Question

我想通过gcc和inline asm将32位值写入索引指针。

这是它应该做的：

this->screenPtr[ x + drawY * this->w ] = col;

这是我的一个尝试：

            asm (
                "movl   %2,     %%eax;"
                "imull  %3;"
                "addl   %1,     %%eax;"
                "movl   %4,     (%0 , %%eax, 4);"

                : "=r" (*this->screenPtr)
                : "r" (x), "r" (drawY), "r" (this->w), "r" (col)
                : "%eax"
                 );

怎么做的？谢谢。

哦，我可以在这里编辑:)（有人要求提供更多代码）

    class raycaster{
        public:
        int     isInit;
        long*   screenPtr;
        int     w;
        int     h;
        int     mapWidth;
        int     mapHeight;
        cam_t   camera;
        ...


    (xLoop)
      (calculations)
        for (int drawY = drawStart; drawY < drawEnd; drawY ++ ) {

            int texY;
            long col;
            asm volatile(
                //int d = (drawY * 2 - h + lineHeight) * 128;
                "movl   %1,         %%eax;" // mov drawY in eax
                "shll   $1,         %%eax;" // * 2
                "subl   %2,         %%eax;" // - h
                "addl   %3,         %%eax;" // + lineHeight
                "movl   $128,       %%ebx;"
                "imull  %%ebx;"             // * 128

                //int texY = ((d * TEX_HEIGHT) / lineHeight) / 256;
                "imull  %4;"                // * TEX_HEIGHT
                "movl   %3,         %%ebx;"
                "idivl  %%ebx;"              // : lineHeight
                "movl   $0,         %%edx;"
                "movl   $256,       %%ebx;"
                "idivl  %%ebx;"             // : 256

                //[texX + texY * TEX_HEIGHT];
                "movl   %4,         %%ebx;"
                "imull  %%ebx;"             // * TEX_HEIGHT
                "addl   %5,         %%eax;" // + lineHeight

                "movl   %%eax,      %0;"
                : "=r" (texY)
                : "r"  (drawY), "m" (h), "r" (lineHeight), "r" (TEX_HEIGHT), "m" (texX)
                : "%eax", "%ebx", "%edx"
                );

            col = this->textureMemory[textureNumber - 1][texY];

这就是我做col的方法，现在我想将它组合起来，这样它就可以直接在缓冲区中编写它。

            /*
            asm (
                "movl   %2,     %%eax;"
                "imull  %3;"
                "addl   %1,     %%eax;"
                "movl   %4,     (%0 , %eax, 4);"
                //"movl   %%eax,  %0;"

                : "=r" (*screenPtr)
                : "r" (x), "r" (drawY), "r" (this->w), "r" (col)
                : "%eax"
                 );
            */

            this->screenPtr[ x + drawY * this->w ] = col;
        }

“额外”代码可能更令人困惑，专注于第一个剪切和问题。

编辑2：

如果有人阅读了评论，我们一直在寻找速度慢的原因。这段代码更好，达到了汇编代码的速度。我改变的是，我删除了临时变量以避免内存访问。

            int texY = texX + (((drawY * 2 - h + lineHeight) * 128* TEX_HEIGHT) / lineHeight) / 256  * TEX_HEIGHT;

            long col = this->textureMemory[textureNumber - 1][texY];
            this->screenPtr[ x + drawY * this->w ] = col;

gcc内联asm和指针索引

0 个答案: