Subzero: Allow per-method controls.

Subzero: Allow per-method controls.

Several aspects of compilation can now be controlled with fine per-function granularity:

- Focus -timing on individual functions
- Only translate certain functions
- Enable verbosity only for certain functions
- Force O2 translation for certain functions (with Om1 default)

In addition, -test-status limits the output of -verbose=status.  This is just used to enable lit testing of the RangeSpec class.

The main motivation here is to enable bisection debugging of a PNaCl application running in the browser.  The initial use is to control O2 versus Om1, and could be extended to control things like address mode inference and advanced phi lowering, possibly even controlling at the granularity of the instruction numbers.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4370
R=eholk@chromium.org, jpp@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=dd6dcfaf765dc93ae64ec45d623106f4b3a3c13a

[Jim Stichnoth, 2016-04-18 07:37:33]

Description was changed from

==========
Subzero: Allow per-method controls.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4370
==========

to

==========
Subzero: Allow per-method controls.

Several aspects of compilation can now be controlled with fine per-function
granularity:

- Focus -timing on individual functions
- Only translate certain functions
- Enable verbosity only for certain functions
- Force O2 translation for certain functions (with Om1 default)

In addition, -test-status limits the output of -verbose=status.  This is just
used to enable lit testing of the RangeSpec class.

The main motivation here is to enable bisection debugging of a PNaCl application
running in the browser.  The initial use is to control O2 versus Om1, and could
be extended to control things like address mode inference and advanced phi
lowering, possibly even controlling at the granularity of the instruction
numbers.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4370
==========

[Jim Stichnoth, 2016-04-18 17:01:43]

stichnot@chromium.org changed reviewers:
+ eholk@chromium.org, jpp@chromium.org, kschimpf@google.com, sehr@chromium.org

[John, 2016-04-18 18:48:39]

lgtm

did you think about how to use this for register include/exclude cl flags?

[Eric Holk, 2016-04-18 18:51:49]

lgtm

https://codereview.chromium.org/1900543002/diff/40001/src/IceClFlags.def
File src/IceClFlags.def (right):

https://codereview.chromium.org/1900543002/diff/40001/src/IceClFlags.def#newc...
src/IceClFlags.def:153: cl::desc("Force -O2 for certain functions (assumes
-Om1)"), cl::init(""))  \
Just to make sure I understand how to use this option, the idea is to isolate
codegen bugs by translating with -Om1 for most things and translating a single
problematic function with -O2, right? The goal is to prevent translation bugs in
other functions from affecting the one we're currently looking at.

My first instinct would have been to use -Om1 for one problematic function and
leave the rest as -O2, but after thinking some more I think -Om1 by default is
better.

https://codereview.chromium.org/1900543002/diff/40001/src/IceStringPool.h
File src/IceStringPool.h (right):

https://codereview.chromium.org/1900543002/diff/40001/src/IceStringPool.h#new...
src/IceStringPool.h:20: #include "IceDefs.h" // Ostream
Where does Ostream get used?

[Jim Stichnoth, 2016-04-18 19:49:39]

On 2016/04/18 18:48:39, John wrote:
> lgtm
> 
> did you think about how to use this for register include/exclude cl flags?

ah, no, I did not consider that at all.  If that is workable, I'll change it in
a separate CL.

[Jim Stichnoth, 2016-04-18 19:49:49]

https://codereview.chromium.org/1900543002/diff/40001/src/IceClFlags.def
File src/IceClFlags.def (right):

https://codereview.chromium.org/1900543002/diff/40001/src/IceClFlags.def#newc...
src/IceClFlags.def:153: cl::desc("Force -O2 for certain functions (assumes
-Om1)"), cl::init(""))  \
On 2016/04/18 18:51:49, Eric Holk wrote:
> Just to make sure I understand how to use this option, the idea is to isolate
> codegen bugs by translating with -Om1 for most things and translating a single
> problematic function with -O2, right? The goal is to prevent translation bugs
in
> other functions from affecting the one we're currently looking at.
> 
> My first instinct would have been to use -Om1 for one problematic function and
> leave the rest as -O2, but after thinking some more I think -Om1 by default is
> better.

Right.  At least in my experience, bisection debugging is easiest when you can
express it as "it breaks when I enable a feature for e.g. methods 0:50000, so
try to reduce that range to a single method that breaks it".  Thus use the
reliable Om1 by default and then for O2 for a narrower and narrower range until
a single offending method is found.

https://codereview.chromium.org/1900543002/diff/40001/src/IceStringPool.h
File src/IceStringPool.h (right):

https://codereview.chromium.org/1900543002/diff/40001/src/IceStringPool.h#new...
src/IceStringPool.h:20: #include "IceDefs.h" // Ostream
On 2016/04/18 18:51:49, Eric Holk wrote:
> Where does Ostream get used?

In dump() and the << operator below.  This came up now because IceRangeSpec.h
includes IceStringPool.h but not IceDefs.h, causing a compile error.

[Jim Stichnoth, 2016-04-18 19:52:14]

Description was changed from

==========
Subzero: Allow per-method controls.

Several aspects of compilation can now be controlled with fine per-function
granularity:

- Focus -timing on individual functions
- Only translate certain functions
- Enable verbosity only for certain functions
- Force O2 translation for certain functions (with Om1 default)

In addition, -test-status limits the output of -verbose=status.  This is just
used to enable lit testing of the RangeSpec class.

The main motivation here is to enable bisection debugging of a PNaCl application
running in the browser.  The initial use is to control O2 versus Om1, and could
be extended to control things like address mode inference and advanced phi
lowering, possibly even controlling at the granularity of the instruction
numbers.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4370
==========

to

==========
Subzero: Allow per-method controls.

Several aspects of compilation can now be controlled with fine per-function
granularity:

- Focus -timing on individual functions
- Only translate certain functions
- Enable verbosity only for certain functions
- Force O2 translation for certain functions (with Om1 default)

In addition, -test-status limits the output of -verbose=status.  This is just
used to enable lit testing of the RangeSpec class.

The main motivation here is to enable bisection debugging of a PNaCl application
running in the browser.  The initial use is to control O2 versus Om1, and could
be extended to control things like address mode inference and advanced phi
lowering, possibly even controlling at the granularity of the instruction
numbers.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4370
R=eholk@chromium.org, jpp@chromium.org

Committed:
https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a...
==========

[Jim Stichnoth, 2016-04-18 19:52:18]

Committed patchset #3 (id:40001) manually as
dd6dcfaf765dc93ae64ec45d623106f4b3a3c13a (presubmit successful).