Change --hydrogen-filter to allow specifying a negative filter for optimized functions.

Change --hydrogen-filter to allow specifying a negative filter for optimized functions.

When prepending a '-' to the function name, the function will not be optimized.

--hydrogen-filter=foo works as before.

--hydrogen-filter=-foo means don't optimize foo.
Committed: https://code.google.com/p/v8/source/detail?r=11036

[Sven Panne, 2012-03-14 07:35:35]

I think the whole code for matching against FLAG_hydrogen_stats should be moved
into a separate function and this should be used everywhere instead of direct
accesses to this flag, something like

   bool MatchAgainstHydrogenFilter(Handle<string> name)

This can handle the normal/inverted logic consistently.

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc
File src/compiler.cc (right):

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc#newcode248
src/compiler.cc:248: if (*FLAG_hydrogen_filter) {
If this is allowed, I'll immediately start writing 'if (ptr) ...' for ptr being
an expression with a pointer type (something I really like, but which is
forbidden under Google law, I think). :-)

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc#newcode250
src/compiler.cc:250: ASSERT(!filter.is_empty());
I don't think we need this assertion when the corresponding test is 2 lines
above.

[fschneider, 2012-03-14 08:34:58]

The hydrogen-filter flag should not be used anywhere else. I removed the filter
from the PrintCode function since it should only concern which function gets
optimized, not which code gets printed.

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc
File src/compiler.cc (right):

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc#newcode248
src/compiler.cc:248: if (*FLAG_hydrogen_filter) {
On 2012/03/14 07:35:36, Sven Panne wrote:
> If this is allowed, I'll immediately start writing 'if (ptr) ...' for ptr
being
> an expression with a pointer type (something I really like, but which is
> forbidden under Google law, I think). :-)

I don't know about that, but I can compare it to \0 instead.

https://chromiumcodereview.appspot.com/9691042/diff/1/src/compiler.cc#newcode250
src/compiler.cc:250: ASSERT(!filter.is_empty());
On 2012/03/14 07:35:36, Sven Panne wrote:
> I don't think we need this assertion when the corresponding test is 2 lines
> above.

Done.

[Sven Panne, 2012-03-14 09:12:06]

LGTM with nits: Remove another print-related occurrence of the flag + improve
the flag description in flag-definitions.h a bit, "hydrogen use/trace filter" is
a bit cryptic and doesn't even mention the "-" prefix.

https://chromiumcodereview.appspot.com/9691042/diff/5001/src/codegen.cc
File src/codegen.cc (right):

https://chromiumcodereview.appspot.com/9691042/diff/5001/src/codegen.cc#newco...
src/codegen.cc:127: if (print_code) {
If we remove the flag here, it should probably be removed in MakeCodePrologue,
too.