Package com.github.tjake.jlama.tensor.operations.cnative

Class NativeSimd

java.lang.Object

com.github.tjake.jlama.tensor.operations.cnative.NativeSimd

public class NativeSimd
extends Object

Constructor Summary

Constructors

Constructor	Description
NativeSimd()

Method Summary

Modifier and Type	Method	Description
static void	gemm_bf16(int flags, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment cr, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_bf16(int flags, short* a, int aoffset, short* b, int boffset, short* cr, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_bf16_batch(int flags, int batch_num, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment cr, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_bf16_batch(int flags, int batch_num, short* a, int aoffset, short** b, int boffset, short** cr, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_f32(int flags, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_f32(int flags, float* a, int aoffset, float* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_f32_batch(int flags, int batch_num, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_f32_batch(int flags, int batch_num, float* a, int aoffset, float** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_f32_bf16(int flags, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment cr, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_f32_bf16(int flags, float* a, int aoffset, short* b, int boffset, short* cr, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_f32_bf16_batch(int flags, int batch_num, MemorySegment a, int aoffset, MemorySegment b, int boffset, MemorySegment cr, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc)	void gemm_f32_bf16_batch(int flags, int batch_num, float* a, int aoffset, short** b, int boffset, short** cr, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);
static void	gemm_f32_q4(int flags, MemorySegment a, int aoffset, MemorySegment bf, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc)	void gemm_f32_q4(int flags, float* a, int aoffset, float* bf, char* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc);
static void	gemm_f32_q4_batch(int flags, int batch_num, MemorySegment a, int aoffset, MemorySegment bf, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc)	void gemm_f32_q4_batch(int flags, int batch_num, float* a, int aoffset, float** bf, char** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc);
static void	gemm_q8_q4(int flags, MemorySegment af, MemorySegment a, int aoffset, MemorySegment bf, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc)	void gemm_q8_q4(int flags, float* af, char* a, int aoffset, float* bf, char* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc);
static void	gemm_q8_q4_batch(int flags, int batch_num, MemorySegment af, MemorySegment a, int aoffset, MemorySegment bf, MemorySegment b, int boffset, MemorySegment r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc)	void gemm_q8_q4_batch(int flags, int batch_num, float* af, char* a, int aoffset, float** bf, char** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc);
static int	HAS_AVX2()	#define HAS_AVX2 4
static int	HAS_F16C()	#define HAS_F16C 2
static int	IS_M_SERIES_MAC()	#define IS_M_SERIES_MAC 8
static int	Q4_BLOCK_SIZE()	#define Q4_BLOCK_SIZE 32
static int	Q8_BLOCK_SIZE()	#define Q8_BLOCK_SIZE 32

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

NativeSimd

public NativeSimd()

Method Details

HAS_F16C

public static int HAS_F16C()

#define HAS_F16C 2

HAS_AVX2

public static int HAS_AVX2()

#define HAS_AVX2 4

IS_M_SERIES_MAC

public static int IS_M_SERIES_MAC()

#define IS_M_SERIES_MAC 8

Q8_BLOCK_SIZE

public static int Q8_BLOCK_SIZE()

#define Q8_BLOCK_SIZE 32

Q4_BLOCK_SIZE

public static int Q4_BLOCK_SIZE()

#define Q4_BLOCK_SIZE 32

gemm_q8_q4

public static void gemm_q8_q4(int flags,
 MemorySegment af,
 MemorySegment a,
 int aoffset,
 MemorySegment bf,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldaf,
 int ldb,
 int ldbf,
 int ldc)

void gemm_q8_q4(int flags, float* af, char* a, int aoffset, float* bf, char* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc);

gemm_q8_q4_batch

public static void gemm_q8_q4_batch(int flags,
 int batch_num,
 MemorySegment af,
 MemorySegment a,
 int aoffset,
 MemorySegment bf,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldaf,
 int ldb,
 int ldbf,
 int ldc)

void gemm_q8_q4_batch(int flags, int batch_num, float* af, char* a, int aoffset, float** bf, char** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldaf, int ldb, int ldbf, int ldc);

gemm_f32

public static void gemm_f32(int flags,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_f32(int flags, float* a, int aoffset, float* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);

gemm_f32_batch

public static void gemm_f32_batch(int flags,
 int batch_num,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_f32_batch(int flags, int batch_num, float* a, int aoffset, float** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);

gemm_f32_q4

public static void gemm_f32_q4(int flags,
 MemorySegment a,
 int aoffset,
 MemorySegment bf,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldbf,
 int ldc)

void gemm_f32_q4(int flags, float* a, int aoffset, float* bf, char* b, int boffset, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc);

gemm_f32_q4_batch

public static void gemm_f32_q4_batch(int flags,
 int batch_num,
 MemorySegment a,
 int aoffset,
 MemorySegment bf,
 MemorySegment b,
 int boffset,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldbf,
 int ldc)

void gemm_f32_q4_batch(int flags, int batch_num, float* a, int aoffset, float** bf, char** b, int boffset, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldbf, int ldc);

gemm_bf16

public static void gemm_bf16(int flags,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment cr,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_bf16(int flags, short* a, int aoffset, short* b, int boffset, short* cr, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);

gemm_bf16_batch

public static void gemm_bf16_batch(int flags,
 int batch_num,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment cr,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_bf16_batch(int flags, int batch_num, short* a, int aoffset, short** b, int boffset, short** cr, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);

gemm_f32_bf16

public static void gemm_f32_bf16(int flags,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment cr,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_f32_bf16(int flags, float* a, int aoffset, short* b, int boffset, short* cr, float* r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);

gemm_f32_bf16_batch

public static void gemm_f32_bf16_batch(int flags,
 int batch_num,
 MemorySegment a,
 int aoffset,
 MemorySegment b,
 int boffset,
 MemorySegment cr,
 MemorySegment r,
 int roffset,
 int m,
 int n0,
 int n,
 int k,
 int lda,
 int ldb,
 int ldc)

void gemm_f32_bf16_batch(int flags, int batch_num, float* a, int aoffset, short** b, int boffset, short** cr, float** r, int roffset, int m, int n0, int n, int k, int lda, int ldb, int ldc);