| … | … | @@ -281,6 +281,14 @@ it's own merits. But does it solve the original issue, which was |
| 281 | 281 | that of interfacing to GPU's, and specifically, being able to |
| 282 | 282 | re-interpret and compile PlusLink, TimesLink in this new setting? |
| 283 | 283 | |
| 284 | Well, sort-of. The BackingStore does have a remote-execute ability. |
| 285 | So requests to execute really do mean "recompile this expressino for |
| 286 | your remote location so that it makes sense, there". So that really |
| 287 | means that the GPU will be a kind-of StorageNode. The GpuStorageNode, |
| 288 | I guess. Maybe the OpenclStorageNode and teh CudaStorageNode. |
| 289 | |
| 290 | OK, so that sounds like a plan. Time to unleash Claude on this mess. |
| 291 | |
| 284 | 292 | |
| 285 | 293 | ### Are SensoryNode's AtomSpaces? |
| 286 | 294 | By logical extension of the above thoughts, SensoryNodes should be |
| … | … | @@ -295,7 +303,7 @@ come back from remote sensing... |
| 295 | 303 | |
| 296 | 304 | TODO List |
| 297 | 305 | --------- |
| 298 | | Writhing the above, the TODO list expands: |
| 306 | Writing the above, the TODO list expands: |
| 299 | 307 | * The ThreadJoinLink should be removed from Atomese. DONE. |
| 300 | 308 | * Benchmark the AtomSpace layers/frame performance. Maybe with a variant |
| 301 | 309 | of the `LargeZipfUTest` layout, but this time with different count |
| … | … | @@ -306,4 +314,6 @@ Spin-off TODO List |
| 306 | 314 | Not immediately relevant, but tangled in: |
| 307 | 315 | * Add MessagesOfLink support to StorageNode |
| 308 | 316 | * Migrate examples and wiki pages away from cog-open, etc. |
| 309 | | * Migrate examples away from TV; |
| 317 | * Migrate examples away from TV ... mostly DONE |
| 318 | (also: cog-evaluate! BindLink, GetLink) |
| 319 | * Remove non-crisp evaluation |