std1.0.0-beta-7 SafeC docs

std :: ml::device

Struct std::DevGemm

#include <std/ml/device.h> source

struct DevGemm {
    int dtype;
    int ctype;
    int ta;
    int tb;
    unsigned long m;
    unsigned long n;
    unsigned long k;
    float alpha;
    unsigned long a;
    unsigned long lda;
    unsigned long sa;
    unsigned long b;
    unsigned long ldb;
    unsigned long sb;
    float beta;
    unsigned long c;
    unsigned long ldc;
    unsigned long sc;
    unsigned long batch;
}

A strided batch of GEMMs: for i < batch, c_i = alpha op(a_i) op(b_i) + beta c_i with x_i = x + i * sx (strides in elements), as gemm(). dtype (DEVICE_FP32 / FP16 / BF16) is a and b's element type; c is fp32 and accumulates in fp32.

Fields

§int dtypesource

a and b's element type

§int ctypesource

c's (DEVICE_FP32 / FP16 / BF16)

§int tasource
§int tbsource
§unsigned long msource
§unsigned long nsource
§unsigned long ksource
§float alphasource
§unsigned long asource
§unsigned long ldasource
§unsigned long sasource
§unsigned long bsource
§unsigned long ldbsource
§unsigned long sbsource
§float betasource
§unsigned long csource
§unsigned long ldcsource
§unsigned long scsource
§unsigned long batchsource