sebsjoberg

seed: 1390938656
1. Intel(R) Core(TM)2 Duo CPU     P8600  @ 2.40GHz (Apple)

Running 11 test cases...
#if defined(cl_khr_fp64)
#  pragma OPENCL EXTENSION cl_khr_fp64: enable
#elif defined(cl_amd_fp64)
#  pragma OPENCL EXTENSION cl_amd_fp64: enable
#endif

sebsjoberg

OpenCL devices:

  Intel(R) Core(TM)2 Duo CPU     P8600  @ 2.40GHz
    CL_PLATFORM_NAME                 = Apple
    CL_DEVICE_VENDOR                 = Intel
    CL_DEVICE_VERSION                = OpenCL 1.2
    CL_DEVICE_MAX_COMPUTE_UNITS      = 2
    CL_DEVICE_HOST_UNIFIED_MEMORY    = 1
    CL_DEVICE_GLOBAL_MEM_SIZE        = 8589934592
    CL_DEVICE_LOCAL_MEM_SIZE         = 32768

sebsjoberg

// Returns char, int, short, or long, or the unsigned variety
// and takes two char, uchar, short, ushort, int, uint, long, ulong inputs.
#define __CLFN_ALL(name) char __OVERLOAD__ name(char x, char y); \
char2 __OVERLOAD__ name(char2 x, char2 y); \
char3 __OVERLOAD__ name(char3 x, char3 y); \
char4 __OVERLOAD__ name(char4 x, char4 y); \
char8 __OVERLOAD__ name(char8 x, char8 y); \
char16 __OVERLOAD__ name(char16 x, char16 y); \
uchar __OVERLOAD__ name(uchar x, uchar y); \
uchar2 __OVERLOAD__ name(uchar2 x, uchar2 y); \

sebsjoberg

#include <vexcl/vexcl.hpp>

vex::vector<cl_float2>
hilbert(vex::vector<float> &in, std::size_t rows, std::size_t cols) {
    // No need to pass context in:
    const auto &ctx = in.queue_list();

    // Use function instead of vector. Should save global memory IO.
    VEX_FUNCTION(h, float(size_t/*prm1 = idx*/, size_t/*prm2 = rows*/),
            VEX_STRINGIZE_SOURCE(

sebsjoberg

#include <vexcl/vexcl.hpp>
#include <vector>
#include <algorithm>

int main() {
  vex::Context ctx(vex::Filter::Env && vex::Filter::Count(1));
  std::cout << ctx << std::endl;

  std::vector<float> v = {
   0, 5,