Textures in RAM (not VRAM)

[AnonymousTipster]

To load textures, I've currently been using this code from the spharm sample:

Code: Select all

#define VRAM_OFFSET ((512*280*4)*3)
// 0x198000
static unsigned int vramaddr = 0;
unsigned char *convertimage(unsigned char *inptr,int size)
{
  // convert our raw image
  // saved as raw. no header .. interleaved and order RGB
  int x;
  unsigned char *input = inptr;
  unsigned char *output,*outptr;
  int tsize = size*size;
  if (vramaddr == 0)
    vramaddr = (0x40000000 | 0x04000000) + VRAM_OFFSET;
  outptr = output = (unsigned char *)vramaddr;
  for &#40;x=0;x<tsize;x++&#41; &#123;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = 0xff;
  &#125;
  vramaddr += tsize * 4;
  if &#40;&#40;vramaddr & 0xff&#41; != 0&#41;
    vramaddr = &#40;vramaddr & 0xffffff00&#41; + 0x100;
  return output;
&#125;

But this funcition loads into VRAM, and I've run out of space, so I'd like to load them to RAM instead. I think that I need to use malloc(), but it returns a void*, and I need a static unsigned char to return. I tried changing the code to:

Code: Select all

unsigned int ramaddr = 0;
unsigned char *convertimageRAM(unsigned char *inptr,int size)
{
  // convert our raw image
  // saved as raw. no header .. interleaved and order RGB
  //this function loads into RAM, not VRAM
  int x;
  unsigned char *input = inptr;
  unsigned char *output,*outptr;
  size_t sizediff;
  sizediff = size;
  int tsize = size*size;
  if (ramaddr == 0)
    ramaddr = (unsigned int)malloc(sizediff);
  outptr = output = (unsigned char *)ramaddr;
  for &#40;x=0;x<tsize;x++&#41; &#123;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = 0xff;
  &#125;
  ramaddr += tsize * 4;
  if &#40;&#40;ramaddr & 0xff&#41; != 0&#41;
    ramaddr = &#40;ramaddr & 0xffffff00&#41; + 0x100;
  return output;
&#125;

But this only displays half the image, the rest speckles of colour. What am I doing wrong, and how can I use the malloc() function like this?
Thanks.


EDIT:
Corrected Code:

Code: Select all

unsigned int ramaddr = 0;
unsigned char *convertimageRAM(unsigned char *inptr,int size)
{
  // convert our raw image
  // saved as raw. no header .. interleaved and order RGB
  //this function loads into RAM, not VRAM
  int x;
  unsigned char *input = inptr;
  unsigned char *output,*outptr;
  size_t sizediff;
  sizediff = size*size*4;
  int tsize = size*size;
	ramaddr = (unsigned int)memalign(16,sizediff);
  outptr = output = (unsigned char *)ramaddr;
  for &#40;x=0;x<tsize;x++&#41; &#123;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = 0xff;
  &#125;
  return output;
&#125;

[urchin]

Just a wild guess, but have you tried memalign instead of malloc?

[AnonymousTipster]

I tried changing the code to the following:

Code: Select all

unsigned int ramaddr = 0;
unsigned char *convertimageRAM(unsigned char *inptr,int size)
{
  // convert our raw image
  // saved as raw. no header .. interleaved and order RGB
  //this function loads into RAM, not VRAM
  int x;
  unsigned char *input = inptr;
  unsigned char *output,*outptr;
  size_t sizediff;
  sizediff = size;
  int tsize = size*size;
  if (ramaddr == 0)
    //ramaddr = (unsigned int)malloc(sizediff);//(0x40000000 | 0x04000000) + VRAM_OFFSET;
	ramaddr = (unsigned int)memalign(RoundUpPow2(sizediff),sizediff);
  outptr = output = (unsigned char *)ramaddr;
  for &#40;x=0;x<tsize;x++&#41; &#123;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = *&#40;input++&#41;;
    *&#40;outptr++&#41; = 0xff;
  &#125;
  ramaddr += tsize * 4;
  if &#40;&#40;ramaddr & 0xff&#41; != 0&#41;
    ramaddr = &#40;ramaddr & 0xffffff00&#41; + 0x100;
  return output;
&#125;

(IE using memalign) and the results are the same. Also, my textures are already power's of 2, so memalign shouldn't change anything. I think the problem is changing void* into unsigned int, but i'm not very sure.

[Paco]

Code: Select all

  // Choppped up
  sizediff = size;
  int tsize = size*size;
  ramaddr = (unsigned int)malloc(sizediff);
  for &#40;x=0;x<tsize;x++&#41; &#123; 

You are allocating "size" bytes and then filling "size*size*4" bytes of memory. That's very wrong.

[Shine]

Also, my textures are already power's of 2, so memalign shouldn't change anything.

memalign has nothing to do with the size of your texture, but assures that the pointer is memory aligned, which is returned.

[ReKleSS]

To follow up on what Shine said, your use of memalign is weird. You're aligning the textures to the size of the texture... for the GU textures need to be 16 byte aligned (i.e. memalign(16, blah)).
-ReK

[AnonymousTipster]

Ok, i've got it, the problem was what Paco mentioned, that I was allocating size, but using size*size*4, so once I change it to alloc size*size*4 bytes, everything works fine.
Thanks.

[patpsp]

It will be nice if you edit the first post to put the corrected code.
So if someone else has the same problem, he could have a look at the bad code and its correction.

Thanks

[AnonymousTipster]

It will be nice if you edit the first post to put the corrected code.
So if someone else has the same problem, he could have a look at the bad code and its correction.

Thanks

Done. ^_^