*/
static void si_emit_cb_render_state(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
struct si_state_blend *blend = sctx->queued.named.blend;
/* CB_COLORn_INFO.FORMAT=INVALID should disable unbound colorbuffers,
* but you never know. */
(sctx->ps_shader.cso->info.colors_written & 0x3) != 0x3)
cb_target_mask = 0;
- radeon_set_context_reg(cs, R_028238_CB_TARGET_MASK, cb_target_mask);
+ radeon_opt_set_context_reg(sctx, R_028238_CB_TARGET_MASK,
+ SI_TRACKED_CB_TARGET_MASK, cb_target_mask);
/* GFX9: Flush DFSM when CB_TARGET_MASK changes.
* I think we don't have to do anything between IBs.
blend->blend_enable_4bit & cb_target_mask &&
sctx->framebuffer.nr_samples >= 2;
- radeon_set_context_reg(cs, R_028424_CB_DCC_CONTROL,
- S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
- S_028424_OVERWRITE_COMBINER_WATERMARK(4) |
- S_028424_OVERWRITE_COMBINER_DISABLE(oc_disable));
+ radeon_opt_set_context_reg(
+ sctx, R_028424_CB_DCC_CONTROL,
+ SI_TRACKED_CB_DCC_CONTROL,
+ S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
+ S_028424_OVERWRITE_COMBINER_WATERMARK(4) |
+ S_028424_OVERWRITE_COMBINER_DISABLE(oc_disable));
}
/* RB+ register settings. */
unsigned sx_blend_opt_control = 0;
for (i = 0; i < sctx->framebuffer.state.nr_cbufs; i++) {
- struct r600_surface *surf =
- (struct r600_surface*)sctx->framebuffer.state.cbufs[i];
+ struct si_surface *surf =
+ (struct si_surface*)sctx->framebuffer.state.cbufs[i];
unsigned format, swap, spi_format, colormask;
bool has_alpha, has_rgb;
}
}
- radeon_set_context_reg_seq(cs, R_028754_SX_PS_DOWNCONVERT, 3);
- radeon_emit(cs, sx_ps_downconvert); /* R_028754_SX_PS_DOWNCONVERT */
- radeon_emit(cs, sx_blend_opt_epsilon); /* R_028758_SX_BLEND_OPT_EPSILON */
- radeon_emit(cs, sx_blend_opt_control); /* R_02875C_SX_BLEND_OPT_CONTROL */
+ /* SX_PS_DOWNCONVERT, SX_BLEND_OPT_EPSILON, SX_BLEND_OPT_CONTROL */
+ radeon_opt_set_context_reg3(sctx, R_028754_SX_PS_DOWNCONVERT,
+ SI_TRACKED_SX_PS_DOWNCONVERT,
+ sx_ps_downconvert, sx_blend_opt_epsilon,
+ sx_blend_opt_control);
}
}
static void si_emit_blend_color(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
radeon_set_context_reg_seq(cs, R_028414_CB_BLEND_RED, 4);
radeon_emit_array(cs, (uint32_t*)sctx->blend_color.state.color, 4);
static void si_emit_clip_state(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
radeon_set_context_reg_seq(cs, R_0285BC_PA_CL_UCP_0_X, 6*4);
radeon_emit_array(cs, (uint32_t*)sctx->clip_state.state.ucp, 6*4);
static void si_emit_clip_regs(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
struct si_shader *vs = si_get_vs_state(sctx);
struct si_shader_selector *vs_sel = vs->selector;
struct tgsi_shader_info *info = &vs_sel->info;
clipdist_mask &= rs->clip_plane_enable;
culldist_mask |= clipdist_mask;
- radeon_set_context_reg(cs, R_02881C_PA_CL_VS_OUT_CNTL,
+ radeon_opt_set_context_reg(sctx, R_02881C_PA_CL_VS_OUT_CNTL,
+ SI_TRACKED_PA_CL_VS_OUT_CNTL,
vs_sel->pa_cl_vs_out_cntl |
S_02881C_VS_OUT_CCDIST0_VEC_ENA((total_mask & 0x0F) != 0) |
S_02881C_VS_OUT_CCDIST1_VEC_ENA((total_mask & 0xF0) != 0) |
clipdist_mask | (culldist_mask << 8));
- radeon_set_context_reg(cs, R_028810_PA_CL_CLIP_CNTL,
+ radeon_opt_set_context_reg(sctx, R_028810_PA_CL_CLIP_CNTL,
+ SI_TRACKED_PA_CL_CLIP_CNTL,
rs->pa_cl_clip_cntl |
ucp_mask |
S_028810_CLIP_DISABLE(window_space));
S_028A0C_REPEAT_COUNT(state->line_stipple_factor) : 0;
rs->pa_cl_clip_cntl =
S_028810_DX_CLIP_SPACE_DEF(state->clip_halfz) |
- S_028810_ZCLIP_NEAR_DISABLE(!state->depth_clip) |
- S_028810_ZCLIP_FAR_DISABLE(!state->depth_clip) |
+ S_028810_ZCLIP_NEAR_DISABLE(!state->depth_clip_near) |
+ S_028810_ZCLIP_FAR_DISABLE(!state->depth_clip_near) |
S_028810_DX_RASTERIZATION_KILL(state->rasterizer_discard) |
S_028810_DX_LINEAR_ATTR_CLIP_ENA(1);
si_update_poly_offset_state(sctx);
if (!old_rs ||
- (old_rs->scissor_enable != rs->scissor_enable ||
- old_rs->line_width != rs->line_width ||
- old_rs->max_point_size != rs->max_point_size)) {
+ old_rs->scissor_enable != rs->scissor_enable) {
sctx->scissors.dirty_mask = (1 << SI_MAX_VIEWPORTS) - 1;
si_mark_atom_dirty(sctx, &sctx->atoms.s.scissors);
}
+ if (!old_rs ||
+ old_rs->line_width != rs->line_width ||
+ old_rs->max_point_size != rs->max_point_size)
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.guardband);
+
if (!old_rs ||
old_rs->clip_halfz != rs->clip_halfz) {
sctx->viewports.depth_range_dirty_mask = (1 << SI_MAX_VIEWPORTS) - 1;
*/
static void si_emit_stencil_ref(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
struct pipe_stencil_ref *ref = &sctx->stencil_ref.state;
struct si_dsa_stencil_ref_part *dsa = &sctx->stencil_ref.dsa_part;
static void si_emit_db_render_state(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
- unsigned db_shader_control;
-
- radeon_set_context_reg_seq(cs, R_028000_DB_RENDER_CONTROL, 2);
+ unsigned db_shader_control, db_render_control, db_count_control;
/* DB_RENDER_CONTROL */
if (sctx->dbcb_depth_copy_enabled ||
sctx->dbcb_stencil_copy_enabled) {
- radeon_emit(cs,
- S_028000_DEPTH_COPY(sctx->dbcb_depth_copy_enabled) |
- S_028000_STENCIL_COPY(sctx->dbcb_stencil_copy_enabled) |
- S_028000_COPY_CENTROID(1) |
- S_028000_COPY_SAMPLE(sctx->dbcb_copy_sample));
+ db_render_control =
+ S_028000_DEPTH_COPY(sctx->dbcb_depth_copy_enabled) |
+ S_028000_STENCIL_COPY(sctx->dbcb_stencil_copy_enabled) |
+ S_028000_COPY_CENTROID(1) |
+ S_028000_COPY_SAMPLE(sctx->dbcb_copy_sample);
} else if (sctx->db_flush_depth_inplace || sctx->db_flush_stencil_inplace) {
- radeon_emit(cs,
- S_028000_DEPTH_COMPRESS_DISABLE(sctx->db_flush_depth_inplace) |
- S_028000_STENCIL_COMPRESS_DISABLE(sctx->db_flush_stencil_inplace));
+ db_render_control =
+ S_028000_DEPTH_COMPRESS_DISABLE(sctx->db_flush_depth_inplace) |
+ S_028000_STENCIL_COMPRESS_DISABLE(sctx->db_flush_stencil_inplace);
} else {
- radeon_emit(cs,
- S_028000_DEPTH_CLEAR_ENABLE(sctx->db_depth_clear) |
- S_028000_STENCIL_CLEAR_ENABLE(sctx->db_stencil_clear));
+ db_render_control =
+ S_028000_DEPTH_CLEAR_ENABLE(sctx->db_depth_clear) |
+ S_028000_STENCIL_CLEAR_ENABLE(sctx->db_stencil_clear);
}
/* DB_COUNT_CONTROL (occlusion queries) */
bool perfect = sctx->num_perfect_occlusion_queries > 0;
if (sctx->chip_class >= CIK) {
- radeon_emit(cs,
- S_028004_PERFECT_ZPASS_COUNTS(perfect) |
- S_028004_SAMPLE_RATE(sctx->framebuffer.log_samples) |
- S_028004_ZPASS_ENABLE(1) |
- S_028004_SLICE_EVEN_ENABLE(1) |
- S_028004_SLICE_ODD_ENABLE(1));
+ unsigned log_sample_rate = sctx->framebuffer.log_samples;
+
+ /* Stoney doesn't increment occlusion query counters
+ * if the sample rate is 16x. Use 8x sample rate instead.
+ */
+ if (sctx->family == CHIP_STONEY)
+ log_sample_rate = MIN2(log_sample_rate, 3);
+
+ db_count_control =
+ S_028004_PERFECT_ZPASS_COUNTS(perfect) |
+ S_028004_SAMPLE_RATE(log_sample_rate) |
+ S_028004_ZPASS_ENABLE(1) |
+ S_028004_SLICE_EVEN_ENABLE(1) |
+ S_028004_SLICE_ODD_ENABLE(1);
} else {
- radeon_emit(cs,
- S_028004_PERFECT_ZPASS_COUNTS(perfect) |
- S_028004_SAMPLE_RATE(sctx->framebuffer.log_samples));
+ db_count_control =
+ S_028004_PERFECT_ZPASS_COUNTS(perfect) |
+ S_028004_SAMPLE_RATE(sctx->framebuffer.log_samples);
}
} else {
/* Disable occlusion queries. */
if (sctx->chip_class >= CIK) {
- radeon_emit(cs, 0);
+ db_count_control = 0;
} else {
- radeon_emit(cs, S_028004_ZPASS_INCREMENT_DISABLE(1));
+ db_count_control = S_028004_ZPASS_INCREMENT_DISABLE(1);
}
}
+ radeon_opt_set_context_reg2(sctx, R_028000_DB_RENDER_CONTROL,
+ SI_TRACKED_DB_RENDER_CONTROL, db_render_control,
+ db_count_control);
+
/* DB_RENDER_OVERRIDE2 */
- radeon_set_context_reg(cs, R_028010_DB_RENDER_OVERRIDE2,
+ radeon_opt_set_context_reg(sctx, R_028010_DB_RENDER_OVERRIDE2,
+ SI_TRACKED_DB_RENDER_OVERRIDE2,
S_028010_DISABLE_ZMASK_EXPCLEAR_OPTIMIZATION(sctx->db_depth_disable_expclear) |
S_028010_DISABLE_SMEM_EXPCLEAR_OPTIMIZATION(sctx->db_stencil_disable_expclear) |
S_028010_DECOMPRESS_Z_ON_FLUSH(sctx->framebuffer.nr_samples >= 4));
}
/* Disable the gl_SampleMask fragment shader output if MSAA is disabled. */
- if (!rs || !rs->multisample_enable)
+ if (!rs->multisample_enable)
db_shader_control &= C_02880C_MASK_EXPORT_ENABLE;
if (sctx->screen->has_rbplus &&
!sctx->screen->rbplus_allowed)
db_shader_control |= S_02880C_DUAL_QUAD_DISABLE(1);
- radeon_set_context_reg(cs, R_02880C_DB_SHADER_CONTROL,
- db_shader_control);
+ radeon_opt_set_context_reg(sctx, R_02880C_DB_SHADER_CONTROL,
+ SI_TRACKED_DB_SHADER_CONTROL, db_shader_control);
}
/*
int first_non_void)
{
struct si_screen *sscreen = (struct si_screen*)screen;
- bool enable_compressed_formats = (sscreen->info.drm_major == 2 &&
- sscreen->info.drm_minor >= 31) ||
- sscreen->info.drm_major == 3;
bool uniform = true;
int i;
}
if (desc->layout == UTIL_FORMAT_LAYOUT_RGTC) {
- if (!enable_compressed_formats)
+ if (!sscreen->info.has_format_bc1_through_bc7)
goto out_unknown;
switch (format) {
}
if (desc->layout == UTIL_FORMAT_LAYOUT_BPTC) {
- if (!enable_compressed_formats)
+ if (!sscreen->info.has_format_bc1_through_bc7)
goto out_unknown;
switch (format) {
}
if (desc->layout == UTIL_FORMAT_LAYOUT_S3TC) {
- if (!enable_compressed_formats)
+ if (!sscreen->info.has_format_bc1_through_bc7)
goto out_unknown;
switch (format) {
}
}
-static unsigned si_tex_dim(struct si_screen *sscreen, struct r600_texture *rtex,
+static unsigned si_tex_dim(struct si_screen *sscreen, struct si_texture *tex,
unsigned view_target, unsigned nr_samples)
{
- unsigned res_target = rtex->buffer.b.b.target;
+ unsigned res_target = tex->buffer.b.b.target;
if (view_target == PIPE_TEXTURE_CUBE ||
view_target == PIPE_TEXTURE_CUBE_ARRAY)
if ((res_target == PIPE_TEXTURE_1D ||
res_target == PIPE_TEXTURE_1D_ARRAY) &&
sscreen->info.chip_class >= GFX9 &&
- rtex->surface.u.gfx9.resource_type == RADEON_RESOURCE_2D) {
+ tex->surface.u.gfx9.resource_type == RADEON_RESOURCE_2D) {
if (res_target == PIPE_TEXTURE_1D)
res_target = PIPE_TEXTURE_2D;
else
enum pipe_format format,
enum pipe_texture_target target,
unsigned sample_count,
+ unsigned storage_sample_count,
unsigned usage)
{
+ struct si_screen *sscreen = (struct si_screen *)screen;
unsigned retval = 0;
if (target >= PIPE_MAX_TEXTURE_TYPES) {
return false;
}
- if (!util_format_is_supported(format, usage))
+ if (MAX2(1, sample_count) < MAX2(1, storage_sample_count))
return false;
if (sample_count > 1) {
if (usage & PIPE_BIND_SHADER_IMAGE)
return false;
- switch (sample_count) {
- case 2:
- case 4:
- case 8:
- break;
- case 16:
- if (format == PIPE_FORMAT_NONE)
- return true;
- else
- return false;
- default:
+ /* Only power-of-two sample counts are supported. */
+ if (!util_is_power_of_two_or_zero(sample_count) ||
+ !util_is_power_of_two_or_zero(storage_sample_count))
return false;
+
+ /* MSAA support without framebuffer attachments. */
+ if (format == PIPE_FORMAT_NONE && sample_count <= 16)
+ return true;
+
+ if (!sscreen->info.has_eqaa_surface_allocator ||
+ util_format_is_depth_or_stencil(format)) {
+ /* Color without EQAA or depth/stencil. */
+ if (sample_count > 8 ||
+ sample_count != storage_sample_count)
+ return false;
+ } else {
+ /* Color with EQAA. */
+ if (sample_count > 16 ||
+ storage_sample_count > 8)
+ return false;
}
}
* framebuffer handling
*/
-static void si_choose_spi_color_formats(struct r600_surface *surf,
+static void si_choose_spi_color_formats(struct si_surface *surf,
unsigned format, unsigned swap,
unsigned ntype, bool is_depth)
{
}
static void si_initialize_color_surface(struct si_context *sctx,
- struct r600_surface *surf)
+ struct si_surface *surf)
{
- struct r600_texture *rtex = (struct r600_texture*)surf->base.texture;
+ struct si_texture *tex = (struct si_texture*)surf->base.texture;
unsigned color_info, color_attrib;
unsigned format, swap, ntype, endian;
const struct util_format_description *desc;
color_attrib = S_028C74_FORCE_DST_ALPHA_1(desc->swizzle[3] == PIPE_SWIZZLE_1 ||
util_format_is_intensity(surf->base.format));
- if (rtex->buffer.b.b.nr_samples > 1) {
- unsigned log_samples = util_logbase2(rtex->buffer.b.b.nr_samples);
+ if (tex->buffer.b.b.nr_samples > 1) {
+ unsigned log_samples = util_logbase2(tex->buffer.b.b.nr_samples);
+ unsigned log_fragments = util_logbase2(tex->buffer.b.b.nr_storage_samples);
color_attrib |= S_028C74_NUM_SAMPLES(log_samples) |
- S_028C74_NUM_FRAGMENTS(log_samples);
+ S_028C74_NUM_FRAGMENTS(log_fragments);
- if (rtex->surface.fmask_size) {
+ if (tex->surface.fmask_size) {
color_info |= S_028C70_COMPRESSION(1);
- unsigned fmask_bankh = util_logbase2(rtex->surface.u.legacy.fmask.bankh);
+ unsigned fmask_bankh = util_logbase2(tex->surface.u.legacy.fmask.bankh);
if (sctx->chip_class == SI) {
/* due to a hw bug, FMASK_BANK_HEIGHT must be set on SI too */
if (!sctx->screen->info.has_dedicated_vram)
min_compressed_block_size = V_028C78_MIN_BLOCK_SIZE_64B;
- if (rtex->buffer.b.b.nr_samples > 1) {
- if (rtex->surface.bpe == 1)
+ if (tex->buffer.b.b.nr_storage_samples > 1) {
+ if (tex->surface.bpe == 1)
max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_64B;
- else if (rtex->surface.bpe == 2)
+ else if (tex->surface.bpe == 2)
max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_128B;
}
}
/* This must be set for fast clear to work without FMASK. */
- if (!rtex->surface.fmask_size && sctx->chip_class == SI) {
- unsigned bankh = util_logbase2(rtex->surface.u.legacy.bankh);
+ if (!tex->surface.fmask_size && sctx->chip_class == SI) {
+ unsigned bankh = util_logbase2(tex->surface.u.legacy.bankh);
color_attrib |= S_028C74_FMASK_BANK_HEIGHT(bankh);
}
S_028C6C_SLICE_MAX(surf->base.u.tex.last_layer);
if (sctx->chip_class >= GFX9) {
- unsigned mip0_depth = util_max_layer(&rtex->buffer.b.b, 0);
+ unsigned mip0_depth = util_max_layer(&tex->buffer.b.b, 0);
color_view |= S_028C6C_MIP_LEVEL(surf->base.u.tex.level);
color_attrib |= S_028C74_MIP0_DEPTH(mip0_depth) |
- S_028C74_RESOURCE_TYPE(rtex->surface.u.gfx9.resource_type);
+ S_028C74_RESOURCE_TYPE(tex->surface.u.gfx9.resource_type);
surf->cb_color_attrib2 = S_028C68_MIP0_WIDTH(surf->width0 - 1) |
S_028C68_MIP0_HEIGHT(surf->height0 - 1) |
- S_028C68_MAX_MIP(rtex->buffer.b.b.last_level);
+ S_028C68_MAX_MIP(tex->buffer.b.b.last_level);
}
surf->cb_color_view = color_view;
surf->cb_color_attrib = color_attrib;
/* Determine pixel shader export format */
- si_choose_spi_color_formats(surf, format, swap, ntype, rtex->is_depth);
+ si_choose_spi_color_formats(surf, format, swap, ntype, tex->is_depth);
surf->color_initialized = true;
}
static void si_init_depth_surface(struct si_context *sctx,
- struct r600_surface *surf)
+ struct si_surface *surf)
{
- struct r600_texture *rtex = (struct r600_texture*)surf->base.texture;
+ struct si_texture *tex = (struct si_texture*)surf->base.texture;
unsigned level = surf->base.u.tex.level;
unsigned format, stencil_format;
uint32_t z_info, s_info;
- format = si_translate_dbformat(rtex->db_render_format);
- stencil_format = rtex->surface.has_stencil ?
+ format = si_translate_dbformat(tex->db_render_format);
+ stencil_format = tex->surface.has_stencil ?
V_028044_STENCIL_8 : V_028044_STENCIL_INVALID;
assert(format != V_028040_Z_INVALID);
if (format == V_028040_Z_INVALID)
- PRINT_ERR("Invalid DB format: %d, disabling DB.\n", rtex->buffer.b.b.format);
+ PRINT_ERR("Invalid DB format: %d, disabling DB.\n", tex->buffer.b.b.format);
surf->db_depth_view = S_028008_SLICE_START(surf->base.u.tex.first_layer) |
S_028008_SLICE_MAX(surf->base.u.tex.last_layer);
surf->db_htile_surface = 0;
if (sctx->chip_class >= GFX9) {
- assert(rtex->surface.u.gfx9.surf_offset == 0);
- surf->db_depth_base = rtex->buffer.gpu_address >> 8;
- surf->db_stencil_base = (rtex->buffer.gpu_address +
- rtex->surface.u.gfx9.stencil_offset) >> 8;
+ assert(tex->surface.u.gfx9.surf_offset == 0);
+ surf->db_depth_base = tex->buffer.gpu_address >> 8;
+ surf->db_stencil_base = (tex->buffer.gpu_address +
+ tex->surface.u.gfx9.stencil_offset) >> 8;
z_info = S_028038_FORMAT(format) |
- S_028038_NUM_SAMPLES(util_logbase2(rtex->buffer.b.b.nr_samples)) |
- S_028038_SW_MODE(rtex->surface.u.gfx9.surf.swizzle_mode) |
- S_028038_MAXMIP(rtex->buffer.b.b.last_level);
+ S_028038_NUM_SAMPLES(util_logbase2(tex->buffer.b.b.nr_samples)) |
+ S_028038_SW_MODE(tex->surface.u.gfx9.surf.swizzle_mode) |
+ S_028038_MAXMIP(tex->buffer.b.b.last_level);
s_info = S_02803C_FORMAT(stencil_format) |
- S_02803C_SW_MODE(rtex->surface.u.gfx9.stencil.swizzle_mode);
- surf->db_z_info2 = S_028068_EPITCH(rtex->surface.u.gfx9.surf.epitch);
- surf->db_stencil_info2 = S_02806C_EPITCH(rtex->surface.u.gfx9.stencil.epitch);
+ S_02803C_SW_MODE(tex->surface.u.gfx9.stencil.swizzle_mode);
+ surf->db_z_info2 = S_028068_EPITCH(tex->surface.u.gfx9.surf.epitch);
+ surf->db_stencil_info2 = S_02806C_EPITCH(tex->surface.u.gfx9.stencil.epitch);
surf->db_depth_view |= S_028008_MIPID(level);
- surf->db_depth_size = S_02801C_X_MAX(rtex->buffer.b.b.width0 - 1) |
- S_02801C_Y_MAX(rtex->buffer.b.b.height0 - 1);
+ surf->db_depth_size = S_02801C_X_MAX(tex->buffer.b.b.width0 - 1) |
+ S_02801C_Y_MAX(tex->buffer.b.b.height0 - 1);
- if (si_htile_enabled(rtex, level)) {
+ if (si_htile_enabled(tex, level)) {
z_info |= S_028038_TILE_SURFACE_ENABLE(1) |
S_028038_ALLOW_EXPCLEAR(1);
- if (rtex->tc_compatible_htile) {
+ if (tex->tc_compatible_htile) {
unsigned max_zplanes = 4;
- if (rtex->db_render_format == PIPE_FORMAT_Z16_UNORM &&
- rtex->buffer.b.b.nr_samples > 1)
+ if (tex->db_render_format == PIPE_FORMAT_Z16_UNORM &&
+ tex->buffer.b.b.nr_samples > 1)
max_zplanes = 2;
z_info |= S_028038_DECOMPRESS_ON_N_ZPLANES(max_zplanes + 1) |
s_info |= S_02803C_ITERATE_FLUSH(1);
}
- if (rtex->surface.has_stencil) {
+ if (tex->surface.has_stencil) {
/* Stencil buffer workaround ported from the SI-CI-VI code.
* See that for explanation.
*/
- s_info |= S_02803C_ALLOW_EXPCLEAR(rtex->buffer.b.b.nr_samples <= 1);
+ s_info |= S_02803C_ALLOW_EXPCLEAR(tex->buffer.b.b.nr_samples <= 1);
} else {
/* Use all HTILE for depth if there's no stencil. */
s_info |= S_02803C_TILE_STENCIL_DISABLE(1);
}
- surf->db_htile_data_base = (rtex->buffer.gpu_address +
- rtex->htile_offset) >> 8;
+ surf->db_htile_data_base = (tex->buffer.gpu_address +
+ tex->htile_offset) >> 8;
surf->db_htile_surface = S_028ABC_FULL_CACHE(1) |
- S_028ABC_PIPE_ALIGNED(rtex->surface.u.gfx9.htile.pipe_aligned) |
- S_028ABC_RB_ALIGNED(rtex->surface.u.gfx9.htile.rb_aligned);
+ S_028ABC_PIPE_ALIGNED(tex->surface.u.gfx9.htile.pipe_aligned) |
+ S_028ABC_RB_ALIGNED(tex->surface.u.gfx9.htile.rb_aligned);
}
} else {
/* SI-CI-VI */
- struct legacy_surf_level *levelinfo = &rtex->surface.u.legacy.level[level];
+ struct legacy_surf_level *levelinfo = &tex->surface.u.legacy.level[level];
assert(levelinfo->nblk_x % 8 == 0 && levelinfo->nblk_y % 8 == 0);
- surf->db_depth_base = (rtex->buffer.gpu_address +
- rtex->surface.u.legacy.level[level].offset) >> 8;
- surf->db_stencil_base = (rtex->buffer.gpu_address +
- rtex->surface.u.legacy.stencil_level[level].offset) >> 8;
+ surf->db_depth_base = (tex->buffer.gpu_address +
+ tex->surface.u.legacy.level[level].offset) >> 8;
+ surf->db_stencil_base = (tex->buffer.gpu_address +
+ tex->surface.u.legacy.stencil_level[level].offset) >> 8;
z_info = S_028040_FORMAT(format) |
- S_028040_NUM_SAMPLES(util_logbase2(rtex->buffer.b.b.nr_samples));
+ S_028040_NUM_SAMPLES(util_logbase2(tex->buffer.b.b.nr_samples));
s_info = S_028044_FORMAT(stencil_format);
- surf->db_depth_info = S_02803C_ADDR5_SWIZZLE_MASK(!rtex->tc_compatible_htile);
+ surf->db_depth_info = S_02803C_ADDR5_SWIZZLE_MASK(!tex->tc_compatible_htile);
if (sctx->chip_class >= CIK) {
struct radeon_info *info = &sctx->screen->info;
- unsigned index = rtex->surface.u.legacy.tiling_index[level];
- unsigned stencil_index = rtex->surface.u.legacy.stencil_tiling_index[level];
- unsigned macro_index = rtex->surface.u.legacy.macro_tile_index;
+ unsigned index = tex->surface.u.legacy.tiling_index[level];
+ unsigned stencil_index = tex->surface.u.legacy.stencil_tiling_index[level];
+ unsigned macro_index = tex->surface.u.legacy.macro_tile_index;
unsigned tile_mode = info->si_tile_mode_array[index];
unsigned stencil_tile_mode = info->si_tile_mode_array[stencil_index];
unsigned macro_mode = info->cik_macrotile_mode_array[macro_index];
z_info |= S_028040_TILE_SPLIT(G_009910_TILE_SPLIT(tile_mode));
s_info |= S_028044_TILE_SPLIT(G_009910_TILE_SPLIT(stencil_tile_mode));
} else {
- unsigned tile_mode_index = si_tile_mode_index(rtex, level, false);
+ unsigned tile_mode_index = si_tile_mode_index(tex, level, false);
z_info |= S_028040_TILE_MODE_INDEX(tile_mode_index);
- tile_mode_index = si_tile_mode_index(rtex, level, true);
+ tile_mode_index = si_tile_mode_index(tex, level, true);
s_info |= S_028044_TILE_MODE_INDEX(tile_mode_index);
}
surf->db_depth_slice = S_02805C_SLICE_TILE_MAX((levelinfo->nblk_x *
levelinfo->nblk_y) / 64 - 1);
- if (si_htile_enabled(rtex, level)) {
+ if (si_htile_enabled(tex, level)) {
z_info |= S_028040_TILE_SURFACE_ENABLE(1) |
S_028040_ALLOW_EXPCLEAR(1);
- if (rtex->surface.has_stencil) {
+ if (tex->surface.has_stencil) {
/* Workaround: For a not yet understood reason, the
* combination of MSAA, fast stencil clear and stencil
* decompress messes with subsequent stencil buffer
* Check piglit's arb_texture_multisample-stencil-clear
* test if you want to try changing this.
*/
- if (rtex->buffer.b.b.nr_samples <= 1)
+ if (tex->buffer.b.b.nr_samples <= 1)
s_info |= S_028044_ALLOW_EXPCLEAR(1);
- } else if (!rtex->tc_compatible_htile) {
+ } else if (!tex->tc_compatible_htile) {
/* Use all of the htile_buffer for depth if there's no stencil.
* This must not be set when TC-compatible HTILE is enabled
* due to a hw bug.
s_info |= S_028044_TILE_STENCIL_DISABLE(1);
}
- surf->db_htile_data_base = (rtex->buffer.gpu_address +
- rtex->htile_offset) >> 8;
+ surf->db_htile_data_base = (tex->buffer.gpu_address +
+ tex->htile_offset) >> 8;
surf->db_htile_surface = S_028ABC_FULL_CACHE(1);
- if (rtex->tc_compatible_htile) {
+ if (tex->tc_compatible_htile) {
surf->db_htile_surface |= S_028ABC_TC_COMPATIBLE(1);
- if (rtex->buffer.b.b.nr_samples <= 1)
+ /* 0 = full compression. N = only compress up to N-1 Z planes. */
+ if (tex->buffer.b.b.nr_samples <= 1)
z_info |= S_028040_DECOMPRESS_ON_N_ZPLANES(5);
- else if (rtex->buffer.b.b.nr_samples <= 4)
+ else if (tex->buffer.b.b.nr_samples <= 4)
z_info |= S_028040_DECOMPRESS_ON_N_ZPLANES(3);
else
z_info |= S_028040_DECOMPRESS_ON_N_ZPLANES(2);
if (sctx->framebuffer.state.zsbuf) {
struct pipe_surface *surf = sctx->framebuffer.state.zsbuf;
- struct r600_texture *rtex = (struct r600_texture *)surf->texture;
+ struct si_texture *tex = (struct si_texture *)surf->texture;
- rtex->dirty_level_mask |= 1 << surf->u.tex.level;
+ tex->dirty_level_mask |= 1 << surf->u.tex.level;
- if (rtex->surface.has_stencil)
- rtex->stencil_dirty_level_mask |= 1 << surf->u.tex.level;
+ if (tex->surface.has_stencil)
+ tex->stencil_dirty_level_mask |= 1 << surf->u.tex.level;
}
unsigned compressed_cb_mask = sctx->framebuffer.compressed_cb_mask;
while (compressed_cb_mask) {
unsigned i = u_bit_scan(&compressed_cb_mask);
struct pipe_surface *surf = sctx->framebuffer.state.cbufs[i];
- struct r600_texture *rtex = (struct r600_texture*)surf->texture;
+ struct si_texture *tex = (struct si_texture*)surf->texture;
- if (rtex->surface.fmask_size)
- rtex->dirty_level_mask |= 1 << surf->u.tex.level;
- if (rtex->dcc_gather_statistics)
- rtex->separate_dcc_dirty = true;
+ if (tex->surface.fmask_size)
+ tex->dirty_level_mask |= 1 << surf->u.tex.level;
+ if (tex->dcc_gather_statistics)
+ tex->separate_dcc_dirty = true;
}
}
static void si_dec_framebuffer_counters(const struct pipe_framebuffer_state *state)
{
for (int i = 0; i < state->nr_cbufs; ++i) {
- struct r600_surface *surf = NULL;
- struct r600_texture *rtex;
+ struct si_surface *surf = NULL;
+ struct si_texture *tex;
if (!state->cbufs[i])
continue;
- surf = (struct r600_surface*)state->cbufs[i];
- rtex = (struct r600_texture*)surf->base.texture;
+ surf = (struct si_surface*)state->cbufs[i];
+ tex = (struct si_texture*)surf->base.texture;
- p_atomic_dec(&rtex->framebuffers_bound);
+ p_atomic_dec(&tex->framebuffers_bound);
}
}
{
struct si_context *sctx = (struct si_context *)ctx;
struct pipe_constant_buffer constbuf = {0};
- struct r600_surface *surf = NULL;
- struct r600_texture *rtex;
+ struct si_surface *surf = NULL;
+ struct si_texture *tex;
bool old_any_dst_linear = sctx->framebuffer.any_dst_linear;
unsigned old_nr_samples = sctx->framebuffer.nr_samples;
unsigned old_colorbuf_enabled_4bit = sctx->framebuffer.colorbuf_enabled_4bit;
bool old_has_zsbuf = !!sctx->framebuffer.state.zsbuf;
bool old_has_stencil =
old_has_zsbuf &&
- ((struct r600_texture*)sctx->framebuffer.state.zsbuf->texture)->surface.has_stencil;
+ ((struct si_texture*)sctx->framebuffer.state.zsbuf->texture)->surface.has_stencil;
bool unbound = false;
int i;
if (!sctx->framebuffer.state.cbufs[i])
continue;
- rtex = (struct r600_texture*)sctx->framebuffer.state.cbufs[i]->texture;
- if (rtex->dcc_gather_statistics)
- vi_separate_dcc_stop_query(sctx, rtex);
+ tex = (struct si_texture*)sctx->framebuffer.state.cbufs[i]->texture;
+ if (tex->dcc_gather_statistics)
+ vi_separate_dcc_stop_query(sctx, tex);
}
/* Disable DCC if the formats are incompatible. */
if (!state->cbufs[i])
continue;
- surf = (struct r600_surface*)state->cbufs[i];
- rtex = (struct r600_texture*)surf->base.texture;
+ surf = (struct si_surface*)state->cbufs[i];
+ tex = (struct si_texture*)surf->base.texture;
if (!surf->dcc_incompatible)
continue;
unbound = true;
}
- if (vi_dcc_enabled(rtex, surf->base.u.tex.level))
- if (!si_texture_disable_dcc(sctx, rtex))
- si_decompress_dcc(sctx, rtex);
+ if (vi_dcc_enabled(tex, surf->base.u.tex.level))
+ if (!si_texture_disable_dcc(sctx, tex))
+ si_decompress_dcc(sctx, tex);
surf->dcc_incompatible = false;
}
sctx->framebuffer.compressed_cb_mask = 0;
sctx->framebuffer.uncompressed_cb_mask = 0;
sctx->framebuffer.nr_samples = util_framebuffer_get_num_samples(state);
+ sctx->framebuffer.nr_color_samples = sctx->framebuffer.nr_samples;
sctx->framebuffer.log_samples = util_logbase2(sctx->framebuffer.nr_samples);
sctx->framebuffer.any_dst_linear = false;
sctx->framebuffer.CB_has_shader_readable_metadata = false;
if (!state->cbufs[i])
continue;
- surf = (struct r600_surface*)state->cbufs[i];
- rtex = (struct r600_texture*)surf->base.texture;
+ surf = (struct si_surface*)state->cbufs[i];
+ tex = (struct si_texture*)surf->base.texture;
if (!surf->color_initialized) {
si_initialize_color_surface(sctx, surf);
if (surf->color_is_int10)
sctx->framebuffer.color_is_int10 |= 1 << i;
- if (rtex->surface.fmask_size)
+ if (tex->surface.fmask_size)
sctx->framebuffer.compressed_cb_mask |= 1 << i;
else
sctx->framebuffer.uncompressed_cb_mask |= 1 << i;
- if (rtex->surface.is_linear)
+ /* Don't update nr_color_samples for non-AA buffers.
+ * (e.g. destination of MSAA resolve)
+ */
+ if (tex->buffer.b.b.nr_samples >= 2 &&
+ tex->buffer.b.b.nr_storage_samples < tex->buffer.b.b.nr_samples) {
+ sctx->framebuffer.nr_color_samples =
+ MIN2(sctx->framebuffer.nr_color_samples,
+ tex->buffer.b.b.nr_storage_samples);
+ sctx->framebuffer.nr_color_samples =
+ MAX2(1, sctx->framebuffer.nr_color_samples);
+ }
+
+ if (tex->surface.is_linear)
sctx->framebuffer.any_dst_linear = true;
- if (vi_dcc_enabled(rtex, surf->base.u.tex.level))
+ if (vi_dcc_enabled(tex, surf->base.u.tex.level))
sctx->framebuffer.CB_has_shader_readable_metadata = true;
si_context_add_resource_size(sctx, surf->base.texture);
- p_atomic_inc(&rtex->framebuffers_bound);
+ p_atomic_inc(&tex->framebuffers_bound);
- if (rtex->dcc_gather_statistics) {
+ if (tex->dcc_gather_statistics) {
/* Dirty tracking must be enabled for DCC usage analysis. */
sctx->framebuffer.compressed_cb_mask |= 1 << i;
- vi_separate_dcc_start_query(sctx, rtex);
+ vi_separate_dcc_start_query(sctx, tex);
}
}
- struct r600_texture *zstex = NULL;
+ struct si_texture *zstex = NULL;
if (state->zsbuf) {
- surf = (struct r600_surface*)state->zsbuf;
- zstex = (struct r600_texture*)surf->base.texture;
+ surf = (struct si_surface*)state->zsbuf;
+ zstex = (struct si_texture*)surf->base.texture;
if (!surf->depth_initialized) {
si_init_depth_surface(sctx, surf);
static void si_emit_framebuffer_state(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
struct pipe_framebuffer_state *state = &sctx->framebuffer.state;
unsigned i, nr_cbufs = state->nr_cbufs;
- struct r600_texture *tex = NULL;
- struct r600_surface *cb = NULL;
+ struct si_texture *tex = NULL;
+ struct si_surface *cb = NULL;
unsigned cb_color_info = 0;
/* Colorbuffers. */
if (!(sctx->framebuffer.dirty_cbufs & (1 << i)))
continue;
- cb = (struct r600_surface*)state->cbufs[i];
+ cb = (struct si_surface*)state->cbufs[i];
if (!cb) {
radeon_set_context_reg(cs, R_028C70_CB_COLOR0_INFO + i * 0x3C,
S_028C70_FORMAT(V_028C70_COLOR_INVALID));
continue;
}
- tex = (struct r600_texture *)cb->base.texture;
+ tex = (struct si_texture *)cb->base.texture;
radeon_add_to_buffer_list(sctx, sctx->gfx_cs,
&tex->buffer, RADEON_USAGE_READWRITE,
tex->buffer.b.b.nr_samples > 1 ?
if (tex->cmask_buffer && tex->cmask_buffer != &tex->buffer) {
radeon_add_to_buffer_list(sctx, sctx->gfx_cs,
tex->cmask_buffer, RADEON_USAGE_READWRITE,
- RADEON_PRIO_CMASK);
+ RADEON_PRIO_SEPARATE_META);
}
if (tex->dcc_separate_buffer)
radeon_add_to_buffer_list(sctx, sctx->gfx_cs,
tex->dcc_separate_buffer,
RADEON_USAGE_READWRITE,
- RADEON_PRIO_DCC);
+ RADEON_PRIO_SEPARATE_META);
/* Compute mutable surface parameters. */
cb_color_base = tex->buffer.gpu_address >> 8;
cb_color_fmask = 0;
- cb_color_cmask = tex->cmask.base_address_reg;
+ cb_color_cmask = tex->cmask_base_address_reg;
cb_dcc_base = 0;
cb_color_info = cb->cb_color_info | tex->cb_color_info;
cb_color_attrib = cb->cb_color_attrib;
radeon_emit(cs, cb_color_attrib); /* CB_COLOR0_ATTRIB */
radeon_emit(cs, cb->cb_dcc_control); /* CB_COLOR0_DCC_CONTROL */
radeon_emit(cs, cb_color_cmask); /* CB_COLOR0_CMASK */
- radeon_emit(cs, tex->cmask.slice_tile_max); /* CB_COLOR0_CMASK_SLICE */
+ radeon_emit(cs, tex->surface.u.legacy.cmask_slice_tile_max); /* CB_COLOR0_CMASK_SLICE */
radeon_emit(cs, cb_color_fmask); /* CB_COLOR0_FMASK */
radeon_emit(cs, cb_color_fmask_slice); /* CB_COLOR0_FMASK_SLICE */
radeon_emit(cs, tex->color_clear_value[0]); /* CB_COLOR0_CLEAR_WORD0 */
/* ZS buffer. */
if (state->zsbuf && sctx->framebuffer.dirty_zsbuf) {
- struct r600_surface *zb = (struct r600_surface*)state->zsbuf;
- struct r600_texture *rtex = (struct r600_texture*)zb->base.texture;
+ struct si_surface *zb = (struct si_surface*)state->zsbuf;
+ struct si_texture *tex = (struct si_texture*)zb->base.texture;
radeon_add_to_buffer_list(sctx, sctx->gfx_cs,
- &rtex->buffer, RADEON_USAGE_READWRITE,
+ &tex->buffer, RADEON_USAGE_READWRITE,
zb->base.texture->nr_samples > 1 ?
RADEON_PRIO_DEPTH_BUFFER_MSAA :
RADEON_PRIO_DEPTH_BUFFER);
radeon_set_context_reg_seq(cs, R_028038_DB_Z_INFO, 10);
radeon_emit(cs, zb->db_z_info | /* DB_Z_INFO */
- S_028038_ZRANGE_PRECISION(rtex->depth_clear_value != 0));
+ S_028038_ZRANGE_PRECISION(tex->depth_clear_value != 0));
radeon_emit(cs, zb->db_stencil_info); /* DB_STENCIL_INFO */
radeon_emit(cs, zb->db_depth_base); /* DB_Z_READ_BASE */
radeon_emit(cs, S_028044_BASE_HI(zb->db_depth_base >> 32)); /* DB_Z_READ_BASE_HI */
radeon_set_context_reg_seq(cs, R_02803C_DB_DEPTH_INFO, 9);
radeon_emit(cs, zb->db_depth_info); /* DB_DEPTH_INFO */
radeon_emit(cs, zb->db_z_info | /* DB_Z_INFO */
- S_028040_ZRANGE_PRECISION(rtex->depth_clear_value != 0));
+ S_028040_ZRANGE_PRECISION(tex->depth_clear_value != 0));
radeon_emit(cs, zb->db_stencil_info); /* DB_STENCIL_INFO */
radeon_emit(cs, zb->db_depth_base); /* DB_Z_READ_BASE */
radeon_emit(cs, zb->db_stencil_base); /* DB_STENCIL_READ_BASE */
}
radeon_set_context_reg_seq(cs, R_028028_DB_STENCIL_CLEAR, 2);
- radeon_emit(cs, rtex->stencil_clear_value); /* R_028028_DB_STENCIL_CLEAR */
- radeon_emit(cs, fui(rtex->depth_clear_value)); /* R_02802C_DB_DEPTH_CLEAR */
+ radeon_emit(cs, tex->stencil_clear_value); /* R_028028_DB_STENCIL_CLEAR */
+ radeon_emit(cs, fui(tex->depth_clear_value)); /* R_02802C_DB_DEPTH_CLEAR */
radeon_set_context_reg(cs, R_028008_DB_DEPTH_VIEW, zb->db_depth_view);
radeon_set_context_reg(cs, R_028ABC_DB_HTILE_SURFACE, zb->db_htile_surface);
static void si_emit_msaa_sample_locs(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
unsigned nr_samples = sctx->framebuffer.nr_samples;
bool has_msaa_sample_loc_bug = sctx->screen->has_msaa_sample_loc_bug;
*/
if (has_msaa_sample_loc_bug &&
sctx->framebuffer.nr_samples > 1 &&
- rs && !rs->multisample_enable)
+ !rs->multisample_enable)
small_prim_filter_cntl &= C_028830_SMALL_PRIM_FILTER_ENABLE;
- radeon_set_context_reg(cs, R_028830_PA_SU_SMALL_PRIM_FILTER_CNTL,
- small_prim_filter_cntl);
+ radeon_opt_set_context_reg(sctx,
+ R_028830_PA_SU_SMALL_PRIM_FILTER_CNTL,
+ SI_TRACKED_PA_SU_SMALL_PRIM_FILTER_CNTL,
+ small_prim_filter_cntl);
}
}
};
if (sctx->framebuffer.state.zsbuf) {
- struct r600_texture *zstex =
- (struct r600_texture*)sctx->framebuffer.state.zsbuf->texture;
+ struct si_texture *zstex =
+ (struct si_texture*)sctx->framebuffer.state.zsbuf->texture;
bool has_stencil = zstex->surface.has_stencil;
dsa_order_invariant = dsa->order_invariance[has_stencil];
if (!dsa_order_invariant.zs)
static void si_emit_msaa_config(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
unsigned num_tile_pipes = sctx->screen->info.num_tile_pipes;
/* 33% faster rendering to linear color buffers */
bool dst_is_linear = sctx->framebuffer.any_dst_linear;
S_028804_INCOHERENT_EQAA_READS(1) |
S_028804_INTERPOLATE_COMP_Z(1) |
S_028804_STATIC_ANCHOR_ASSOCIATIONS(1);
+ unsigned coverage_samples, color_samples, z_samples;
- int setup_samples = sctx->framebuffer.nr_samples > 1 ? sctx->framebuffer.nr_samples :
- sctx->smoothing_enabled ? SI_NUM_SMOOTH_AA_SAMPLES : 0;
+ /* S: Coverage samples (up to 16x):
+ * - Scan conversion samples (PA_SC_AA_CONFIG.MSAA_NUM_SAMPLES)
+ * - CB FMASK samples (CB_COLORi_ATTRIB.NUM_SAMPLES)
+ *
+ * Z: Z/S samples (up to 8x, must be <= coverage samples and >= color samples):
+ * - Value seen by DB (DB_Z_INFO.NUM_SAMPLES)
+ * - Value seen by CB, must be correct even if Z/S is unbound (DB_EQAA.MAX_ANCHOR_SAMPLES)
+ * # Missing samples are derived from Z planes if Z is compressed (up to 16x quality), or
+ * # from the closest defined sample if Z is uncompressed (same quality as the number of
+ * # Z samples).
+ *
+ * F: Color samples (up to 8x, must be <= coverage samples):
+ * - CB color samples (CB_COLORi_ATTRIB.NUM_FRAGMENTS)
+ * - PS iter samples (DB_EQAA.PS_ITER_SAMPLES)
+ *
+ * Can be anything between coverage and color samples:
+ * - SampleMaskIn samples (PA_SC_AA_CONFIG.MSAA_EXPOSED_SAMPLES)
+ * - SampleMaskOut samples (DB_EQAA.MASK_EXPORT_NUM_SAMPLES)
+ * - Alpha-to-coverage samples (DB_EQAA.ALPHA_TO_MASK_NUM_SAMPLES)
+ * - Occlusion query samples (DB_COUNT_CONTROL.SAMPLE_RATE)
+ * # All are currently set the same as coverage samples.
+ *
+ * If color samples < coverage samples, FMASK has a higher bpp to store an "unknown"
+ * flag for undefined color samples. A shader-based resolve must handle unknowns
+ * or mask them out with AND. Unknowns can also be guessed from neighbors via
+ * an edge-detect shader-based resolve, which is required to make "color samples = 1"
+ * useful. The CB resolve always drops unknowns.
+ *
+ * Sensible AA configurations:
+ * EQAA 16s 8z 8f - might look the same as 16x MSAA if Z is compressed
+ * EQAA 16s 8z 4f - might look the same as 16x MSAA if Z is compressed
+ * EQAA 16s 4z 4f - might look the same as 16x MSAA if Z is compressed
+ * EQAA 8s 8z 8f = 8x MSAA
+ * EQAA 8s 8z 4f - might look the same as 8x MSAA
+ * EQAA 8s 8z 2f - might look the same as 8x MSAA with low-density geometry
+ * EQAA 8s 4z 4f - might look the same as 8x MSAA if Z is compressed
+ * EQAA 8s 4z 2f - might look the same as 8x MSAA with low-density geometry if Z is compressed
+ * EQAA 4s 4z 4f = 4x MSAA
+ * EQAA 4s 4z 2f - might look the same as 4x MSAA with low-density geometry
+ * EQAA 2s 2z 2f = 2x MSAA
+ */
+ if (sctx->framebuffer.nr_samples > 1) {
+ coverage_samples = sctx->framebuffer.nr_samples;
+ color_samples = sctx->framebuffer.nr_color_samples;
+
+ if (sctx->framebuffer.state.zsbuf) {
+ z_samples = sctx->framebuffer.state.zsbuf->texture->nr_samples;
+ z_samples = MAX2(1, z_samples);
+ } else {
+ z_samples = coverage_samples;
+ }
+ } else if (sctx->smoothing_enabled) {
+ coverage_samples = color_samples = z_samples = SI_NUM_SMOOTH_AA_SAMPLES;
+ } else {
+ coverage_samples = color_samples = z_samples = 1;
+ }
/* Required by OpenGL line rasterization.
*
* endcaps.
*/
unsigned sc_line_cntl = S_028BDC_DX10_DIAMOND_TEST_ENA(1);
+ unsigned sc_aa_config = 0;
- if (setup_samples > 1) {
+ if (coverage_samples > 1) {
/* distance from the pixel center, indexed by log2(nr_samples) */
static unsigned max_dist[] = {
0, /* unused */
7, /* 8x MSAA */
8, /* 16x MSAA */
};
- unsigned log_samples = util_logbase2(setup_samples);
+ unsigned log_samples = util_logbase2(coverage_samples);
+ unsigned log_z_samples = util_logbase2(z_samples);
unsigned ps_iter_samples = si_get_ps_iter_samples(sctx);
- unsigned log_ps_iter_samples =
- util_logbase2(util_next_power_of_two(ps_iter_samples));
+ unsigned log_ps_iter_samples = util_logbase2(ps_iter_samples);
- radeon_set_context_reg_seq(cs, R_028BDC_PA_SC_LINE_CNTL, 2);
- radeon_emit(cs, sc_line_cntl |
- S_028BDC_EXPAND_LINE_WIDTH(1)); /* R_028BDC_PA_SC_LINE_CNTL */
- radeon_emit(cs, S_028BE0_MSAA_NUM_SAMPLES(log_samples) |
- S_028BE0_MAX_SAMPLE_DIST(max_dist[log_samples]) |
- S_028BE0_MSAA_EXPOSED_SAMPLES(log_samples)); /* R_028BE0_PA_SC_AA_CONFIG */
+ sc_line_cntl |= S_028BDC_EXPAND_LINE_WIDTH(1);
+ sc_aa_config = S_028BE0_MSAA_NUM_SAMPLES(log_samples) |
+ S_028BE0_MAX_SAMPLE_DIST(max_dist[log_samples]) |
+ S_028BE0_MSAA_EXPOSED_SAMPLES(log_samples);
if (sctx->framebuffer.nr_samples > 1) {
- radeon_set_context_reg(cs, R_028804_DB_EQAA,
- db_eqaa |
- S_028804_MAX_ANCHOR_SAMPLES(log_samples) |
- S_028804_PS_ITER_SAMPLES(log_ps_iter_samples) |
- S_028804_MASK_EXPORT_NUM_SAMPLES(log_samples) |
- S_028804_ALPHA_TO_MASK_NUM_SAMPLES(log_samples));
- radeon_set_context_reg(cs, R_028A4C_PA_SC_MODE_CNTL_1,
- S_028A4C_PS_ITER_SAMPLE(ps_iter_samples > 1) |
- sc_mode_cntl_1);
+ db_eqaa |= S_028804_MAX_ANCHOR_SAMPLES(log_z_samples) |
+ S_028804_PS_ITER_SAMPLES(log_ps_iter_samples) |
+ S_028804_MASK_EXPORT_NUM_SAMPLES(log_samples) |
+ S_028804_ALPHA_TO_MASK_NUM_SAMPLES(log_samples);
+ sc_mode_cntl_1 |= S_028A4C_PS_ITER_SAMPLE(ps_iter_samples > 1);
} else if (sctx->smoothing_enabled) {
- radeon_set_context_reg(cs, R_028804_DB_EQAA,
- db_eqaa |
- S_028804_OVERRASTERIZATION_AMOUNT(log_samples));
- radeon_set_context_reg(cs, R_028A4C_PA_SC_MODE_CNTL_1,
- sc_mode_cntl_1);
+ db_eqaa |= S_028804_OVERRASTERIZATION_AMOUNT(log_samples);
}
- } else {
- radeon_set_context_reg_seq(cs, R_028BDC_PA_SC_LINE_CNTL, 2);
- radeon_emit(cs, sc_line_cntl); /* R_028BDC_PA_SC_LINE_CNTL */
- radeon_emit(cs, 0); /* R_028BE0_PA_SC_AA_CONFIG */
-
- radeon_set_context_reg(cs, R_028804_DB_EQAA, db_eqaa);
- radeon_set_context_reg(cs, R_028A4C_PA_SC_MODE_CNTL_1,
- sc_mode_cntl_1);
}
+ /* R_028BDC_PA_SC_LINE_CNTL, R_028BE0_PA_SC_AA_CONFIG */
+ radeon_opt_set_context_reg2(sctx, R_028BDC_PA_SC_LINE_CNTL,
+ SI_TRACKED_PA_SC_LINE_CNTL, sc_line_cntl,
+ sc_aa_config);
+ /* R_028804_DB_EQAA */
+ radeon_opt_set_context_reg(sctx, R_028804_DB_EQAA, SI_TRACKED_DB_EQAA,
+ db_eqaa);
+ /* R_028A4C_PA_SC_MODE_CNTL_1 */
+ radeon_opt_set_context_reg(sctx, R_028A4C_PA_SC_MODE_CNTL_1,
+ SI_TRACKED_PA_SC_MODE_CNTL_1, sc_mode_cntl_1);
+
/* GFX9: Flush DFSM when the AA mode changes. */
if (sctx->screen->dfsm_allowed) {
radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
{
struct si_context *sctx = (struct si_context *)ctx;
+ /* The hardware can only do sample shading with 2^n samples. */
+ min_samples = util_next_power_of_two(min_samples);
+
if (sctx->ps_iter_samples == min_samples)
return;
*/
void
si_make_texture_descriptor(struct si_screen *screen,
- struct r600_texture *tex,
+ struct si_texture *tex,
bool sampler,
enum pipe_texture_target target,
enum pipe_format pipe_format,
const struct util_format_description *desc;
unsigned char swizzle[4];
int first_non_void;
- unsigned num_format, data_format, type;
+ unsigned num_format, data_format, type, num_samples;
uint64_t va;
desc = util_format_description(pipe_format);
+ num_samples = desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS ?
+ MAX2(1, res->nr_samples) :
+ MAX2(1, res->nr_storage_samples);
+
if (desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS) {
const unsigned char swizzle_xxxx[4] = {0, 0, 0, 0};
const unsigned char swizzle_yyyy[4] = {1, 1, 1, 1};
assert(res->target != PIPE_TEXTURE_3D || (first_level == 0 && last_level == 0));
} else {
- type = si_tex_dim(screen, tex, target, res->nr_samples);
+ type = si_tex_dim(screen, tex, target, num_samples);
}
if (type == V_008F1C_SQ_RSRC_IMG_1D_ARRAY) {
S_008F1C_DST_SEL_Y(si_map_swizzle(swizzle[1])) |
S_008F1C_DST_SEL_Z(si_map_swizzle(swizzle[2])) |
S_008F1C_DST_SEL_W(si_map_swizzle(swizzle[3])) |
- S_008F1C_BASE_LEVEL(res->nr_samples > 1 ?
- 0 : first_level) |
- S_008F1C_LAST_LEVEL(res->nr_samples > 1 ?
- util_logbase2(res->nr_samples) :
+ S_008F1C_BASE_LEVEL(num_samples > 1 ? 0 : first_level) |
+ S_008F1C_LAST_LEVEL(num_samples > 1 ?
+ util_logbase2(num_samples) :
last_level) |
S_008F1C_TYPE(type));
state[4] = 0;
state[4] |= S_008F20_DEPTH(last_layer);
state[4] |= S_008F20_BC_SWIZZLE(bc_swizzle);
- state[5] |= S_008F24_MAX_MIP(res->nr_samples > 1 ?
- util_logbase2(res->nr_samples) :
+ state[5] |= S_008F24_MAX_MIP(num_samples > 1 ?
+ util_logbase2(num_samples) :
tex->buffer.b.b.last_level);
} else {
state[3] |= S_008F1C_POW2_PAD(res->last_level > 0);
va = tex->buffer.gpu_address + tex->fmask_offset;
+#define FMASK(s,f) (((unsigned)(MAX2(1, s)) * 16) + (MAX2(1, f)))
if (screen->info.chip_class >= GFX9) {
data_format = V_008F14_IMG_DATA_FORMAT_FMASK;
- switch (res->nr_samples) {
- case 2:
+ switch (FMASK(res->nr_samples, res->nr_storage_samples)) {
+ case FMASK(2,1):
+ num_format = V_008F14_IMG_FMASK_8_2_1;
+ break;
+ case FMASK(2,2):
num_format = V_008F14_IMG_FMASK_8_2_2;
break;
- case 4:
+ case FMASK(4,1):
+ num_format = V_008F14_IMG_FMASK_8_4_1;
+ break;
+ case FMASK(4,2):
+ num_format = V_008F14_IMG_FMASK_8_4_2;
+ break;
+ case FMASK(4,4):
num_format = V_008F14_IMG_FMASK_8_4_4;
break;
- case 8:
+ case FMASK(8,1):
+ num_format = V_008F14_IMG_FMASK_8_8_1;
+ break;
+ case FMASK(8,2):
+ num_format = V_008F14_IMG_FMASK_16_8_2;
+ break;
+ case FMASK(8,4):
+ num_format = V_008F14_IMG_FMASK_32_8_4;
+ break;
+ case FMASK(8,8):
num_format = V_008F14_IMG_FMASK_32_8_8;
break;
+ case FMASK(16,1):
+ num_format = V_008F14_IMG_FMASK_16_16_1;
+ break;
+ case FMASK(16,2):
+ num_format = V_008F14_IMG_FMASK_32_16_2;
+ break;
+ case FMASK(16,4):
+ num_format = V_008F14_IMG_FMASK_64_16_4;
+ break;
+ case FMASK(16,8):
+ num_format = V_008F14_IMG_FMASK_64_16_8;
+ break;
default:
unreachable("invalid nr_samples");
}
} else {
- switch (res->nr_samples) {
- case 2:
+ switch (FMASK(res->nr_samples, res->nr_storage_samples)) {
+ case FMASK(2,1):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S2_F1;
+ break;
+ case FMASK(2,2):
data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S2_F2;
break;
- case 4:
+ case FMASK(4,1):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S4_F1;
+ break;
+ case FMASK(4,2):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S4_F2;
+ break;
+ case FMASK(4,4):
data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S4_F4;
break;
- case 8:
+ case FMASK(8,1):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S8_F1;
+ break;
+ case FMASK(8,2):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK16_S8_F2;
+ break;
+ case FMASK(8,4):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK32_S8_F4;
+ break;
+ case FMASK(8,8):
data_format = V_008F14_IMG_DATA_FORMAT_FMASK32_S8_F8;
break;
+ case FMASK(16,1):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK16_S16_F1;
+ break;
+ case FMASK(16,2):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK32_S16_F2;
+ break;
+ case FMASK(16,4):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK64_S16_F4;
+ break;
+ case FMASK(16,8):
+ data_format = V_008F14_IMG_DATA_FORMAT_FMASK64_S16_F8;
+ break;
default:
unreachable("invalid nr_samples");
}
num_format = V_008F14_IMG_NUM_FORMAT_UINT;
}
+#undef FMASK
fmask_state[0] = (va >> 8) | tex->surface.fmask_tile_swizzle;
fmask_state[1] = S_008F14_BASE_ADDRESS_HI(va >> 40) |
{
struct si_context *sctx = (struct si_context*)ctx;
struct si_sampler_view *view = CALLOC_STRUCT(si_sampler_view);
- struct r600_texture *tmp = (struct r600_texture*)texture;
+ struct si_texture *tex = (struct si_texture*)texture;
unsigned base_level, first_level, last_level;
unsigned char state_swizzle[4];
unsigned height, depth, width;
pipe_format = state->format;
/* Depth/stencil texturing sometimes needs separate texture. */
- if (tmp->is_depth && !si_can_sample_zs(tmp, view->is_stencil_sampler)) {
- if (!tmp->flushed_depth_texture &&
+ if (tex->is_depth && !si_can_sample_zs(tex, view->is_stencil_sampler)) {
+ if (!tex->flushed_depth_texture &&
!si_init_flushed_depth_texture(ctx, texture, NULL)) {
pipe_resource_reference(&view->base.texture, NULL);
FREE(view);
return NULL;
}
- assert(tmp->flushed_depth_texture);
+ assert(tex->flushed_depth_texture);
/* Override format for the case where the flushed texture
* contains only Z or only S.
*/
- if (tmp->flushed_depth_texture->buffer.b.b.format != tmp->buffer.b.b.format)
- pipe_format = tmp->flushed_depth_texture->buffer.b.b.format;
+ if (tex->flushed_depth_texture->buffer.b.b.format != tex->buffer.b.b.format)
+ pipe_format = tex->flushed_depth_texture->buffer.b.b.format;
- tmp = tmp->flushed_depth_texture;
+ tex = tex->flushed_depth_texture;
}
- surflevel = tmp->surface.u.legacy.level;
+ surflevel = tex->surface.u.legacy.level;
- if (tmp->db_compatible) {
+ if (tex->db_compatible) {
if (!view->is_stencil_sampler)
- pipe_format = tmp->db_render_format;
+ pipe_format = tex->db_render_format;
switch (pipe_format) {
case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
case PIPE_FORMAT_S8X24_UINT:
case PIPE_FORMAT_X32_S8X24_UINT:
pipe_format = PIPE_FORMAT_S8_UINT;
- surflevel = tmp->surface.u.legacy.stencil_level;
+ surflevel = tex->surface.u.legacy.stencil_level;
break;
default:;
}
state->u.tex.first_level,
state->format);
- si_make_texture_descriptor(sctx->screen, tmp, true,
+ si_make_texture_descriptor(sctx->screen, tex, true,
state->target, pipe_format, state_swizzle,
first_level, last_level,
state->u.tex.first_layer, last_layer,
static void si_emit_sample_mask(struct si_context *sctx)
{
- struct radeon_winsys_cs *cs = sctx->gfx_cs;
+ struct radeon_cmdbuf *cs = sctx->gfx_cs;
unsigned mask = sctx->sample_mask;
/* Needed for line and polygon smoothing as well as for the Polaris
si_pm4_set_reg(pm4, R_008A14_PA_CL_ENHANCE, S_008A14_NUM_CLIP_SEQ(3) |
S_008A14_CLIP_VTX_REORDER_ENA(1));
- si_pm4_set_reg(pm4, R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210);
- si_pm4_set_reg(pm4, R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98);
-
if (!has_clear_state)
si_pm4_set_reg(pm4, R_02882C_PA_SU_PRIM_FILTER_CNTL, 0);
}
if (!has_clear_state) {
- si_pm4_set_reg(pm4, R_02820C_PA_SC_CLIPRECT_RULE, 0xFFFF);
si_pm4_set_reg(pm4, R_028230_PA_SC_EDGERULE,
S_028230_ER_TRI(0xA) |
S_028230_ER_POINT(0xA) |
switch (sctx->family) {
case CHIP_VEGA10:
case CHIP_VEGA12:
+ case CHIP_VEGA20:
pc_lines = 4096;
break;
case CHIP_RAVEN: