OpenBLAS: Multi-Thread & Speed Problem

Hi, everyone I am not sure that if the speed I got is the quickest result, so I wanna make for sure here, and in my opinion, It can be faster. I ran a test.c like this:

#include <stdio.h>
#include "/usr/local/openblas_a9/include/cblas.h"
#include <time.h>
#include <stdlib.h>
#define random(x)(rand()%x)

void main()
{
   int M=160*160,K=32*9,N=64;
   int i;
   float* A=(float*)malloc(sizeof(float)*M*K);
   for(i=0;i<M*K;i++)
   {
      A[K]=(float)random(10);
    }

   float* B=(float*)malloc(sizeof(float)*N*K);
   for(i=0;i<N*K;i++)
   {
      B[K]=(float)random(10);
    }
   float* C=(float*)malloc(sizeof(float)*N*M);
   
   float alpha=1;
   float beta=0;
   int lda=K;
   int ldb=N;
   int ldc=N;
   
   clock_t start1,end1;
   start1=clock();
   cblas_sgemm(CblasRowMajor,CblasNoTrans,CblasNoTrans,M,N,K,alpha,A,lda,B,ldb,beta,C,ldc);
   end1=clock();
   printf("timeusing:%lf",((double)end1 - (double)start1)/ CLOCKS_PER_SEC);
   printf("\n");
}

Before I compiled it I tried to use the multi-threads, I inputed this in my ubuntu teminal export OPENBLAS_NUM_THREADS=4 Then I compiled this test.c by using this instruction : arm-linux-gnueabihf-gcc -static -o test test.c -I /usr/local/openblas_a9/include/ -L /usr/local/openblas_a9/lib /usr/local/openblas_a9/lib/libopenblas.a -lpthread -lgfortran -std=c99 -mcpu=cortex-a9 -mfpu=neon-fp16 -mfloat-abi=hard -O3 -ffast-math

The CPU that run the executable file is exynos4412 cortex-a9

the time that cblas_sgemm used is around 0.48s. If I didn’t input export OPENBLAS_NUM_THREADS=4 the time is 0.68s

I wanna know that if I did something wrong when I use openblas to caculate the matrix and if the speed is reasonable, if not , how can I speed up?

Thanks for your attention!