PCL RFC 0002: Better type for indices

Details

• Author: Kunal Tyagi
• Title: PCL-RFC-0002: Better type for indices
• Gitter room: PointCloudLibrary/PCL-RFC-02
• Past discussion: #3351
  • People of Interest: jasjuang, aPonza, taketwo, SergioRAgostinho

Motivation

Currently, there's mixed use of int, long, unsigned int, unsigned long, unsigned long long and std::size_t for indices. Some effort has been made to unify towards std::size_t to reduce the limit of 2 billions points due to int.

Pros

• Follows a best practice
• Allows operations on actually limitless points (if using 64 bit systems)

Cons

• One of the following complication if using GSL
  • Extra build-time dependency on internet
  • Copy-pasted source file
  • git sub-tree/sub-module
• A new typedef using pcl::index_t = std::ptrdiff_t;
• Not platform independent
• Wastes memory: Largest RAM supported is ~32TB (2^48). That's a clear 25% loss of efficiency for extreme cases, and larger for average use-case.

ABI/API Breakage

• None for existing internal migration
• Major API breakage for changing Indices from std::vector<int> to std::vector<index_t>

Effort Required

Minor to medium, split over multiple phases:

1. Internal API modification is underway, would need modification from std::size_t to pcl::index_t
2. Modification of public API will be faster since it is easily identifiable (function declaration, data members)

Migration Path

Multi-step migration:

1. Replace std::vector<int> with IndicesVec
2. Replace std::size_t with pcl::index_t
3. Replace IndicesVec = std::vector<int> with IndicesVec = std::vector<pcl::index_t>

Question

Is there anyone who has been adversely impacted with size increase and cache misses due to std::size_t instead of int? If so, there might be merit in discussing alternatives to std::ptrdiff_t (such as std::int32_t instead, which is the same but platform independent and brings back performance)