Remove unused constants from optimized code objects

[abalkin]

BPO	2493
Nosy	@rhettinger, @abalkin
Files	compress-consts.diff: Patch against revision 61948 compress-consts-1.diff: Patch against revision 61948

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/rhettinger'
closed_at = <Date 2008-06-12.22:08:50.084>
created_at = <Date 2008-03-26.14:47:35.009>
labels = ['interpreter-core', 'performance']
title = 'Remove unused constants from optimized code objects'
updated_at = <Date 2008-06-12.22:08:50.062>
user = 'https://github.com/abalkin'

bugs.python.org fields:

activity = <Date 2008-06-12.22:08:50.062>
actor = 'rhettinger'
assignee = 'rhettinger'
closed = True
closed_date = <Date 2008-06-12.22:08:50.084>
closer = 'rhettinger'
components = ['Interpreter Core']
creation = <Date 2008-03-26.14:47:35.009>
creator = 'belopolsky'
dependencies = []
files = ['9867', '9868']
hgrepos = []
issue_num = 2493
keywords = ['patch']
message_count = 3.0
messages = ['64558', '64561', '68101']
nosy_count = 2.0
nosy_names = ['rhettinger', 'belopolsky']
pr_nums = []
priority = 'normal'
resolution = 'rejected'
stage = None
status = 'closed'
superseder = None
type = 'resource usage'
url = 'https://bugs.python.org/issue2493'
versions = ['Python 2.6']

[abalkin]

When peephole optimizer folds multiple constants into one, the old
constants remain in co_consts. Attached patch removes such unused
constants.

[abalkin]

I've noticed that the original patch does not handle the error condition
from failed consts resize correctly. Please see compress-consts-1.diff
for a fix.

[rhettinger]

The issue with unused constants is an artifact of the peepholer running
after the bytecode generation phase. When constant folding was
written, I intentionally left-out a step to remove unused constants
because of the code complexity, the fragility of the process, because
the benefits were inconsequential, and because it added to compilation
time.

It would be better to wait for the AST optimizer to replace the
peepholer. AST optimizations are upstream from bytecode generation, so
the unused constant issue simply disappears.