小波的世界

生活、教育、技术、旅游

日志

关于我

nkwavelet

文章分类

SSE命令示例代码（算术、逻辑、比较）

2014-09-10 20:01:37| 分类： SIMD | 标签： |举报 |字号大中小订阅

下载LOFTER 我的照片书 |

以下示例代码为原创，主要是为了测试SSE各种命令的实际效果，为了更好的理解各种命令。

1. SSE的算术操作

// 测试SSE的算术操作
void TestSSEArith()
{
	__m128 a;
	a.m128_f32[0] = 10.0;
	a.m128_f32[1] = -20.0;
	a.m128_f32[2] = 30.0;
	a.m128_f32[3] = -40.0;

	__m128 b;
	b.m128_f32[0] = 90.0;
	b.m128_f32[1] = 75.0;
	b.m128_f32[2] = -50.0;
	b.m128_f32[3] = 35.0;

	printf("a: (%.2f, %.2f, %.2f, %.2f)\n\n", a.m128_f32[0], a.m128_f32[1], a.m128_f32[2], a.m128_f32[3]);
	printf("b: (%.2f, %.2f, %.2f, %.2f)\n\n", b.m128_f32[0], b.m128_f32[1], b.m128_f32[2], b.m128_f32[3]);

	// 加、减、乘、除
	__m128 v1 = _mm_add_ss(a, b);		// a[0] + b[0], a[1], a[2], a[3]
	__m128 v2 = _mm_add_ps(a, b);		// a[0] + b[0], a[1] + b[1], a[2] + b[2], a[3] + b[3]
	__m128 v3 = _mm_sub_ss(a, b);		// a[0] - b[0], a[1], a[2], a[3]
	__m128 v4 = _mm_sub_ps(a, b);		// a[0] - b[0], a[1] - b[1], a[2] - b[2], a[3] - b[3]
	__m128 v5 = _mm_mul_ss(a, b);		// a[0] * b[0], a[1], a[2], a[3]
	__m128 v6 = _mm_mul_ps(a, b);		// a[0] * b[0], a[1] * b[1], a[2] * b[2], a[3] * b[3]
	__m128 v7 = _mm_div_ss(a, b);		// a[0] / b[0], a[1], a[2], a[3]
	__m128 v8 = _mm_div_ps(a, b);		// a[0] / b[0], a[1] / b[1], a[2] / b[2], a[3] / b[3]

	printf("v1 = a0 + b0: (%.2f, %.2f, %.2f, %.2f)\n\n", v1.m128_f32[0], v1.m128_f32[1], v1.m128_f32[2], v1.m128_f32[3]);
	printf("v2 = a  + b : (%.2f, %.2f, %.2f, %.2f)\n\n", v2.m128_f32[0], v2.m128_f32[1], v2.m128_f32[2], v2.m128_f32[3]);
	printf("v3 = a0 - b0: (%.2f, %.2f, %.2f, %.2f)\n\n", v3.m128_f32[0], v3.m128_f32[1], v3.m128_f32[2], v3.m128_f32[3]);
	printf("v4 = a  - b : (%.2f, %.2f, %.2f, %.2f)\n\n", v4.m128_f32[0], v4.m128_f32[1], v4.m128_f32[2], v4.m128_f32[3]);
	printf("v5 = a0 * b0: (%.2f, %.2f, %.2f, %.2f)\n\n", v5.m128_f32[0], v5.m128_f32[1], v5.m128_f32[2], v5.m128_f32[3]);
	printf("v6 = a  * b : (%.2f, %.2f, %.2f, %.2f)\n\n", v6.m128_f32[0], v6.m128_f32[1], v6.m128_f32[2], v6.m128_f32[3]);
	printf("v7 = a0 / b0: (%.2f, %.2f, %.2f, %.2f)\n\n", v7.m128_f32[0], v7.m128_f32[1], v7.m128_f32[2], v7.m128_f32[3]);
	printf("v8 = a  / b : (%.2f, %.2f, %.2f, %.2f)\n\n", v8.m128_f32[0], v8.m128_f32[1], v8.m128_f32[2], v8.m128_f32[3]);

	// 开方、倒数
	__m128 v9  = _mm_sqrt_ss(a);			// sqrt(a[0]), a[1], a[2], a[3]
	__m128 v10 = _mm_sqrt_ps(a);			// sqrt(a[0]), sqrt(a[1]), sqrt(a[2]), sqrt(a[3])
	__m128 v11 = _mm_rcp_ss(a);			// 1/a[0], a[1], a[2], a[3]
	__m128 v12 = _mm_rcp_ps(a);			// 1/a[0], 1/a[1], 1/a[2], 1/a[3]
	__m128 v13 = _mm_rsqrt_ss(a);			// 1/(sqrt(a[0])), a[1], a[2], a[3]
	__m128 v14 = _mm_rsqrt_ps(a);			// 1/(sqrt(a[0])), 1/(sqrt(a[1])), 1/(sqrt(a[2])), 1/(sqrt(a[3]))
	printf("v9  = sqrt(a0) : (%.2f, %.2f, %.2f, %.2f)\n\n", v9.m128_f32[0],  v9.m128_f32[1],  v9.m128_f32[2],  v9.m128_f32[3]);
	printf("v10 = sqrt(a)  : (%.2f, %.2f, %.2f, %.2f)\n\n", v10.m128_f32[0], v10.m128_f32[1], v10.m128_f32[2], v10.m128_f32[3]);
	printf("v11 =  1 / a0  : (%.2f, %.2f, %.2f, %.2f)\n\n", v11.m128_f32[0], v11.m128_f32[1], v11.m128_f32[2], v11.m128_f32[3]);
	printf("v12 =  1 / a   : (%.2f, %.2f, %.2f, %.2f)\n\n", v12.m128_f32[0], v12.m128_f32[1], v12.m128_f32[2], v12.m128_f32[3]);
	printf("v13 =1/sqrt(a0): (%.2f, %.2f, %.2f, %.2f)\n\n", v13.m128_f32[0], v13.m128_f32[1], v13.m128_f32[2], v13.m128_f32[3]);
	printf("v14 =1/sqrt(a) : (%.2f, %.2f, %.2f, %.2f)\n\n", v14.m128_f32[0], v14.m128_f32[1], v14.m128_f32[2], v14.m128_f32[3]);

	// 最大值、最小值
	__m128 v15 = _mm_min_ss(a, b);			// min(a[0], b[0]), a[1], a[2], a[3]
	__m128 v16 = _mm_min_ps(a, b);			// min(a[0], b[0]), min(a[1], b[1]), min(a[2], b[2]), min(a[3], b[3])
	__m128 v17 = _mm_max_ss(a, b);			// max(a[0], b[0]), a[1], a[2], a[3]
	__m128 v18 = _mm_max_ps(a, b);			// max(a[0], b[0]), max(a[1], b[1]), max(a[2], b[2]), max(a[3], b[3])
	printf("v15 = min(a0, b0): (%.2f, %.2f, %.2f, %.2f)\n\n", v15.m128_f32[0], v15.m128_f32[1], v15.m128_f32[2], v15.m128_f32[3]);
	printf("v16 = min(a,  b) : (%.2f, %.2f, %.2f, %.2f)\n\n", v16.m128_f32[0], v16.m128_f32[1], v16.m128_f32[2], v16.m128_f32[3]);
	printf("v17 = max(a0, b0): (%.2f, %.2f, %.2f, %.2f)\n\n", v17.m128_f32[0], v17.m128_f32[1], v17.m128_f32[2], v17.m128_f32[3]);
	printf("v18 = max(a , b) : (%.2f, %.2f, %.2f, %.2f)\n\n", v18.m128_f32[0], v18.m128_f32[1], v18.m128_f32[2], v18.m128_f32[3]);
}

测试结果：

Remark: 注意以上的 -1.#J 是因为负数开方出错导致的。

2. SSE的逻辑操作

// 测试SSE的逻辑操作
void TestSSELogic()
{
	__m128 a;
	a.m128_i32[0] = 0x96;
	a.m128_i32[1] = 0x15;
	a.m128_i32[2] = 0xB4;
	a.m128_i32[3] = 0x7A;

	__m128 b;
	b.m128_i32[0] = 0x34;
	b.m128_i32[1] = 0x57;
	b.m128_i32[2] = 0x9C;
	b.m128_i32[3] = 0xD1;

	__m128 v1 = _mm_and_ps(a, b);			// a & b
	__m128 v2 = _mm_andnot_ps(a, b);		// ~a & b
	__m128 v3 = _mm_or_ps(a, b);			// a | b
	__m128 v4 = _mm_xor_ps(a, b);			// a ^ b

        printf("a : (0x%x, 0x%x, 0x%x, 0x%x)\n\n", a.m128_i32[0], a.m128_i32[1], a.m128_i32[2], a.m128_i32[3]);
	printf("b : (0x%x, 0x%x, 0x%x, 0x%x)\n\n", b.m128_i32[0], b.m128_i32[1], b.m128_i32[2], b.m128_i32[3]);
	printf("v1 = a & b: (%d, %d, %d, %d)\n\n", v1.m128_i32[0], v1.m128_i32[1], v1.m128_i32[2], v1.m128_i32[3]);
	printf("v2 =~a & b: (%d, %d, %d, %d)\n\n", v2.m128_i32[0], v2.m128_i32[1], v2.m128_i32[2], v2.m128_i32[3]);
	printf("v3 = a | b: (%d, %d, %d, %d)\n\n", v3.m128_i32[0], v3.m128_i32[1], v3.m128_i32[2], v3.m128_i32[3]);
	printf("v4 = a ^ b: (%d, %d, %d, %d)\n\n", v4.m128_i32[0], v4.m128_i32[1], v4.m128_i32[2], v4.m128_i32[3]);
}

测试结果：

3. SSE的比较操作

// 测试SSE的比较操作
void TestSSECompare()
{
	__m128 a;
	a.m128_i32[0] = 1200;
	a.m128_i32[1] = 100;
	a.m128_i32[2] = 350;
	a.m128_i32[3] = 800;

	__m128 b;
	b.m128_i32[0] = 900;
	b.m128_i32[1] = 400;
	b.m128_i32[2] = 350;
	b.m128_i32[3] = 600;

	__m128 v[30];
	v[0]  = _mm_cmpeq_ss(a, b);
	v[1]  = _mm_cmpeq_ps(a, b);
	v[2]  = _mm_cmplt_ss(a, b);
	v[3]  = _mm_cmplt_ps(a, b);
	v[4]  = _mm_cmple_ss(a, b);
	v[5]  = _mm_cmple_ps(a, b);
	v[6]  = _mm_cmpgt_ss(a, b);
	v[7]  = _mm_cmpgt_ps(a, b);
	v[8]  = _mm_cmpge_ss(a, b);
	v[9]  = _mm_cmpge_ps(a, b);
	v[10] = _mm_cmpneq_ss(a, b);
	v[11] = _mm_cmpneq_ps(a, b);
	v[12] = _mm_cmpnlt_ss(a, b);
	v[13] = _mm_cmpnlt_ps(a, b);
	v[14] = _mm_cmpnle_ss(a, b);
	v[15] = _mm_cmpnle_ps(a, b);
	v[16] = _mm_cmpngt_ss(a, b);
	v[17] = _mm_cmpngt_ps(a, b);
	v[18] = _mm_cmpnge_ss(a, b);
	v[19] = _mm_cmpnge_ps(a, b);

	v[20] = _mm_cmpord_ss(a, b);
	v[21] = _mm_cmpord_ps(a, b);
	v[22] = _mm_cmpunord_ss(a, b);
	v[23] = _mm_cmpunord_ps(a, b);

	int r[12];
	r[0] = _mm_comieq_ss(a, b);
	r[1] = _mm_comilt_ss(a, b);
	r[2] = _mm_comile_ss(a, b);
	r[3] = _mm_comigt_ss(a, b);
	r[4] = _mm_comige_ss(a, b);
	r[5] = _mm_comineq_ss(a, b);
	r[6] = _mm_ucomieq_ss(a, b);
	r[7] = _mm_ucomilt_ss(a, b);
	r[8] = _mm_ucomile_ss(a, b);
	r[9] = _mm_ucomigt_ss(a, b);
	r[10] = _mm_ucomige_ss(a, b);
	r[11] = _mm_ucomineq_ss(a, b);

	for(int i = 0; i < 24; i += 2)
	{
		printf("v[%2d]: (%d, %d, %d, %d)\t v[%2d]: (%d, %d, %d, %d)\n", i, v[i].m128_i32[0], v[i].m128_i32[1], v[i].m128_i32[2], 
                        v[i].m128_i32[3], i+1, v[i+1].m128_i32[0], v[i+1].m128_i32[1], v[i+1].m128_i32[2], v[i+1].m128_i32[3]);
	}
	printf("\n");
	for(int i = 0; i < 12; i += 4)
		printf("r[%2d]: %d\t r[%2d]: %d\t r[%2d]: %d\t r[%2d]: %d\n", i, r[i], i+1, r[i+1], i+2, r[i+2], i+3, r[i+3]);
}

测试结果：

评论这张

转发至微博

阅读(596)| 评论(0)

历史上的今天

this.p={  m:2,
              b:2,
              loftPermalink:'',
              id:'fks_087071093080087066092095085071072084083068083083080067087094',
              blogTitle:'SSE命令示例代码（算术、逻辑、比较）',
              blogAbstract:'<font size=\"2\"   \>以下示例代码为原创，主要是为了测试SSE各种命令的实际效果，为了更好的理解各种命令。<wbr\></font\><div\><font size=\"2\"   \><br\></font\></div\><div\><font size=\"2\"   color=\"#ff0000\"   \><b\>1.  SSE的算术操作</b\></font\></div\><pre class=\"prettyprint\"   \><p\><font size=\"2\"   \><font color=\"#008000\"   \>// 测试SSE的算术操作</font\><br\>void TestSSEArith()<br\>{<br\>\t__m128 a;<br\>\ta.m128_f32[0] = 10.0;<br\>\ta.m128_f32[1] = -20.0;<br\>\ta.m128_f32[2] = 30.0;<br\>\ta.m128_f32[3] = -40.0;<br\><br\>\t__m128 b;<br\>\tb.m128_f32[0] = 90.0;<br\>\tb.m128_f32[1] = 75.0;<br\>\tb.m128_f32[2] = -50.0;<br\>\tb.m128_f32[3] = 35.0;<br\><br\>\tprintf(\"a: (%.2f, %.2f, %.2f, %.2f)\\n\\n\",</font\></p\></pre\>',
              blogTag:'simd',
              blogUrl:'blog/static/22775603820148108057906',
              isPublished:1,
              istop:false,
              type:2,
              modifyTime:1410404893241,
              publishTime:1410350497029,
              permalink:'blog/static/22775603820148108057906',
              commentCount:0,
              mainCommentCount:0,
              recommendCount:0,
              bsrk:-100,
              publisherId:0,
              recomBlogHome:false,
              currentRecomBlog:false,
              attachmentsFileIds:[],
              vote:{},
              groupInfo:{},
              friendstatus:'none',
              followstatus:'unFollow',
              pubSucc:'',
              visitorProvince:'',
              visitorCity:'',
              visitorNewUser:false,
              postAddInfo:{},
              mset:'000',
              mcon:'',
              srk:-100,
              remindgoodnightblog:false,
              isBlackVisitor:false,
              isShowYodaoAd:false,
              hostIntro:'',
              hmcon:'0',
              selfRecomBlogCount:'0',
              lofter_single:'<iframe width="140" height="560" style="overflow:hidden;" src="http://www.lofter.com/mailEntry.do?blogad=1&blog" frameBorder="0"></iframe>'
            }

{list a as x}
    {if !!x}
    <div class="iblock nbw-fce nbw-f40">
      <a class="fc03 noul" target="_blank" hidefocus="true" href="http://blog.163.com/${x.visitorName}/">
      {if x.visitorName==visitor.userName}
      <img alt="${x.visitorNickname|escape}" onerror="this.src=location.f40" class="cwd bdwa bdc0" src="${fn1(x.visitorName)}&r=${visitor.imageUpdateTime}"/>
      {else}
      <img alt="${x.visitorNickname|escape}" onerror="this.src=location.f40" class="cwd bdwa bdc0" src="${fn1(x.visitorName)}"/>
      {/if}
      </a>
      <div class="cwd vname thide">
        {if x.moveFrom=='wap'}
          <a class="noul pnt" target="_blank" href="http://blog.163.com/services/wapblog.html?frompersonalbloghome"><span title="来自网易手机博客" class="iblock wapIcon"> </span></a>
        {elseif x.moveFrom=='iphone'}
          <a class="noul pnt" target="_blank"><span title="来自iPhone客户端" class="iblock iphoneIcon"> </span></a>
        {elseif x.moveFrom=='android'}
          <a class="noul pnt" target="_blank"><span title="来自Android客户端" class="iblock androidIcon"> </span></a>
        {elseif x.moveFrom=='mobile'}
          <a class="noul pnt" target="_blank" href="http://blog.163.com/services/emsblog.html?frompersonalbloghome"><span title="来自网易短信写博" class="iblock wapIcon"> </span></a>
        {/if}
        <a class="fc03 m2a"  target="_blank" hidefocus="true" href="http://blog.163.com/${x.visitorName}/">
          ${fn(x.visitorNickname,8)|escape}
        </a>
      </div>
    </div>
    {/if}
    {/list}

<#--最新日志，群博日志--> <#--推荐日志-->

<p class="fc06">推荐过这篇日志的人：</p>
    <div>
      {list a as x}
      {if !!x}
      <div class="iblock nbw-fce nbw-f40">
        <a class="fc03 noul" target="_blank" hidefocus="true" href="http://blog.163.com/${x.recommenderName}/">
        <img alt="${x.recommenderNickname|escape}" onerror="this.src=location.f40" class="cwd bdwa bdc0" src="${fn1(x.recommenderName)}"/>
        </a>
        <div class="cwd thide">
          <a class="fc03 m2a" target="_blank" hidefocus="true" href="http://blog.163.com/${x.recommenderName}/">
            ${fn(x.recommenderNickname,6)|escape}
          </a>
        </div>
      </div>
      {/if}
      {/list}
    </div>
    {if !!b&&b.length>0}
    <p  class="fc06">他们还推荐了：</p>
    <ul>
    {list b as y}
      {if !!y}
        <li class="rrb"><span class="iblock">·</span><a class="fc03 m2a" target="_blank" href="http://blog.163.com/${y.recommendBlogPermalink}/?from=blog/static/22775603820148108057906">${y.recommendBlogTitle|escape}</a></li>
      {/if}
    {/list}
    </ul>
    {/if}

<#--引用记录--> <#--博主推荐--> <#--随机阅读--> <#--首页推荐--> <#--历史上的今天--> <#--被推荐日志--> <#--上一篇，下一篇--> <#-- 热度 -->

{list a as x}
    {if !!x}
    <div class="hotItem iblock nbw-fce nbw-f40">
      <a class="fc03 noul" target="_blank" hidefocus="true" href="http://blog.163.com/${x.publisherUsername}/">
      {if x.publisherUsername==visitor.userName}
      <img alt="${x.publisherNickname|escape}" onerror="this.src=location.f40" class="cwd bdwa bdc0" src="${fn1(x.publisherUsername)}&r=${visitor.imageUpdateTime}"/>
      {else}
      <img alt="${x.publisherNickname|escape}" onerror="this.src=location.f40" class="cwd bdwa bdc0" src="${fn1(x.publisherUsername)}"/>
      {/if}
      </a>
      <div class="cwd vname thide">
        <a class="fc03 m2a"  target="_blank" hidefocus="true" href="http://blog.163.com/${x.publisherUsername}/">
          ${fn(x.publisherNickname,8)|escape}
        </a>
      </div>
      <a class="f-myLikeIcons hottype {if x.type==1} js-liketype{elseif x.type==2} js-reblogtype{elseif x.type==3} js-sharetype{else}{/if}" target="_blank" hidefocus="true" href="http://blog.163.com/${x.publisherUsername}/"> </a>
    </div>
    {/if}
    {/list}

<#-- 网易新闻广告 --> <#--右边模块结构--> <#--评论模块结构--> <#--引用模块结构--> <#--博主发起的投票-->

页脚

我的照片书 - 手机博客 - 下载LOFTER APP - 订阅此博客

小波的世界

导航

日志

SSE命令示例代码（算术、逻辑、比较）

历史上的今天

最近读者

热度

评论

页脚