Upgrading the integer data type of index recorders and arrays

[chaseshyu]

Upgrading the integer data type of index arrays and recorders is necessary to accommodate the increasing model scale with DES. A 250x50x125 km 3D model with a 2 km resolution contains 1 million elements and 8 million markers, which numbers exceed the upper limit of the int data type (2 million). Seeking your suggestions.

[echoi]

I did a quick google search and learned that we have at least two options.
One is to change the int type to int64_t; and the other is to use size_t.

size_t looks like a better option in terms of sustainability and portability.
Below is the result of my google search with size_t c++.

Key characteristics and uses of size_t:

Unsigned:

It is always a non-negative value, as sizes and counts cannot be negative.

Guaranteed to be large enough:

size_t is guaranteed by the C++ standard to be able to represent the size of the largest possible object on the target system. This makes it a portable type for handling memory sizes and array indexing, as its underlying size (e.g., 32-bit or 64-bit) adapts to the architecture.

Used for array indexing and loop counters:

It is commonly used as an index for arrays and for loop counters when iterating over collections, especially when the number of elements might be large.

Portability:

Using size_t promotes code portability across different systems with varying integer sizes, avoiding potential overflow issues that might arise from using fixed-size types like unsigned int or unsigned long directly.

[chaseshyu]

Unsigned type is a good option for recording bigger numbers with the same memory. But right now somewhere in the code has used '-1' as n/a (ex. neighbors of elements at boundaries). If we switch to use unsigned types, will we need to create an additional bool array to keep track of the availability of each index? If so, maybe an array of struct with unsigned type and bool is a good option?

DynEarthSol/mesh.cxx

Line 2649 in 32e6140

var.neighbor = new conn_t(var.nelem, -1);

[echoi]

Your solution sounds good to me, but I don't have a definite opinion. Let's wait to hear from @tan2 and others.

[tan2]

I agree to switch to size_t for the array size. For the sentinel value (-1), we can use `std::numeric_limits<size_t>::max()` instead. As a precaution, after remeshing, we need to check that the number of nodes/elements/segments is less than or equal to `std::numeric_limits<size_t>::max() - 1`.

…

On Tue, Sep 2, 2025 at 11:40 PM Eunseo Choi ***@***.***> wrote: *echoi* left a comment (GeoFLAC/DynEarthSol#27) <#27 (comment)> Your solution sounds good to me, but I don't have a definite opinion. Let's wait to hear from @tan2 <https://github.com/tan2> and others. — Reply to this email directly, view it on GitHub <#27 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABWQQT5UI7E65NN5ZS5NQD3QW3ANAVCNFSM6AAAAACE3DOZEOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTENBVHA3DQNJZGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[echoi]

@chaseshyu This upgrade has been made, right? If you agree, let's close this issue.