Data Structures
struct	hipsparseLtHandle_t
	Handle to the hipSPARSELt library context queue. More...

struct	hipsparseLtMatDescriptor_t
	Descriptor of the matrix. More...

struct	hipsparseLtMatmulDescriptor_t
	Descriptor of the matrix multiplication operation. More...

struct	hipsparseLtMatmulAlgSelection_t
	Descriptor of the matrix multiplication algorithm. More...

struct	hipsparseLtMatmulPlan_t
	Descriptor of the matrix multiplication execution plan. More...

Enumerations
enum	hipsparseLtDatatype_t { HIPSPARSELT_R_16F = 150 , HIPSPARSELT_R_32F = 151 , HIPSPARSELT_R_8I = 160 , HIPSPARSELT_R_16BF = 168 , HIPSPARSELT_R_8F = 170 , HIPSPARSELT_R_8BF = 171 }
	List of hipsparselt data types. More...

enum	hipsparseLtSparsity_t { HIPSPARSELT_SPARSITY_50_PERCENT }
	Specify the sparsity of the structured matrix. More...

enum	hipsparseLtMatDescAttribute_t { HIPSPARSELT_MAT_NUM_BATCHES , HIPSPARSELT_MAT_BATCH_STRIDE }
	Specify the additional attributes of a matrix descriptor. More...

enum	hipsparseLtComputetype_t { HIPSPARSELT_COMPUTE_16F = 0 , HIPSPARSELT_COMPUTE_32I , HIPSPARSELT_COMPUTE_32F , HIPSPARSELT_COMPUTE_TF32 , HIPSPARSELT_COMPUTE_TF32_FAST }
	Specify the compute precision modes of the matrix. More...

enum	hipsparseLtMatmulDescAttribute_t { HIPSPARSELT_MATMUL_ACTIVATION_RELU = 0 , HIPSPARSELT_MATMUL_ACTIVATION_RELU_UPPERBOUND = 1 , HIPSPARSELT_MATMUL_ACTIVATION_RELU_THRESHOLD = 2 , HIPSPARSELT_MATMUL_ACTIVATION_GELU = 3 , HIPSPARSELT_MATMUL_ACTIVATION_GELU_SCALING = 4 , HIPSPARSELT_MATMUL_ALPHA_VECTOR_SCALING = 5 , HIPSPARSELT_MATMUL_BETA_VECTOR_SCALING = 6 , HIPSPARSELT_MATMUL_BIAS_STRIDE = 7 , HIPSPARSELT_MATMUL_BIAS_POINTER = 8 , HIPSPARSELT_MATMUL_ACTIVATION_ABS = 9 , HIPSPARSELT_MATMUL_ACTIVATION_LEAKYRELU = 10 , HIPSPARSELT_MATMUL_ACTIVATION_LEAKYRELU_ALPHA = 11 , HIPSPARSELT_MATMUL_ACTIVATION_SIGMOID = 12 , HIPSPARSELT_MATMUL_ACTIVATION_TANH = 13 , HIPSPARSELT_MATMUL_ACTIVATION_TANH_ALPHA = 14 , HIPSPARSELT_MATMUL_ACTIVATION_TANH_BETA = 15 , HIPSPARSELT_MATMUL_BIAS_TYPE = 16 }
	Specify the additional attributes of a matrix multiplication descriptor. More...

enum	hipsparseLtMatmulAlg_t { HIPSPARSELT_MATMUL_ALG_DEFAULT }
	Specify the algorithm for matrix-matrix multiplication. More...

enum	hipsparseLtMatmulAlgAttribute_t { HIPSPARSELT_MATMUL_ALG_CONFIG_ID = 0 , HIPSPARSELT_MATMUL_ALG_CONFIG_MAX_ID = 1 , HIPSPARSELT_MATMUL_SEARCH_ITERATIONS = 2 , HIPSPARSELT_MATMUL_SPLIT_K = 3 , HIPSPARSELT_MATMUL_SPLIT_K_MODE = 4 , HIPSPARSELT_MATMUL_SPLIT_K_BUFFERS = 5 }
	Specify the matrix multiplication algorithm attributes. More...

enum	hipsparseLtPruneAlg_t { HIPSPARSELT_PRUNE_SPMMA_TILE = 0 , HIPSPARSELT_PRUNE_SPMMA_STRIP = 1 }
	Specify the pruning algorithm to apply to the structured matrix before the compression. More...

enum	hipsparseLtSplitKMode_t { HIPSPARSELT_SPLIT_K_MODE_ONE_KERNEL = 0 , HIPSPARSELT_SPLIT_K_MODE_TWO_KERNELS = 1 }
	Specify the split k mode value. More...

Detailed Description

HIP = Heterogeneous-compute Interface for Portability

Define a extremely thin runtime layer that allows source code to be compiled unmodified through either AMD HCC or NVCC. Key features tend to be in the spirit and terminology of CUDA, but with a portable path to other accelerators as well.

This is the master include file for hipSPARSELt, wrapping around hipSPARSELt and cuSPARSELt "version 0.3".

Enumeration Type Documentation

◆ hipsparseLtComputetype_t

enum hipsparseLtComputetype_t

Specify the compute precision modes of the matrix.

Enumerator
HIPSPARSELT_COMPUTE_16F	16-bit floating-point precision. CUDA backend only.
HIPSPARSELT_COMPUTE_32I	32-bit integer precision
HIPSPARSELT_COMPUTE_32F	32-bit floating-point precision. HIP backend only.
HIPSPARSELT_COMPUTE_TF32	32-bit floating point value are rounded to TF32 before the computation. CUDA backend only.
HIPSPARSELT_COMPUTE_TF32_FAST	32-bit floating point value are truncated to TF32 before the computation. CUDA backend only.

◆ hipsparseLtDatatype_t

enum hipsparseLtDatatype_t

List of hipsparselt data types.

Indicates the precision width of data stored in a hipsparselt type. Should use hipDatatype_t instead in the furture.

Enumerator
HIPSPARSELT_R_16F	16 bit floating point, real
HIPSPARSELT_R_32F	32 bit floating point, real
HIPSPARSELT_R_8I	8 bit signed integer, real
HIPSPARSELT_R_16BF	16 bit bfloat, real
HIPSPARSELT_R_8F	8 bit floating point, real
HIPSPARSELT_R_8BF	8 bit bfloat, real

◆ hipsparseLtMatDescAttribute_t

enum hipsparseLtMatDescAttribute_t

Specify the additional attributes of a matrix descriptor.

The hipsparseLtMatDescAttribute_t is used in the hipsparseLtMatDescSetAttribute and hipsparseLtMatDescGetAttribute functions

Enumerator
HIPSPARSELT_MAT_NUM_BATCHES	number of matrices in a batch. READ/WRITE
HIPSPARSELT_MAT_BATCH_STRIDE	stride between consecutive matrices in a batch expressed in terms of matrix elements. READ/WRITE

◆ hipsparseLtMatmulAlg_t

enum hipsparseLtMatmulAlg_t

Specify the algorithm for matrix-matrix multiplication.

The hipsparseLtMatmulAlg_t is used in the hipsparseLtMatmulAlgSelectionInit function.

Enumerator
HIPSPARSELT_MATMUL_ALG_DEFAULT

◆ hipsparseLtMatmulAlgAttribute_t

enum hipsparseLtMatmulAlgAttribute_t

Specify the matrix multiplication algorithm attributes.

The hipsparseLtMatmulAlgAttribute_t is used in the hipsparseLtMatmulAlgGetAttribute and hipsparseLtMatmulAlgSetAttribute functions.

Enumerator
HIPSPARSELT_MATMUL_ALG_CONFIG_ID
HIPSPARSELT_MATMUL_ALG_CONFIG_MAX_ID
HIPSPARSELT_MATMUL_SEARCH_ITERATIONS
HIPSPARSELT_MATMUL_SPLIT_K
HIPSPARSELT_MATMUL_SPLIT_K_MODE
HIPSPARSELT_MATMUL_SPLIT_K_BUFFERS

◆ hipsparseLtMatmulDescAttribute_t

enum hipsparseLtMatmulDescAttribute_t

Specify the additional attributes of a matrix multiplication descriptor.

The hipsparseLtMatmulDescAttribute_t is used in the hipsparseLtMatmulDescSetAttribute and hipsparseLtMatmulDescGetAttribute functions

Enumerator
HIPSPARSELT_MATMUL_ACTIVATION_RELU	ReLU activation function.
HIPSPARSELT_MATMUL_ACTIVATION_RELU_UPPERBOUND	Upper bound of the ReLU activation function.
HIPSPARSELT_MATMUL_ACTIVATION_RELU_THRESHOLD	Lower threshold of the ReLU activation function.
HIPSPARSELT_MATMUL_ACTIVATION_GELU	GeLU activation function.
HIPSPARSELT_MATMUL_ACTIVATION_GELU_SCALING	Scaling coefficient for the GeLU activation function. It implies gelu is endable
HIPSPARSELT_MATMUL_ALPHA_VECTOR_SCALING	Enable/Disable alpha vector (per-channel) scaling
HIPSPARSELT_MATMUL_BETA_VECTOR_SCALING	Enable/Disable beta vector (per-channel) scaling
HIPSPARSELT_MATMUL_BIAS_STRIDE	Bias pointer. The bias vector size must equal to the number of rows of the output matrix (D).
HIPSPARSELT_MATMUL_BIAS_POINTER	Bias stride between consecutive bias vectors. 0 means broadcast the first bias vector.
HIPSPARSELT_MATMUL_ACTIVATION_ABS	ABS activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_LEAKYRELU	LeakyReLU activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_LEAKYRELU_ALPHA	Alpha value of the LeakyReLU activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_SIGMOID	Sigmoid activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_TANH	Tanh activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_TANH_ALPHA	Alpha value of the Tanh activation function. HIP backend only
HIPSPARSELT_MATMUL_ACTIVATION_TANH_BETA	Beta value of the Tanh activation function. HIP backend only
HIPSPARSELT_MATMUL_BIAS_TYPE	Precision of Bias. HIP backend only, When Input's datatype is FP16 - Bias type can be FP16 or FP32. (default FP16) When Input's datatype is BF16 - Bias type can be BF16 or FP32. (default BF16) In other cases - Bias type is FP32.

◆ hipsparseLtPruneAlg_t

enum hipsparseLtPruneAlg_t

Specify the pruning algorithm to apply to the structured matrix before the compression.

The hipsparseLtPruneAlg_t is used in the hipsparseLtSpMMAPrune and hipsparseLtSpMMAPrune2 function.

Enumerator
HIPSPARSELT_PRUNE_SPMMA_TILE	Zero out eight elements in a 4x4 tile, nonzero elements have the maximum L1-norm value in all combinations in the tile. Exactly two elements in each row and column.
HIPSPARSELT_PRUNE_SPMMA_STRIP	Zero out two elements in a 1x4 strip, nonzero elements have the maximum L1-norm value in all combinations in the strip.

◆ hipsparseLtSparsity_t

enum hipsparseLtSparsity_t

Specify the sparsity of the structured matrix.

The enumerator specifies the sparsity ratio of the structured matrix as sparsity = nnz / total elements The sparsity property is used in the hipsparseLtStructuredDescriptorInit function.

Enumerator
HIPSPARSELT_SPARSITY_50_PERCENT	50% sparsity ratio - 1:2 for tf32 and float, 2:4 for half, bfloat16, int

◆ hipsparseLtSplitKMode_t

enum hipsparseLtSplitKMode_t

Specify the split k mode value.

The hipsparseLtSplitKMode_t is used by HIPSPARSELT_MATMUL_SPLIT_K_MODE attribute in hipsparseLtMatmulAlgAttribute_t.

Enumerator
HIPSPARSELT_SPLIT_K_MODE_ONE_KERNEL	Use the same SP-MM kernel to do the final reduction
HIPSPARSELT_SPLIT_K_MODE_TWO_KERNELS	Use another kernel to do the final reduction

Data types

Data types#

Data Structures

Enumerations