gallium: remove PIPE_CAP_TEXTURE_SHADOW_MAP
[mesa.git] / src / gallium / drivers / radeonsi / si_state.c
index 46c45eae99cc5560c0acecab38a7ccf9379993eb..f88bf29e2020edb58c90c6d7bacf62a5947f991b 100644 (file)
@@ -120,8 +120,25 @@ static void si_emit_cb_render_state(struct si_context *sctx, struct r600_atom *a
                radeon_emit(cs, EVENT_TYPE(V_028A90_FLUSH_DFSM) | EVENT_INDEX(0));
        }
 
+       if (sctx->b.chip_class >= VI) {
+               /* DCC MSAA workaround for blending.
+                * Alternatively, we can set CB_COLORi_DCC_CONTROL.OVERWRITE_-
+                * COMBINER_DISABLE, but that would be more complicated.
+                */
+               bool oc_disable = (sctx->b.chip_class == VI ||
+                                  sctx->b.chip_class == GFX9) &&
+                                 blend &&
+                                 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));
+       }
+
        /* RB+ register settings. */
-       if (sctx->screen->b.rbplus_allowed) {
+       if (sctx->screen->rbplus_allowed) {
                unsigned spi_shader_col_format =
                        sctx->ps_shader.cso ?
                        sctx->ps_shader.current->key.part.ps.epilog.spi_shader_col_format : 0;
@@ -249,7 +266,7 @@ static void si_emit_cb_render_state(struct si_context *sctx, struct r600_atom *a
                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 */
-       } else if (sctx->screen->b.has_rbplus) {
+       } else if (sctx->screen->has_rbplus) {
                radeon_set_context_reg_seq(cs, R_028754_SX_PS_DOWNCONVERT, 3);
                radeon_emit(cs, 0);     /* R_028754_SX_PS_DOWNCONVERT */
                radeon_emit(cs, 0);     /* R_028758_SX_BLEND_OPT_EPSILON */
@@ -612,7 +629,7 @@ static void *si_create_blend_state_mode(struct pipe_context *ctx,
                color_control |= S_028808_MODE(V_028808_CB_DISABLE);
        }
 
-       if (sctx->screen->b.has_rbplus) {
+       if (sctx->screen->has_rbplus) {
                /* Disable RB+ blend optimizations for dual source blending.
                 * Vulkan does this.
                 */
@@ -653,13 +670,16 @@ static void si_bind_blend_state(struct pipe_context *ctx, void *state)
        if (!state)
                return;
 
+       si_pm4_bind_state(sctx, blend, state);
+
        if (!old_blend ||
-            old_blend->cb_target_mask != blend->cb_target_mask ||
-            old_blend->dual_src_blend != blend->dual_src_blend)
+           old_blend->cb_target_mask != blend->cb_target_mask ||
+           old_blend->dual_src_blend != blend->dual_src_blend ||
+           (old_blend->blend_enable_4bit != blend->blend_enable_4bit &&
+            sctx->framebuffer.nr_samples >= 2 &&
+            sctx->screen->dcc_msaa_allowed))
                si_mark_atom_dirty(sctx, &sctx->cb_render_state);
 
-       si_pm4_bind_state(sctx, blend, state);
-
        if (!old_blend ||
            old_blend->cb_target_mask != blend->cb_target_mask ||
            old_blend->alpha_to_coverage != blend->alpha_to_coverage ||
@@ -876,7 +896,7 @@ static void *si_create_rs_state(struct pipe_context *ctx,
 
        si_pm4_set_reg(pm4, R_0286D4_SPI_INTERP_CONTROL_0,
                S_0286D4_FLAT_SHADE_ENA(1) |
-               S_0286D4_PNT_SPRITE_ENA(1) |
+               S_0286D4_PNT_SPRITE_ENA(state->point_quad_rasterization) |
                S_0286D4_PNT_SPRITE_OVRD_X(V_0286D4_SPI_PNT_SPRITE_SEL_S) |
                S_0286D4_PNT_SPRITE_OVRD_Y(V_0286D4_SPI_PNT_SPRITE_SEL_T) |
                S_0286D4_PNT_SPRITE_OVRD_Z(V_0286D4_SPI_PNT_SPRITE_SEL_0) |
@@ -910,7 +930,7 @@ static void *si_create_rs_state(struct pipe_context *ctx,
                                            state->poly_smooth ||
                                            state->line_smooth) |
                       S_028A48_VPORT_SCISSOR_ENABLE(1) |
-                      S_028A48_ALTERNATE_RBS_PER_TILE(sscreen->b.chip_class >= GFX9));
+                      S_028A48_ALTERNATE_RBS_PER_TILE(sscreen->info.chip_class >= GFX9));
 
        si_pm4_set_reg(pm4, R_028BE4_PA_SU_VTX_CNTL,
                       S_028BE4_PIX_CENTER(state->half_pixel_center) |
@@ -1313,11 +1333,11 @@ static void si_set_active_query_state(struct pipe_context *ctx, boolean enable)
 
        /* Pipeline stat & streamout queries. */
        if (enable) {
-               sctx->b.flags &= ~R600_CONTEXT_STOP_PIPELINE_STATS;
-               sctx->b.flags |= R600_CONTEXT_START_PIPELINE_STATS;
+               sctx->b.flags &= ~SI_CONTEXT_STOP_PIPELINE_STATS;
+               sctx->b.flags |= SI_CONTEXT_START_PIPELINE_STATS;
        } else {
-               sctx->b.flags &= ~R600_CONTEXT_START_PIPELINE_STATS;
-               sctx->b.flags |= R600_CONTEXT_STOP_PIPELINE_STATS;
+               sctx->b.flags &= ~SI_CONTEXT_START_PIPELINE_STATS;
+               sctx->b.flags |= SI_CONTEXT_STOP_PIPELINE_STATS;
        }
 
        /* Occlusion queries. */
@@ -1421,8 +1441,8 @@ static void si_emit_db_render_state(struct si_context *sctx, struct r600_atom *s
        if (!rs || !rs->multisample_enable)
                db_shader_control &= C_02880C_MASK_EXPORT_ENABLE;
 
-       if (sctx->screen->b.has_rbplus &&
-           !sctx->screen->b.rbplus_allowed)
+       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,
@@ -1583,9 +1603,9 @@ static uint32_t si_translate_texformat(struct pipe_screen *screen,
                                       int first_non_void)
 {
        struct si_screen *sscreen = (struct si_screen*)screen;
-       bool enable_compressed_formats = (sscreen->b.info.drm_major == 2 &&
-                                         sscreen->b.info.drm_minor >= 31) ||
-                                        sscreen->b.info.drm_major == 3;
+       bool enable_compressed_formats = (sscreen->info.drm_major == 2 &&
+                                         sscreen->info.drm_minor >= 31) ||
+                                        sscreen->info.drm_major == 3;
        bool uniform = true;
        int i;
 
@@ -1654,8 +1674,8 @@ static uint32_t si_translate_texformat(struct pipe_screen *screen,
        }
 
        if (desc->layout == UTIL_FORMAT_LAYOUT_ETC &&
-           (sscreen->b.family == CHIP_STONEY ||
-            sscreen->b.chip_class >= GFX9)) {
+           (sscreen->info.family == CHIP_STONEY ||
+            sscreen->info.chip_class >= GFX9)) {
                switch (format) {
                case PIPE_FORMAT_ETC1_RGB8:
                case PIPE_FORMAT_ETC2_RGB8:
@@ -1910,7 +1930,7 @@ static unsigned si_tex_dim(struct si_screen *sscreen, struct r600_texture *rtex,
        /* GFX9 allocates 1D textures as 2D. */
        if ((res_target == PIPE_TEXTURE_1D ||
             res_target == PIPE_TEXTURE_1D_ARRAY) &&
-           sscreen->b.chip_class >= GFX9 &&
+           sscreen->info.chip_class >= GFX9 &&
            rtex->surface.u.gfx9.resource_type == RADEON_RESOURCE_2D) {
                if (res_target == PIPE_TEXTURE_1D)
                        res_target = PIPE_TEXTURE_2D;
@@ -2327,15 +2347,12 @@ static void si_initialize_color_surface(struct si_context *sctx,
                                        struct r600_surface *surf)
 {
        struct r600_texture *rtex = (struct r600_texture*)surf->base.texture;
-       unsigned color_info, color_attrib, color_view;
+       unsigned color_info, color_attrib;
        unsigned format, swap, ntype, endian;
        const struct util_format_description *desc;
        int firstchan;
        unsigned blend_clamp = 0, blend_bypass = 0;
 
-       color_view = S_028C6C_SLICE_START(surf->base.u.tex.first_layer) |
-                    S_028C6C_SLICE_MAX(surf->base.u.tex.last_layer);
-
        desc = util_format_description(surf->base.format);
        for (firstchan = 0; firstchan < 4; firstchan++) {
                if (desc->channel[firstchan].type != UTIL_FORMAT_TYPE_VOID) {
@@ -2432,41 +2449,53 @@ static void si_initialize_color_surface(struct si_context *sctx,
                }
        }
 
-       surf->cb_color_view = color_view;
-       surf->cb_color_info = color_info;
-       surf->cb_color_attrib = color_attrib;
-
        if (sctx->b.chip_class >= VI) {
-               unsigned max_uncompressed_block_size = 2;
+               unsigned max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_256B;
+               unsigned min_compressed_block_size = V_028C78_MIN_BLOCK_SIZE_32B;
+
+               /* amdvlk: [min-compressed-block-size] should be set to 32 for dGPU and
+                  64 for APU because all of our APUs to date use DIMMs which have
+                  a request granularity size of 64B while all other chips have a
+                  32B request size */
+               if (!sctx->screen->info.has_dedicated_vram)
+                       min_compressed_block_size = V_028C78_MIN_BLOCK_SIZE_64B;
 
                if (rtex->resource.b.b.nr_samples > 1) {
                        if (rtex->surface.bpe == 1)
-                               max_uncompressed_block_size = 0;
+                               max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_64B;
                        else if (rtex->surface.bpe == 2)
-                               max_uncompressed_block_size = 1;
+                               max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_128B;
                }
 
                surf->cb_dcc_control = S_028C78_MAX_UNCOMPRESSED_BLOCK_SIZE(max_uncompressed_block_size) |
+                                      S_028C78_MIN_COMPRESSED_BLOCK_SIZE(min_compressed_block_size) |
                                       S_028C78_INDEPENDENT_64B_BLOCKS(1);
        }
 
        /* This must be set for fast clear to work without FMASK. */
        if (!rtex->fmask.size && sctx->b.chip_class == SI) {
                unsigned bankh = util_logbase2(rtex->surface.u.legacy.bankh);
-               surf->cb_color_attrib |= S_028C74_FMASK_BANK_HEIGHT(bankh);
+               color_attrib |= S_028C74_FMASK_BANK_HEIGHT(bankh);
        }
 
+       unsigned color_view = S_028C6C_SLICE_START(surf->base.u.tex.first_layer) |
+                             S_028C6C_SLICE_MAX(surf->base.u.tex.last_layer);
+
        if (sctx->b.chip_class >= GFX9) {
                unsigned mip0_depth = util_max_layer(&rtex->resource.b.b, 0);
 
-               surf->cb_color_view |= S_028C6C_MIP_LEVEL(surf->base.u.tex.level);
-               surf->cb_color_attrib |= S_028C74_MIP0_DEPTH(mip0_depth) |
-                                        S_028C74_RESOURCE_TYPE(rtex->surface.u.gfx9.resource_type);
+               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);
                surf->cb_color_attrib2 = S_028C68_MIP0_WIDTH(surf->width0 - 1) |
                                         S_028C68_MIP0_HEIGHT(surf->height0 - 1) |
                                         S_028C68_MAX_MIP(rtex->resource.b.b.last_level);
        }
 
+       surf->cb_color_view = color_view;
+       surf->cb_color_info = color_info;
+       surf->cb_color_attrib = color_attrib;
+
        /* Determine pixel shader export format */
        si_choose_spi_color_formats(surf, format, swap, ntype, rtex->is_depth);
 
@@ -2511,7 +2540,7 @@ static void si_init_depth_surface(struct si_context *sctx,
                surf->db_depth_size = S_02801C_X_MAX(rtex->resource.b.b.width0 - 1) |
                                      S_02801C_Y_MAX(rtex->resource.b.b.height0 - 1);
 
-               if (r600_htile_enabled(rtex, level)) {
+               if (si_htile_enabled(rtex, level)) {
                        z_info |= S_028038_TILE_SURFACE_ENABLE(1) |
                                  S_028038_ALLOW_EXPCLEAR(1);
 
@@ -2560,7 +2589,7 @@ static void si_init_depth_surface(struct si_context *sctx,
                surf->db_depth_info = S_02803C_ADDR5_SWIZZLE_MASK(!rtex->tc_compatible_htile);
 
                if (sctx->b.chip_class >= CIK) {
-                       struct radeon_info *info = &sctx->screen->b.info;
+                       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;
@@ -2589,7 +2618,7 @@ static void si_init_depth_surface(struct si_context *sctx,
                surf->db_depth_slice = S_02805C_SLICE_TILE_MAX((levelinfo->nblk_x *
                                                                levelinfo->nblk_y) / 64 - 1);
 
-               if (r600_htile_enabled(rtex, level)) {
+               if (si_htile_enabled(rtex, level)) {
                        z_info |= S_028040_TILE_SURFACE_ENABLE(1) |
                                  S_028040_ALLOW_EXPCLEAR(1);
 
@@ -2841,7 +2870,7 @@ static void si_set_framebuffer_state(struct pipe_context *ctx,
                if (vi_dcc_enabled(rtex, surf->base.u.tex.level))
                        sctx->framebuffer.CB_has_shader_readable_metadata = true;
 
-               r600_context_add_resource_size(ctx, surf->base.texture);
+               si_context_add_resource_size(ctx, surf->base.texture);
 
                p_atomic_inc(&rtex->framebuffers_bound);
 
@@ -2865,7 +2894,7 @@ static void si_set_framebuffer_state(struct pipe_context *ctx,
                if (vi_tc_compat_htile_enabled(zstex, surf->base.u.tex.level))
                        sctx->framebuffer.DB_has_shader_readable_metadata = true;
 
-               r600_context_add_resource_size(ctx, surf->base.texture);
+               si_context_add_resource_size(ctx, surf->base.texture);
        }
 
        si_update_poly_offset_state(sctx);
@@ -3290,7 +3319,7 @@ static bool si_out_of_order_rasterization(struct si_context *sctx)
 static void si_emit_msaa_config(struct si_context *sctx, struct r600_atom *atom)
 {
        struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
-       unsigned num_tile_pipes = sctx->screen->b.info.num_tile_pipes;
+       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;
        bool out_of_order_rast = si_out_of_order_rasterization(sctx);
@@ -3417,8 +3446,8 @@ si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
        desc = util_format_description(format);
        first_non_void = util_format_get_first_non_void_channel(format);
        stride = desc->block.bits / 8;
-       num_format = si_translate_buffer_numformat(&screen->b.b, desc, first_non_void);
-       data_format = si_translate_buffer_dataformat(&screen->b.b, desc, first_non_void);
+       num_format = si_translate_buffer_numformat(&screen->b, desc, first_non_void);
+       data_format = si_translate_buffer_dataformat(&screen->b, desc, first_non_void);
 
        num_records = size / stride;
        num_records = MIN2(num_records, (buf->b.b.width0 - offset) / stride);
@@ -3446,7 +3475,7 @@ si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
         * - For VMEM and inst.IDXEN == 0 or STRIDE == 0, it's in byte units.
         * - For VMEM and inst.IDXEN == 1 and STRIDE != 0, it's in units of STRIDE.
         */
-       if (screen->b.chip_class >= GFX9)
+       if (screen->info.chip_class >= GFX9)
                /* When vindex == 0, LLVM sets IDXEN = 0, thus changing units
                 * from STRIDE to bytes. This works around it by setting
                 * NUM_RECORDS to at least the size of one element, so that
@@ -3456,7 +3485,7 @@ si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
                 *       IDXEN is enforced?
                 */
                num_records = num_records ? MAX2(num_records, stride) : 0;
-       else if (screen->b.chip_class == VI)
+       else if (screen->info.chip_class == VI)
                num_records *= stride;
 
        state[4] = 0;
@@ -3618,13 +3647,13 @@ si_make_texture_descriptor(struct si_screen *screen,
                }
        }
 
-       data_format = si_translate_texformat(&screen->b.b, pipe_format, desc, first_non_void);
+       data_format = si_translate_texformat(&screen->b, pipe_format, desc, first_non_void);
        if (data_format == ~0) {
                data_format = 0;
        }
 
        /* S8 with Z32 HTILE needs a special format. */
-       if (screen->b.chip_class >= GFX9 &&
+       if (screen->info.chip_class >= GFX9 &&
            pipe_format == PIPE_FORMAT_S8_UINT &&
            tex->tc_compatible_htile)
                data_format = V_008F14_IMG_DATA_FORMAT_S8_32;
@@ -3632,7 +3661,7 @@ si_make_texture_descriptor(struct si_screen *screen,
        if (!sampler &&
            (res->target == PIPE_TEXTURE_CUBE ||
             res->target == PIPE_TEXTURE_CUBE_ARRAY ||
-            (screen->b.chip_class <= VI &&
+            (screen->info.chip_class <= VI &&
              res->target == PIPE_TEXTURE_3D))) {
                /* For the purpose of shader images, treat cube maps and 3D
                 * textures as 2D arrays. For 3D textures, the address
@@ -3677,7 +3706,7 @@ si_make_texture_descriptor(struct si_screen *screen,
        state[6] = 0;
        state[7] = 0;
 
-       if (screen->b.chip_class >= GFX9) {
+       if (screen->info.chip_class >= GFX9) {
                unsigned bc_swizzle = gfx9_border_color_swizzle(desc->swizzle);
 
                /* Depth is the the last accessible layer on Gfx9.
@@ -3706,7 +3735,7 @@ si_make_texture_descriptor(struct si_screen *screen,
                /* The last dword is unused by hw. The shader uses it to clear
                 * bits in the first dword of sampler state.
                 */
-               if (screen->b.chip_class <= CIK && res->nr_samples <= 1) {
+               if (screen->info.chip_class <= CIK && res->nr_samples <= 1) {
                        if (first_level == last_level)
                                state[7] = C_008F30_MAX_ANISO_RATIO;
                        else
@@ -3720,7 +3749,7 @@ si_make_texture_descriptor(struct si_screen *screen,
 
                va = tex->resource.gpu_address + tex->fmask.offset;
 
-               if (screen->b.chip_class >= GFX9) {
+               if (screen->info.chip_class >= GFX9) {
                        data_format = V_008F14_IMG_DATA_FORMAT_FMASK;
                        switch (res->nr_samples) {
                        case 2:
@@ -3768,7 +3797,7 @@ si_make_texture_descriptor(struct si_screen *screen,
                fmask_state[6] = 0;
                fmask_state[7] = 0;
 
-               if (screen->b.chip_class >= GFX9) {
+               if (screen->info.chip_class >= GFX9) {
                        fmask_state[3] |= S_008F1C_SW_MODE(tex->surface.u.gfx9.fmask.swizzle_mode);
                        fmask_state[4] |= S_008F20_DEPTH(last_layer) |
                                          S_008F20_PITCH_GFX9(tex->surface.u.gfx9.fmask.epitch);
@@ -3873,7 +3902,7 @@ si_create_sampler_view_custom(struct pipe_context *ctx,
        pipe_format = state->format;
 
        /* Depth/stencil texturing sometimes needs separate texture. */
-       if (tmp->is_depth && !r600_can_sample_zs(tmp, view->is_stencil_sampler)) {
+       if (tmp->is_depth && !si_can_sample_zs(tmp, view->is_stencil_sampler)) {
                if (!tmp->flushed_depth_texture &&
                    !si_init_flushed_depth_texture(ctx, texture, NULL)) {
                        pipe_resource_reference(&view->base.texture, NULL);
@@ -4033,15 +4062,43 @@ do { \
               S_008F3C_BORDER_COLOR_TYPE(V_008F3C_SQ_TEX_BORDER_COLOR_REGISTER);
 }
 
+static inline int S_FIXED(float value, unsigned frac_bits)
+{
+       return value * (1 << frac_bits);
+}
+
+static inline unsigned si_tex_filter(unsigned filter, unsigned max_aniso)
+{
+       if (filter == PIPE_TEX_FILTER_LINEAR)
+               return max_aniso > 1 ? V_008F38_SQ_TEX_XY_FILTER_ANISO_BILINEAR
+                                    : V_008F38_SQ_TEX_XY_FILTER_BILINEAR;
+       else
+               return max_aniso > 1 ? V_008F38_SQ_TEX_XY_FILTER_ANISO_POINT
+                                    : V_008F38_SQ_TEX_XY_FILTER_POINT;
+}
+
+static inline unsigned si_tex_aniso_filter(unsigned filter)
+{
+       if (filter < 2)
+               return 0;
+       if (filter < 4)
+               return 1;
+       if (filter < 8)
+               return 2;
+       if (filter < 16)
+               return 3;
+       return 4;
+}
+
 static void *si_create_sampler_state(struct pipe_context *ctx,
                                     const struct pipe_sampler_state *state)
 {
        struct si_context *sctx = (struct si_context *)ctx;
-       struct r600_common_screen *rscreen = sctx->b.screen;
+       struct si_screen *sscreen = sctx->screen;
        struct si_sampler_state *rstate = CALLOC_STRUCT(si_sampler_state);
-       unsigned max_aniso = rscreen->force_aniso >= 0 ? rscreen->force_aniso
+       unsigned max_aniso = sscreen->force_aniso >= 0 ? sscreen->force_aniso
                                                       : state->max_anisotropy;
-       unsigned max_aniso_ratio = r600_tex_aniso_filter(max_aniso);
+       unsigned max_aniso_ratio = si_tex_aniso_filter(max_aniso);
        union pipe_color_union clamped_border_color;
 
        if (!rstate) {
@@ -4065,8 +4122,8 @@ static void *si_create_sampler_state(struct pipe_context *ctx,
                          S_008F34_MAX_LOD(S_FIXED(CLAMP(state->max_lod, 0, 15), 8)) |
                          S_008F34_PERF_MIP(max_aniso_ratio ? max_aniso_ratio + 6 : 0));
        rstate->val[2] = (S_008F38_LOD_BIAS(S_FIXED(CLAMP(state->lod_bias, -16, 16), 8)) |
-                         S_008F38_XY_MAG_FILTER(eg_tex_filter(state->mag_img_filter, max_aniso)) |
-                         S_008F38_XY_MIN_FILTER(eg_tex_filter(state->min_img_filter, max_aniso)) |
+                         S_008F38_XY_MAG_FILTER(si_tex_filter(state->mag_img_filter, max_aniso)) |
+                         S_008F38_XY_MIN_FILTER(si_tex_filter(state->min_img_filter, max_aniso)) |
                          S_008F38_MIP_FILTER(si_tex_mipfilter(state->min_mip_filter)) |
                          S_008F38_MIP_POINT_PRECLAMP(0) |
                          S_008F38_DISABLE_LSB_CEIL(sctx->b.chip_class <= VI) |
@@ -4200,8 +4257,8 @@ static void *si_create_vertex_elements(struct pipe_context *ctx,
                 * chips are VI and older except Stoney (GFX8.1).
                 */
                if (data_format == V_008F0C_BUF_DATA_FORMAT_2_10_10_10 &&
-                   sscreen->b.chip_class <= VI &&
-                   sscreen->b.family != CHIP_STONEY) {
+                   sscreen->info.chip_class <= VI &&
+                   sscreen->info.family != CHIP_STONEY) {
                        if (num_format == V_008F0C_BUF_NUM_FORMAT_SNORM) {
                                v->fix_fetch[i] = SI_FIX_FETCH_A2_SNORM;
                        } else if (num_format == V_008F0C_BUF_NUM_FORMAT_SSCALED) {
@@ -4345,7 +4402,7 @@ static void si_set_vertex_buffers(struct pipe_context *ctx,
                        pipe_resource_reference(&dsti->buffer.resource, buf);
                        dsti->buffer_offset = src->buffer_offset;
                        dsti->stride = src->stride;
-                       r600_context_add_resource_size(ctx, buf);
+                       si_context_add_resource_size(ctx, buf);
                        if (buf)
                                r600_resource(buf)->bind_history |= PIPE_BIND_VERTEX_BUFFER;
                }
@@ -4427,7 +4484,7 @@ static void si_memory_barrier(struct pipe_context *ctx, unsigned flags)
                /* Indices are read through TC L2 since VI.
                 * L1 isn't used.
                 */
-               if (sctx->screen->b.chip_class <= CIK)
+               if (sctx->screen->info.chip_class <= CIK)
                        sctx->b.flags |= SI_CONTEXT_WRITEBACK_GLOBAL_L2;
        }
 
@@ -4444,7 +4501,7 @@ static void si_memory_barrier(struct pipe_context *ctx, unsigned flags)
        }
 
        /* Indirect buffers use TC L2 on GFX9, but not older hw. */
-       if (sctx->screen->b.chip_class <= VI &&
+       if (sctx->screen->info.chip_class <= VI &&
            flags & PIPE_BARRIER_INDIRECT_BUFFER)
                sctx->b.flags |= SI_CONTEXT_WRITEBACK_GLOBAL_L2;
 }
@@ -4539,121 +4596,23 @@ void si_init_state_functions(struct si_context *sctx)
        si_init_config(sctx);
 }
 
-static uint32_t si_get_bo_metadata_word1(struct r600_common_screen *rscreen)
-{
-       return (ATI_VENDOR_ID << 16) | rscreen->info.pci_id;
-}
-
-static void si_query_opaque_metadata(struct r600_common_screen *rscreen,
-                                    struct r600_texture *rtex,
-                                    struct radeon_bo_metadata *md)
-{
-       struct si_screen *sscreen = (struct si_screen*)rscreen;
-       struct pipe_resource *res = &rtex->resource.b.b;
-       static const unsigned char swizzle[] = {
-               PIPE_SWIZZLE_X,
-               PIPE_SWIZZLE_Y,
-               PIPE_SWIZZLE_Z,
-               PIPE_SWIZZLE_W
-       };
-       uint32_t desc[8], i;
-       bool is_array = util_resource_is_array_texture(res);
-
-       /* DRM 2.x.x doesn't support this. */
-       if (rscreen->info.drm_major != 3)
-               return;
-
-       assert(rtex->dcc_separate_buffer == NULL);
-       assert(rtex->fmask.size == 0);
-
-       /* Metadata image format format version 1:
-        * [0] = 1 (metadata format identifier)
-        * [1] = (VENDOR_ID << 16) | PCI_ID
-        * [2:9] = image descriptor for the whole resource
-        *         [2] is always 0, because the base address is cleared
-        *         [9] is the DCC offset bits [39:8] from the beginning of
-        *             the buffer
-        * [10:10+LAST_LEVEL] = mipmap level offset bits [39:8] for each level
-        */
-
-       md->metadata[0] = 1; /* metadata image format version 1 */
-
-       /* TILE_MODE_INDEX is ambiguous without a PCI ID. */
-       md->metadata[1] = si_get_bo_metadata_word1(rscreen);
-
-       si_make_texture_descriptor(sscreen, rtex, true,
-                                  res->target, res->format,
-                                  swizzle, 0, res->last_level, 0,
-                                  is_array ? res->array_size - 1 : 0,
-                                  res->width0, res->height0, res->depth0,
-                                  desc, NULL);
-
-       si_set_mutable_tex_desc_fields(sscreen, rtex, &rtex->surface.u.legacy.level[0],
-                                      0, 0, rtex->surface.blk_w, false, desc);
-
-       /* Clear the base address and set the relative DCC offset. */
-       desc[0] = 0;
-       desc[1] &= C_008F14_BASE_ADDRESS_HI;
-       desc[7] = rtex->dcc_offset >> 8;
-
-       /* Dwords [2:9] contain the image descriptor. */
-       memcpy(&md->metadata[2], desc, sizeof(desc));
-       md->size_metadata = 10 * 4;
-
-       /* Dwords [10:..] contain the mipmap level offsets. */
-       if (rscreen->chip_class <= VI) {
-               for (i = 0; i <= res->last_level; i++)
-                       md->metadata[10+i] = rtex->surface.u.legacy.level[i].offset >> 8;
-
-               md->size_metadata += (1 + res->last_level) * 4;
-       }
-}
-
-static void si_apply_opaque_metadata(struct r600_common_screen *rscreen,
-                                    struct r600_texture *rtex,
-                                    struct radeon_bo_metadata *md)
-{
-       uint32_t *desc = &md->metadata[2];
-
-       if (rscreen->chip_class < VI)
-               return;
-
-       /* Return if DCC is enabled. The texture should be set up with it
-        * already.
-        */
-       if (md->size_metadata >= 11 * 4 &&
-           md->metadata[0] != 0 &&
-           md->metadata[1] == si_get_bo_metadata_word1(rscreen) &&
-           G_008F28_COMPRESSION_EN(desc[6])) {
-               rtex->dcc_offset = (uint64_t)desc[7] << 8;
-               return;
-       }
-
-       /* Disable DCC. These are always set by texture_from_handle and must
-        * be cleared here.
-        */
-       rtex->dcc_offset = 0;
-}
-
 void si_init_screen_state_functions(struct si_screen *sscreen)
 {
-       sscreen->b.b.is_format_supported = si_is_format_supported;
-       sscreen->b.query_opaque_metadata = si_query_opaque_metadata;
-       sscreen->b.apply_opaque_metadata = si_apply_opaque_metadata;
+       sscreen->b.is_format_supported = si_is_format_supported;
 }
 
 static void si_set_grbm_gfx_index(struct si_context *sctx,
                                  struct si_pm4_state *pm4,  unsigned value)
 {
        unsigned reg = sctx->b.chip_class >= CIK ? R_030800_GRBM_GFX_INDEX :
-                                                  GRBM_GFX_INDEX;
+                                                  R_00802C_GRBM_GFX_INDEX;
        si_pm4_set_reg(pm4, reg, value);
 }
 
 static void si_set_grbm_gfx_index_se(struct si_context *sctx,
                                     struct si_pm4_state *pm4, unsigned se)
 {
-       assert(se == ~0 || se < sctx->screen->b.info.max_se);
+       assert(se == ~0 || se < sctx->screen->info.max_se);
        si_set_grbm_gfx_index(sctx, pm4,
                              (se == ~0 ? S_030800_SE_BROADCAST_WRITES(1) :
                                          S_030800_SE_INDEX(se)) |
@@ -4667,10 +4626,10 @@ si_write_harvested_raster_configs(struct si_context *sctx,
                                  unsigned raster_config,
                                  unsigned raster_config_1)
 {
-       unsigned sh_per_se = MAX2(sctx->screen->b.info.max_sh_per_se, 1);
-       unsigned num_se = MAX2(sctx->screen->b.info.max_se, 1);
-       unsigned rb_mask = sctx->screen->b.info.enabled_rb_mask;
-       unsigned num_rb = MIN2(sctx->screen->b.info.num_render_backends, 16);
+       unsigned sh_per_se = MAX2(sctx->screen->info.max_sh_per_se, 1);
+       unsigned num_se = MAX2(sctx->screen->info.max_se, 1);
+       unsigned rb_mask = sctx->screen->info.enabled_rb_mask;
+       unsigned num_rb = MIN2(sctx->screen->info.num_render_backends, 16);
        unsigned rb_per_pkr = MIN2(num_rb / num_se / sh_per_se, 2);
        unsigned rb_per_se = num_rb / num_se;
        unsigned se_mask[4];
@@ -4789,8 +4748,8 @@ si_write_harvested_raster_configs(struct si_context *sctx,
 static void si_set_raster_config(struct si_context *sctx, struct si_pm4_state *pm4)
 {
        struct si_screen *sscreen = sctx->screen;
-       unsigned num_rb = MIN2(sctx->screen->b.info.num_render_backends, 16);
-       unsigned rb_mask = sctx->screen->b.info.enabled_rb_mask;
+       unsigned num_rb = MIN2(sctx->screen->info.num_render_backends, 16);
+       unsigned rb_mask = sctx->screen->info.enabled_rb_mask;
        unsigned raster_config, raster_config_1;
 
        switch (sctx->b.family) {
@@ -4820,7 +4779,7 @@ static void si_set_raster_config(struct si_context *sctx, struct si_pm4_state *p
                raster_config_1 = 0x0000002e;
                break;
        case CHIP_FIJI:
-               if (sscreen->b.info.cik_macrotile_mode_array[0] == 0x000000e8) {
+               if (sscreen->info.cik_macrotile_mode_array[0] == 0x000000e8) {
                        /* old kernels with old tiling config */
                        raster_config = 0x16000012;
                        raster_config_1 = 0x0000002a;
@@ -4893,7 +4852,7 @@ static void si_init_config(struct si_context *sctx)
        struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
 
        /* Only SI can disable CLEAR_STATE for now. */
-       assert(has_clear_state || sscreen->b.chip_class == SI);
+       assert(has_clear_state || sscreen->info.chip_class == SI);
 
        if (!pm4)
                return;
@@ -5013,9 +4972,9 @@ static void si_init_config(struct si_context *sctx)
                               S_00B21C_CU_EN(0xffff) | S_00B21C_WAVE_LIMIT(0x3F));
 
                /* Compute LATE_ALLOC_VS.LIMIT. */
-               unsigned num_cu_per_sh = sscreen->b.info.num_good_compute_units /
-                                        (sscreen->b.info.max_se *
-                                         sscreen->b.info.max_sh_per_se);
+               unsigned num_cu_per_sh = sscreen->info.num_good_compute_units /
+                                        (sscreen->info.max_se *
+                                         sscreen->info.max_sh_per_se);
                unsigned late_alloc_limit; /* The limit is per SH. */
 
                if (sctx->b.family == CHIP_KABINI) {
@@ -5053,10 +5012,6 @@ static void si_init_config(struct si_context *sctx)
        if (sctx->b.chip_class >= VI) {
                unsigned vgt_tess_distribution;
 
-               si_pm4_set_reg(pm4, R_028424_CB_DCC_CONTROL,
-                              S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
-                              S_028424_OVERWRITE_COMBINER_WATERMARK(4));
-
                vgt_tess_distribution =
                        S_028B50_ACCUM_ISOLINE(32) |
                        S_028B50_ACCUM_TRI(11) |
@@ -5083,7 +5038,7 @@ static void si_init_config(struct si_context *sctx)
                      RADEON_PRIO_BORDER_COLORS);
 
        if (sctx->b.chip_class >= GFX9) {
-               unsigned num_se = sscreen->b.info.max_se;
+               unsigned num_se = sscreen->info.max_se;
                unsigned pc_lines = 0;
 
                switch (sctx->b.family) {