https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86078
Bug ID: 86078
Summary: Documentation: missing param default values
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: web
Assignee: unassigned at gcc dot gnu.org
Reporter: ptdrnvqd at 10mail dot org
Target Milestone: ---
in
https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
or https://gcc.gnu.org/onlinedocs/gcc-5.5.0/gcc/Optimize-Options.html
or whatever number,
the params section at the bottom does not link/reference
source-code e.g. https://github.com/gcc-mirror/gcc/blob/master/gcc/params.def
it contains max,min and default values, along with description, includes
default values not mentioned in documentation, along with disagreements
e.g.
sched-state-edge-prob-cutoff&selsched-insns-to-rename in docs(gcc6+) changes
name, err correction, but not reflected|updated in older doc versions
could this be automated/scripted (link to source-code) to reduce human errors?
missing default doc
max-crossjump-edges=100
max-delay-slot-insn-search=100
max-delay-slot-live-search=333
max-gcse-memory=50*1024*1024 (2015)
max-gcse-memory=128*1024*1024 (2018)
max-modulo-backtrack-attempts=40
iv-consider-all-candidates-bound=30(2015)
iv-consider-all-candidates-bound=40(2018)
iv-max-considered-uses=250
scev-max-expr-size=100
scev-max-expr-complexity=10
max-partial-antic-length=100
devirt-type-list-size(4.6),ipa-cp-value-list-size(4.7+)=8
max-slsr-cand-scan=50(4.8+)
ipa-max-agg-items=16(4.9+)
ipa-max-aa-steps=25000(5+)
max-ssa-name-query-depth=3(6+)
max-speculative-devirt-maydefs=50(6+)
max-stores-to-merge=64(7+)
dse-max-object-size=256(7+)
Not mentioned @docs except SourceCode
max-tracked-strlens=10000
what impact do they have,
anyway to emit debug info to show whether compilation hit limits?
to generalize, for benchmarks should an average of params be used?
should gcc params better reflect the host-compile-machine instead of a
one-size-fits all (compiling on mobile|desktop|workstation|server, new|old)
another question: when compiling using link-time-optimization-LTO,
since it's bunched into a single unit would it hit the limits sooner and should
it be compensated for against
any mechanism for compile more/less aggressively based on usage patters, how
much (CPU-time|usage/day) a program uses (relevance)
