int main(int argc, char **argv)
{
int rv;
int width, height;
int width_s, height_s;
int padded_width, padded_height;
int backbuffer = 0;
int supersample_x = 2; // 1 or 2
int supersample_y = 2; // 1 or 2
int extra_ps_inputs = 0;
fb_info fb;
rv = fb_open(0, &fb);
if(rv!=0)
{
exit(1);
}
width = fb.fb_var.xres;
height = fb.fb_var.yres;
width_s = width * supersample_x;
height_s = height * supersample_y;
padded_width = etna_align_up(width_s, 64);
padded_height = etna_align_up(height_s, 64);
extra_ps_inputs = (supersample_x > 1 || supersample_y > 1); // in case of supersample, there is an extra input to PS
printf("padded_width %i padded_height %i\n", padded_width, padded_height);
rv = viv_open();
if(rv!=0)
{
fprintf(stderr, "Error opening device\n");
exit(1);
}
printf("Succesfully opened device\n");
etna_vidmem *rt = 0; /* main render target */
etna_vidmem *rt_ts = 0; /* tile status for main render target */
etna_vidmem *z = 0; /* depth for main render target */
etna_vidmem *z_ts = 0; /* depth ts for main render target */
etna_vidmem *vtx = 0; /* vertex buffer */
etna_vidmem *idx = 0; /* index buffer */
etna_vidmem *aux_rt = 0; /* auxilary render target */
etna_vidmem *aux_rt_ts = 0; /* tile status for auxilary render target */
etna_vidmem *tex = 0; /* texture */
etna_vidmem *bmp = 0; /* bitmap */
/* TODO: anti aliasing (doubles width/height) */
size_t rt_size = padded_width * padded_height * 4;
size_t rt_ts_size = etna_align_up((padded_width * padded_height * 4)/0x100, 0x100);
size_t z_size = padded_width * padded_height * 2;
size_t z_ts_size = etna_align_up((padded_width * padded_height * 2)/0x100, 0x100);
size_t bmp_size = width * height * 4;
if(etna_vidmem_alloc_linear(&rt, rt_size, gcvSURF_RENDER_TARGET, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&rt_ts, rt_ts_size, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&z, z_size, gcvSURF_DEPTH, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&z_ts, z_ts_size, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&vtx, VERTEX_BUFFER_SIZE, gcvSURF_VERTEX, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&idx, INDEX_BUFFER_SIZE, gcvSURF_INDEX, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&aux_rt, 0x4000, gcvSURF_RENDER_TARGET, gcvPOOL_SYSTEM, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&aux_rt_ts, 0x100, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&tex, 0x100000, gcvSURF_TEXTURE, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&bmp, bmp_size, gcvSURF_BITMAP, gcvPOOL_DEFAULT, true)!=ETNA_OK
)
{
fprintf(stderr, "Error allocating video memory\n");
exit(1);
}
/* Phew, now we got all the memory we need.
* Write interleaved attribute vertex stream.
* Unlike the GL example we only do this once, not every time glDrawArrays is called, the same would be accomplished
* from GL by using a vertex buffer object.
*/
memset(vtx->logical, 0, 0x5ef80);
#ifndef INDEXED
printf("Interleaving vertices...\n");
float *vertices_array = companion_vertices_array();
float *texture_coordinates_array =
companion_texture_coordinates_array();
float *normals_array = companion_normals_array();
assert(COMPANION_ARRAY_COUNT*(3+3+2)*sizeof(float) < VERTEX_BUFFER_SIZE);
for(int vert=0; vert<COMPANION_ARRAY_COUNT; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+0] = vertices_array[vert*3 + comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+3] = normals_array[vert*3 + comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
((float*)vtx->logical)[dest_idx+comp+6] = texture_coordinates_array[vert*2 + comp]; /* 2 */
}
#else
printf("Interleaving vertices and copying index buffer...\n");
assert(COMPANION_VERTEX_COUNT*(3+3+2)*sizeof(float) < VERTEX_BUFFER_SIZE);
for(int vert=0; vert<COMPANION_VERTEX_COUNT; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+0] = companion_vertices[vert][comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+3] = companion_normals[vert][comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
((float*)vtx->logical)[dest_idx+comp+6] = companion_texture_coordinates[vert][comp]; /* 2 */
}
assert(COMPANION_TRIANGLE_COUNT*3*sizeof(unsigned short) < INDEX_BUFFER_SIZE);
memcpy(idx->logical, &companion_triangles[0][0], COMPANION_TRIANGLE_COUNT*3*sizeof(unsigned short));
#endif
/* Fill in texture (convert from RGB linear to tiled) */
#define TILE_WIDTH (4)
#define TILE_HEIGHT (4)
#define TILE_WORDS (TILE_WIDTH*TILE_HEIGHT)
unsigned ytiles = COMPANION_TEXTURE_HEIGHT / TILE_HEIGHT;
unsigned xtiles = COMPANION_TEXTURE_WIDTH / TILE_WIDTH;
unsigned dst_stride = xtiles * TILE_WORDS;
for(unsigned ty=0; ty<ytiles; ++ty)
{
for(unsigned tx=0; tx<xtiles; ++tx)
{
unsigned ofs = ty * dst_stride + tx * TILE_WORDS;
for(unsigned y=0; y<TILE_HEIGHT; ++y)
{
for(unsigned x=0; x<TILE_WIDTH; ++x)
{
unsigned srcy = ty*TILE_HEIGHT + y;
unsigned srcx = tx*TILE_WIDTH + x;
unsigned src_ofs = (srcy*COMPANION_TEXTURE_WIDTH+srcx)*3;
unsigned r,g,b,a;
#ifndef TEST_PATTERN /* actual texture */
r = ((uint8_t*)companion_texture)[src_ofs+0];
g = ((uint8_t*)companion_texture)[src_ofs+1];
b = ((uint8_t*)companion_texture)[src_ofs+2];
#else /* test pattern */
r = srcx; g = srcy; b = 0;
#endif
a = 255;
((uint32_t*)tex->logical)[ofs] = ((a&0xFF) << 24) | ((b&0xFF) << 16) | ((g&0xFF) << 8) | (r&0xFF);
ofs += 1;
}
}
}
}
etna_ctx *ctx = 0;
if(etna_create(&ctx) != ETNA_OK)
{
printf("Unable to create context\n");
exit(1);
}
for(int frame=0; frame<1000; ++frame)
{
printf("*** FRAME %i ****\n", frame);
/* XXX part of this can be put outside the loop, but until we have usable context management
* this is safest.
*/
etna_set_state(ctx, VIVS_GL_VERTEX_ELEMENT_CONFIG, 0x1);
etna_set_state(ctx, VIVS_RA_CONTROL, 0x1);
etna_set_state(ctx, VIVS_PA_W_CLIP_LIMIT, 0x34000001);
etna_set_state(ctx, VIVS_PA_SYSTEM_MODE, 0x11);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_UNK22, 0));
etna_set_state(ctx, VIVS_SE_CONFIG, 0x0);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_PE_ALPHA_CONFIG,
ETNA_MASKED_BIT(VIVS_PE_ALPHA_CONFIG_BLEND_ENABLE_COLOR, 0) &
ETNA_MASKED_BIT(VIVS_PE_ALPHA_CONFIG_BLEND_ENABLE_ALPHA, 0) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_SRC_FUNC_COLOR, BLEND_FUNC_ONE) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_SRC_FUNC_ALPHA, BLEND_FUNC_ONE) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_DST_FUNC_COLOR, BLEND_FUNC_ZERO) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_DST_FUNC_ALPHA, BLEND_FUNC_ZERO) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_EQ_COLOR, BLEND_EQ_ADD) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_EQ_ALPHA, BLEND_EQ_ADD));
etna_set_state(ctx, VIVS_PE_ALPHA_BLEND_COLOR,
VIVS_PE_ALPHA_BLEND_COLOR_B(0) |
VIVS_PE_ALPHA_BLEND_COLOR_G(0) |
VIVS_PE_ALPHA_BLEND_COLOR_R(0) |
VIVS_PE_ALPHA_BLEND_COLOR_A(0));
etna_set_state(ctx, VIVS_PE_ALPHA_OP, ETNA_MASKED_BIT(VIVS_PE_ALPHA_OP_ALPHA_TEST, 0));
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_CULL_FACE_MODE, CCW));
etna_set_state(ctx, VIVS_PE_STENCIL_CONFIG, ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_REF_FRONT, 0) &
ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_MASK_FRONT, 0xff) &
ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_WRITE_MASK, 0xff) &
ETNA_MASKED_INL(VIVS_PE_STENCIL_CONFIG_MODE, DISABLED));
etna_set_state(ctx, VIVS_PE_STENCIL_OP, ETNA_MASKED(VIVS_PE_STENCIL_OP_FUNC_FRONT, COMPARE_FUNC_ALWAYS) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FUNC_BACK, COMPARE_FUNC_ALWAYS) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FAIL_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FAIL_BACK, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_DEPTH_FAIL_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_DEPTH_FAIL_BACK, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_PASS_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_PASS_BACK, STENCIL_OP_KEEP));
etna_set_state(ctx, VIVS_SE_DEPTH_SCALE, 0x0);
etna_set_state(ctx, VIVS_SE_DEPTH_BIAS, 0x0);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_FILL_MODE, SOLID) &
ETNA_MASKED_INL(VIVS_PA_CONFIG_SHADE_MODEL, SMOOTH));
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT,
ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_PARTIAL, 0) &
ETNA_MASKED(VIVS_PE_COLOR_FORMAT_COMPONENTS, 0xf) &
ETNA_MASKED(VIVS_PE_COLOR_FORMAT_FORMAT, RS_FORMAT_X8R8G8B8) &
ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_SUPER_TILED, 1));
etna_set_state(ctx, VIVS_PE_COLOR_ADDR, rt->address);
etna_set_state(ctx, VIVS_PE_COLOR_STRIDE, padded_width * 4);
uint32_t ts_msaa_config;
if(supersample_x == 2 && supersample_y == 2)
{
// 4X MSAA
etna_set_state(ctx, VIVS_GL_MULTI_SAMPLE_CONFIG,
ETNA_MASKED_INL(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES, 4X) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_ENABLES, 0xf) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK12, 0x0) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK16, 0x0)
);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E04, 0x0);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E10(2), 0xaaa22a22);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(8), 0x262a2288);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(9), 0x886688a2);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(10), 0x888866aa);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(11), 0x668888a6);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E10(1), 0xe6ae622a);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(4), 0x46622a88);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(5), 0x888888ae);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(6), 0x888888e6);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(7), 0x888888ca);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E10(0), 0xeaa26e26);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(0), 0x4a6e2688);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(1), 0x888888a2);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(2), 0x888888ea);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(3), 0x888888c6);
ts_msaa_config = VIVS_TS_MEM_CONFIG_MSAA | VIVS_TS_MEM_CONFIG_MSAA_FORMAT_A8R8G8B8;
} else if(supersample_x == 2 && supersample_y == 1)
{
// 2X MSAA
etna_set_state(ctx, VIVS_GL_MULTI_SAMPLE_CONFIG,
ETNA_MASKED_INL(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES, 2X) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_ENABLES, 0xf) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK12, 0x0) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK16, 0x0)
);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E04, 0x0);
etna_set_state(ctx, VIVS_RA_MULTISAMPLE_UNK00E10(0), 0xaa22);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(0), 0x66aa2288);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(1), 0x88558800);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(2), 0x88881100);
etna_set_state(ctx, VIVS_RA_CENTROID_TABLE(3), 0x33888800);
ts_msaa_config = VIVS_TS_MEM_CONFIG_MSAA | VIVS_TS_MEM_CONFIG_MSAA_FORMAT_A8R8G8B8;
} else { // No multisampling
etna_set_state(ctx, VIVS_GL_MULTI_SAMPLE_CONFIG,
ETNA_MASKED_INL(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES, NONE) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_ENABLES, 0xf) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK12, 0x0) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK16, 0x0)
);
ts_msaa_config = 0;
}
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG,
ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE, 0) &
ETNA_MASKED_INL(VIVS_PE_DEPTH_CONFIG_DEPTH_FORMAT, D16) &
ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_SUPER_TILED, 1) &
ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_EARLY_Z, 1));
etna_set_state(ctx, VIVS_PE_DEPTH_ADDR, z->address);
etna_set_state(ctx, VIVS_PE_DEPTH_STRIDE, padded_width * 2);
etna_set_state(ctx, VIVS_PE_HDEPTH_CONTROL, VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_TS_DEPTH_CLEAR_VALUE, 0xffffffff);
etna_set_state(ctx, VIVS_TS_DEPTH_STATUS_BASE, z_ts->address);
etna_set_state(ctx, VIVS_TS_DEPTH_SURFACE_BASE, z->address);
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION |
ts_msaa_config);
#ifdef EXTRA_DELAYS
/* Warm up RS on aux render target (is this needed?) */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts_physical);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt_physical);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt_physical, aux_rt_ts_physical);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts_physical);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt_physical);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt_physical, aux_rt_ts_physical);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts_physical);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt_physical);
#endif
/* sync rasterizer to pixel engine after changes to PE config */
etna_stall(ctx, SYNC_RECIPIENT_RA, SYNC_RECIPIENT_PE);
/* Set up the resolve to clear tile status for main render target and depth
* Regard the TS as an image of width 16 with 4 bytes per pixel (64 bytes per row)
* */
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8)
);
etna_set_state_multi(ctx, VIVS_RS_DITHER(0), 2, (uint32_t[]){0xffffffff, 0xffffffff});
etna_set_state(ctx, VIVS_RS_FILL_VALUE(0), 0x55555555);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL,
VIVS_RS_CLEAR_CONTROL_MODE_ENABLED1 |
(0xffff << VIVS_RS_CLEAR_CONTROL_BITS__SHIFT));
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0);
/* clear color ts */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt_ts->address);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, 0x40);
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
((rt_ts_size/0x40) << VIVS_RS_WINDOW_SIZE_HEIGHT__SHIFT) |
(16 << VIVS_RS_WINDOW_SIZE_WIDTH__SHIFT));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
/* clear depth ts */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, z_ts->address);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, 0x40);
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
((z_ts_size/0x40) << VIVS_RS_WINDOW_SIZE_HEIGHT__SHIFT) |
(16 << VIVS_RS_WINDOW_SIZE_WIDTH__SHIFT));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
/** Done */
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0xff7f7f7f);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
/* depth setup */
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE, 1) &
ETNA_MASKED(VIVS_PE_DEPTH_CONFIG_DEPTH_FUNC, COMPARE_FUNC_LESS) &
ETNA_MASKED_INL(VIVS_PE_DEPTH_CONFIG_DEPTH_MODE, Z) &
ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH, 0));
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NEAR, 0.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_FAR, 1.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
/* set up primitive assembly and setup engine */
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_OFFSET_Z, 0.0);
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_SCALE_Z, 1.0);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_Y, height << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_Y, height << 15);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_LEFT, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_TOP, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_RIGHT, (width << 16) | 5);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_BOTTOM, (height << 16) | 5);
/* set up texture unit */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_TEXTURE);
etna_set_state(ctx, VIVS_TE_SAMPLER_SIZE(0),
VIVS_TE_SAMPLER_SIZE_WIDTH(512)|VIVS_TE_SAMPLER_SIZE_HEIGHT(512));
etna_set_state(ctx, VIVS_TE_SAMPLER_LOG_SIZE(0),
VIVS_TE_SAMPLER_LOG_SIZE_WIDTH(9<<5) |
VIVS_TE_SAMPLER_LOG_SIZE_HEIGHT(9<<5));
etna_set_state(ctx, VIVS_TE_SAMPLER_LOD_ADDR(0,0), tex->address);
etna_set_state(ctx, VIVS_TE_SAMPLER_CONFIG0(0),
VIVS_TE_SAMPLER_CONFIG0_TYPE(TEXTURE_TYPE_2D)|
VIVS_TE_SAMPLER_CONFIG0_UWRAP(TEXTURE_WRAPMODE_CLAMP_TO_EDGE)|
VIVS_TE_SAMPLER_CONFIG0_VWRAP(TEXTURE_WRAPMODE_CLAMP_TO_EDGE)|
VIVS_TE_SAMPLER_CONFIG0_MIN(TEXTURE_FILTER_LINEAR)|
VIVS_TE_SAMPLER_CONFIG0_MIP(TEXTURE_FILTER_NONE)|
VIVS_TE_SAMPLER_CONFIG0_MAG(TEXTURE_FILTER_LINEAR)|
VIVS_TE_SAMPLER_CONFIG0_FORMAT(TEXTURE_FORMAT_X8R8G8B8));
etna_set_state(ctx, VIVS_TE_SAMPLER_LOD_CONFIG(0), 0x00000000); /* TE.SAMPLER[0].LOD_CONFIG := 0x0 */
/* shader setup */
etna_set_state(ctx, VIVS_VS_END_PC, vs_size/16);
etna_set_state_multi(ctx, VIVS_VS_INPUT_COUNT, 3, (uint32_t[]){
/* VIVS_VS_INPUT_COUNT */ VIVS_VS_INPUT_COUNT_UNK8(1) | VIVS_VS_INPUT_COUNT_COUNT(3),
/* VIVS_VS_TEMP_REGISTER_CONTROL */ VIVS_VS_TEMP_REGISTER_CONTROL_NUM_TEMPS(6),
/* VIVS_VS_OUTPUT(0) */ 0x10004});
etna_set_state(ctx, VIVS_VS_START_PC, 0x0);
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(45), 0.5); /* u11.y */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(44), 1.0); /* u11.x */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(27), 0.0); /* u6.w */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(23), 20.0); /* u5.w */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(19), 2.0); /* u4.w */
/* Now load the shader itself */
etna_set_state_multi(ctx, VIVS_VS_INST_MEM(0), vs_size/4, vs);
etna_set_state(ctx, VIVS_RA_CONTROL, 0x3);
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(0), 1.0); /* u0.x */
etna_set_state_multi(ctx, VIVS_PS_END_PC, 2, (uint32_t[]){
/* VIVS_PS_END_PC */ ps_size/16,
/* VIVS_PS_OUTPUT_REG */ 0x1});
etna_set_state(ctx, VIVS_PS_START_PC, 0x0);
etna_set_state(ctx, VIVS_PA_ATTRIBUTE_ELEMENT_COUNT, 0x200);
etna_set_state(ctx, VIVS_PA_SHADER_ATTRIBUTES(0), 0x200);
etna_set_state(ctx, VIVS_PA_SHADER_ATTRIBUTES(1), 0x200);
etna_set_state(ctx, VIVS_GL_VARYING_NUM_COMPONENTS,
VIVS_GL_VARYING_NUM_COMPONENTS_VAR0(4)| /* position */
VIVS_GL_VARYING_NUM_COMPONENTS_VAR1(2) /* texture coordinate */
);
etna_set_state_multi(ctx, VIVS_GL_VARYING_COMPONENT_USE(0), 2, (uint32_t[]){
VIVS_GL_VARYING_COMPONENT_USE_COMP0(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP1(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP2(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP3(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP4(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP5(VARYING_COMPONENT_USE_USED)
, 0
});
etna_set_state_multi(ctx, VIVS_PS_INST_MEM(0), ps_size/4, ps);
etna_set_state(ctx, VIVS_PS_INPUT_COUNT,
VIVS_PS_INPUT_COUNT_UNK8(31)| VIVS_PS_INPUT_COUNT_COUNT(3 + extra_ps_inputs));
etna_set_state(ctx, VIVS_PS_TEMP_REGISTER_CONTROL,
VIVS_PS_TEMP_REGISTER_CONTROL_NUM_TEMPS(3 + extra_ps_inputs));
etna_set_state(ctx, VIVS_PS_CONTROL, VIVS_PS_CONTROL_UNK1);
etna_set_state(ctx, VIVS_GL_VARYING_TOTAL_COMPONENTS,
VIVS_GL_VARYING_TOTAL_COMPONENTS_NUM(6)); /* 4+2=6 total varying components */
etna_set_state(ctx, VIVS_VS_LOAD_BALANCING, 0xf3f0542); /* depends on number of inputs/outputs/varyings? XXX how exactly */
etna_set_state(ctx, VIVS_VS_OUTPUT_COUNT, 3);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_POINT_SIZE_ENABLE, 0));
/* Compute transform matrices in the same way as cube egl demo */
ESMatrix modelview;
esMatrixLoadIdentity(&modelview);
esTranslate(&modelview, 0.0f, 0.0f, -9.0f);
esRotate(&modelview, 45.0f, 1.0f, 0.0f, 0.0f);
esRotate(&modelview, 45.0f, 0.0f, 1.0f, 0.0f);
esRotate(&modelview, frame*0.5f, 0.0f, 0.0f, 1.0f);
esScale(&modelview, 0.475f, 0.475f, 0.475f);
GLfloat aspect = (GLfloat)(height) / (GLfloat)(width);
ESMatrix projection;
esMatrixLoadIdentity(&projection);
esFrustum(&projection, -2.8f, +2.8f, -2.8f * aspect, +2.8f * aspect, 6.0f, 10.0f);
ESMatrix modelviewprojection;
esMatrixLoadIdentity(&modelviewprojection);
esMatrixMultiply(&modelviewprojection, &modelview, &projection);
ESMatrix inverse, normal; /* compute inverse transpose normal transformation matrix */
esMatrixInverse3x3(&inverse, &modelview);
esMatrixTranspose(&normal, &inverse);
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(0), 16, (uint32_t*)&modelviewprojection.m[0][0]);
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(16), 3, (uint32_t*)&normal.m[0][0]); /* u4.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(20), 3, (uint32_t*)&normal.m[1][0]); /* u5.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(24), 3, (uint32_t*)&normal.m[2][0]); /* u6.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(28), 16, (uint32_t*)&modelview.m[0][0]);
#ifdef INDEXED
etna_set_state(ctx, VIVS_FE_INDEX_STREAM_BASE_ADDR, idx->address);
etna_set_state(ctx, VIVS_FE_INDEX_STREAM_CONTROL, VIVS_FE_INDEX_STREAM_CONTROL_TYPE_UNSIGNED_SHORT);
#endif
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_BASE_ADDR, vtx->address);
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_CONTROL,
VIVS_FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE((3 + 3 + 2)*4));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(0),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM_3 |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0x0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0xc));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(1),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM_3 |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0xc) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0x18));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(2),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM_2 |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0x18) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0x20));
etna_set_state(ctx, VIVS_VS_INPUT(0), 0x20100);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_POINT_SPRITE_ENABLE, 0));
#ifdef INDEXED
etna_draw_indexed_primitives(ctx, PRIMITIVE_TYPE_TRIANGLES, 0, COMPANION_TRIANGLE_COUNT, 0);
#else
etna_draw_primitives(ctx, PRIMITIVE_TYPE_TRIANGLES, 0, COMPANION_TRIANGLE_COUNT);
#endif
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
#ifdef EXTRA_DELAYS
etna_flush(ctx);
#endif
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_DEST_TILED);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, (padded_width * 4 * 4) | VIVS_RS_DEST_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0);
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address);
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt->address);
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(padded_height) |
VIVS_RS_WINDOW_SIZE_WIDTH(padded_width));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
#ifdef EXTRA_DELAYS
etna_flush(ctx);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
#endif
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
/* Wait for pixel engine to finish */
#ifdef EXTRA_DELAYS
etna_finish(ctx);
#else
etna_stall(ctx, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE);
#endif
/* Copy result to framebuffer */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
((supersample_x>1)?VIVS_RS_CONFIG_DOWNSAMPLE_X:0) |
((supersample_y>1)?VIVS_RS_CONFIG_DOWNSAMPLE_Y:0) |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_SWAP_RB);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, fb.fb_fix.line_length);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0);
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address);
etna_set_state(ctx, VIVS_RS_DEST_ADDR, fb.physical[backbuffer]);
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(height * supersample_y) |
VIVS_RS_WINDOW_SIZE_WIDTH(width * supersample_x));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
etna_finish(ctx);
/* switch buffers */
fb_set_buffer(&fb, backbuffer);
backbuffer = 1-backbuffer;
}
etna_free(ctx);
viv_close();
return 0;
}
static boolean
etna_get_specs(struct etna_screen *screen)
{
uint64_t val;
uint32_t instruction_count;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_INSTRUCTION_COUNT, &val)) {
DBG("could not get ETNA_GPU_INSTRUCTION_COUNT");
goto fail;
}
instruction_count = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_VERTEX_OUTPUT_BUFFER_SIZE,
&val)) {
DBG("could not get ETNA_GPU_VERTEX_OUTPUT_BUFFER_SIZE");
goto fail;
}
screen->specs.vertex_output_buffer_size = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_VERTEX_CACHE_SIZE, &val)) {
DBG("could not get ETNA_GPU_VERTEX_CACHE_SIZE");
goto fail;
}
screen->specs.vertex_cache_size = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_SHADER_CORE_COUNT, &val)) {
DBG("could not get ETNA_GPU_SHADER_CORE_COUNT");
goto fail;
}
screen->specs.shader_core_count = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_STREAM_COUNT, &val)) {
DBG("could not get ETNA_GPU_STREAM_COUNT");
goto fail;
}
screen->specs.stream_count = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_REGISTER_MAX, &val)) {
DBG("could not get ETNA_GPU_REGISTER_MAX");
goto fail;
}
screen->specs.max_registers = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_PIXEL_PIPES, &val)) {
DBG("could not get ETNA_GPU_PIXEL_PIPES");
goto fail;
}
screen->specs.pixel_pipes = val;
if (etna_gpu_get_param(screen->gpu, ETNA_GPU_NUM_CONSTANTS, &val)) {
DBG("could not get %s", "ETNA_GPU_NUM_CONSTANTS");
goto fail;
}
if (val == 0) {
fprintf(stderr, "Warning: zero num constants (update kernel?)\n");
val = 168;
}
screen->specs.num_constants = val;
/* Figure out gross GPU architecture. See rnndb/common.xml for a specific
* description of the differences. */
if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI5))
screen->specs.halti = 5; /* New GC7000/GC8x00 */
else if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI4))
screen->specs.halti = 4; /* Old GC7000/GC7400 */
else if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI3))
screen->specs.halti = 3; /* None? */
else if (VIV_FEATURE(screen, chipMinorFeatures4, HALTI2))
screen->specs.halti = 2; /* GC2500/GC3000/GC5000/GC6400 */
else if (VIV_FEATURE(screen, chipMinorFeatures2, HALTI1))
screen->specs.halti = 1; /* GC900/GC4000/GC7000UL */
else if (VIV_FEATURE(screen, chipMinorFeatures1, HALTI0))
screen->specs.halti = 0; /* GC880/GC2000/GC7000TM */
else
screen->specs.halti = -1; /* GC7000nanolite / pre-GC2000 except GC880 */
if (screen->specs.halti >= 0)
DBG("etnaviv: GPU arch: HALTI%d", screen->specs.halti);
else
DBG("etnaviv: GPU arch: pre-HALTI");
screen->specs.can_supertile =
VIV_FEATURE(screen, chipMinorFeatures0, SUPER_TILED);
screen->specs.bits_per_tile =
VIV_FEATURE(screen, chipMinorFeatures0, 2BITPERTILE) ? 2 : 4;
screen->specs.ts_clear_value =
VIV_FEATURE(screen, chipMinorFeatures0, 2BITPERTILE) ? 0x55555555
: 0x11111111;
/* vertex and fragment samplers live in one address space */
screen->specs.vertex_sampler_offset = 8;
screen->specs.fragment_sampler_count = 8;
screen->specs.vertex_sampler_count = 4;
screen->specs.vs_need_z_div =
screen->model < 0x1000 && screen->model != 0x880;
screen->specs.has_sin_cos_sqrt =
VIV_FEATURE(screen, chipMinorFeatures0, HAS_SQRT_TRIG);
screen->specs.has_sign_floor_ceil =
VIV_FEATURE(screen, chipMinorFeatures0, HAS_SIGN_FLOOR_CEIL);
screen->specs.has_shader_range_registers =
screen->model >= 0x1000 || screen->model == 0x880;
screen->specs.npot_tex_any_wrap =
VIV_FEATURE(screen, chipMinorFeatures1, NON_POWER_OF_TWO);
screen->specs.has_new_transcendentals =
VIV_FEATURE(screen, chipMinorFeatures3, HAS_FAST_TRANSCENDENTALS);
screen->specs.has_halti2_instructions =
VIV_FEATURE(screen, chipMinorFeatures4, HALTI2);
if (screen->specs.halti >= 5) {
/* GC7000 - this core must load shaders from memory. */
screen->specs.vs_offset = 0;
screen->specs.ps_offset = 0;
screen->specs.max_instructions = 0; /* Do not program shaders manually */
screen->specs.has_icache = true;
} else if (VIV_FEATURE(screen, chipMinorFeatures3, INSTRUCTION_CACHE)) {
/* GC3000 - this core is capable of loading shaders from
* memory. It can also run shaders from registers, as a fallback, but
* "max_instructions" does not have the correct value. It has place for
* 2*256 instructions just like GC2000, but the offsets are slightly
* different.
*/
screen->specs.vs_offset = 0xC000;
/* State 08000-0C000 mirrors 0C000-0E000, and the Vivante driver uses
* this mirror for writing PS instructions, probably safest to do the
* same.
*/
screen->specs.ps_offset = 0x8000 + 0x1000;
screen->specs.max_instructions = 256; /* maximum number instructions for non-icache use */
screen->specs.has_icache = true;
} else {
if (instruction_count > 256) { /* unified instruction memory? */
screen->specs.vs_offset = 0xC000;
screen->specs.ps_offset = 0xD000; /* like vivante driver */
screen->specs.max_instructions = 256;
} else {
screen->specs.vs_offset = 0x4000;
screen->specs.ps_offset = 0x6000;
screen->specs.max_instructions = instruction_count / 2;
}
screen->specs.has_icache = false;
}
if (VIV_FEATURE(screen, chipMinorFeatures1, HALTI0)) {
screen->specs.max_varyings = 12;
screen->specs.vertex_max_elements = 16;
} else {
screen->specs.max_varyings = 8;
/* Etna_viv documentation seems confused over the correct value
* here so choose the lower to be safe: HALTI0 says 16 i.s.o.
* 10, but VERTEX_ELEMENT_CONFIG register says 16 i.s.o. 12. */
screen->specs.vertex_max_elements = 10;
}
/* Etna_viv documentation does not indicate where varyings above 8 are
* stored. Moreover, if we are passed more than 8 varyings, we will
* walk off the end of some arrays. Limit the maximum number of varyings. */
if (screen->specs.max_varyings > ETNA_NUM_VARYINGS)
screen->specs.max_varyings = ETNA_NUM_VARYINGS;
/* from QueryShaderCaps in kernel driver */
if (screen->model < chipModel_GC4000) {
screen->specs.max_vs_uniforms = 168;
screen->specs.max_ps_uniforms = 64;
} else {
screen->specs.max_vs_uniforms = 256;
screen->specs.max_ps_uniforms = 256;
}
if (screen->specs.halti >= 5) {
screen->specs.has_unified_uniforms = true;
screen->specs.vs_uniforms_offset = VIVS_SH_HALTI5_UNIFORMS_MIRROR(0);
screen->specs.ps_uniforms_offset = VIVS_SH_HALTI5_UNIFORMS(screen->specs.max_vs_uniforms*4);
} else if (screen->specs.halti >= 1) {
/* unified uniform memory on GC3000 - HALTI1 feature bit is just a guess
*/
screen->specs.has_unified_uniforms = true;
screen->specs.vs_uniforms_offset = VIVS_SH_UNIFORMS(0);
/* hardcode PS uniforms to start after end of VS uniforms -
* for more flexibility this offset could be variable based on the
* shader.
*/
screen->specs.ps_uniforms_offset = VIVS_SH_UNIFORMS(screen->specs.max_vs_uniforms*4);
} else {
screen->specs.has_unified_uniforms = false;
screen->specs.vs_uniforms_offset = VIVS_VS_UNIFORMS(0);
screen->specs.ps_uniforms_offset = VIVS_PS_UNIFORMS(0);
}
screen->specs.max_texture_size =
VIV_FEATURE(screen, chipMinorFeatures0, TEXTURE_8K) ? 8192 : 2048;
screen->specs.max_rendertarget_size =
VIV_FEATURE(screen, chipMinorFeatures0, RENDERTARGET_8K) ? 8192 : 2048;
screen->specs.single_buffer = VIV_FEATURE(screen, chipMinorFeatures4, SINGLE_BUFFER);
if (screen->specs.single_buffer)
DBG("etnaviv: Single buffer mode enabled with %d pixel pipes", screen->specs.pixel_pipes);
screen->specs.tex_astc = VIV_FEATURE(screen, chipMinorFeatures4, TEXTURE_ASTC);
screen->specs.use_blt = VIV_FEATURE(screen, chipMinorFeatures5, BLT_ENGINE);
return true;
fail:
return false;
}
int main(int argc, char **argv)
{
int rv;
int width = 256;
int height = 256;
int padded_width = etna_align_up(width, 64);
int padded_height = etna_align_up(height, 64);
printf("padded_width %i padded_height %i\n", padded_width, padded_height);
struct viv_conn *conn = 0;
rv = viv_open(VIV_HW_3D, &conn);
if(rv!=0)
{
fprintf(stderr, "Error opening device\n");
exit(1);
}
printf("Succesfully opened device\n");
struct etna_vidmem *rt = 0; /* main render target */
struct etna_vidmem *rt_ts = 0; /* tile status for main render target */
struct etna_vidmem *z = 0; /* depth for main render target */
struct etna_vidmem *z_ts = 0; /* depth ts for main render target */
struct etna_vidmem *vtx = 0; /* vertex buffer */
struct etna_vidmem *aux_rt = 0; /* auxilary render target */
struct etna_vidmem *aux_rt_ts = 0; /* tile status for auxilary render target */
struct etna_vidmem *bmp = 0; /* bitmap */
size_t rt_size = padded_width * padded_height * 4;
size_t rt_ts_size = etna_align_up((padded_width * padded_height * 4)/0x100, 0x100);
size_t z_size = padded_width * padded_height * 2;
size_t z_ts_size = etna_align_up((padded_width * padded_height * 2)/0x100, 0x100);
size_t bmp_size = width * height * 4;
if(etna_vidmem_alloc_linear(conn, &rt, rt_size, VIV_SURF_RENDER_TARGET, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &rt_ts, rt_ts_size, VIV_SURF_TILE_STATUS, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &z, z_size, VIV_SURF_DEPTH, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &z_ts, z_ts_size, VIV_SURF_TILE_STATUS, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &vtx, VERTEX_BUFFER_SIZE, VIV_SURF_VERTEX, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &aux_rt, 0x4000, VIV_SURF_RENDER_TARGET, VIV_POOL_SYSTEM, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &aux_rt_ts, 0x100, VIV_SURF_TILE_STATUS, VIV_POOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(conn, &bmp, bmp_size, VIV_SURF_BITMAP, VIV_POOL_DEFAULT, true)!=ETNA_OK
)
{
fprintf(stderr, "Error allocating video memory\n");
exit(1);
}
/* Phew, now we got all the memory we need.
* Write interleaved attribute vertex stream.
* Unlike the GL example we only do this once, not every time glDrawArrays is called, the same would be accomplished
* from GL by using a vertex buffer object.
*/
for(int vert=0; vert<NUM_VERTICES; ++vert)
{
int dest_idx = vert * (3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+0] = vVertices[vert*3 + comp]; /* 0 */
for(int comp=0; comp<2; ++comp)
((float*)vtx->logical)[dest_idx+comp+3] = vTexCoords[vert*2 + comp]; /* 1 */
}
struct etna_ctx *ctx = 0;
if(etna_create(conn, &ctx) != ETNA_OK)
{
printf("Unable to create context\n");
exit(1);
}
/* Now load the shader itself */
uint32_t vs[] = {
0x02001001, 0x2a800800, 0x00000000, 0x003fc008,
0x02001003, 0x2a800800, 0x00000040, 0x00000002,
};
uint32_t vs_size = sizeof(vs);
uint32_t *ps;
uint32_t ps_size;
if(argc < 2)
{
perror("provide shader on command line");
exit(1);
}
int fd = open(argv[1], O_RDONLY);
if(fd == -1)
{
perror("opening shader");
exit(1);
}
ps_size = lseek(fd, 0, SEEK_END);
ps = malloc(ps_size);
lseek(fd, 0, SEEK_SET);
if(ps_size == 0 || ps_size>8192 || read(fd, ps, ps_size) != ps_size)
{
perror("empty or unreadable shader");
exit(1);
}
close(fd);
/* XXX how important is the ordering? I suppose we could group states (except the flushes, kickers, semaphores etc)
* and simply submit them at once. Especially for consecutive states and masked stated this could be a big win
* in DMA command buffer size. */
/* Build first command buffer */
etna_set_state(ctx, VIVS_GL_VERTEX_ELEMENT_CONFIG, 0x1);
etna_set_state(ctx, VIVS_RA_CONTROL, 0x1);
etna_set_state(ctx, VIVS_PA_W_CLIP_LIMIT, 0x34000001);
etna_set_state(ctx, VIVS_PA_SYSTEM_MODE, 0x11);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_UNK22, 0));
etna_set_state(ctx, VIVS_SE_CONFIG, 0x0);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
/* Set up pixel engine */
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT,
ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_OVERWRITE, 0));
etna_set_state(ctx, VIVS_PE_ALPHA_CONFIG,
ETNA_MASKED_BIT(VIVS_PE_ALPHA_CONFIG_BLEND_ENABLE_COLOR, 0) &
ETNA_MASKED_BIT(VIVS_PE_ALPHA_CONFIG_BLEND_SEPARATE_ALPHA, 0) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_SRC_FUNC_COLOR, BLEND_FUNC_ONE) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_SRC_FUNC_ALPHA, BLEND_FUNC_ONE) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_DST_FUNC_COLOR, BLEND_FUNC_ZERO) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_DST_FUNC_ALPHA, BLEND_FUNC_ZERO) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_EQ_COLOR, BLEND_EQ_ADD) &
ETNA_MASKED(VIVS_PE_ALPHA_CONFIG_EQ_ALPHA, BLEND_EQ_ADD));
etna_set_state(ctx, VIVS_PE_ALPHA_BLEND_COLOR,
VIVS_PE_ALPHA_BLEND_COLOR_B(0) |
VIVS_PE_ALPHA_BLEND_COLOR_G(0) |
VIVS_PE_ALPHA_BLEND_COLOR_R(0) |
VIVS_PE_ALPHA_BLEND_COLOR_A(0));
etna_set_state(ctx, VIVS_PE_ALPHA_OP, ETNA_MASKED_BIT(VIVS_PE_ALPHA_OP_ALPHA_TEST, 0));
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_CULL_FACE_MODE, OFF));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE, 0));
etna_set_state(ctx, VIVS_PE_STENCIL_CONFIG, ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_REF_FRONT, 0) &
ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_MASK_FRONT, 0xff) &
ETNA_MASKED(VIVS_PE_STENCIL_CONFIG_WRITE_MASK, 0xff));
etna_set_state(ctx, VIVS_PE_STENCIL_OP, ETNA_MASKED(VIVS_PE_STENCIL_OP_FUNC_FRONT, COMPARE_FUNC_ALWAYS) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FUNC_BACK, COMPARE_FUNC_ALWAYS) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FAIL_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_FAIL_BACK, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_DEPTH_FAIL_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_DEPTH_FAIL_BACK, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_PASS_FRONT, STENCIL_OP_KEEP) &
ETNA_MASKED(VIVS_PE_STENCIL_OP_PASS_BACK, STENCIL_OP_KEEP));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_EARLY_Z, 0));
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT, ETNA_MASKED(VIVS_PE_COLOR_FORMAT_COMPONENTS, 0xf));
etna_set_state(ctx, VIVS_SE_DEPTH_SCALE, 0x0);
etna_set_state(ctx, VIVS_SE_DEPTH_BIAS, 0x0);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_FILL_MODE, SOLID));
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_SHADE_MODEL, SMOOTH));
/* Set up render target */
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT,
ETNA_MASKED(VIVS_PE_COLOR_FORMAT_FORMAT, RS_FORMAT_A8R8G8B8) &
ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_SUPER_TILED, 1));
for(int frame=0; frame<1; ++frame)
{
printf("*** FRAME %i ****\n", frame);
etna_set_state(ctx, VIVS_PE_COLOR_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_PE_COLOR_STRIDE, padded_width * 4);
etna_set_state(ctx, VIVS_GL_MULTI_SAMPLE_CONFIG,
ETNA_MASKED_INL(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES, NONE) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_ENABLES, 0xf) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK12, 0x0) &
ETNA_MASKED(VIVS_GL_MULTI_SAMPLE_CONFIG_UNK16, 0x0)
);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT, ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_OVERWRITE, 1));
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_TS_MEM_CONFIG, VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR); /* ADDR_A */
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG,
ETNA_MASKED_INL(VIVS_PE_DEPTH_CONFIG_DEPTH_FORMAT, D16) &
ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_SUPER_TILED, 1)
);
etna_set_state(ctx, VIVS_PE_DEPTH_ADDR, z->address); /* ADDR_C */
etna_set_state(ctx, VIVS_PE_DEPTH_STRIDE, padded_width * 2);
etna_set_state(ctx, VIVS_PE_STENCIL_CONFIG, ETNA_MASKED_INL(VIVS_PE_STENCIL_CONFIG_MODE, DISABLED));
etna_set_state(ctx, VIVS_PE_HDEPTH_CONTROL, VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_EARLY_Z, 0));
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_TS_DEPTH_CLEAR_VALUE, 0xffffffff);
etna_set_state(ctx, VIVS_TS_DEPTH_STATUS_BASE, z_ts->address); /* ADDR_D */
etna_set_state(ctx, VIVS_TS_DEPTH_SURFACE_BASE, z->address); /* ADDR_C */
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_EARLY_Z, 1)); /* flip-flopping once again */
/* Warm up RS on aux render target */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
/* Phew, now that's one hell of a setup; the serious rendering starts now */
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
/* ... or so we thought */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
/* maybe now? */
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
/* nope, not really... */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_stall(ctx, SYNC_RECIPIENT_RA, SYNC_RECIPIENT_PE);
/* Set up the resolve to clear tile status for main render target
* What the blob does is regard the TS as an image of width N, height 4, with 4 bytes per pixel
* Looks like the height always stays the same. I don't think it matters as long as the entire memory are is covered.
* XXX need to clear the depth ts too.
* */
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8)
);
etna_set_state_multi(ctx, VIVS_RS_DITHER(0), 2, (uint32_t[]){0xffffffff, 0xffffffff});
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, 0x100); /* 0x100 iso 0x40! seems it uses a width of 256 if width divisible by 256, XXX need to figure out these rules */
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(rt_ts_size/0x100) |
VIVS_RS_WINDOW_SIZE_WIDTH(64));
etna_set_state(ctx, VIVS_RS_FILL_VALUE(0), 0x55555555);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL,
VIVS_RS_CLEAR_CONTROL_MODE_ENABLED1 |
VIVS_RS_CLEAR_CONTROL_BITS(0xffff));
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG,
0); /* no AA, no endian switch */
etna_set_state(ctx, VIVS_RS_KICKER,
0xbeebbeeb);
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0xff7f7f7f);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0xff7f7f7f);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
//etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_INL(VIVS_PA_CONFIG_CULL_FACE_MODE, CCW));
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE, 0));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_INL(VIVS_PE_DEPTH_CONFIG_DEPTH_MODE, NONE));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE, 0));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED(VIVS_PE_DEPTH_CONFIG_DEPTH_FUNC, COMPARE_FUNC_ALWAYS));
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_INL(VIVS_PE_DEPTH_CONFIG_DEPTH_MODE, Z));
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NEAR, 0.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_FAR, 1.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
/* set up primitive assembly */
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_OFFSET_Z, 0.0);
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_SCALE_Z, 1.0);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG, ETNA_MASKED_BIT(VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH, 0));
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_Y, height << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_Y, height << 15);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_LEFT, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_TOP, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_RIGHT, (width << 16) | 5);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_BOTTOM, (height << 16) | 5);
/* shader setup */
etna_set_state(ctx, VIVS_VS_END_PC, vs_size/16);
etna_set_state_multi(ctx, VIVS_VS_INPUT_COUNT, 3, (uint32_t[]){
/* VIVS_VS_INPUT_COUNT */ (1<<8) | 2,
/* VIVS_VS_TEMP_REGISTER_CONTROL */ VIVS_VS_TEMP_REGISTER_CONTROL_NUM_TEMPS(2),
/* VIVS_VS_OUTPUT(0) */ 0x100});
etna_set_state(ctx, VIVS_VS_START_PC, 0x0);
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(0), 0.5); /* u0.x */
etna_set_state_multi(ctx, VIVS_VS_INST_MEM(0), vs_size/4, vs);
etna_set_state(ctx, VIVS_RA_CONTROL, 0x3); /* huh, this is 1 for the cubes */
etna_set_state_multi(ctx, VIVS_PS_END_PC, 2, (uint32_t[]){
/* VIVS_PS_END_PC */ ps_size/16,
/* VIVS_PS_OUTPUT_REG */ 0x1});
etna_set_state(ctx, VIVS_PS_START_PC, 0x0);
etna_set_state(ctx, VIVS_PA_SHADER_ATTRIBUTES(0), 0x200);
etna_set_state(ctx, VIVS_GL_VARYING_NUM_COMPONENTS, /* one varying, with two components */
VIVS_GL_VARYING_NUM_COMPONENTS_VAR0(2)
);
etna_set_state_multi(ctx, VIVS_GL_VARYING_COMPONENT_USE(0), 2, (uint32_t[]){ /* one varying, with four components */
VIVS_GL_VARYING_COMPONENT_USE_COMP0(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP1(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP2(VARYING_COMPONENT_USE_UNUSED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP3(VARYING_COMPONENT_USE_UNUSED)
, 0
});
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(0), 0.0); /* u0.x */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(1), 1.0); /* u0.y */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(2), 0.5); /* u0.z */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(3), 2.0); /* u0.w */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(4), 1/256.0); /* u1.x */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(5), 16.0); /* u1.y */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(6), 10.0); /* u1.z */
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(8), frame); /* u2.x */
etna_set_state_multi(ctx, VIVS_PS_INST_MEM(0), ps_size/4, ps);
etna_set_state(ctx, VIVS_PS_INPUT_COUNT, (31<<8)|2);
etna_set_state(ctx, VIVS_PS_TEMP_REGISTER_CONTROL,
VIVS_PS_TEMP_REGISTER_CONTROL_NUM_TEMPS(4));
etna_set_state(ctx, VIVS_PS_CONTROL,
VIVS_PS_CONTROL_UNK1
);
etna_set_state(ctx, VIVS_PA_ATTRIBUTE_ELEMENT_COUNT, 0x100);
etna_set_state(ctx, VIVS_GL_VARYING_TOTAL_COMPONENTS, /* one varying, with two components, must be
changed together with GL_VARYING_NUM_COMPONENTS */
VIVS_GL_VARYING_TOTAL_COMPONENTS_NUM(2)
);
etna_set_state(ctx, VIVS_VS_LOAD_BALANCING, 0xf3f0582);
etna_set_state(ctx, VIVS_VS_OUTPUT_COUNT, 2);
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_POINT_SIZE_ENABLE, 0));
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_BASE_ADDR, vtx->address); /* ADDR_E */
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_CONTROL,
VIVS_FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE(0x14));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(0),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN(ENDIAN_MODE_NO_SWAP) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(3) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0x0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0xc));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(1),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN(ENDIAN_MODE_NO_SWAP) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(2) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0xc) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0x14));
etna_set_state(ctx, VIVS_VS_INPUT(0), 0x00100); /* 0x20000 in etna_cube */
etna_set_state(ctx, VIVS_PA_CONFIG, ETNA_MASKED_BIT(VIVS_PA_CONFIG_POINT_SPRITE_ENABLE, 0));
etna_draw_primitives(ctx, PRIMITIVE_TYPE_TRIANGLE_STRIP, 0, 2);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
/* Submit first command buffer */
etna_flush(ctx);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_DEST_TILED);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, (padded_width * 4 * 4) | VIVS_RS_DEST_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0); /* no AA, no endian switch */
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(padded_height) |
VIVS_RS_WINDOW_SIZE_WIDTH(padded_width));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
/* Submit second command buffer */
etna_flush(ctx);
etna_warm_up_rs(ctx, aux_rt->address, aux_rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT,
ETNA_MASKED_BIT(VIVS_PE_COLOR_FORMAT_OVERWRITE, 0));
/* Submit third command buffer, wait for pixel engine to finish */
etna_finish(ctx);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_A8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_A8R8G8B8) /*|
VIVS_RS_CONFIG_SWAP_RB*/);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, width * 4);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG,
0); /* no AA, no endian switch */
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, bmp->address); /* ADDR_J */
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(height) |
VIVS_RS_WINDOW_SIZE_WIDTH(width));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
etna_finish(ctx);
}
if(argc>2)
{
printf("Dumping image to %s\n", argv[2]);
bmp_dump32(bmp->logical, width, height, true, argv[2]);
}
/* Unlock video memory */
if(etna_vidmem_unlock(conn, bmp) != 0)
{
fprintf(stderr, "Cannot unlock vidmem\n");
exit(1);
}
etna_free(ctx);
viv_close(conn);
return 0;
}
int main(int argc, char **argv)
{
int rv;
int width = 256;
int height = 256;
int padded_width, padded_height;
fb_info fb;
rv = fb_open(0, &fb);
if(rv!=0)
{
exit(1);
}
width = fb.fb_var.xres;
height = fb.fb_var.yres;
padded_width = etna_align_up(width, 64);
padded_height = etna_align_up(height, 64);
printf("padded_width %i padded_height %i\n", padded_width, padded_height);
rv = viv_open();
if(rv!=0)
{
fprintf(stderr, "Error opening device\n");
exit(1);
}
printf("Succesfully opened device\n");
etna_ctx *ctx = 0;
if(etna_create(&ctx) != ETNA_OK)
{
printf("Unable to create context\n");
exit(1);
}
/* Initialize buffers synchronization structure */
etna_bswap_buffers *buffers = 0;
if(etna_bswap_create(ctx, &buffers, (int (*)(void *, int))&fb_set_buffer, NULL, &fb) < 0)
{
fprintf(stderr, "Error creating buffer swapper\n");
exit(1);
}
/* Allocate video memory */
etna_vidmem *rt = 0; /* main render target */
etna_vidmem *rt_ts = 0; /* tile status for main render target */
etna_vidmem *z = 0; /* depth for main render target */
etna_vidmem *z_ts = 0; /* depth ts for main render target */
etna_vidmem *vtx = 0; /* vertex buffer */
etna_vidmem *aux_rt = 0; /* auxilary render target */
etna_vidmem *aux_rt_ts = 0; /* tile status for auxilary render target */
etna_vidmem *tex = 0; /* texture */
size_t rt_size = padded_width * padded_height * 4;
size_t rt_ts_size = etna_align_up((padded_width * padded_height * 4)/0x100, 0x100);
size_t z_size = padded_width * padded_height * 2;
size_t z_ts_size = etna_align_up((padded_width * padded_height * 2)/0x100, 0x100);
dds_texture *dds = 0;
if(argc<2 || !dds_load(argv[1], &dds))
{
printf("Error loading texture\n");
exit(1);
}
if(etna_vidmem_alloc_linear(&rt, rt_size, gcvSURF_RENDER_TARGET, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&rt_ts, rt_ts_size, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&z, z_size, gcvSURF_DEPTH, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&z_ts, z_ts_size, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&vtx, VERTEX_BUFFER_SIZE, gcvSURF_VERTEX, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&aux_rt, 0x4000, gcvSURF_RENDER_TARGET, gcvPOOL_SYSTEM, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&aux_rt_ts, 0x100, gcvSURF_TILE_STATUS, gcvPOOL_DEFAULT, true)!=ETNA_OK ||
etna_vidmem_alloc_linear(&tex, dds->size, gcvSURF_TEXTURE, gcvPOOL_DEFAULT, true)!=ETNA_OK
)
{
fprintf(stderr, "Error allocating video memory\n");
exit(1);
}
uint32_t tex_format = 0;
uint32_t tex_base_width = dds->slices[0][0].width;
uint32_t tex_base_height = dds->slices[0][0].height;
uint32_t tex_base_log_width = (int)(logf(tex_base_width) * RCPLOG2 * 32.0f + 0.5f);
uint32_t tex_base_log_height = (int)(logf(tex_base_height) * RCPLOG2 * 32.0f + 0.5f);
printf("Loading compressed texture (format %i, %ix%i) log_width=%i log_height=%i\n", dds->fmt, tex_base_width, tex_base_height, tex_base_log_width, tex_base_log_height);
if(dds->fmt == FMT_X8R8G8B8 || dds->fmt == FMT_A8R8G8B8)
{
for(int ix=0; ix<dds->num_mipmaps; ++ix)
{
printf("%08x: Tiling mipmap %i (%ix%i)\n", dds->slices[0][ix].offset, ix, dds->slices[0][ix].width, dds->slices[0][ix].height);
etna_texture_tile((void*)((size_t)tex->logical + dds->slices[0][ix].offset),
dds->slices[0][ix].data, dds->slices[0][ix].width, dds->slices[0][ix].height, dds->slices[0][ix].stride, 4);
}
tex_format = TEXTURE_FORMAT_X8R8G8B8;
} else if(dds->fmt == FMT_DXT1 || dds->fmt == FMT_DXT3 || dds->fmt == FMT_DXT5 || dds->fmt == FMT_ETC1)
{
printf("Uploading compressed texture\n");
memcpy(tex->logical, dds->data, dds->size);
switch(dds->fmt)
{
case FMT_DXT1: tex_format = TEXTURE_FORMAT_DXT1; break;
case FMT_DXT3: tex_format = TEXTURE_FORMAT_DXT2_DXT3; break;
case FMT_DXT5: tex_format = TEXTURE_FORMAT_DXT4_DXT5; break;
case FMT_ETC1: tex_format = TEXTURE_FORMAT_ETC1; break;
}
} else
{
printf("Unknown texture format\n");
exit(1);
}
/* Phew, now we got all the memory we need.
* Write interleaved attribute vertex stream.
* Unlike the GL example we only do this once, not every time glDrawArrays is called, the same would be accomplished
* from GL by using a vertex buffer object.
*/
for(int vert=0; vert<NUM_VERTICES; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+0] = vVertices[vert*3 + comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
((float*)vtx->logical)[dest_idx+comp+3] = vNormals[vert*3 + comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
((float*)vtx->logical)[dest_idx+comp+6] = vTexCoords[vert*2 + comp]; /* 2 */
}
for(int frame=0; frame<1000; ++frame)
{
if(frame%50 == 0)
printf("*** FRAME %i ****\n", frame);
/* Compute transform matrices in the same way as cube egl demo */
ESMatrix modelview, projection, modelviewprojection;
ESMatrix inverse, normal;
esMatrixLoadIdentity(&modelview);
esTranslate(&modelview, 0.0f, 0.0f, -8.0f);
esRotate(&modelview, 45.0f, 1.0f, 0.0f, 0.0f);
esRotate(&modelview, 45.0f, 0.0f, 1.0f, 0.0f);
esRotate(&modelview, frame*0.5f, 0.0f, 0.0f, 1.0f);
GLfloat aspect = (GLfloat)(height) / (GLfloat)(width);
esMatrixLoadIdentity(&projection);
esFrustum(&projection, -2.8f, +2.8f, -2.8f * aspect, +2.8f * aspect, 6.0f, 10.0f);
esMatrixLoadIdentity(&modelviewprojection);
esMatrixMultiply(&modelviewprojection, &modelview, &projection);
esMatrixInverse3x3(&inverse, &modelview);
esMatrixTranspose(&normal, &inverse);
/* XXX part of this can be put outside the loop, but until we have usable context management
* this is safest.
*/
etna_set_state(ctx, VIVS_RA_CONTROL, 0x3);
etna_set_state(ctx, VIVS_GL_MULTI_SAMPLE_CONFIG,
VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_NONE |
VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_ENABLES(0xf) /*|
VIVS_GL_MULTI_SAMPLE_CONFIG_UNK12 |
VIVS_GL_MULTI_SAMPLE_CONFIG_UNK16 */
);
etna_set_state(ctx, VIVS_GL_VERTEX_ELEMENT_CONFIG, 0x1);
etna_set_state(ctx, VIVS_GL_VARYING_NUM_COMPONENTS,
VIVS_GL_VARYING_NUM_COMPONENTS_VAR0(4)| /* position */
VIVS_GL_VARYING_NUM_COMPONENTS_VAR1(2) /* texture coordinate */
);
etna_set_state(ctx, VIVS_GL_VARYING_TOTAL_COMPONENTS,
VIVS_GL_VARYING_TOTAL_COMPONENTS_NUM(4 + 2)
);
etna_set_state_multi(ctx, VIVS_GL_VARYING_COMPONENT_USE(0), 2, (uint32_t[]){
VIVS_GL_VARYING_COMPONENT_USE_COMP0(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP1(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP2(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP3(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP4(VARYING_COMPONENT_USE_USED) |
VIVS_GL_VARYING_COMPONENT_USE_COMP5(VARYING_COMPONENT_USE_USED)
, 0
});
etna_set_state(ctx, VIVS_PA_W_CLIP_LIMIT, 0x34000001);
etna_set_state(ctx, VIVS_PA_SYSTEM_MODE, 0x11);
etna_set_state(ctx, VIVS_PA_CONFIG, /* VIVS_PA_CONFIG_UNK22 | */
VIVS_PA_CONFIG_CULL_FACE_MODE_CCW |
VIVS_PA_CONFIG_FILL_MODE_SOLID |
VIVS_PA_CONFIG_SHADE_MODEL_SMOOTH /* |
VIVS_PA_CONFIG_POINT_SIZE_ENABLE |
VIVS_PA_CONFIG_POINT_SPRITE_ENABLE*/);
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_OFFSET_Z, 0.0);
etna_set_state_f32(ctx, VIVS_PA_VIEWPORT_SCALE_Z, 1.0);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_OFFSET_Y, height << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_X, width << 15);
etna_set_state_fixp(ctx, VIVS_PA_VIEWPORT_SCALE_Y, height << 15);
etna_set_state(ctx, VIVS_PA_ATTRIBUTE_ELEMENT_COUNT, 0x200);
etna_set_state(ctx, VIVS_PA_SHADER_ATTRIBUTES(0), 0x200);
etna_set_state(ctx, VIVS_PA_SHADER_ATTRIBUTES(1), 0x200);
etna_set_state(ctx, VIVS_SE_CONFIG, 0x0);
etna_set_state(ctx, VIVS_SE_DEPTH_SCALE, 0x0);
etna_set_state(ctx, VIVS_SE_DEPTH_BIAS, 0x0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_LEFT, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_TOP, 0);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_RIGHT, (width << 16) | 5);
etna_set_state_fixp(ctx, VIVS_SE_SCISSOR_BOTTOM, (height << 16) | 5);
etna_set_state(ctx, VIVS_PE_ALPHA_CONFIG,
/* VIVS_PE_ALPHA_CONFIG_BLEND_ENABLE_COLOR | */
/* VIVS_PE_ALPHA_CONFIG_BLEND_ENABLE_ALPHA | */
VIVS_PE_ALPHA_CONFIG_SRC_FUNC_COLOR(BLEND_FUNC_ONE) |
VIVS_PE_ALPHA_CONFIG_SRC_FUNC_ALPHA(BLEND_FUNC_ONE) |
VIVS_PE_ALPHA_CONFIG_DST_FUNC_COLOR(BLEND_FUNC_ZERO) |
VIVS_PE_ALPHA_CONFIG_DST_FUNC_ALPHA(BLEND_FUNC_ZERO) |
VIVS_PE_ALPHA_CONFIG_EQ_COLOR(BLEND_EQ_ADD) |
VIVS_PE_ALPHA_CONFIG_EQ_ALPHA(BLEND_EQ_ADD));
etna_set_state(ctx, VIVS_PE_ALPHA_BLEND_COLOR,
VIVS_PE_ALPHA_BLEND_COLOR_B(0) |
VIVS_PE_ALPHA_BLEND_COLOR_G(0) |
VIVS_PE_ALPHA_BLEND_COLOR_R(0) |
VIVS_PE_ALPHA_BLEND_COLOR_A(0));
etna_set_state(ctx, VIVS_PE_ALPHA_OP, /* VIVS_PE_ALPHA_OP_ALPHA_TEST */ 0);
etna_set_state(ctx, VIVS_PE_STENCIL_CONFIG, VIVS_PE_STENCIL_CONFIG_REF_FRONT(0) |
VIVS_PE_STENCIL_CONFIG_MASK_FRONT(0xff) |
VIVS_PE_STENCIL_CONFIG_WRITE_MASK(0xff) |
VIVS_PE_STENCIL_CONFIG_MODE_DISABLED);
etna_set_state(ctx, VIVS_PE_STENCIL_OP, VIVS_PE_STENCIL_OP_FUNC_FRONT(COMPARE_FUNC_ALWAYS) |
VIVS_PE_STENCIL_OP_FUNC_BACK(COMPARE_FUNC_ALWAYS) |
VIVS_PE_STENCIL_OP_FAIL_FRONT(STENCIL_OP_KEEP) |
VIVS_PE_STENCIL_OP_FAIL_BACK(STENCIL_OP_KEEP) |
VIVS_PE_STENCIL_OP_DEPTH_FAIL_FRONT(STENCIL_OP_KEEP) |
VIVS_PE_STENCIL_OP_DEPTH_FAIL_BACK(STENCIL_OP_KEEP) |
VIVS_PE_STENCIL_OP_PASS_FRONT(STENCIL_OP_KEEP) |
VIVS_PE_STENCIL_OP_PASS_BACK(STENCIL_OP_KEEP));
etna_set_state(ctx, VIVS_PE_COLOR_FORMAT,
VIVS_PE_COLOR_FORMAT_COMPONENTS(0xf) |
VIVS_PE_COLOR_FORMAT_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_PE_COLOR_FORMAT_SUPER_TILED |
VIVS_PE_COLOR_FORMAT_OVERWRITE);
etna_set_state(ctx, VIVS_PE_DEPTH_CONFIG,
VIVS_PE_DEPTH_CONFIG_DEPTH_FORMAT_D16 |
VIVS_PE_DEPTH_CONFIG_SUPER_TILED |
VIVS_PE_DEPTH_CONFIG_EARLY_Z |
/* VIVS_PE_DEPTH_CONFIG_WRITE_ENABLE | */
VIVS_PE_DEPTH_CONFIG_DEPTH_FUNC(COMPARE_FUNC_ALWAYS) |
VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_Z
/* VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH */
);
etna_set_state(ctx, VIVS_PE_DEPTH_ADDR, z->address);
etna_set_state(ctx, VIVS_PE_DEPTH_STRIDE, padded_width * 2);
etna_set_state(ctx, VIVS_PE_HDEPTH_CONTROL, VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
etna_set_state(ctx, VIVS_PE_COLOR_ADDR, rt->address);
etna_set_state(ctx, VIVS_PE_COLOR_STRIDE, padded_width * 4);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NEAR, 0.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_FAR, 1.0);
etna_set_state_f32(ctx, VIVS_PE_DEPTH_NORMALIZE, 65535.0);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_FLUSH_CACHE, VIVS_RS_FLUSH_CACHE_FLUSH);
etna_stall(ctx, SYNC_RECIPIENT_RA, SYNC_RECIPIENT_PE);
/* Set up the resolve to clear tile status for main render target
* Regard the TS plane as an image of width 16 with 4 bytes per pixel (64 bytes per row)
* XXX need to clear the depth ts too? we don't really use depth buffer in this sample
* */
etna_set_state(ctx, VIVS_TS_MEM_CONFIG, 0);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_X8R8G8B8)
);
etna_set_state_multi(ctx, VIVS_RS_DITHER(0), 2, (uint32_t[]){0xffffffff, 0xffffffff});
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, 0x40);
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
((rt_ts_size/0x40) << VIVS_RS_WINDOW_SIZE_HEIGHT__SHIFT) |
(16 << VIVS_RS_WINDOW_SIZE_WIDTH__SHIFT));
etna_set_state(ctx, VIVS_RS_FILL_VALUE(0), 0x55555555);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL,
VIVS_RS_CLEAR_CONTROL_MODE_ENABLED1 |
(0xffff << VIVS_RS_CLEAR_CONTROL_BITS__SHIFT));
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0);
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
/** Done */
/* Now set up TS */
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
etna_set_state(ctx, VIVS_TS_DEPTH_CLEAR_VALUE, 0xffffffff);
etna_set_state(ctx, VIVS_TS_DEPTH_STATUS_BASE, z_ts->address);
etna_set_state(ctx, VIVS_TS_DEPTH_SURFACE_BASE, z->address);
etna_set_state(ctx, VIVS_TS_COLOR_CLEAR_VALUE, 0xff303030);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address);
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
/* set up texture unit */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_TEXTURE);
etna_set_state(ctx, VIVS_TE_SAMPLER_SIZE(0),
VIVS_TE_SAMPLER_SIZE_WIDTH(tex_base_width)|
VIVS_TE_SAMPLER_SIZE_HEIGHT(tex_base_height));
etna_set_state(ctx, VIVS_TE_SAMPLER_LOG_SIZE(0),
VIVS_TE_SAMPLER_LOG_SIZE_WIDTH(tex_base_log_width) |
VIVS_TE_SAMPLER_LOG_SIZE_HEIGHT(tex_base_log_height));
for(int ix=0; ix<dds->num_mipmaps; ++ix)
{
etna_set_state(ctx, VIVS_TE_SAMPLER_LOD_ADDR(0,ix), tex->address + dds->slices[0][ix].offset);
}
etna_set_state(ctx, VIVS_TE_SAMPLER_CONFIG0(0),
VIVS_TE_SAMPLER_CONFIG0_TYPE(TEXTURE_TYPE_2D)|
VIVS_TE_SAMPLER_CONFIG0_UWRAP(TEXTURE_WRAPMODE_CLAMP_TO_EDGE)|
VIVS_TE_SAMPLER_CONFIG0_VWRAP(TEXTURE_WRAPMODE_CLAMP_TO_EDGE)|
VIVS_TE_SAMPLER_CONFIG0_MIN(TEXTURE_FILTER_LINEAR)|
VIVS_TE_SAMPLER_CONFIG0_MIP(TEXTURE_FILTER_LINEAR)|
VIVS_TE_SAMPLER_CONFIG0_MAG(TEXTURE_FILTER_LINEAR)|
VIVS_TE_SAMPLER_CONFIG0_FORMAT(tex_format));
etna_set_state(ctx, VIVS_TE_SAMPLER_LOD_CONFIG(0),
VIVS_TE_SAMPLER_LOD_CONFIG_MAX((dds->num_mipmaps - 1)<<5) | VIVS_TE_SAMPLER_LOD_CONFIG_MIN(0));
//etna_set_state(ctx, VIVS_TE_SAMPLER_UNK2100(0), 0);
//etna_set_state(ctx, VIVS_TE_SAMPLER_UNK2140(0), 0);
/* shader setup */
etna_set_state(ctx, VIVS_VS_START_PC, 0x0);
etna_set_state(ctx, VIVS_VS_END_PC, vs_size/16);
etna_set_state_multi(ctx, VIVS_VS_INPUT_COUNT, 3, (uint32_t[]){
/* VIVS_VS_INPUT_COUNT */ VIVS_VS_INPUT_COUNT_UNK8(1) | VIVS_VS_INPUT_COUNT_COUNT(3),
/* VIVS_VS_TEMP_REGISTER_CONTROL */ VIVS_VS_TEMP_REGISTER_CONTROL_NUM_TEMPS(6),
/* VIVS_VS_OUTPUT(0) */ VIVS_VS_OUTPUT_O0(4) |
VIVS_VS_OUTPUT_O1(0) |
VIVS_VS_OUTPUT_O2(1)});
etna_set_state_multi(ctx, VIVS_VS_INST_MEM(0), vs_size/4, vs);
etna_set_state(ctx, VIVS_VS_OUTPUT_COUNT, 3);
etna_set_state(ctx, VIVS_VS_LOAD_BALANCING, 0xf3f0542); /* depends on number of inputs/outputs/varyings? XXX how exactly */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(0), 16, (uint32_t*)&modelviewprojection.m[0][0]);
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(16), 3, (uint32_t*)&normal.m[0][0]); /* u4.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(20), 3, (uint32_t*)&normal.m[1][0]); /* u5.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(24), 3, (uint32_t*)&normal.m[2][0]); /* u6.xyz */
etna_set_state_multi(ctx, VIVS_VS_UNIFORMS(28), 16, (uint32_t*)&modelview.m[0][0]);
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(19), 2.0); /* u4.w */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(23), 20.0); /* u5.w */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(27), 0.0); /* u6.w */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(45), 0.5); /* u11.y */
etna_set_state_f32(ctx, VIVS_VS_UNIFORMS(44), 1.0); /* u11.x */
etna_set_state(ctx, VIVS_VS_INPUT(0), VIVS_VS_INPUT_I0(0) |
VIVS_VS_INPUT_I1(1) |
VIVS_VS_INPUT_I2(2));
etna_set_state(ctx, VIVS_PS_START_PC, 0x0);
etna_set_state_multi(ctx, VIVS_PS_END_PC, 2, (uint32_t[]){
/* VIVS_PS_END_PC */ ps_size/16,
/* VIVS_PS_OUTPUT_REG */ 0x1});
etna_set_state_multi(ctx, VIVS_PS_INST_MEM(0), ps_size/4, ps);
etna_set_state(ctx, VIVS_PS_INPUT_COUNT, VIVS_PS_INPUT_COUNT_UNK8(31) | VIVS_PS_INPUT_COUNT_COUNT(3));
etna_set_state(ctx, VIVS_PS_TEMP_REGISTER_CONTROL, VIVS_PS_TEMP_REGISTER_CONTROL_NUM_TEMPS(3));
etna_set_state(ctx, VIVS_PS_CONTROL, VIVS_PS_CONTROL_UNK1);
etna_set_state_f32(ctx, VIVS_PS_UNIFORMS(0), 1.0); /* u0.x */
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_BASE_ADDR, vtx->address); /* ADDR_E */
etna_set_state(ctx, VIVS_FE_VERTEX_STREAM_CONTROL,
VIVS_FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE((3 + 3 + 2)*4));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(0),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(3) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0x0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0xc));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(1),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(3) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0xc) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0x18));
etna_set_state(ctx, VIVS_FE_VERTEX_ELEMENT_CONFIG(2),
VIVS_FE_VERTEX_ELEMENT_CONFIG_TYPE_FLOAT |
(ENDIAN_MODE_NO_SWAP << VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN__SHIFT) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE |
VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(0) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(2) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_NORMALIZE_OFF |
VIVS_FE_VERTEX_ELEMENT_CONFIG_START(0x18) |
VIVS_FE_VERTEX_ELEMENT_CONFIG_END(0x20));
for(int prim=0; prim<6; ++prim)
{
etna_draw_primitives(ctx, PRIMITIVE_TYPE_TRIANGLE_STRIP, prim*4, 2);
}
#if 0
/* resolve to self */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_RS_CONFIG_DEST_TILED);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, (padded_width * 4 * 4) | VIVS_RS_DEST_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG, 0); /* no AA, no endian switch */
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(padded_height) |
VIVS_RS_WINDOW_SIZE_WIDTH(padded_width));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
etna_set_state(ctx, VIVS_RS_FLUSH_CACHE, VIVS_RS_FLUSH_CACHE_FLUSH);
etna_set_state(ctx, VIVS_TS_COLOR_STATUS_BASE, rt_ts->address); /* ADDR_B */
etna_set_state(ctx, VIVS_TS_COLOR_SURFACE_BASE, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
etna_set_state(ctx, VIVS_TS_MEM_CONFIG,
VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR |
VIVS_TS_MEM_CONFIG_DEPTH_16BPP |
VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR);
#endif
etna_stall(ctx, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE);
/* copy to screen */
etna_bswap_wait_available(buffers);
etna_set_state(ctx, VIVS_GL_FLUSH_CACHE, VIVS_GL_FLUSH_CACHE_COLOR | VIVS_GL_FLUSH_CACHE_DEPTH);
etna_set_state(ctx, VIVS_RS_CONFIG,
VIVS_RS_CONFIG_SOURCE_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_RS_CONFIG_SOURCE_TILED |
VIVS_RS_CONFIG_DEST_FORMAT(RS_FORMAT_X8R8G8B8) |
VIVS_RS_CONFIG_SWAP_RB);
etna_set_state(ctx, VIVS_RS_SOURCE_STRIDE, (padded_width * 4 * 4) | VIVS_RS_SOURCE_STRIDE_TILING);
etna_set_state(ctx, VIVS_RS_DEST_STRIDE, fb.fb_fix.line_length);
etna_set_state(ctx, VIVS_RS_DITHER(0), 0xffffffff);
etna_set_state(ctx, VIVS_RS_DITHER(1), 0xffffffff);
etna_set_state(ctx, VIVS_RS_CLEAR_CONTROL, VIVS_RS_CLEAR_CONTROL_MODE_DISABLED);
etna_set_state(ctx, VIVS_RS_EXTRA_CONFIG,
0); /* no AA, no endian switch */
etna_set_state(ctx, VIVS_RS_SOURCE_ADDR, rt->address); /* ADDR_A */
etna_set_state(ctx, VIVS_RS_DEST_ADDR, fb.physical[buffers->backbuffer]); /* ADDR_J */
etna_set_state(ctx, VIVS_RS_WINDOW_SIZE,
VIVS_RS_WINDOW_SIZE_HEIGHT(height) |
VIVS_RS_WINDOW_SIZE_WIDTH(width));
etna_set_state(ctx, VIVS_RS_KICKER, 0xbeebbeeb);
etna_flush(ctx);
etna_bswap_queue_swap(buffers);
}
#ifdef DUMP
bmp_dump32(fb.logical[1-backbuffer], width, height, false, "/mnt/sdcard/fb.bmp");
printf("Dump complete\n");
#endif
etna_bswap_free(buffers);
etna_free(ctx);
viv_close();
return 0;
}