openproj — detail

No No-gos section. The pitch template asks for one — it is what keeps the appetite honest.

Problem

The dycore runs in double precision throughout. A single precision option would halve the memory traffic of the parts that can take it, which is where the dycore spends its time.

starkphi has had this open since 29.09.2025 as #886, and since 05.12.2025 as #970, which its own title calls an intermediate state. Neither has merged. The cycle-37 brainstorming note raised it as "Merge single precision (Philipp)" but it did not reach that cycle's betting table, nor cycle 38's, so this record carries no cycle: the work is real and is written down, but nobody has bet on it.

Appetite

Eight person-weeks. Two person-cycles is what eleven months across two unmerged PRs suggests this actually costs, rather than what anyone would like it to.

Solution

See #970. It is the live branch; #886 is its predecessor.

Rabbit holes

The reason this has not landed is worth establishing before more is spent on it. An eleven-month-old branch against a dycore that has been refactored repeatedly in that time may cost more to rebase than to redo.

No-gos

Progress

#886 opened 29.09.2025, superseded
#970 opened 05.12.2025, the intermediate state, reviewed by OngChia, muellch, egparedes, iomaganaris and msimberg
starkphi kept the branch alive during cycle 38 with #1469 (closed) and landed an unrelated rename as #1472

## Problem

The dycore runs in double precision throughout. A single precision option would
halve the memory traffic of the parts that can take it, which is where the dycore
spends its time.

starkphi has had this open since 29.09.2025 as #886, and since 05.12.2025 as
#970, which its own title calls an intermediate state. Neither has merged. The
cycle-37 brainstorming note raised it as "Merge single precision (Philipp)" but
it did not reach that cycle's betting table, nor cycle 38's, so this record
carries no cycle: the work is real and is written down, but nobody has bet on it.

## Appetite

Eight person-weeks. Two person-cycles is what eleven months across two unmerged
PRs suggests this actually costs, rather than what anyone would like it to.

## Solution

See #970. It is the live branch; #886 is its predecessor.

## Rabbit holes

The reason this has not landed is worth establishing before more is spent on it.
An eleven-month-old branch against a dycore that has been refactored repeatedly
in that time may cost more to rebase than to redo.

## No-gos

## Progress

- [ ] #886 opened 29.09.2025, superseded
- [ ] #970 opened 05.12.2025, the intermediate state, reviewed by OngChia,
      muellch, egparedes, iomaganaris and msimberg
- [ ] starkphi kept the branch alive during cycle 38 with #1469 (closed) and
      landed an unrelated rename as #1472