Reduce the amount of llvm IR instantiated#205

Merged

bors merged 33 commits intorust-lang:masterfrom Marwes:smaller_ir

Jan 26, 2021

+689-507

Commits on Jan 23, 2021

chore: Add a benchmark for growing a map while inserting to it

Markus Westerlind
authored and
Marwes
committed
refactor: Extract a non generic part of RawTable

Markus Westerlind
authored and
Marwes
committed
perf(compile): Extract part of rehash_in_place

Markus Westerlind
authored and
Marwes
committed
perf(compile): find_insert_slot does not depend on T

Markus Westerlind
authored and
Marwes
committed
perf(compile): Extract more parts out of rehash_in_place

Markus Westerlind
authored and
Marwes
committed
perf(compile): Move part of fallible_with_capacity out

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make calculate_layout non-generic

Markus Westerlind
authored and
Marwes
committed
perf(compiler): Make the panic guard non-generic on rehash

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make raw_iter_hash less generic

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make resize less generic

Markus Westerlind
authored and
Marwes
committed
perf(compiler): Make RawIterRange less generic

Markus Westerlind
authored and
Marwes
committed
clear_no_drop does not need to be generic

Markus Westerlind
authored and
Marwes
committed
refactor: Merge the drop loop into a function

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make the resize panic guard less generic

Markus Westerlind
authored and
Marwes
committed
perf(compiler): Shrink rehash_in_place a bit more

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make erase less generic

Markus Westerlind
authored and
Marwes
committed
cleanup

Markus Westerlind
authored and
Marwes
committed
Restore performance of the RawIterRange

Markus Westerlind
authored and
Marwes
committed
Fix performance on raw_iter_hash

Markus Westerlind
authored and
Marwes
committed
chore: Ensure the raw feature is compiled on CI

Markus Westerlind
authored and
Marwes
committed
Revert the iteration to be generic again

Markus Westerlind
authored and
Marwes
committed
perf(compile): Make rehash_in_place smaller

Markus Westerlind
authored and
Marwes
committed
fix rayon feature

Markus Westerlind
authored and
Marwes
committed
perf: Avoid some re-hashing in rehash_in_place

Markus Westerlind
authored and
Marwes
committed
perf(compile): Parameterize by Layout instead of T

Markus Westerlind
authored and
Marwes
committed
perf(compile): Shrink resize

Markus Westerlind
authored and
Marwes
committed
perf(compile): Move h2 calls into a dedicated set_ctrl_h2 function

Markus Westerlind
authored and
Marwes
committed
refactor: Merge set_ctrl_h2 into find_insert_slot when possible

Markus Westerlind
authored and
Marwes
committed
Finish up into something that does not regress performance

Markus Westerlind
authored and
Marwes
committed

Commits on Jan 24, 2021

Commits on Jan 26, 2021

Try specializing the drop in rehash_in_place again
Markus Westerlind
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Reduce the amount of llvm IR instantiated#205

Commits on Jan 23, 2021

chore: Add a benchmark for growing a map while inserting to it

refactor: Extract a non generic part of RawTable

perf(compile): Extract part of rehash_in_place

perf(compile): find_insert_slot does not depend on T

perf(compile): Extract more parts out of rehash_in_place

perf(compile): Move part of fallible_with_capacity out

perf(compile): Make calculate_layout non-generic

perf(compiler): Make the panic guard non-generic on rehash

perf(compile): Make raw_iter_hash less generic

perf(compile): Make resize less generic

perf(compiler): Make RawIterRange less generic

clear_no_drop does not need to be generic

refactor: Merge the drop loop into a function

perf(compile): Make the resize panic guard less generic

perf(compiler): Shrink rehash_in_place a bit more

perf(compile): Make erase less generic

cleanup

Restore performance of the RawIterRange

Fix performance on raw_iter_hash

chore: Ensure the raw feature is compiled on CI

Revert the iteration to be generic again

perf(compile): Make rehash_in_place smaller

fix rayon feature

perf: Avoid some re-hashing in rehash_in_place

perf(compile): Parameterize by Layout instead of T

perf(compile): Shrink resize

perf(compile): Move h2 calls into a dedicated set_ctrl_h2 function

refactor: Merge set_ctrl_h2 into find_insert_slot when possible

Finish up into something that does not regress performance

Commits on Jan 24, 2021

refactor: Remove calcualuate_layout for the nightly feature

Add inline on prepare_rehash_in_place

Explain the fn argument

Commits on Jan 26, 2021

Try specializing the drop in rehash_in_place again