feat(accel): version floors, a probe channel, constrained device globs, and a CUDA example that takes two routes by Sunrisepeak · Pull Request #563 · mcpp-community/mcpp

feat(doctor): read the device toolkit from a payload before the host, and state the driver relation without probing for it

speak-agent

committed

ad167f4

feat(example): the CUDA example takes its toolkit from a payload, not from the host

speak-agent

committed

f431285

feat(build): version floors, so a machine's limits are known before anything is compiled

speak-agent

committed

d5a933c

test(e2e): give 317 the window its bound needs, and run 602 where the report exists

speak-agent

committed

0f84867

fix(config): offline mode skips the first-use sandbox bootstrap

speak-agent

committed

6d80c39

feat(build): the probe channel, and the device axis as a build program sees it

speak-agent

committed

0ac6397

feat(build): a source glob may carry the accel it is for, and the build narrows to it

speak-agent

committed

96fda4e

docs(plan): record the engine batch, the fingerprint drift, and the local 168 reading

speak-agent

committed

5eae6cd

feat(build): what a second compiler needs, and the device axis on run and test

speak-agent

committed

89c986d

refactor(doctor): the device toolkit report belongs to the package that owns the tool

speak-agent

committed

8a3a18d

feat(example): two routes for the CUDA example, and the pairings it refuses

speak-agent

committed

6115c35

fix(build): key a package's device sources by its root, not by its bare name

speak-agent

committed

1317625

release: 2026.9.5.2

speak-agent

committed

88ed2e5

fix(build): "no accelerator" is the empty string, not the display spelling

speak-agent

committed

79c3d51

docs(plan): batch ④ landed, and three readings the engine batch produced

speak-agent

committed

49c4ebc

refactor(example): the CUDA rule package moves into the mcpplibs namespace

speak-agent

committed

dfc02e8

docs(design): §12 — what the implementation overturned, produced, and left

speak-agent

committed

e01f4d4

feat(build): an object action's outputs reach a static library, not only a linked image

speak-agent

committed

7d08c76

docs(plan): T5.1 as a gate paid for itself, and what it left

speak-agent

committed

91d5222

fix(example): the clang route passes NVIDIA's own libc++ escape

speak-agent

committed

eea34c1

docs(plan): where T5.1's CUDA lane actually stops, and what it proved on the way

speak-agent

committed

6d4d498

test(e2e): 188's no-consumer case needs a target that does not exist, not a library

speak-agent

committed

0b265c2

docs(design): §12.5 — what the framework gate actually measured

speak-agent

committed

e5b3867

docs(example): the clang route does pass one flag, and the README should say which

speak-agent

committed

edf579f

test(verify): section E asserted a silence that a fresh sandbox gives for free

speak-agent

committed

29328b4

Navigation Menu

Uh oh!

feat(accel): version floors, a probe channel, constrained device globs, and a CUDA example that takes two routes - #563

feat(accel): version floors, a probe channel, constrained device globs, and a CUDA example that takes two routes#563
Sunrisepeak merged 25 commits into
mainfrom
feat/multi-device

Commits on Sep 5, 2026

feat(doctor): read the device toolkit from a payload before the host, and state the driver relation without probing for it

feat(example): the CUDA example takes its toolkit from a payload, not from the host

feat(build): version floors, so a machine's limits are known before anything is compiled

test(e2e): give 317 the window its bound needs, and run 602 where the report exists

fix(config): offline mode skips the first-use sandbox bootstrap

feat(build): the probe channel, and the device axis as a build program sees it

feat(build): a source glob may carry the accel it is for, and the build narrows to it

docs(plan): record the engine batch, the fingerprint drift, and the local 168 reading

feat(build): what a second compiler needs, and the device axis on run and test

refactor(doctor): the device toolkit report belongs to the package that owns the tool

feat(example): two routes for the CUDA example, and the pairings it refuses

fix(build): key a package's device sources by its root, not by its bare name

release: 2026.9.5.2

fix(build): "no accelerator" is the empty string, not the display spelling

docs(plan): batch ④ landed, and three readings the engine batch produced

refactor(example): the CUDA rule package moves into the mcpplibs namespace

docs(design): §12 — what the implementation overturned, produced, and left

feat(build): an object action's outputs reach a static library, not only a linked image

docs(plan): T5.1 as a gate paid for itself, and what it left

fix(example): the clang route passes NVIDIA's own libc++ escape

docs(plan): where T5.1's CUDA lane actually stops, and what it proved on the way

test(e2e): 188's no-consumer case needs a target that does not exist, not a library

docs(design): §12.5 — what the framework gate actually measured

docs(example): the clang route does pass one flag, and the README should say which

test(verify): section E asserted a silence that a fresh sandbox gives for free