static inline void
util_format_dxtn_pack_rgba_float(uint8_t *dst_row, unsigned dst_stride,
const float *src, unsigned src_stride,
unsigned width, unsigned height,
enum util_format_dxtn format,
unsigned block_size, boolean srgb)
{
unsigned x, y, i, j, k;
for(y = 0; y < height; y += 4) {
uint8_t *dst = dst_row;
for(x = 0; x < width; x += 4) {
uint8_t tmp[4][4][4];
for(j = 0; j < 4; ++j) {
for(i = 0; i < 4; ++i) {
float src_tmp;
for(k = 0; k < 3; ++k) {
src_tmp = src[(y + j)*src_stride/sizeof(*src) + (x+i)*4 + k];
if (srgb) {
tmp[j][i][k] = util_format_linear_float_to_srgb_8unorm(src_tmp);
}
else {
tmp[j][i][k] = float_to_ubyte(src_tmp);
}
}
/* for sake of simplicity there's an unneeded 4th component for dxt1_rgb */
src_tmp = src[(y + j)*src_stride/sizeof(*src) + (x+i)*4 + 3];
tmp[j][i][3] = float_to_ubyte(src_tmp);
}
}
util_format_dxtn_pack(4, 4, 4, &tmp[0][0][0], format, dst, 0);
dst += block_size;
}
dst_row += 4*dst_stride/sizeof(*dst_row);
}
}
void *
nv50_sampler_state_create(struct pipe_context *pipe,
const struct pipe_sampler_state *cso)
{
struct nv50_tsc_entry *so = MALLOC_STRUCT(nv50_tsc_entry);
float f[2];
so->id = -1;
so->tsc[0] = (0x00026000 |
(nv50_tsc_wrap_mode(cso->wrap_s) << 0) |
(nv50_tsc_wrap_mode(cso->wrap_t) << 3) |
(nv50_tsc_wrap_mode(cso->wrap_r) << 6));
switch (cso->mag_img_filter) {
case PIPE_TEX_FILTER_LINEAR:
so->tsc[1] = NV50_TSC_1_MAGF_LINEAR;
break;
case PIPE_TEX_FILTER_NEAREST:
default:
so->tsc[1] = NV50_TSC_1_MAGF_NEAREST;
break;
}
switch (cso->min_img_filter) {
case PIPE_TEX_FILTER_LINEAR:
so->tsc[1] |= NV50_TSC_1_MINF_LINEAR;
break;
case PIPE_TEX_FILTER_NEAREST:
default:
so->tsc[1] |= NV50_TSC_1_MINF_NEAREST;
break;
}
switch (cso->min_mip_filter) {
case PIPE_TEX_MIPFILTER_LINEAR:
so->tsc[1] |= NV50_TSC_1_MIPF_LINEAR;
break;
case PIPE_TEX_MIPFILTER_NEAREST:
so->tsc[1] |= NV50_TSC_1_MIPF_NEAREST;
break;
case PIPE_TEX_MIPFILTER_NONE:
default:
so->tsc[1] |= NV50_TSC_1_MIPF_NONE;
break;
}
if (nouveau_screen(pipe->screen)->class_3d >= NVE4_3D_CLASS) {
if (cso->seamless_cube_map)
so->tsc[1] |= NVE4_TSC_1_CUBE_SEAMLESS;
if (!cso->normalized_coords)
so->tsc[1] |= NVE4_TSC_1_FORCE_NONNORMALIZED_COORDS;
}
if (cso->max_anisotropy >= 16)
so->tsc[0] |= (7 << 20);
else
if (cso->max_anisotropy >= 12)
so->tsc[0] |= (6 << 20);
else {
so->tsc[0] |= (cso->max_anisotropy >> 1) << 20;
if (cso->max_anisotropy >= 4)
so->tsc[1] |= NV50_TSC_1_UNKN_ANISO_35;
else
if (cso->max_anisotropy >= 2)
so->tsc[1] |= NV50_TSC_1_UNKN_ANISO_15;
}
if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
/* NOTE: must be deactivated for non-shadow textures */
so->tsc[0] |= (1 << 9);
so->tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7) << 10;
}
f[0] = CLAMP(cso->lod_bias, -16.0f, 15.0f);
so->tsc[1] |= ((int)(f[0] * 256.0f) & 0x1fff) << 12;
f[0] = CLAMP(cso->min_lod, 0.0f, 15.0f);
f[1] = CLAMP(cso->max_lod, 0.0f, 15.0f);
so->tsc[2] =
(((int)(f[1] * 256.0f) & 0xfff) << 12) | ((int)(f[0] * 256.0f) & 0xfff);
so->tsc[2] |=
util_format_linear_float_to_srgb_8unorm(cso->border_color.f[0]) << 24;
so->tsc[3] =
util_format_linear_float_to_srgb_8unorm(cso->border_color.f[1]) << 12;
so->tsc[3] |=
util_format_linear_float_to_srgb_8unorm(cso->border_color.f[2]) << 20;
so->tsc[4] = fui(cso->border_color.f[0]);
so->tsc[5] = fui(cso->border_color.f[1]);
so->tsc[6] = fui(cso->border_color.f[2]);
so->tsc[7] = fui(cso->border_color.f[3]);
return (void *)so;
}
bool radv_format_pack_clear_color(VkFormat format,
uint32_t clear_vals[2],
VkClearColorValue *value)
{
uint8_t r = 0, g = 0, b = 0, a = 0;
const struct vk_format_description *desc = vk_format_description(format);
if (vk_format_get_component_bits(format, VK_FORMAT_COLORSPACE_RGB, 0) <= 8) {
if (desc->colorspace == VK_FORMAT_COLORSPACE_RGB) {
r = float_to_ubyte(value->float32[0]);
g = float_to_ubyte(value->float32[1]);
b = float_to_ubyte(value->float32[2]);
a = float_to_ubyte(value->float32[3]);
} else if (desc->colorspace == VK_FORMAT_COLORSPACE_SRGB) {
r = util_format_linear_float_to_srgb_8unorm(value->float32[0]);
g = util_format_linear_float_to_srgb_8unorm(value->float32[1]);
b = util_format_linear_float_to_srgb_8unorm(value->float32[2]);
a = float_to_ubyte(value->float32[3]);
}
}
switch (format) {
case VK_FORMAT_R8_UNORM:
case VK_FORMAT_R8_SRGB:
clear_vals[0] = r;
clear_vals[1] = 0;
break;
case VK_FORMAT_R8G8_UNORM:
case VK_FORMAT_R8G8_SRGB:
clear_vals[0] = r | g << 8;
clear_vals[1] = 0;
break;
case VK_FORMAT_R8G8B8A8_SRGB:
case VK_FORMAT_R8G8B8A8_UNORM:
clear_vals[0] = r | g << 8 | b << 16 | a << 24;
clear_vals[1] = 0;
break;
case VK_FORMAT_B8G8R8A8_SRGB:
case VK_FORMAT_B8G8R8A8_UNORM:
clear_vals[0] = b | g << 8 | r << 16 | a << 24;
clear_vals[1] = 0;
break;
case VK_FORMAT_A8B8G8R8_UNORM_PACK32:
case VK_FORMAT_A8B8G8R8_SRGB_PACK32:
clear_vals[0] = r | g << 8 | b << 16 | a << 24;
clear_vals[1] = 0;
break;
case VK_FORMAT_R8_UINT:
clear_vals[0] = value->uint32[0] & 0xff;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16_UINT:
clear_vals[0] = value->uint32[0] & 0xffff;
clear_vals[1] = 0;
break;
case VK_FORMAT_R8G8_UINT:
clear_vals[0] = value->uint32[0] & 0xff;
clear_vals[0] |= (value->uint32[1] & 0xff) << 8;
clear_vals[1] = 0;
break;
case VK_FORMAT_R8G8B8A8_UINT:
clear_vals[0] = value->uint32[0] & 0xff;
clear_vals[0] |= (value->uint32[1] & 0xff) << 8;
clear_vals[0] |= (value->uint32[2] & 0xff) << 16;
clear_vals[0] |= (value->uint32[3] & 0xff) << 24;
clear_vals[1] = 0;
break;
case VK_FORMAT_A8B8G8R8_UINT_PACK32:
clear_vals[0] = value->uint32[0] & 0xff;
clear_vals[0] |= (value->uint32[1] & 0xff) << 8;
clear_vals[0] |= (value->uint32[2] & 0xff) << 16;
clear_vals[0] |= (value->uint32[3] & 0xff) << 24;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16_UINT:
clear_vals[0] = value->uint32[0] & 0xffff;
clear_vals[0] |= (value->uint32[1] & 0xffff) << 16;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16B16A16_UINT:
clear_vals[0] = value->uint32[0] & 0xffff;
clear_vals[0] |= (value->uint32[1] & 0xffff) << 16;
clear_vals[1] = value->uint32[2] & 0xffff;
clear_vals[1] |= (value->uint32[3] & 0xffff) << 16;
break;
case VK_FORMAT_R32_UINT:
clear_vals[0] = value->uint32[0];
clear_vals[1] = 0;
break;
case VK_FORMAT_R32G32_UINT:
clear_vals[0] = value->uint32[0];
clear_vals[1] = value->uint32[1];
break;
case VK_FORMAT_R32_SINT:
clear_vals[0] = value->int32[0];
clear_vals[1] = 0;
break;
case VK_FORMAT_R16_SFLOAT:
clear_vals[0] = util_float_to_half(value->float32[0]);
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16_SFLOAT:
clear_vals[0] = util_float_to_half(value->float32[0]);
clear_vals[0] |= (uint32_t)util_float_to_half(value->float32[1]) << 16;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16B16A16_SFLOAT:
clear_vals[0] = util_float_to_half(value->float32[0]);
clear_vals[0] |= (uint32_t)util_float_to_half(value->float32[1]) << 16;
clear_vals[1] = util_float_to_half(value->float32[2]);
clear_vals[1] |= (uint32_t)util_float_to_half(value->float32[3]) << 16;
break;
case VK_FORMAT_R16_UNORM:
clear_vals[0] = ((uint16_t)util_iround(CLAMP(value->float32[0], 0.0f, 1.0f) * 0xffff)) & 0xffff;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16_UNORM:
clear_vals[0] = ((uint16_t)util_iround(CLAMP(value->float32[0], 0.0f, 1.0f) * 0xffff)) & 0xffff;
clear_vals[0] |= ((uint16_t)util_iround(CLAMP(value->float32[1], 0.0f, 1.0f) * 0xffff)) << 16;
clear_vals[1] = 0;
break;
case VK_FORMAT_R16G16B16A16_UNORM:
clear_vals[0] = ((uint16_t)util_iround(CLAMP(value->float32[0], 0.0f, 1.0f) * 0xffff)) & 0xffff;
clear_vals[0] |= ((uint16_t)util_iround(CLAMP(value->float32[1], 0.0f, 1.0f) * 0xffff)) << 16;
clear_vals[1] = ((uint16_t)util_iround(CLAMP(value->float32[2], 0.0f, 1.0f) * 0xffff)) & 0xffff;
clear_vals[1] |= ((uint16_t)util_iround(CLAMP(value->float32[3], 0.0f, 1.0f) * 0xffff)) << 16;
break;
case VK_FORMAT_A2B10G10R10_UNORM_PACK32:
/* TODO */
return false;
case VK_FORMAT_R32G32_SFLOAT:
clear_vals[0] = fui(value->float32[0]);
clear_vals[1] = fui(value->float32[1]);
break;
case VK_FORMAT_R32_SFLOAT:
clear_vals[1] = 0;
clear_vals[0] = fui(value->float32[0]);
break;
default:
fprintf(stderr, "failed to fast clear %d\n", format);
return false;
}
return true;
}
