fft_execute.cpp

#include "fft_internal.h" 
#include <clFFT.h> 
#include <stdlib.h> 
#include <stdio.h> 
#include <math.h> 
  
#define max(a,b) (((a)>(b)) ? (a) : (b)) 
#define min(a,b) (((a)<(b)) ? (a) : (b)) 
  
static cl_int 
allocateTemporaryBufferInterleaved(cl_fft_plan *plan, cl_uint batchSize) 
{ 
    cl_int err = CL_SUCCESS; 
    if(plan->temp_buffer_needed && plan->last_batch_size != batchSize)  
    { 
        plan->last_batch_size = batchSize;  
        size_t tmpLength = plan->n.x * plan->n.y * plan->n.z * batchSize * 2 * sizeof(cl_float); 
         
        if(plan->tempmemobj) 
            clReleaseMemObject(plan->tempmemobj); 
             
        plan->tempmemobj = clCreateBuffer(plan->context, CL_MEM_READ_WRITE, tmpLength, NULL, &err); 
    } 
    return err;  
} 
  
static cl_int 
allocateTemporaryBufferPlannar(cl_fft_plan *plan, cl_uint batchSize) 
{ 
    cl_int err = CL_SUCCESS; 
    cl_int terr; 
    if(plan->temp_buffer_needed && plan->last_batch_size != batchSize)  
    { 
        plan->last_batch_size = batchSize;  
        size_t tmpLength = plan->n.x * plan->n.y * plan->n.z * batchSize * sizeof(cl_float); 
         
        if(plan->tempmemobj_real) 
            clReleaseMemObject(plan->tempmemobj_real); 
  
        if(plan->tempmemobj_imag) 
            clReleaseMemObject(plan->tempmemobj_imag);           
             
        plan->tempmemobj_real = clCreateBuffer(plan->context, CL_MEM_READ_WRITE, tmpLength, NULL, &err); 
        plan->tempmemobj_imag = clCreateBuffer(plan->context, CL_MEM_READ_WRITE, tmpLength, NULL, &terr); 
        err |= terr; 
    }    
    return err; 
} 
  
void 
getKernelWorkDimensions(cl_fft_plan *plan, cl_fft_kernel_info *kernelInfo, cl_int *batchSize, size_t *gWorkItems, size_t *lWorkItems) 
{ 
    *lWorkItems = kernelInfo->num_workitems_per_workgroup; 
    int numWorkGroups = kernelInfo->num_workgroups; 
    int numXFormsPerWG = kernelInfo->num_xforms_per_workgroup; 
     
    switch(kernelInfo->dir) 
    { 
        case cl_fft_kernel_x: 
            *batchSize *= (plan->n.y * plan->n.z); 
            numWorkGroups = (*batchSize % numXFormsPerWG) ? (*batchSize/numXFormsPerWG + 1) : (*batchSize/numXFormsPerWG); 
            numWorkGroups *= kernelInfo->num_workgroups; 
            break; 
        case cl_fft_kernel_y: 
            *batchSize *= plan->n.z; 
            numWorkGroups *= *batchSize; 
            break; 
        case cl_fft_kernel_z: 
            numWorkGroups *= *batchSize; 
            break;