#include "../platform.h"
#include "sh4.h"

#define TA_SQ_ADDR (unsigned int *)(void *) \
    (0xe0000000 | (((unsigned long)0x10000000) & 0x03ffffe0))

#define QACRTA ((((unsigned int)0x10000000)>>26)<<2)&0x1c

#define PVR_VERTEX_BUF_SIZE 2560 * 256

void InitGPU(_Bool autosort, _Bool fsaa) {
    pvr_init_params_t params = {
        /* Enable opaque and translucent polygons with size 32 and 32 */
        {PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32},
        PVR_VERTEX_BUF_SIZE, /* Vertex buffer size */
        0, /* No DMA */
        fsaa, /* No FSAA */
        (autosort) ? 0 : 1 /* Disable translucent auto-sorting to match traditional GL */
    };

    pvr_init(&params);
}

void SceneBegin() {
    pvr_wait_ready();
    pvr_scene_begin();
    QACR0 = QACRTA;
    QACR1 = QACRTA;
}

void SceneListBegin(GPUList list) {
    pvr_list_begin(list);
}

void SceneListSubmit(void* src, int n) {
    uint32_t *d = (uint32_t*) TA_SQ_ADDR;
    uint32_t *s = src;

    /* fill/write queues as many times necessary */
    while(n--) {
        __asm__("pref @%0" : : "r"(s + 8));  /* prefetch 32 bytes for next loop */
        d[0] = *(s++);
        d[1] = *(s++);
        d[2] = *(s++);
        d[3] = *(s++);
        d[4] = *(s++);
        d[5] = *(s++);
        d[6] = *(s++);
        d[7] = *(s++);
        __asm__("pref @%0" : : "r"(d));
        d += 8;
    }

    /* Wait for both store queues to complete */
    d = (uint32_t *)0xe0000000;
    d[0] = d[8] = 0;
}

void SceneListFinish() {
    pvr_list_finish();
}

void SceneFinish() {
    pvr_scene_finish();
}

const VideoMode* GetVideoMode() {
    static VideoMode mode;
    mode.width = vid_mode->width;
    mode.height = vid_mode->height;
    return &mode;
}