Module nvptx
Platform-specific intrinsics for the NVPTX platform.
See the module documentation for more details.
Modules
- packed NVPTX Packed data types (SIMD)
Structs
- f16x2 PTX-specific 32-bit wide floating point (f16 x 2) vector type
Functions
-
__assert_fail
Syscall to be used whenever the assert expression produces a
falsevalue. - __assertfail
- _block_dim_x x-th thread-block dimension.
- _block_dim_y y-th thread-block dimension.
- _block_dim_z z-th thread-block dimension.
- _block_idx_x x-th thread-block index.
- _block_idx_y y-th thread-block index.
- _block_idx_z z-th thread-block index.
- _grid_dim_x x-th block-grid dimension.
- _grid_dim_y y-th block-grid dimension.
- _grid_dim_z z-th block-grid dimension.
- _syncthreads Synchronizes all threads in the block.
- _thread_idx_x x-th thread index.
- _thread_idx_y y-th thread index.
- _thread_idx_z z-th thread index.
- block_dim_x
- block_dim_y
- block_dim_z
- block_idx_x
- block_idx_y
- block_idx_z
- f16x2_add Add two values, round to nearest even
- f16x2_fma Fused multiply-add, round to nearest even
- f16x2_max Find the maximum of two values
- f16x2_max_nan Find the maximum of two values, NaNs pass through.
- f16x2_min Find the minimum of two values
- f16x2_min_nan Find the minimum of two values, NaNs pass through.
- f16x2_mul Multiply two values, round to nearest even
- f16x2_neg Arithmetic negate
- f16x2_sub Subtract two values, round to nearest even
- free Free previously dynamically allocated memory.
- grid_dim_x
- grid_dim_y
- grid_dim_z
- llvm_f16x2_maximum
- llvm_f16x2_minimum
- malloc Allocate memory dynamically from a fixed-size heap in global memory.
- syncthreads
- thread_idx_x
- thread_idx_y
- thread_idx_z
-
trap
Generates the trap instruction
TRAP - vprintf Print formatted output from a kernel to a host-side output stream.