基本特点

平行向量处理机最大的特点是系统中的CPU是专门定制的向量处理器（VP）。系统还提供共享存储器以及与VP相连的高速交叉开关。

来自现实世界的例子: x86 架构中的向量指令应用

// 改自英文維基 /wiki/Vector_processor//SSE simd function for vectorized multiplication of 2 arrays with single-precision floatingpoint numbers//t param pointer on source/destination array, 2nd param 2. source array, 3rd param number of floats per arrayvoidmul_asm(float*out,float*in,unsignedintleng){unsignedintcount,rest;//compute if array is big enough for vector operationrest=(leng*4)%16;count=(leng*4)-rest;// vectorized part; 4 floats per loop iterationif(count>0){__asm__volatile__(".intel_syntax noprefix\n\t""loop: \n\t""sub ecx,16 \n\t"// decrease address pointer by 4 floats"movups xmm0,[ebx+ecx] \n\t"// loads 4 floats in first register (xmm0)"movups xmm1,[eax+ecx] \n\t"// loads 4 floats in second register (xmm1)"mulps xmm0,xmm1 \n\t"// multiplies both vector registers"movups [eax+ecx],xmm0 \n\t"// write back the result to memory"jnz loop \n\t"".att_syntax prefix \n\t"::"a"(out),"b"(in),"c"(count),"d"(rest):"xmm0","xmm1");}// scalar part; 1 float per loop iterationif(rest!=0){__asm__volatile__(".intel_syntax noprefix\n\t""add eax,ecx \n\t""add ebx,ecx \n\t""rest: \n\t""sub edx,4 \n\t""movss xmm0,[ebx+edx] \n\t"// load 1 float in first register (xmm0)"movss xmm1,[eax+edx] \n\t"// load 1 float in second register (xmm1)"mulss xmm0,xmm1 \n\t"// multiplies both scalar parts of registers"movss [eax+edx],xmm0 \n\t"// write back the result\n\t""jnz rest \n\t"".att_syntax prefix \n\t"::"a"(out),"b"(in),"c"(count),"d"(rest):"xmm0","xmm1");}return;}

参阅

并行计算

免责声明：以上内容版权归原作者所有，如有侵犯您的原创版权请告知，我们将尽快删除相关内容。感谢每一位辛勤著写的作者，感谢每一位的分享。