV>a < b => (a # b) && ((a & b) == a)</DIV>
<DIV>a >= b => (a & b) == b</DIV>
<DIV>a > b => (a # b) && ((a & b) == b)</DIV>
<DIV><BR></DIV>
<BLOCKQUOTE><SPAN class=EC_Apple-style-span style="WORD-SPACING: 0px; FONT: 12px Helvetica; TEXT-TRANSFORM: none; COLOR: rgb(0,0,0); TEXT-INDENT: 0px; WHITE-SPACE: normal; LETTER-SPACING: normal; BORDER-COLLAPSE: separate">
<DIV class=EC_hmmessage style="FONT-SIZE: 10pt; FONT-FAMILY: Tahoma"><FONT size=2><BR> <BR>The bug is in the frontend.<BR> <BR>m3-sys\m3front\src\misc\CG.m3.<BR> <BR>The code should be /something/ like:<BR> <BR>PROCEDURE Set_compare (s: Size; op: Cmp) =<BR>VAR tmp: Val;<BR> swap := FALSE;<BR>BEGIN<BR> (* a op b => BOOLEAN *)<BR> IF Force_pair (commute := TRUE) THEN<BR> op := M3CG.SwappedCompare [op];<BR> END;<BR> IF (s <= Target.Integer.size) THEN<BR> IF (op = Cmp.EQ) OR (op = Cmp.NE) THEN<BR> cg.compare (Target.Word.cg_type, Target.Integer.cg_type, op);<BR> ELSE<BR> (* set a is less than or equal to set b, if all of set a's members are in set b. *)<BR> IF (op = Cmp.LT) OR (op = Cmp.LE) THEN<BR> swap := TRUE;<BR> Swap ();<BR> END;<BR><STRONG> tmp := Pop ();</STRONG><BR><STRONG> Push (tmp);</STRONG><BR> IF swap THEN<BR> Swap ();<BR> END;<BR><STRONG> cg.and (Target.Integer.cg_type);</STRONG><BR><STRONG> Push (tmp);</STRONG><BR> SimpleIndirectLoad (tmp^, Target.Word.cg_type);<BR> EVAL Force_pair (commute := TRUE);<BR><STRONG> cg.compare (Target.Word.cg_type, Target.Integer.cg_type, Cmp.EQ);</STRONG><BR> SPop (1, "Set_compare");<BR> Free (tmp);<BR> END;<BR> ELSE<BR> cg.set_compare (AsBytes (s), op, Target.Integer.cg_type);<BR> END;<BR> SPop (2, "Set_compare");<BR> SPush (Type.Int32);<BR>END Set_compare;<BR> <BR>though this doesn't quite drive all the machinery correctly, since it yields assertion failures in the compiler due to an unbalanced software stack.<BR>upgrade works, but the test case (p155) fails assertions in the integrated backend.<BR> <BR>I'd love to figure this out but have to do other stuff for now.<BR>Anyone (if there is anyone) familiar with what all is being pushed and popped around here should be able to figure it out easily from this mail.<BR>Otherwise I'll stare at more later.<BR> <BR>ps: "small" sets should probably be anything up to the number of bits in longint, rather than int or pointer, since that is probably efficient enough at the next level down. This is tangential.<BR> <BR>- Jay<BR></FONT></DIV></SPAN></BLOCKQUOTE></DIV><BR></BLOCKQUOTE></body>
</html>