@@ -491,51 +491,6 @@ namespace core {
491
491
return 0 ;
492
492
}
493
493
494
- template <class Kernel >
495
- int CUB_RUNTIME_FUNCTION
496
- get_max_block_size (Kernel k)
497
- {
498
- int devId;
499
- cuda_cub::throw_on_error (cudaGetDevice (&devId),
500
- " get_max_block_size :"
501
- " failed to cudaGetDevice" );
502
-
503
- cudaOccDeviceProp occ_prop;
504
- cuda_cub::throw_on_error (get_occ_device_properties (occ_prop, devId),
505
- " get_max_block_size: "
506
- " failed to cudaGetDeviceProperties" );
507
-
508
-
509
- cudaFuncAttributes attribs;
510
- cuda_cub::throw_on_error (cudaFuncGetAttributes (&attribs, reinterpret_cast <void *>(k)),
511
- " get_max_block_size: "
512
- " failed to cudaFuncGetAttributes" );
513
- cudaOccFuncAttributes occ_attrib (attribs);
514
-
515
-
516
- cudaFuncCache cacheConfig;
517
- cuda_cub::throw_on_error (cudaDeviceGetCacheConfig (&cacheConfig),
518
- " get_max_block_size: "
519
- " failed to cudaDeviceGetCacheConfig" );
520
-
521
- cudaOccDeviceState occ_state;
522
- occ_state.cacheConfig = (cudaOccCacheConfig)cacheConfig;
523
- int block_size = 0 ;
524
- int min_grid_size = 0 ;
525
- cudaOccError occ_status = cudaOccMaxPotentialOccupancyBlockSize (&min_grid_size,
526
- &block_size,
527
- &occ_prop,
528
- &occ_attrib,
529
- &occ_state,
530
- 0 );
531
- if (CUDA_OCC_SUCCESS != occ_status || block_size <= 0 )
532
- cuda_cub::throw_on_error (cudaErrorInvalidConfiguration,
533
- " get_max_block_size: "
534
- " failed to cudaOccMaxPotentialOccupancyBlockSize" );
535
-
536
- return block_size;
537
- }
538
-
539
494
// LoadIterator
540
495
// ------------
541
496
// if trivial iterator is passed, wrap loads into LDG
0 commit comments