[WIP] Profiling-guided optimisations#8906

Draft

tautschnig wants to merge 33 commits intodiffblue:developfrom

tautschnig:experiment/hash-optimization

Commits on Mar 28, 2026

Add profiling analysis results and optimization plan

tautschnig
and
kiro-agent
committed
Add string hashing investigation results

tautschnig
and
kiro-agent
committed
Merge hashing investigation into profiling analysis

tautschnig
and
kiro-agent
committed
Replace ostringstream with string concatenation in SSA identifier building

tautschnig
and
kiro-agent
committed
Update profiling analysis with SSA identifier optimization results

tautschnig
and
kiro-agent
committed
Optimize set_level_2 to avoid full SSA identifier rebuild

tautschnig
and
kiro-agent
committed
Optimize set_level_0 and set_level_1 to append suffixes directly

tautschnig
and
kiro-agent
committed
Update profiling analysis with post-optimization investigation

tautschnig
and
kiro-agent
committed
Add alternative allocator investigation: 18-25% speedup with tcmalloc/jemalloc

tautschnig
and
kiro-agent
committed
Add CMake option to auto-detect and link tcmalloc or jemalloc

tautschnig
and
kiro-agent
committed
Remove unused `non_sharing_tree`

thomasspriggs
authored and
tautschnig
committed
Make `irept` hold a union of a pointer or an id

thomasspriggs
authored and
tautschnig
committed
Update profiling analysis with final combined results

tautschnig
and
kiro-agent
committed
Install tcmalloc in Linux CI workflows

tautschnig
and
kiro-agent
committed
Add remaining optimization plans (A-D)

tautschnig
and
kiro-agent
committed
Complete profiling analysis documentation

tautschnig
and
kiro-agent
committed
Cache suffix counter in get_new_name to avoid O(n) symbol table probing

tautschnig
and
kiro-agent
committed
Add free-list pool allocator for tree_nodet

tautschnig
and
kiro-agent
committed
Optimize field_sensitivityt::get_fields loop invariants

tautschnig
and
kiro-agent
committed
Update profiling analysis with post-verification hotspot analysis

tautschnig
and
kiro-agent
committed
Document field_sensitivityt::apply optimization investigation

tautschnig
and
kiro-agent
committed
Document get_fields loop invariant hoisting results

tautschnig
and
kiro-agent
committed
Profile after get_fields optimization: field_sensitivity down to 1.6%

tautschnig
and
kiro-agent
committed
Compare irep hash functions: BASIC, MURMURHASH2A, MURMURHASH3

tautschnig
and
kiro-agent
committed
Investigate reducing operator== calls in merge_irept

tautschnig
and
kiro-agent
committed
Profiling: pre-compile C sources to goto binaries via goto-cc

tautschnig
and
kiro-agent
committed
Analysis: merge frequency reduction and goto-cc workflow

tautschnig
and
kiro-agent
committed
Update profiling analysis with re-verification results

tautschnig
and
kiro-agent
committed
Definitive allocator analysis: tcmalloc 21%, pool 3.2%, combined 22%

tautschnig
and
kiro-agent
committed
Fix contradictions in PROFILING_ANALYSIS.md

tautschnig
and
kiro-agent
committed
Use unordered_set in simplify_inequality_pointer_object

tautschnig
and
kiro-agent
committed
Construct object_descriptor_exprt directly in value_sett::to_expr

tautschnig
and
kiro-agent
committed
Update PROFILING_ANALYSIS.md with comprehensive results and latest profile

tautschnig
and
kiro-agent
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] Profiling-guided optimisations#8906

[WIP] Profiling-guided optimisations#8906
tautschnig wants to merge 33 commits intodiffblue:developfrom
tautschnig:experiment/hash-optimization

Commits on Mar 28, 2026

Add profiling analysis results and optimization plan

Add string hashing investigation results

Merge hashing investigation into profiling analysis

Replace ostringstream with string concatenation in SSA identifier building

Update profiling analysis with SSA identifier optimization results

Optimize set_level_2 to avoid full SSA identifier rebuild

Optimize set_level_0 and set_level_1 to append suffixes directly

Update profiling analysis with post-optimization investigation

Add alternative allocator investigation: 18-25% speedup with tcmalloc/jemalloc

Add CMake option to auto-detect and link tcmalloc or jemalloc

Remove unused `non_sharing_tree`

Make `irept` hold a union of a pointer or an id

Update profiling analysis with final combined results

Install tcmalloc in Linux CI workflows

Add remaining optimization plans (A-D)

Complete profiling analysis documentation

Cache suffix counter in get_new_name to avoid O(n) symbol table probing

Add free-list pool allocator for tree_nodet

Optimize field_sensitivityt::get_fields loop invariants

Update profiling analysis with post-verification hotspot analysis

Document field_sensitivityt::apply optimization investigation

Document get_fields loop invariant hoisting results

Profile after get_fields optimization: field_sensitivity down to 1.6%

Compare irep hash functions: BASIC, MURMURHASH2A, MURMURHASH3

Investigate reducing operator== calls in merge_irept

Profiling: pre-compile C sources to goto binaries via goto-cc

Analysis: merge frequency reduction and goto-cc workflow

Update profiling analysis with re-verification results

Definitive allocator analysis: tcmalloc 21%, pool 3.2%, combined 22%

Fix contradictions in PROFILING_ANALYSIS.md

Use unordered_set in simplify_inequality_pointer_object

Construct object_descriptor_exprt directly in value_sett::to_expr

Update PROFILING_ANALYSIS.md with comprehensive results and latest profile