[miscco]

This uses the new machinery only on find

---

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[coderabbitai]

📝 Summary

Summary by CodeRabbit

• Refactor
  • Updated internal input processing for certain eligible data types. This is an implementation-level change; no user-facing behavior changes are noted. The release does not add or remove any documented capabilities, and no changes to public interfaces are reported.

AgentFind now uses CUDA’s byte-oriented vector type for vectorized loads. Vectorization eligibility excludes volatile input values. The alignment check and vectorized load use the class-level vector type.

Changes: AgentFind vectorization

Layer / File(s)	Summary
Vector type and vectorization conditions cub/cub/agent/agent_find.cuh	VectorT now uses cuda::__as_bytes_vectorized_t. Vectorization eligibility requires contiguous iterators and trivially copyable, non-volatile input values. The alignment check and vectorized-load path use the class-level VectorT.

Suggested reviewers: bernhardmgruber

Priority: Low

Change: Refactor

Merge Risk: Low · up to 96300

AgentFind can fail to compile for volatile pointer inputs; guard the iterator reference before merging.

- [ ] Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

---

ℹ️ Review info

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: vectorized&#95;agent&#95;reduce
• File: .coderabbit.yaml
• Commit: 07211fb
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 0b4792af-ec19-4c19-bf93-845fab5030d1

📥 Commits

Reviewing files that changed from the base of the PR and between 07211fb and fcbff2f.

📒 Files selected for processing: 1

• cub/cub/agent/agent_find.cuh

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 4 remain after this review.

[coderabbitai]

🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

🔎 Supported by static analysis

🏁 Script executed:

git diff --find-renames 07211fb006a72bcb8a4e628d78214d1fed311934 fcbff2fed03c788fd594491093f7137b49274a39 -- cub/cub/agent/agent_find.cuh
sed -n '1,125p' cub/cub/agent/agent_find.cuh
rg -n "__as_bytes_vectorized_t|__is_vectorizable|LoadDirectBlockedVectorized" cub cuda include 2>/dev/null | head -120

Repository: NVIDIA/cccl

Length of output: 9215

---

🏁 Script executed:

set -eu
printf '%s\n' '--- as_bytes_vectorized and trait definitions ---'
rg -n -F --glob '*.h' --glob '*.cuh' -e '__as_bytes_vectorized_t' -e '__is_vectorizable' cuda libcudacxx cub | head -120
for f in $(rg -l -F --glob '*.h' --glob '*.cuh' -e '__as_bytes_vectorized_t' cuda libcudacxx 2>/dev/null | head -10); do
  echo "--- $f"
  sed -n '1,240p' "$f"
done
printf '%s\n' '--- CubVector definition ---'
rg -n -F -- 'struct CubVector' cub/cub
rg -n -F -- 'CubVector<' cub/cub/detail cub/cub/agent/agent_find.cuh | head -40
printf '%s\n' '--- complete AgentFind implementation ---'
nl -ba cub/cub/agent/agent_find.cuh | sed -n '1,210p'
printf '%s\n' '--- AgentFind instantiations and iterator contracts ---'
rg -n -F --glob '*.cuh' --glob '*.cu' -e 'detail::find::agent_t' -e 'find::agent_t' -e 'AgentFind' cub/test cub/cub | head -160
printf '%s\n' '--- vectorized load implementation ---'
nl -ba cub/cub/block/block_load.cuh | sed -n '175,295p'

Repository: NVIDIA/cccl

Length of output: 35784

---

🏁 Script executed:

set -eu
printf '%s\n' '--- CacheModifiedInputIterator ---'
rg -n -F -- 'class CacheModifiedInputIterator' cub/cub/iterator
f=$(rg -l -F -- 'class CacheModifiedInputIterator' cub/cub/iterator | head -1)
nl -ba "$f" | sed -n '1,230p'
printf '%s\n' '--- alignment predicate ---'
rg -n -F -- 'is_aligned' libcudacxx/include/cuda/__memory cub/cub | head -80
for f in $(rg -l -F -- 'is_aligned' libcudacxx/include/cuda/__memory 2>/dev/null | head -10); do
  echo "--- $f"
  nl -ba "$f" | sed -n '1,180p'
done
printf '%s\n' '--- comparable AgentReduce vector load ---'
nl -ba cub/cub/agent/agent_reduce.cuh | sed -n '180,285p'
printf '%s\n' '--- AgentFind base and current diff ---'
git show 07211fb006a72bcb8a4e628d78214d1fed311934:cub/cub/agent/agent_find.cuh | nl -ba | sed -n '25,155p'
git diff --find-renames 07211fb006a72bcb8a4e628d78214d1fed311934 fcbff2fed03c788fd594491093f7137b49274a39 -- cub/cub/agent/agent_find.cuh
printf '%s\n' '--- vector alias uses and AgentFind callers ---'
rg -n -F --glob '*.cuh' --glob '*.cu' -e 'agent_t<' -e 'agent_t::' -e 'AgentFind' cub | head -200

Repository: NVIDIA/cccl

Length of output: 38706

---

🏁 Script executed:

set -eu
printf '%s\n' '--- AgentFind policy and dispatch ---'
nl -ba cub/cub/device/dispatch/dispatch_find.cuh | sed -n '1,150p'
rg -n -F --glob '*.cuh' --glob '*.cu' -e 'AgentFindPolicy' -e 'VECTOR_LOAD_LENGTH' -e 'agent_t<' -e 'PolicyHub' cub/cub/device cub/cub/agent | head -240
printf '%s\n' '--- CubVector definition ---'
nl -ba cub/cub/detail/cub_vector.cuh | sed -n '1,210p'
printf '%s\n' '--- ThreadLoad implementation ---'
rg -n -F -- 'ThreadLoad' cub/cub/thread/thread_load.cuh | head -40
nl -ba cub/cub/thread/thread_load.cuh | sed -n '1,260p'
printf '%s\n' '--- alignment validation ---'
nl -ba libcudacxx/include/cuda/__memory/is_valid_alignment.h | sed -n '1,130p'
printf '%s\n' '--- AgentFind policy declarations ---'
rg -n -F --glob '*.cuh' --glob '*.cu' -e 'find_policy' -e 'FindPolicy' -e 'AgentFind' cub | head -240

Repository: NVIDIA/cccl

Length of output: 41935

---

🏁 Script executed:

set -eu
printf '%s\n' '--- Find tuning policies ---'
nl -ba cub/cub/device/dispatch/tuning/tuning_find.cuh | sed -n '1,180p'
printf '%s\n' '--- full CubVector type declarations ---'
nl -ba cub/cub/detail/cub_vector.cuh | sed -n '20,190p'
printf '%s\n' '--- remaining ThreadLoad implementation ---'
nl -ba cub/cub/thread/thread_load.cuh | sed -n '255,390p'
printf '%s\n' '--- Find policy tests/custom policy values ---'
nl -ba cub/test/catch2_test_device_find_env_api.cu | sed -n '95,160p'
rg -n -F --glob '*.cuh' --glob '*.cu' -e 'FindIfPolicy<' -e 'vec_size' cub/cub/device/dispatch/tuning cub/test/catch2_test_device_find* | head -180

Repository: NVIDIA/cccl

Length of output: 28769

---

important: Use alignof(VectorT) for the alignment check.

If a custom FindIfPolicy sets vec_size to 3, InputT = float passes the vectorization gate and VectorT has size 12. cuda::is_aligned requires a power-of-two alignment, so passing sizeof(VectorT) can trigger its assertion and applies an invalid alignment mask.

Suggested fix

-      return full_tile && ::cuda::is_aligned(d_in + tile_offset, sizeof(VectorT));
+      return full_tile && ::cuda::is_aligned(d_in + tile_offset, alignof(VectorT));

[coderabbitai]

Actionable comments posted: 1

---

Review info

Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: vectorized&#95;agent&#95;reduce
• File: .coderabbit.yaml
• Commit: b01a64d
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: f4f161a5-9e9e-4ba7-9c42-52940511ac30

Commits

Reviewing files that changed from the base of the PR and between 8c405cb and 9630079.

Files selected for processing: 1

• cub/cub/agent/agent_find.cuh

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 9 remain after this review.

[coderabbitai]

Minor · Functional Correctness · Quick win

🔎 Supported by static analysis

Script executed:

git diff --find-renames b01a64d1a43bc665f3973058e583474c61c08eaf3 963007903d25758fbed9cd269c35faf3b5e7f22f -- cub/cub/agent/agent_find.cuh
sed -n '1,125p' cub/cub/agent/agent_find.cuh
rg -n 'CacheModifiedInputIterator|LoadDirectBlockedVectorized|__as_bytes_vectorized_t|iterator_traits' cub/cub/agent libcudacxx/include/cuda/std/__iterator libcudacxx/include/cuda 2>/dev/null | head -240

Repository: NVIDIA/cccl

Length of output: 39114

---

Script executed:

set -eu
printf '%s\n' '--- iterator_traits raw pointer specialization ---'
nl -ba libcudacxx/include/cuda/std/__iterator/iterator_traits.h | sed -n '480,515p'
printf '%s\n' '--- contiguous iterator trait binding ---'
rg -n -F --glob '*.{h,hpp,cuh,cpp}' -e 'is_contiguous_iterator' thrust libcudacxx cub | head -80
printf '%s\n' '--- CacheModifiedInputIterator declaration and load path ---'
rg -n -F --glob '*.{h,hpp,cuh}' -e 'class CacheModifiedInputIterator' -e 'struct CacheModifiedInputIterator' -e 'operator[]' cub/cub/iterator/cache_modified_input_iterator.cuh cub/cub/thread/thread_load.cuh
nl -ba cub/cub/iterator/cache_modified_input_iterator.cuh | sed -n '1,220p'
nl -ba cub/cub/thread/thread_load.cuh | sed -n '1,180p'
printf '%s\n' '--- byte-vectorized type definition ---'
nl -ba libcudacxx/include/cuda/__memory/as_bytes_vectorized.h | sed -n '1,180p'
printf '%s\n' '--- AgentFind complete relevant block ---'
nl -ba cub/cub/agent/agent_find.cuh | sed -n '30,125p'

Repository: NVIDIA/cccl

Length of output: 38739

---

Script executed:

set -eu
printf '%s\n' '--- contiguous iterator implementation ---'
nl -ba thrust/thrust/type_traits/is_contiguous_iterator.h | sed -n '70,140p'
printf '%s\n' '--- ThreadLoad overloads and generic dispatch ---'
nl -ba cub/cub/thread/thread_load.cuh | sed -n '170,390p'
printf '%s\n' '--- CacheModifiedInputIterator contiguous unwrap helper ---'
nl -ba cub/cub/iterator/cache_modified_input_iterator.cuh | sed -n '220,250p'

Repository: NVIDIA/cccl

Length of output: 19586

---

Script executed:

set -eu
if command -v g++ >/dev/null 2>&1; then
  cat >/tmp/volatile_reinterpret_cast_probe.cpp <<'EOF'
struct V {};
void f(volatile int* p) {
  auto q = reinterpret_cast<const V*>(p);
  (void)q;
}
EOF
  if g++ -std=c++17 -fsyntax-only /tmp/volatile_reinterpret_cast_probe.cpp >/tmp/volatile_probe.out 2>&1; then
    echo 'g++ accepted the cv-removing reinterpret_cast'
  else
    echo 'g++ rejected the cv-removing reinterpret_cast'
    cat /tmp/volatile_probe.out
  fi
else
  echo 'g++ unavailable'
fi
printf '%s\n' '--- AgentFind dispatch to ConsumeTile ---'
rg -n -F -- 'ConsumeTile(' cub/cub/agent/agent_find.cuh
nl -ba cub/cub/agent/agent_find.cuh | sed -n '120,230p'

Repository: NVIDIA/cccl

Length of output: 4104

---

important: Guard against volatile iterator references.

For volatile T*, value_type is non-volatile, but reference is volatile T&, so attempt_vectorization remains true. Process can select the vectorized ConsumeTile branch, where reinterpret_cast<const VectorT*>(d_in + ...) removes volatile. This cast is ill-formed, so the volatile specialization can fail to compile rather than perform a non-volatile read.

Check remove_reference_t<typename iterator_traits<InputIteratorT>::reference> instead. Checking the pointee type is equivalent for native pointers, but the iterator reference is the correct generic check.

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 3h 15m: Pass: 100%/210 | Total: 9d 21h | Max: 3h 15m | Hits: 40%/1288695

See results here.