Prefer Your Language

Search This Blog

CUDA complete | Complete reference on CUDA


CUDA syntax

Source code is in .cu files, which contain mixture of host (CPU) and device (GPU) code.

Declaring functions

__global__ declares kernel, which is called on host and executed on device
__device__ declares device function, which is called and executed on device
__host__ declares host function, which is called and executed on host
__noinline__ to avoid inlining
__forceinline__to force inlining

Declaring variables

__device__ declares device variable in global memory, accessible from all threads, with lifetime of application
__constant__declares device variable in constant memory, accessible from all threads, with lifetime of application
__shared__ declares device varibale in block's shared memory, accessible from all threads within a block, with lifetime of block
__restrict__standard C definition that pointers are not aliased

Types

Most routines return an error code of type cudaError_t.

Vector types

char1, uchar1, short1, ushort1, int1, uint1, long1, ulong1, float1
char2, uchar2, short2, ushort2, int2, uint2, long2, ulong2, float2
char3, uchar3, short3, ushort3, int3, uint3, long3, ulong3, float3
char4, uchar4, short4, ushort4, int4, uint4, long4, ulong4, float4

longlong1, ulonglong1, double1
longlong2, ulonglong2, double2

dim3
Components are accessible as variable.x,  variable.y,  variable.z,  variable.w.
Constructor is make_<type>( x, ... ), for example:
float2 xx = make_float2( 1., 2. );
dim3 can take 1, 2, or 3 argumetns:
dim3 blocks1D( 5       );
dim3 blocks2D( 5, 5    );
dim3 blocks3D( 5, 5, 5 );

Pre-defined variables

dim3 gridDim dimensions of grid
dim3 blockDim dimensions of block
uint3 blockIdx block index within grid
uint3 threadIdxthread index within block
int warpSize number of threads in warp

Kernel invocation

__global__ void kernel( ... ) { ... }

dim3 blocks( nx, ny, nz );           // cuda 1.x has 1D and 2D grids, cuda 2.x adds 3D grids
dim3 threadsPerBlock( mx, my, mz );  // cuda 1.x has 1D, 2D, and 3D blocks

kernel<<< blocks, threadsPerBlock >>>( ... );

Thread management

__threadfence_block(); wait until memory accesses are visible to block
__threadfence(); wait until memory accesses are visible to block and device
__threadfence_system();wait until memory accesses are visible to block and device and host (2.x)
__syncthreads(); wait until all threads reach sync

Memory management

__device__ float* pointer;
cudaMalloc( &pointer, size );
cudaFree( pointer );

// direction is one of cudaMemcpyHostToDevice or cudaMemcpyDeviceToHost
cudaMemcpy( dst_pointer, src_pointer, size, direction );

__constant__ float dev_data[n];
float host_data[n];
cudaMemcpyToSymbol  ( dev_data,  host_data, sizeof(host_data) );  // dev_data  = host_data
cudaMemcpyFromSymbol( host_data, dev_data,  sizeof(host_data) );  // host_data = dev_data
Also, malloc and free work inside a kernel (2.x), but memory allocated in a kernel must be deallocated in a kernel (not the host). It can be freed in a different kernel, though.

Atomic functions

old = atomicAdd ( &addr, value );  // old = *addr;  *addr += value
old = atomicSub ( &addr, value );  // old = *addr;  *addr –= value
old = atomicExch( &addr, value );  // old = *addr;  *addr  = value

old = atomicMin ( &addr, value );  // old = *addr;  *addr = min( old, value )
old = atomicMax ( &addr, value );  // old = *addr;  *addr = max( old, value )

// increment up to value, then reset to 0  
// decrement down to 0, then reset to value
old = atomicInc ( &addr, value );  // old = *addr;  *addr = ((old >= value) ? 0 : old+1 )
old = atomicDec ( &addr, value );  // old = *addr;  *addr = ((old == 0) or (old > val) ? val : old–1 )

old = atomicAnd ( &addr, value );  // old = *addr;  *addr &= value
old = atomicOr  ( &addr, value );  // old = *addr;  *addr |= value
old = atomicXor ( &addr, value );  // old = *addr;  *addr ^= value

// compare-and-store
old = atomicCAS ( &addr, compare, value );  // old = *addr;  *addr = ((old == compare) ? value : old)

Warp vote

int __all   ( predicate );
int __any   ( predicate );
int __ballot( predicate );  // nth thread sets nth bit to predicate

Timer

wall clock cycle counter
clock_t clock();

Texture

can also return float2 or float4, depending on texRef.
// integer index
float tex1Dfetch( texRef, ix );

// float index
float tex1D( texRef, x       );
float tex2D( texRef, x, y    );
float tex3D( texRef, x, y, z );

float tex1DLayered( texRef, x    );
float tex2DLayered( texRef, x, y );

Low-level Driver API

#include <cuda.h>

CUdevice dev;
CUdevprop properties;
char name[n];
int major, minor;
size_t bytes;

cuInit( 0 );  // takes flags for future use
cuDeviceGetCount         ( &cnt );
cuDeviceGet              ( &dev, index );
cuDeviceGetName          ( name, sizeof(name), dev );
cuDeviceComputeCapability( &major, &minor,     dev );
cuDeviceTotalMem         ( &bytes,             dev );
cuDeviceGetProperties    ( &properties,        dev );  // max threads, etc.

cuBLAS

Matrices are column-major. Indices are 1-based; this affects result of i<t>amax and i<t>amin.
#include <cublas_v2.h>

cublasHandle_t handle;
cudaStream_t   stream;

cublasCreate( &handle );
cublasDestroy( handle );
cublasGetVersion( handle, &version );
cublasSetStream( handle,  stream );
cublasGetStream( handle, &stream );
cublasSetPointerMode( handle,  mode );
cublasGetPointerMode( handle, &mode );

// copy x => y
cublasSetVector     ( n, elemSize, x_src_host, incx, y_dst_dev,  incy );
cublasGetVector     ( n, elemSize, x_src_dev,  incx, y_dst_host, incy );
cublasSetVectorAsync( n, elemSize, x_src_host, incx, y_dst_dev,  incy, stream );
cublasGetVectorAsync( n, elemSize, x_src_dev,  incx, y_dst_host, incy, stream );

// copy A => B
cublasSetMatrix     ( rows, cols, elemSize, A_src_host, lda, B_dst_dev,  ldb );
cublasGetMatrix     ( rows, cols, elemSize, A_src_dev,  lda, B_dst_host, ldb );
cublasSetMatrixAsync( rows, cols, elemSize, A_src_host, lda, B_dst_dev,  ldb, stream );
cublasGetMatrixAsync( rows, cols, elemSize, A_src_dev,  lda, B_dst_host, ldb, stream );

Constants

argumentconstantsdescription (Fortran letter)
transCUBLAS_OP_N non-transposed ('N')
CUBLAS_OP_T transposed ('T')
CUBLAS_OP_C conjugate transposed ('C')
 
uploCUBLAS_FILL_MODE_LOWER lower part filled ('L')
CUBLAS_FILL_MODE_UPPER upper part filled ('U')
 
sideCUBLAS_SIDE_LEFT matrix on left ('L')
CUBLAS_SIDE_RIGHT matrix on right ('R')
 
modeCUBLAS_POINTER_MODE_HOST alpha and beta scalars passed on host
CUBLAS_POINTER_MODE_DEVICEalpha and beta scalars passed on device
BLAS functions have cublas prefix and first letter of usual BLAS function name is capitalized. Arguments are the same as standard BLAS, with these exceptions:
  • All functions add handle as first argument.
  • All functions return cublasStatus_t error code.
  • Constants alpha and beta are passed by pointer. All other scalars (n, incx, etc.) are bassed by value.
  • Functions that return a value, such as ddot, add result as last argument, and save value to result.
  • Constants are given in table above, instead of using characters.
Examples:
cublasDdot ( handle, n, x, incx, y, incy, &result );  // result = ddot( n, x, incx, y, incy );
cublasDaxpy( handle, n, &alpha, x, incx, y, incy );   // daxpy( n, alpha, x, incx, y, incy );

Compiler

nvcc, often found in /usr/local/cuda/bin
Defines __CUDACC__

Flags common with cc

Short flagLong flagOutput or Description
-c--compile.o object file
-E--preprocesson standard output
-M--generate-dependencieson standard output
-o file--output-file file
-I directory--include-path directoryheader search path
-L directory--library-path directorylibrary search path
-l lib--library liblink with library
-libgenerate library
-sharedgenerate shared library
-pg--profilefor gprof
-g level--debug level
-G--device-debug
-O level--optimize level
 
Undocumented (but in sample makefiles)
-m64compile x86_64 host CPU code

Flags specific to nvcc

-vlist compilation commands as they are executed
-dryrunlist compilation commands, without executing
-keepsaves intermediate files (e.g., pre-processed) for debugging
-cleanremoves output files (with same exact compiler options)
-arch=<compute_xy>generate PTX for capability x.y
-code=<sm_xy>generate binary for capability x.y, by default same as -arch
-gencode arch=...,code=...same as -arch and -code, but may be repeated

Argumenents for -arch and -code

It makes most sense (to me) to give -arch a virtual architecture and -code a real architecture, though both flags accept both virtual and real architectures (at times).
Virtual architectureReal architectureFeatures
Teslacompute_10sm_10Basic features
compute_11sm_11+ atomic memory ops on global memory
compute_12sm_12+ atomic memory ops on shared memory
+ vote instructions
compute_13sm_13+ double precision
Fermicompute_20sm_20+ Fermi

Some hardware constraints

1.x2.x
max x- or y-dimension of block5121024
max z-dimension of block6464
max threads per block5121024
warp size3232
max blocks per MP88
max warps per MP3248
max threads per MP10241536
max 32-bit registers per MP16k32k
max shared memory per MP16 KB48 KB
shared memory banks1632
local memory per thread16 KB512 KB
const memory64 KB64 KB
const cache8 KB8 KB
texture cache8 KB8 KB




Got Questions? Feel free to ask me any question because I'd be happy to walk you through step by step! References For Contact us….. Click on Contact us Tab

0 comments:

Help us to improve our quality and become contributor to our blog

Become a contributor to this blog. Click on contact us tab
Blogger Template by Clairvo