Float to bf16*2 conversion

Accelerated Computing CUDA CUDA Programming and Performance
1

hello i would like to know if there is any instruction to convert BF16*2 to float, thanks

i tried this, but I don’t know how to specify the value 0 or 1 of bf16 * 2

#include <mma.h>
#include "math.h"
#include <iostream>
#include <stdio.h>
#include <cuda_pipeline.h>

__global__ void Test(){
    uint32_t a = 0;
    float x = 0;
    float x1 = 1;
    float x2 = 2;

    asm("cvt.rn.bf16x2.f32 %0, %1, %2;\n" : "=r"(a) : "f"(x1), "f"(x2));

    asm ("mov.b32 %0, %1;\n" : "=f"(x) : "r"(a));
    printf("x = %f \n", x);


}

int main() {
    Test<<<1, 32>>>();
    cudaDeviceSynchronize();
}

thank for help

2

there are intrinsics provided for this