c++ - Output of cuda program is not what was expected - 


#include<cuda_runtime.h> #include<stdio.h> #include<cuda.h> #include<stdlib.h>   __global__ void setval(char **c){  c[(blockidx.y * griddim.x) + blockidx.x] = "hello\0";  }   int main(){  char **gpu = null; cudamalloc((void**)&gpu, 6 * sizeof(char *)); int i; /* cannot access second level directly for( =0 ; < 6 ;i++){     cudamalloc((void**)&gpu[i], 10 * sizeof(char)); }*/   dim3 grid(3,2); setval<<<grid, 1>>>(gpu); char *p = (char*)malloc(10 * sizeof(char)); char *x[6];  cudamemcpy(x, gpu, 6*sizeof(char*), cudamemcpydevicetohost); for( =0 ; i< 6; i++){     cudamemcpy(p, x[i], 10*sizeof(char), cudamemcpydevicetohost);     //put synchronize here if problem     printf("%s\n",p);  }   getchar(); return 0; }

based on suggestions, revised code make concept correct. but, code still not working :(. appreciated

try -- tested on gtx 285 under cuda 3.2 -- it's bit more restrictive current version, works.

#include<stdio.h> #include<string.h>  __global__ void setvalues(char** word) {     volatile char* myword = word[blockidx.x];      myword[0] = 'h';     myword[1] = 'o';     myword[2] = 'l';     myword[3] = 'a';     myword[4] = '\0'; }  int main() {     const size_t buffersize = 32;     const int nobjects = 10;      char*  h_x[nobjects];     char** d_x = 0;      cudamalloc( (void**)(&d_x), nobjects * sizeof(char*) );      ( int i=0; < nobjects; i++ )     {         h_x[i] = null;         cudamalloc( (void**)(&h_x[i]), buffersize * sizeof(char) );         printf("h_x[%d] = %lx\n",i,(unsigned long)h_x[i]);     }      cudamemcpy( d_x, h_x, nobjects*sizeof(char*), cudamemcpyhosttodevice);     printf("copied h_x[] d_x[]\n");      char msg[] = "hello world!";     cudamemcpy( h_x[0], msg, 13*sizeof(char), cudamemcpyhosttodevice );      /*  force thread synchronization  */     cudaerror err = cudathreadsynchronize();      /*  check , display error  */     if ( cudasuccess != err )     {         fprintf( stderr, "cuda error in file '%s' in line %i : %s.\n",                 __file__, __line__, cudageterrorstring( err) );     }      setvalues<<<nobjects,1>>>(d_x);      /*  force thread synchronization  */     err = cudathreadsynchronize();      /*  check , display error  */     if ( cudasuccess != err )     {         fprintf( stderr, "cuda error in file '%s' in line %i : %s.\n",                 __file__, __line__, cudageterrorstring( err) );     }      printf("kernel completed successfully.  woot.\n\n");      char p[buffersize];      printf("d_x = %lx\n", (unsigned long)d_x );     printf("h_x = %lx\n", (unsigned long)h_x );      cudamemcpy( h_x, d_x, nobjects*sizeof(char*), cudamemcpydevicetohost);      printf("d_x = %lx\n", (unsigned long)d_x );     printf("h_x = %lx\n", (unsigned long)h_x );      ( int i=0; < nobjects; i++ )     {         cudamemcpy( &p, h_x[i], buffersize*sizeof(char), cudamemcpydevicetohost);         printf("%d p[] = %s\n",i,p);     }      /*  force thread synchronization  */     err = cudathreadsynchronize();      /*  check , display error  */     if ( cudasuccess != err )     {         fprintf( stderr, "cuda error in file '%s' in line %i : %s.\n",                 __file__, __line__, cudageterrorstring( err) );     }      getchar();      return 0; }

as @jon notes, can't pass x (as had declared) gpu, because it's address lives on cpu. in code above, create array of char*'s , pass them char** allocated on gpu. hope helps!


Comments

Post a Comment

Popular posts from this blog

c++ - Is it possible to compile a VST on linux? -

for class project i'm attempting write vst plugin backed cuda. current cuda workflow on linux box, i'd prefer compile , link there. according wikipedia , should possible (i couldn't find steinberg documentation relevant linux) can't find makefile or instructions on how build if aren't using xcode or visual studio. i'm vst 3 sdk doesn't support linux. when try compile plugin under linux, error: ./base/source/fatomic.cpp:39:30: fatal error: libkern/osatomic.h: no such file or directory this issue caused following code in "vst3 sdk/base/source/fatomic.cpp" #if mac #include <libkern/osatomic.h> #if mac_os_x_version_min_required > mac_os_x_version_10_4 #define native_atomic_type (volatile int32_t*) #else #define native_atomic_type (int32_t*) #endif #elif windows #include <windows.h> #endif but hope compiling under linux work vst sdk 2.4. reading. jvstwrapper seems run on lin...
Read more

java - Output of Eclipse is rubbish -

i want write program in eclipse language different english. when display output using system.out program, displays output correctly in console editor. when accept input console , output it, output rubbish. how can solve it. many thnaks i'm not sure if problem, have character encoding problem. make sure you're using unicode, utf-8. you can read more @ official docs converting non-unicode text .
Read more

jquery - Confused with JSON data and normal data in Django ajax request -

i read json internet still have not got grasp of it. reading article http://webcloud.se/log/ajax-in-django-with-jquery/ i not understood first part function using json def xhr_test(request, format): if request.is_ajax(): if format == 'xml': mimetype = 'application/xml' if format == 'json': mimetype = 'application/javascript' data = serializers.serialize(format, examplemodel.objects.all()) return httpresponse(data,mimetype) # if want prevent non xhr calls else: return httpresponse(status=400) my main problems are from function getting format variable does format json mean data given function json or data recived json can give me simple example ouput of function data = serializers.serialize(format, examplemodel.objects.all()) how use data when response in jquery function if don't use json in above function how input , response chnage thanks ...
Read more