Repository navigation

Add online CUDA lookup and GPU/cuCollections benchmark scaffolding - #83

Merged

tpn merged 38 commits into

mainfrom

78-online-cuda-gpu-bench

Apr 18, 2026

Merged

Add online CUDA lookup and GPU/cuCollections benchmark scaffolding#83
tpn merged 38 commits into
mainfrom
78-online-cuda-gpu-bench

Commits on Apr 9, 2026
- Add online CUDA lookup and GPU benchmark scaffolding
  tpn
  committed
  4feb671
- Fix benchmark review findings
  tpn
  committed
  28d8058
- Fix CUDA emitter and verifier bugs
  tpn
  committed
  a8c6bce
- Fix portability and emitter edge cases
  tpn
  committed
  690f2c7
- Fix emitter string and empty-build guards
  tpn
  committed
  48bd4a9
- Tighten benchmark verification and accounting
  tpn
  committed
  53fa8d0
- Fix exports and emitted CUDA naming
  tpn
  committed
  ae21cf3
- Fix query extraction and probe stream limits
  tpn
  committed
  3e80d18
- Fix runtime compatibility edge cases
  tpn
  committed
  51fed78
- Fix host probe arithmetic parity
  tpn
  committed
  bb9f56d
- Fix benchmark runtime edge cases
  tpn
  committed
  1541cb3
- Make CPU verification opt-in by default
  tpn
  committed
  3e8983d
- Fix public metadata mapping and ratio guards
  tpn
  committed
  9caf4d9
- Fix example portability includes
  tpn
  committed
  dd8398b
- Harden extractor inputs and emitter contract
  tpn
  committed
  f36e8e6
- Polish review follow-ups
  tpn
  committed
  06ad297
- Apply final review nits
  tpn
  committed
  a1dff39
- Address PR review thread feedback
  tpn
  committed
  06826de
- Tighten experimental lookup mode handling
  tpn
  committed
  acaf720
- Fix CI and verifier assumptions
  tpn
  committed
  de24af1
- Make direct probe validation bounds-safe
  tpn
  committed
  05a6fa4
- Relax benchmark verification assumptions
  tpn
  committed
  2226fd5
- Enforce unique build domains in extractors
  tpn
  committed
  c56b0e2
- Handle tail items and subset metadata explicitly
  tpn
  committed
  ab47462
- Fix experimental kernel tail handling and Q21 extraction
  tpn
  committed
  59e8884
- Harden tail handling and split timing
  tpn
  committed
  5ace6ee
- Restrict verification to safe modes
  tpn
  committed
  86b6145
- Guard split gather and CPU baseline assumptions
  tpn
  committed
  8878af3
- Fix emitted namespace and Q21 grouping
  tpn
  committed
  b640ba9
- Use table hash metadata for host analysis
  tpn
  committed
  dd78eb4
- Fix blocksort shared memory accounting
  tpn
  committed
  9145240
- Tighten CUDA source export sizing and copies
  tpn
  committed
  b4640a7
- Fix Windows export path and raise CUDA source reserve
  tpn
  committed
  ea72ebd
Commits on Apr 18, 2026
- Fix Windows copy macros and CUDA file-work source access
  tpn
  committed
  0240f39
- Address Copilot review comments
  tpn
  committed
  94422ff
- Fix current Windows warnings and assigned buffer fallback
  tpn
  committed
  2f19cf6
- Prefer local tree for example fallback builds
  tpn
  committed
  dcdefe3
- examples: prefer online core in jit finder
  tpn
  committed
  4658f37

Commits

Commits on Apr 9, 2026

Commits on Apr 18, 2026