小编jca*_*zac的帖子

编辑:我编辑了问题及其标题更精确.

考虑以下源代码:

#include <vector>
struct xyz {
    xyz() { } // empty constructor, but the compiler doesn't care
    xyz(const xyz& o): v(o.v) { } 
    xyz& operator=(const xyz& o) { v=o.v; return *this; }
    int v; // <will be initialized to int(), which means 0
};

std::vector<xyz> test() {
    return std::vector<xyz>(1024); // will do a memset() :-(
}

...我怎么能避免vector <>分配的内存用它的第一个元素的副本进行初始化,这是一个O(n)操作我宁愿为了速度而跳过,因为我的默认构造函数什么都不做？

如果不存在通用的解决方案,那么g ++特定的解决方案就可以做到(但我找不到任何属性来执行此操作).

编辑:生成的代码如下(命令行:arm-elf-g ++ - 4.5 -O3 -S -fno-verbose-asm -o-test.cpp | arm-elf-c ++ filt | grep -vE'^ [[: space:]] …

有人可以访问iPhone 3GS或Pandora,请测试我刚写的下面的汇编程序吗？

它应该在NEON矢量FPU上非常快速地计算正弦和余弦.我知道它编译得很好,但没有足够的硬件我就无法测试它.如果你可以计算一些正弦和余弦,并将结果与​​sinf()和cosf()的结果进行比较,那将会有所帮助.

谢谢!

#include <math.h>

/// Computes the sine and cosine of two angles
/// in: angles = Two angles, expressed in radians, in the [-PI,PI] range.
/// out: results = vector containing [sin(angles[0]),cos(angles[0]),sin(angles[1]),cos(angles[1])]
static inline void vsincos(const float angles[2], float results[4]) {
    static const float constants[]  = { 
    /* q1 */  0,                M_PI_2,           0,                M_PI_2,
    /* q2 */  M_PI,             M_PI,             M_PI,             M_PI,
    /* q3 */  4.f/M_PI,         4.f/M_PI,         4.f/M_PI,         4.f/M_PI,
    /* q4 */ -4.f/(M_PI*M_PI), -4.f/(M_PI*M_PI), -4.f/(M_PI*M_PI), -4.f/(M_PI*M_PI),
    /* q5 …

使用arm-elf-gcc-4.5 -O3 -march = armv7-a -mthumb -mfpu = neon -mfloat-abi = softfp编译该程序时:

#include <arm_neon.h>
extern float32x4_t cross(const float32x4_t& v1, const float32x4_t& v2) {
    float32x4x2_t
        xxyyzz1(vzipq_f32(v1, v1)),
        xxyyzz2(vzipq_f32(v2, v2));

    float32x2_t
        xx1(vget_low_f32(xxyyzz1.val[0])),
        yy1(vget_high_f32(xxyyzz1.val[0])),
        zz1(vget_low_f32(xxyyzz1.val[1])),
        xx2(vget_low_f32(xxyyzz2.val[0])),
        yy2(vget_high_f32(xxyyzz2.val[0])),
        zz2(vget_low_f32(xxyyzz2.val[1]));

    float32x2_t
        x(vmls_f32(vmul_f32(yy1, zz2), zz1, yy2)),
        y(vmls_f32(vmul_f32(zz1, xx2), xx1, zz2)),
        z(vmls_f32(vmul_f32(xx1, yy2), yy1, xx2));

    return vcombine_f32(vuzp_f32(x, y).val[0], z);
}

......这就是我得到的.注意两个用@ <<<标记的无用指令

_Z5crossRK19__simd128_float32_tS1_:
    vldmia  r0, {d16-d17}
    vldmia  r1, {d22-d23}
    vmov    q10, q8  @ v4sf
    vmov    q9, q11  @ v4sf
    vzip.32 q8, q10
    vzip.32 q11, q9
    vmov    d24, …

我如何指示python在内部存储我的字符串的预先散列版本,以便在使用我的字符串作为键执行dict/set查找时它将使用该值？

我记得几周前读过它,但目前在python docs中找不到它: - /

考虑以下NEON优化功能:

void mat44_multiply_neon(float32x4x4_t& result, const float32x4x4_t& a, const float32x4x4_t& b) {
    // Make sure "a" is mapped to registers in the d0-d15 range,
    // as requested by NEON multiply operations below:
    register float32x4_t a0 asm("q0") = a.val[0];
    register float32x4_t a1 asm("q1") = a.val[1];
    register float32x4_t a2 asm("q2") = a.val[2];
    register float32x4_t a3 asm("q3") = a.val[3];
    asm volatile (
    "\n\t# multiply two matrices...\n\t"
    "# result (%q0,%q1,%q2,%q3)  = first column of B (%q4) * first row of A (q0-q3)\n\t"
    "vmul.f32 %q0, %q4, …