i965: Use ISL for emitting buffer surface states
[mesa.git] / src / mesa / drivers / dri / i965 / gen8_surface_state.c
index 72983f54ffdc4b50b3b17c8bf7213501b95e812a..1f86557620d105423b1cb03816c0032135cb548a 100644 (file)
 #include "main/mtypes.h"
 #include "main/samplerobj.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "program/prog_parameter.h"
+#include "program/prog_instruction.h"
 
 #include "intel_mipmap_tree.h"
 #include "intel_batchbuffer.h"
 #include "intel_tex.h"
 #include "intel_fbo.h"
 #include "intel_buffer_objects.h"
+#include "intel_image.h"
 
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
 #include "brw_wm.h"
+#include "isl/isl.h"
 
-static uint32_t
-surface_tiling_mode(uint32_t tiling)
+static uint32_t *
+gen8_allocate_surface_state(struct brw_context *brw,
+                            uint32_t *out_offset, int index)
 {
-   switch (tiling) {
-   case I915_TILING_X:
-      return GEN8_SURFACE_TILING_X;
-   case I915_TILING_Y:
-      return GEN8_SURFACE_TILING_Y;
-   default:
-      return GEN8_SURFACE_TILING_NONE;
-   }
-}
-
-static unsigned
-vertical_alignment(struct intel_mipmap_tree *mt)
-{
-   switch (mt->align_h) {
-   case 4:
-      return GEN8_SURFACE_VALIGN_4;
-   case 8:
-      return GEN8_SURFACE_VALIGN_8;
-   case 16:
-      return GEN8_SURFACE_VALIGN_16;
-   default:
-      assert(!"Unsupported vertical surface alignment.");
-      return GEN8_SURFACE_VALIGN_4;
-   }
-}
-
-static unsigned
-horizontal_alignment(struct intel_mipmap_tree *mt)
-{
-   switch (mt->align_w) {
-   case 4:
-      return GEN8_SURFACE_HALIGN_4;
-   case 8:
-      return GEN8_SURFACE_HALIGN_8;
-   case 16:
-      return GEN8_SURFACE_HALIGN_16;
-   default:
-      assert(!"Unsupported horizontal surface alignment.");
-      return GEN8_SURFACE_HALIGN_4;
-   }
-}
-
-static void
-gen8_emit_buffer_surface_state(struct brw_context *brw,
-                               uint32_t *out_offset,
-                               drm_intel_bo *bo,
-                               unsigned buffer_offset,
-                               unsigned surface_format,
-                               unsigned buffer_size,
-                               unsigned pitch,
-                               unsigned mocs,
-                               bool rw)
-{
-   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
-                                    13 * 4, 64, out_offset);
-   memset(surf, 0, 13 * 4);
-
-   surf[0] = BRW_SURFACE_BUFFER << BRW_SURFACE_TYPE_SHIFT |
-             surface_format << BRW_SURFACE_FORMAT_SHIFT |
-             BRW_SURFACE_RC_READ_WRITE;
-   surf[1] = SET_FIELD(mocs, GEN8_SURFACE_MOCS);
-
-   surf[2] = SET_FIELD((buffer_size - 1) & 0x7f, GEN7_SURFACE_WIDTH) |
-             SET_FIELD(((buffer_size - 1) >> 7) & 0x3fff, GEN7_SURFACE_HEIGHT);
-   surf[3] = SET_FIELD(((buffer_size - 1) >> 21) & 0x3f, BRW_SURFACE_DEPTH) |
-             (pitch - 1);
-   surf[7] = SET_FIELD(HSW_SCS_RED,   GEN7_SURFACE_SCS_R) |
-             SET_FIELD(HSW_SCS_GREEN, GEN7_SURFACE_SCS_G) |
-             SET_FIELD(HSW_SCS_BLUE,  GEN7_SURFACE_SCS_B) |
-             SET_FIELD(HSW_SCS_ALPHA, GEN7_SURFACE_SCS_A);
-   /* reloc */
-   *((uint64_t *) &surf[8]) = (bo ? bo->offset64 : 0) + buffer_offset;
-
-   /* Emit relocation to surface contents. */
-   if (bo) {
-      drm_intel_bo_emit_reloc(brw->batch.bo, *out_offset + 8 * 4,
-                              bo, buffer_offset, I915_GEM_DOMAIN_SAMPLER,
-                              rw ? I915_GEM_DOMAIN_SAMPLER : 0);
-   }
-}
-
-static void
-gen8_update_texture_surface(struct gl_context *ctx,
-                            unsigned unit,
-                            uint32_t *surf_offset,
-                            bool for_gather)
-{
-   struct brw_context *brw = brw_context(ctx);
-   struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
-   struct intel_texture_object *intelObj = intel_texture_object(tObj);
-   struct intel_mipmap_tree *mt = intelObj->mt;
-   struct gl_texture_image *firstImage = tObj->Image[0][tObj->BaseLevel];
-   struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
-   struct intel_mipmap_tree *aux_mt = NULL;
-   uint32_t aux_mode = 0;
-   mesa_format format = intelObj->_Format;
-
-   if (tObj->Target == GL_TEXTURE_BUFFER) {
-      brw_update_buffer_texture_surface(ctx, unit, surf_offset);
-      return;
-   }
-
-   if (tObj->StencilSampling && firstImage->_BaseFormat == GL_DEPTH_STENCIL) {
-      mt = mt->stencil_mt;
-      format = MESA_FORMAT_S_UINT8;
-   }
-
-   unsigned tiling_mode, pitch;
-   if (format == MESA_FORMAT_S_UINT8) {
-      tiling_mode = GEN8_SURFACE_TILING_W;
-      pitch = 2 * mt->pitch;
-   } else {
-      tiling_mode = surface_tiling_mode(mt->tiling);
-      pitch = mt->pitch;
-   }
-
-   if (mt->mcs_mt) {
-      aux_mt = mt->mcs_mt;
-      aux_mode = GEN8_SURFACE_AUX_MODE_MCS;
-   }
-
-   /* If this is a view with restricted NumLayers, then our effective depth
-    * is not just the miptree depth.
-    */
-   uint32_t effective_depth =
-      (tObj->Immutable && tObj->Target != GL_TEXTURE_3D) ? tObj->NumLayers
-                                                         : mt->logical_depth0;
-
-   uint32_t tex_format = translate_tex_format(brw, format, sampler->sRGBDecode);
-
-   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
-                                    13 * 4, 64, surf_offset);
-
-   surf[0] = translate_tex_target(tObj->Target) << BRW_SURFACE_TYPE_SHIFT |
-             tex_format << BRW_SURFACE_FORMAT_SHIFT |
-             vertical_alignment(mt) |
-             horizontal_alignment(mt) |
-             tiling_mode;
-
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP ||
-       tObj->Target == GL_TEXTURE_CUBE_MAP_ARRAY) {
-      surf[0] |= BRW_SURFACE_CUBEFACE_ENABLES;
-   }
-
-   if (mt->logical_depth0 > 1 && tObj->Target != GL_TEXTURE_3D)
-      surf[0] |= GEN8_SURFACE_IS_ARRAY;
-
-   surf[1] = SET_FIELD(BDW_MOCS_WB, GEN8_SURFACE_MOCS) | mt->qpitch >> 2;
-
-   surf[2] = SET_FIELD(mt->logical_width0 - 1, GEN7_SURFACE_WIDTH) |
-             SET_FIELD(mt->logical_height0 - 1, GEN7_SURFACE_HEIGHT);
-
-   surf[3] = SET_FIELD(effective_depth - 1, BRW_SURFACE_DEPTH) | (pitch - 1);
-
-   surf[4] = gen7_surface_msaa_bits(mt->num_samples, mt->msaa_layout) |
-             SET_FIELD(tObj->MinLayer, GEN7_SURFACE_MIN_ARRAY_ELEMENT) |
-             SET_FIELD(effective_depth - 1,
-                       GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT);
-
-   surf[5] = SET_FIELD(tObj->MinLevel + tObj->BaseLevel - mt->first_level,
-                       GEN7_SURFACE_MIN_LOD) |
-             (intelObj->_MaxLevel - tObj->BaseLevel); /* mip count */
-
-   if (aux_mt) {
-      surf[6] = SET_FIELD(mt->qpitch / 4, GEN8_SURFACE_AUX_QPITCH) |
-                SET_FIELD((aux_mt->pitch / 128) - 1, GEN8_SURFACE_AUX_PITCH) |
-                aux_mode;
-   } else {
-      surf[6] = 0;
-   }
-
-   /* Handling GL_ALPHA as a surface format override breaks 1.30+ style
-    * texturing functions that return a float, as our code generation always
-    * selects the .x channel (which would always be 0).
-    */
-   const bool alpha_depth = tObj->DepthMode == GL_ALPHA &&
-      (firstImage->_BaseFormat == GL_DEPTH_COMPONENT ||
-       firstImage->_BaseFormat == GL_DEPTH_STENCIL);
-
-   surf[7] = mt->fast_clear_color_value;
-
-   const int swizzle =
-      unlikely(alpha_depth) ? SWIZZLE_XYZW : brw_get_texture_swizzle(ctx, tObj);
-   surf[7] |=
-      SET_FIELD(brw_swizzle_to_scs(GET_SWZ(swizzle, 0), false), GEN7_SURFACE_SCS_R) |
-      SET_FIELD(brw_swizzle_to_scs(GET_SWZ(swizzle, 1), false), GEN7_SURFACE_SCS_G) |
-      SET_FIELD(brw_swizzle_to_scs(GET_SWZ(swizzle, 2), false), GEN7_SURFACE_SCS_B) |
-      SET_FIELD(brw_swizzle_to_scs(GET_SWZ(swizzle, 3), false), GEN7_SURFACE_SCS_A);
-
-   *((uint64_t *) &surf[8]) = mt->bo->offset64 + mt->offset; /* reloc */
-
-   if (aux_mt) {
-      *((uint64_t *) &surf[10]) = aux_mt->bo->offset64;
-      drm_intel_bo_emit_reloc(brw->batch.bo, *surf_offset + 10 * 4,
-                              aux_mt->bo, 0,
-                              I915_GEM_DOMAIN_SAMPLER, 0);
-   } else {
-      surf[10] = 0;
-      surf[11] = 0;
-   }
-   surf[12] = 0;
-
-   /* Emit relocation to surface contents */
-   drm_intel_bo_emit_reloc(brw->batch.bo,
-                           *surf_offset + 8 * 4,
-                           mt->bo,
-                           mt->offset,
-                           I915_GEM_DOMAIN_SAMPLER, 0);
-}
-
-static void
-gen8_create_raw_surface(struct brw_context *brw, drm_intel_bo *bo,
-                        uint32_t offset, uint32_t size,
-                        uint32_t *out_offset, bool rw)
-{
-   gen8_emit_buffer_surface_state(brw,
-                                  out_offset,
-                                  bo,
-                                  offset,
-                                  BRW_SURFACEFORMAT_RAW,
-                                  size,
-                                  1,
-                                  0 /* mocs */,
-                                  true /* rw */);
+   int dwords = brw->gen >= 9 ? 16 : 13;
+   uint32_t *surf = __brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
+                                      dwords * 4, 64, index, out_offset);
+   memset(surf, 0, dwords * 4);
+   return surf;
 }
 
 /**
- * Create the constant buffer surface.  Vertex/fragment shader constants will be
- * read from this buffer with Data Port Read instructions/messages.
+ * Creates a null surface.
+ *
+ * This is used when the shader doesn't write to any color output.  An FB
+ * write to target 0 will still be emitted, because that's how the thread is
+ * terminated (and computed depth is returned), so we need to have the
+ * hardware discard the target 0 color output..
  */
 static void
-gen8_update_null_renderbuffer_surface(struct brw_context *brw, unsigned unit)
+gen8_emit_null_surface_state(struct brw_context *brw,
+                             unsigned width,
+                             unsigned height,
+                             unsigned samples,
+                             uint32_t *out_offset)
 {
-   struct gl_context *ctx = &brw->ctx;
-
-   /* _NEW_BUFFERS */
-   const struct gl_framebuffer *fb = ctx->DrawBuffer;
-   uint32_t surf_index =
-      brw->wm.prog_data->binding_table.render_target_start + unit;
-
-   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 13 * 4, 64,
-                                    &brw->wm.base.surf_offset[surf_index]);
-   memset(surf, 0, 13 * 4);
+   uint32_t *surf = gen8_allocate_surface_state(brw, out_offset, -1);
 
    surf[0] = BRW_SURFACE_NULL << BRW_SURFACE_TYPE_SHIFT |
              BRW_SURFACEFORMAT_B8G8R8A8_UNORM << BRW_SURFACE_FORMAT_SHIFT |
              GEN8_SURFACE_TILING_Y;
-   surf[2] = SET_FIELD(fb->Width - 1, GEN7_SURFACE_WIDTH) |
-             SET_FIELD(fb->Height - 1, GEN7_SURFACE_HEIGHT);
-}
-
-/**
- * Sets up a surface state structure to point at the given region.
- * While it is only used for the front/back buffer currently, it should be
- * usable for further buffers when doing ARB_draw_buffer support.
- */
-static void
-gen8_update_renderbuffer_surface(struct brw_context *brw,
-                                 struct gl_renderbuffer *rb,
-                                 bool layered,
-                                 unsigned unit)
-{
-   struct gl_context *ctx = &brw->ctx;
-   struct intel_renderbuffer *irb = intel_renderbuffer(rb);
-   struct intel_mipmap_tree *mt = irb->mt;
-   struct intel_mipmap_tree *aux_mt = NULL;
-   uint32_t aux_mode = 0;
-   unsigned width = mt->logical_width0;
-   unsigned height = mt->logical_height0;
-   unsigned pitch = mt->pitch;
-   uint32_t tiling = mt->tiling;
-   uint32_t format = 0;
-   uint32_t surf_type;
-   bool is_array = false;
-   int depth = MAX2(irb->layer_count, 1);
-   const int min_array_element = (mt->format == MESA_FORMAT_S_UINT8) ?
-      irb->mt_layer : (irb->mt_layer / MAX2(mt->num_samples, 1));
-   GLenum gl_target =
-      rb->TexImage ? rb->TexImage->TexObject->Target : GL_TEXTURE_2D;
-
-   uint32_t surf_index =
-      brw->wm.prog_data->binding_table.render_target_start + unit;
-
-   intel_miptree_used_for_rendering(mt);
-
-   switch (gl_target) {
-   case GL_TEXTURE_CUBE_MAP_ARRAY:
-   case GL_TEXTURE_CUBE_MAP:
-      surf_type = BRW_SURFACE_2D;
-      is_array = true;
-      depth *= 6;
-      break;
-   case GL_TEXTURE_3D:
-      depth = MAX2(irb->mt->logical_depth0, 1);
-      /* fallthrough */
-   default:
-      surf_type = translate_tex_target(gl_target);
-      is_array = _mesa_tex_target_is_array(gl_target);
-      break;
-   }
-
-   /* _NEW_BUFFERS */
-   /* Render targets can't use IMS layout. Stencil in turn gets configured as
-    * single sampled and indexed manually by the program.
-    */
-   if (mt->format == MESA_FORMAT_S_UINT8) {
-      brw_configure_w_tiled(mt, true, &width, &height, &pitch,
-                            &tiling, &format);
-   } else {
-      assert(mt->msaa_layout != INTEL_MSAA_LAYOUT_IMS);
-      assert(brw_render_target_supported(brw, rb));
-      mesa_format rb_format = _mesa_get_render_format(ctx,
-                                                      intel_rb_format(irb));
-      format = brw->render_target_format[rb_format];
-      if (unlikely(!brw->format_supported_as_render_target[rb_format]))
-         _mesa_problem(ctx, "%s: renderbuffer format %s unsupported\n",
-                       __FUNCTION__, _mesa_get_format_name(rb_format));
-   }
-
-   if (mt->mcs_mt) {
-      aux_mt = mt->mcs_mt;
-      aux_mode = GEN8_SURFACE_AUX_MODE_MCS;
-   }
-
-   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 13 * 4, 64,
-                                    &brw->wm.base.surf_offset[surf_index]);
-
-   surf[0] = (surf_type << BRW_SURFACE_TYPE_SHIFT) |
-             (is_array ? GEN7_SURFACE_IS_ARRAY : 0) |
-             (format << BRW_SURFACE_FORMAT_SHIFT) |
-             vertical_alignment(mt) |
-             horizontal_alignment(mt) |
-             surface_tiling_mode(tiling);
-
-   surf[1] = SET_FIELD(BDW_MOCS_WT, GEN8_SURFACE_MOCS) | mt->qpitch >> 2;
-
    surf[2] = SET_FIELD(width - 1, GEN7_SURFACE_WIDTH) |
              SET_FIELD(height - 1, GEN7_SURFACE_HEIGHT);
-
-   surf[3] = (depth - 1) << BRW_SURFACE_DEPTH_SHIFT |
-             (pitch - 1); /* Surface Pitch */
-
-   surf[4] = min_array_element << GEN7_SURFACE_MIN_ARRAY_ELEMENT_SHIFT |
-             (depth - 1) << GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT_SHIFT;
-
-   if (mt->format != MESA_FORMAT_S_UINT8)
-      surf[4] |= gen7_surface_msaa_bits(mt->num_samples, mt->msaa_layout);
-
-   surf[5] = irb->mt_level - irb->mt->first_level;
-
-   if (aux_mt) {
-      surf[6] = SET_FIELD(mt->qpitch / 4, GEN8_SURFACE_AUX_QPITCH) |
-                SET_FIELD((aux_mt->pitch / 128) - 1, GEN8_SURFACE_AUX_PITCH) |
-                aux_mode;
-   } else {
-      surf[6] = 0;
-   }
-
-   surf[7] = mt->fast_clear_color_value |
-             SET_FIELD(HSW_SCS_RED,   GEN7_SURFACE_SCS_R) |
-             SET_FIELD(HSW_SCS_GREEN, GEN7_SURFACE_SCS_G) |
-             SET_FIELD(HSW_SCS_BLUE,  GEN7_SURFACE_SCS_B) |
-             SET_FIELD(HSW_SCS_ALPHA, GEN7_SURFACE_SCS_A);
-
-   *((uint64_t *) &surf[8]) = mt->bo->offset64; /* reloc */
-
-   if (aux_mt) {
-      *((uint64_t *) &surf[10]) = aux_mt->bo->offset64;
-      drm_intel_bo_emit_reloc(brw->batch.bo,
-                              brw->wm.base.surf_offset[surf_index] + 10 * 4,
-                              aux_mt->bo, 0,
-                              I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER);
-   } else {
-      surf[10] = 0;
-      surf[11] = 0;
-   }
-   surf[12] = 0;
-
-   drm_intel_bo_emit_reloc(brw->batch.bo,
-                           brw->wm.base.surf_offset[surf_index] + 8 * 4,
-                           mt->bo,
-                           0,
-                           I915_GEM_DOMAIN_RENDER,
-                           I915_GEM_DOMAIN_RENDER);
 }
 
 void
 gen8_init_vtable_surface_functions(struct brw_context *brw)
 {
-   brw->vtbl.update_texture_surface = gen8_update_texture_surface;
-   brw->vtbl.update_renderbuffer_surface = gen8_update_renderbuffer_surface;
-   brw->vtbl.update_null_renderbuffer_surface =
-      gen8_update_null_renderbuffer_surface;
-   brw->vtbl.create_raw_surface = gen8_create_raw_surface;
-   brw->vtbl.emit_buffer_surface_state = gen8_emit_buffer_surface_state;
+   brw->vtbl.update_texture_surface = brw_update_texture_surface;
+   brw->vtbl.update_renderbuffer_surface = brw_update_renderbuffer_surface;
+   brw->vtbl.emit_null_surface_state = gen8_emit_null_surface_state;
 }