Hello,
I noticed the huge commit 2f8a2487597092e2a7df04f176f7b03e5788dd59 ("Merge
master-with-mute branch, set version to 0.12.0.") which does not have any
documentation, so I tried to look at this patch to understand what it does
and now I have a couple of questions. Plus I have some concerns about the
changes in compile_scal.cpp.
------------------------------------------------------------------------------
Firtsly, the change in generatePrefix() is obviously wrong. I understand that
nobody cares, this primitive was always broken, but now even this faust code
process = prefix(1,2);
can't be compiled.
--- a/compiler/generator/compile_scal.cpp
+++ b/compiler/generator/compile_scal.cpp
@@ -1094,7 +1094,7 @@ string ScalarCompiler::generatePrefix (Tree sig,
Tree x, Tree e)
fClass->addDeclCode(subst("$0 \t$1;", type, vperm));
fClass->addInitCode(subst("$0 = $1;", vperm, CS(x)));
- fClass->addInitCode(subst("$0 \t$1;", type, vtemp));
+ fClass->addDeclCode(subst("$0 \t$1;", type, vtemp));
fClass->addExecCode(Statement(getConditionCode(sig), subst("$0 =
$1;", vtemp, vperm)));
fixes this particular problem. But as I said it is still broken.
-------------------------------------------------------------------------------
Now lets look at the change in ScalarCompiler::generateVariableStore(). I guess
it connects to the "strange" semantics of control primitive (see below) but this
looks obviously unfriendly performance-wise. Even this trivial example
import("stdfaust.lib");
process = ma.nextafter <: +(1),-(1);
shows the difference. "faust -a bench.cpp" reports
before "Merge master-with-mute branch":
53.143 425.144 MB/s inputs 425.144 MB/s outputs 51.8919 38.596
53.3681 426.945 MB/s inputs 426.945 MB/s outputs 52.0998 41.9565
53.0742 424.593 MB/s inputs 424.593 MB/s outputs 51.8304 42.4126
after:
52.3618 418.894 MB/s inputs 418.894 MB/s outputs 51.1408 32.3092
52.2949 418.359 MB/s inputs 418.359 MB/s outputs 51.3026 39.413
52.4288 419.43 MB/s inputs 419.43 MB/s outputs 51.5565 46.3348
-------------------------------------------------------------------------------
Finally, could you please explain control/enable primitives added by that
commit?
OK, I understand what they do. Say, control(x,y) is roughly
control(x,y) = select2(y, _, x) ~ _;
_except_ "x" is not computed when y == F and nothing else needs the value of
"x".
OK, looks useful, but:
- why it depends on the new "-es" option added by the same patch?
"-es 0" turns control() into identity function, and this changes
its semantics completely.
Not to mention this contradicts with "faust --help" which says
"use --enable-semantics 1|0 when 1 and simple multiplication
otherwise"
- why it can not be implemented without penalizing
generateVariableStore ?
Thanks,
Oleg.
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Faudiostream-devel mailing list
Faudiostream-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/faudiostream-devel
