Replace O(N) recursive sequence_map_inverse with O(1) pack expansion#4288
Closed
assistant-librarian[bot] wants to merge 8 commits intodevelopfrom
Closed
Replace O(N) recursive sequence_map_inverse with O(1) pack expansion#4288assistant-librarian[bot] wants to merge 8 commits intodevelopfrom
assistant-librarian[bot] wants to merge 8 commits intodevelopfrom
Conversation
Co-Authored-By: Claude <noreply@anthropic.com>
…evelop/ROCm_composable_kernel/pr-3596
assistant-librarian
bot
requested review from
andriy-ca,
aosewski,
bartekxk,
carlushuang,
geyyer,
illsilin,
poyenc and
qianfengz
as code owners
assistant-librarian
bot
requested review from
ThomasNing,
afagaj,
asleepzzz,
coderfeli and
tenpercent
as code owners
assistant-librarian
bot
added
the
external contribution
This was referenced Feb 10, 2026
Use explicit for loop instead of fold expression to build the inverse permutation array. This avoids potential issues with fold expression depth limits while maintaining the same O(1) template instantiation depth optimization. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
tenpercent
force-pushed
the
import/develop/ROCm_composable_kernel/pr-3596
branch
from
c7bcd1f to
027528f
Compare
tenpercent
reviewed
Feb 10, 2026
| }; | ||
| InverseArray result{}; | ||
| constexpr index_t input[] = {Is...}; | ||
| for(index_t pos = 0; pos < static_cast<index_t>(sizeof...(Is)); ++pos) |
Contributor
There was a problem hiding this comment.
discussed with @cgmillette and for-loop is more readable while impact on build time is non-measurable
cgmillette
pushed a commit
that referenced
this pull request
Feb 11, 2026
Use @tenpercent's implementation from #4288. This change uses a constexpr for loop to build the inverse in O(N) operations with O(1) template instantiation depth. Co-authored-by: Max Podkorytov <4273004+tenpercent@users.noreply.github.com>
Contributor
|
Merged to #4447 . The new PR has proper CI running |
Contributor
|
Merge with #4447 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replace the O(N) recursive
sequence_map_inverseimplementation with O(1) template depth using pack expansion to reduce compile time (#4229).Approach
constexprloop infind_source_indexto locate permutation inverse indicesWhy It Works
Template recursion requires N template instantiations for N iterations, each with its own overhead. Constexpr loops execute within a single template instantiation, avoiding per-instantiation overhead.
Build Performance Impact
Template Instantiation Reduction (measured on
device_grouped_conv3d_fwd_bias_bnorm_clamp_instancetarget, 248 files):This confirms the optimization successfully reduces template instantiation overhead by eliminating recursive template patterns in favor of pack expansion.
Test Plan
SequenceMapInverse.InverseMapandSequenceMapInverse.InverseIdentityMaptests validate correctness🔁 Imported from ROCm/composable_kernel#3596
🧑💻 Originally authored by @tenpercent