[SDL] Hardware per pixel alpha blitting

[Amir Cicak]

Hello, can someone show me code for loading PNG with alpha in HW surface and 
blitting it using hwaccel? I am using SDL_image library (IMG_Load) to load 
bmp, and although I open display in HW doublebuf mode, surfaces are never 
loaded in video mem. Also on my geforce 6600 I get these caps:

vi->hw_available: 1
vi->wm_available: 1
vi->blit_hw: 1
vi->blit_hw_CC: 1
vi->blit_hw_A: 0
vi->blit_sw: 1
vi->blit_sw_CC: 1
vi->blit_sw_A: 0
vi->blit_fill: 1
vi->video_mem: 124284
vi->vfmt->BitsPerPixel: 16
vi->vfmt->BytesPerPixel: 2
vi->vfmt->colorkey: 0
vi->vfmt->alpha: 255
vi->vfmt->Rmask, Gmask, Bmask, Amask: 63488, 2016, 31, 0
vi->vfmt->Rshift, Gshift, Bshift, Ashift: 11, 5, 0, 0
vi->vfmt->Rmask, Rloss, Gloss, Bloss, Aloss: 3, 2, 3, 8

I also tried SDL_ConvertSurface(img, img->format, SDL_HWSURFACE|SDL_HWACCEL);, 
with no success.

Why is blit_hw_A 0, and blit_sw_A 0, on this card? What is the simplest way to 
load png alpha (per pixel) image into video surface (and/or system mem) and 
blit it using hardware acceleration? Is it possible at all? If not whats the 
fastest way to blit alpha PNG?

Thank you in advance.