|
34 | 34 |
|
35 | 35 | #include "RAJA/util/macros.hpp" |
36 | 36 | #include "RAJA/util/types.hpp" |
| 37 | +#include "RAJA/util/Jit.hpp" |
37 | 38 |
|
38 | 39 | #include "RAJA/internal/fault_tolerance.hpp" |
39 | 40 |
|
@@ -380,11 +381,12 @@ template<typename EXEC_POL, |
380 | 381 | IterationMapping>::value && |
381 | 382 | (IterationGetter::block_size > 0), |
382 | 383 | size_t> BlockSize = IterationGetter::block_size> |
383 | | -__launch_bounds__(BlockSize, BlocksPerSM) __global__ RAJA_JIT_COMPILE(3) |
384 | | - void forallp_cuda_kernel(const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
385 | | - const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
386 | | - const RAJA_CUDA_GRID_CONSTANT IndexType length, |
387 | | - ForallParam f_params) |
| 384 | +__launch_bounds__(BlockSize, BlocksPerSM) __global__ |
| 385 | + RAJA_JIT_COMPILE_ARGS(3) void forallp_cuda_kernel( |
| 386 | + const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
| 387 | + const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
| 388 | + const RAJA_CUDA_GRID_CONSTANT IndexType length, |
| 389 | + ForallParam f_params) |
388 | 390 | { |
389 | 391 | using RAJA::internal::thread_privatize; |
390 | 392 | auto privatizer = thread_privatize(loop_body); |
@@ -412,7 +414,7 @@ template<typename EXEC_POL, |
412 | 414 | IterationMapping>::value && |
413 | 415 | (IterationGetter::block_size <= 0), |
414 | 416 | size_t> RAJA_UNUSED_ARG(BlockSize) = 0> |
415 | | -__global__ RAJA_JIT_COMPILE(3) void forallp_cuda_kernel( |
| 417 | +__global__ RAJA_JIT_COMPILE_ARGS(3) void forallp_cuda_kernel( |
416 | 418 | const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
417 | 419 | const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
418 | 420 | const RAJA_CUDA_GRID_CONSTANT IndexType length, |
@@ -447,11 +449,12 @@ template< |
447 | 449 | IterationMapping>::value && |
448 | 450 | (IterationGetter::block_size > 0), |
449 | 451 | size_t> BlockSize = IterationGetter::block_size> |
450 | | -__launch_bounds__(BlockSize, BlocksPerSM) __global__ RAJA_JIT_COMPILE(3) |
451 | | - void forallp_cuda_kernel(const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
452 | | - const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
453 | | - const RAJA_CUDA_GRID_CONSTANT IndexType length, |
454 | | - ForallParam f_params) |
| 452 | +__launch_bounds__(BlockSize, BlocksPerSM) __global__ |
| 453 | + RAJA_JIT_COMPILE_ARGS(3) void forallp_cuda_kernel( |
| 454 | + const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
| 455 | + const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
| 456 | + const RAJA_CUDA_GRID_CONSTANT IndexType length, |
| 457 | + ForallParam f_params) |
455 | 458 | { |
456 | 459 | using RAJA::internal::thread_privatize; |
457 | 460 | auto privatizer = thread_privatize(loop_body); |
@@ -482,7 +485,7 @@ template< |
482 | 485 | IterationMapping>::value && |
483 | 486 | (IterationGetter::block_size <= 0), |
484 | 487 | size_t> RAJA_UNUSED_ARG(BlockSize) = 0> |
485 | | -__global__ RAJA_JIT_COMPILE(3) void forallp_cuda_kernel( |
| 488 | +__global__ RAJA_JIT_COMPILE_ARGS(3) void forallp_cuda_kernel( |
486 | 489 | const RAJA_CUDA_GRID_CONSTANT LOOP_BODY loop_body, |
487 | 490 | const RAJA_CUDA_GRID_CONSTANT Iterator idx, |
488 | 491 | const RAJA_CUDA_GRID_CONSTANT IndexType length, |
|
0 commit comments