i965/miptree: Rename align_w,align_h -> halign,valign
[mesa.git] / src / mesa / drivers / dri / i965 / intel_mipmap_tree.c
index f218a2a196ac039eeb428e62b991b769c088947a..ffc356c9240ea55df9e962f6c1ed7806762f76cb 100644 (file)
@@ -1,5 +1,4 @@
-/**************************************************************************
- *
+/*
  * Copyright 2006 VMware, Inc.
  * All Rights Reserved.
  *
@@ -7,7 +6,7 @@
  * copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
  * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
+ * distribute, sublicense, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
  *
  *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
  * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
+ */
 
 #include <GL/gl.h>
 #include <GL/internal/dri_interface.h>
 
 #define FILE_DEBUG_FLAG DEBUG_MIPTREE
 
+static void *intel_miptree_map_raw(struct brw_context *brw,
+                                   struct intel_mipmap_tree *mt);
+
+static void intel_miptree_unmap_raw(struct intel_mipmap_tree *mt);
+
 static bool
 intel_miptree_alloc_mcs(struct brw_context *brw,
                         struct intel_mipmap_tree *mt,
@@ -59,7 +62,7 @@ intel_miptree_alloc_mcs(struct brw_context *brw,
  * created, based on the chip generation and the surface type.
  */
 static enum intel_msaa_layout
-compute_msaa_layout(struct brw_context *brw, mesa_format format, GLenum target,
+compute_msaa_layout(struct brw_context *brw, mesa_format format,
                     bool disable_aux_buffers)
 {
    /* Prior to Gen7, all MSAA surfaces used IMS layout. */
@@ -140,8 +143,7 @@ compute_msaa_layout(struct brw_context *brw, mesa_format format, GLenum target,
  *   by half the block width, and Y coordinates by half the block height.
  */
 void
-intel_get_non_msrt_mcs_alignment(struct brw_context *brw,
-                                 struct intel_mipmap_tree *mt,
+intel_get_non_msrt_mcs_alignment(struct intel_mipmap_tree *mt,
                                  unsigned *width_px, unsigned *height)
 {
    switch (mt->tiling) {
@@ -272,7 +274,6 @@ intel_miptree_create_layout(struct brw_context *brw,
                             GLuint height0,
                             GLuint depth0,
                             GLuint num_samples,
-                            enum intel_miptree_tiling_mode requested,
                             uint32_t layout_flags)
 {
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
@@ -280,7 +281,7 @@ intel_miptree_create_layout(struct brw_context *brw,
       return NULL;
 
    DBG("%s target %s format %s level %d..%d slices %d <-- %p\n", __func__,
-       _mesa_lookup_enum_by_nr(target),
+       _mesa_enum_to_string(target),
        _mesa_get_format_name(format),
        first_level, last_level, depth0, mt);
 
@@ -314,15 +315,7 @@ intel_miptree_create_layout(struct brw_context *brw,
    mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_NO_MCS;
    mt->disable_aux_buffers = (layout_flags & MIPTREE_LAYOUT_DISABLE_AUX) != 0;
    exec_list_make_empty(&mt->hiz_map);
-
-   /* The cpp is bytes per (1, blockheight)-sized block for compressed
-    * textures.  This is why you'll see divides by blockheight all over
-    */
-   unsigned bw, bh;
-   _mesa_get_format_block_size(format, &bw, &bh);
-   assert(_mesa_get_format_bytes(mt->format) % bw == 0);
-   mt->cpp = _mesa_get_format_bytes(mt->format) / bw;
-
+   mt->cpp = _mesa_get_format_bytes(format);
    mt->num_samples = num_samples;
    mt->compressed = _mesa_is_format_compressed(format);
    mt->msaa_layout = INTEL_MSAA_LAYOUT_NONE;
@@ -331,7 +324,7 @@ intel_miptree_create_layout(struct brw_context *brw,
    if (num_samples > 1) {
       /* Adjust width/height/depth for MSAA */
       mt->msaa_layout = compute_msaa_layout(brw, format,
-                                            mt->target, mt->disable_aux_buffers);
+                                            mt->disable_aux_buffers);
       if (mt->msaa_layout == INTEL_MSAA_LAYOUT_IMS) {
          /* From the Ivybridge PRM, Volume 1, Part 1, page 108:
           * "If the surface is multisampled and it is a depth or stencil
@@ -454,8 +447,10 @@ intel_miptree_create_layout(struct brw_context *brw,
        (brw->has_separate_stencil &&
          intel_miptree_wants_hiz_buffer(brw, mt)))) {
       uint32_t stencil_flags = MIPTREE_LAYOUT_ACCELERATED_UPLOAD;
-      if (brw->gen == 6)
-         stencil_flags |= MIPTREE_LAYOUT_FORCE_ALL_SLICE_AT_LOD;
+      if (brw->gen == 6) {
+         stencil_flags |= MIPTREE_LAYOUT_FORCE_ALL_SLICE_AT_LOD |
+                          MIPTREE_LAYOUT_TILING_ANY;
+      }
 
       mt->stencil_mt = intel_miptree_create(brw,
                                             mt->target,
@@ -466,7 +461,6 @@ intel_miptree_create_layout(struct brw_context *brw,
                                             mt->logical_height0,
                                             mt->logical_depth0,
                                             num_samples,
-                                            INTEL_MIPTREE_TILING_ANY,
                                             stencil_flags);
 
       if (!mt->stencil_mt) {
@@ -489,12 +483,28 @@ intel_miptree_create_layout(struct brw_context *brw,
    if (layout_flags & MIPTREE_LAYOUT_FORCE_ALL_SLICE_AT_LOD)
       mt->array_layout = ALL_SLICES_AT_EACH_LOD;
 
-   /* Use HALIGN_16 if MCS is enabled for non-MSRT */
-   if (brw->gen >= 8 && num_samples < 2 &&
-       intel_miptree_is_fast_clear_capable(brw, mt))
+   /*
+    * Obey HALIGN_16 constraints for Gen8 and Gen9 buffers which are
+    * multisampled or have an AUX buffer attached to it.
+    *
+    * GEN  |    MSRT        | AUX_CCS_* or AUX_MCS
+    *  -------------------------------------------
+    *  9   |  HALIGN_16     |    HALIGN_16
+    *  8   |  HALIGN_ANY    |    HALIGN_16
+    *  7   |      ?         |        ?
+    *  6   |      ?         |        ?
+    */
+   if (intel_miptree_is_fast_clear_capable(brw, mt)) {
+      if (brw->gen >= 9 || (brw->gen == 8 && num_samples <= 1))
+         layout_flags |= MIPTREE_LAYOUT_FORCE_HALIGN16;
+   } else if (brw->gen >= 9 && num_samples > 1) {
       layout_flags |= MIPTREE_LAYOUT_FORCE_HALIGN16;
+   } else {
+      /* For now, nothing else has this requirement */
+      assert((layout_flags & MIPTREE_LAYOUT_FORCE_HALIGN16) == 0);
+   }
 
-   brw_miptree_layout(brw, mt, requested, layout_flags);
+   brw_miptree_layout(brw, mt, layout_flags);
 
    if (mt->disable_aux_buffers)
       assert(mt->msaa_layout != INTEL_MSAA_LAYOUT_CMS);
@@ -542,6 +552,33 @@ intel_lower_compressed_format(struct brw_context *brw, mesa_format format)
    }
 }
 
+/* This function computes Yf/Ys tiled bo size, alignment and pitch. */
+static unsigned long
+intel_get_yf_ys_bo_size(struct intel_mipmap_tree *mt, unsigned *alignment,
+                        unsigned long *pitch)
+{
+   uint32_t tile_width, tile_height;
+   unsigned long stride, size, aligned_y;
+
+   assert(mt->tr_mode != INTEL_MIPTREE_TRMODE_NONE);
+   intel_get_tile_dims(mt->tiling, mt->tr_mode, mt->cpp,
+                       &tile_width, &tile_height);
+
+   aligned_y = ALIGN(mt->total_height, tile_height);
+   stride = mt->total_width * mt->cpp;
+   stride = ALIGN(stride, tile_width);
+   size = stride * aligned_y;
+
+   if (mt->tr_mode == INTEL_MIPTREE_TRMODE_YF) {
+      assert(size % 4096 == 0);
+      *alignment = 4096;
+   } else {
+      assert(size % (64 * 1024) == 0);
+      *alignment = 64 * 1024;
+   }
+   *pitch = stride;
+   return size;
+}
 
 struct intel_mipmap_tree *
 intel_miptree_create(struct brw_context *brw,
@@ -553,7 +590,6 @@ intel_miptree_create(struct brw_context *brw,
                      GLuint height0,
                      GLuint depth0,
                      GLuint num_samples,
-                     enum intel_miptree_tiling_mode requested_tiling,
                      uint32_t layout_flags)
 {
    struct intel_mipmap_tree *mt;
@@ -571,7 +607,7 @@ intel_miptree_create(struct brw_context *brw,
    mt = intel_miptree_create_layout(brw, target, format,
                                     first_level, last_level, width0,
                                     height0, depth0, num_samples,
-                                    requested_tiling, layout_flags);
+                                    layout_flags);
    /*
     * pitch == 0 || height == 0  indicates the null texture
     */
@@ -600,10 +636,22 @@ intel_miptree_create(struct brw_context *brw,
       alloc_flags |= BO_ALLOC_FOR_RENDER;
 
    unsigned long pitch;
-   mt->bo = drm_intel_bo_alloc_tiled(brw->bufmgr, "miptree", total_width,
-                                     total_height, mt->cpp, &mt->tiling,
-                                     &pitch, alloc_flags);
    mt->etc_format = etc_format;
+
+   if (mt->tr_mode != INTEL_MIPTREE_TRMODE_NONE) {
+      unsigned alignment = 0;
+      unsigned long size;
+      size = intel_get_yf_ys_bo_size(mt, &alignment, &pitch);
+      assert(size);
+      mt->bo = drm_intel_bo_alloc_for_render(brw->bufmgr, "miptree",
+                                             size, alignment);
+   } else {
+      mt->bo = drm_intel_bo_alloc_tiled(brw->bufmgr, "miptree",
+                                        total_width, total_height, mt->cpp,
+                                        &mt->tiling, &pitch,
+                                        alloc_flags);
+   }
+
    mt->pitch = pitch;
 
    /* If the BO is too large to fit in the aperture, we need to use the
@@ -646,7 +694,7 @@ intel_miptree_create(struct brw_context *brw,
    if (intel_tiling_supports_non_msrt_mcs(brw, mt->tiling) &&
        intel_miptree_is_fast_clear_capable(brw, mt)) {
       mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_RESOLVED;
-      assert(brw->gen < 8 || mt->align_w == 16 || num_samples <= 1);
+      assert(brw->gen < 8 || mt->halign == 16 || num_samples <= 1);
    }
 
    return mt;
@@ -682,17 +730,16 @@ intel_miptree_create_for_bo(struct brw_context *brw,
 
    target = depth > 1 ? GL_TEXTURE_2D_ARRAY : GL_TEXTURE_2D;
 
-   /* 'requested' parameter of intel_miptree_create_layout() is relevant
-    * only for non bo miptree. Tiling for bo is already computed above.
-    * So, the tiling requested (INTEL_MIPTREE_TILING_ANY) below is
-    * just a place holder and will not make any change to the miptree
-    * tiling format.
+   /* The BO already has a tiling format and we shouldn't confuse the lower
+    * layers by making it try to find a tiling format again.
     */
+   assert((layout_flags & MIPTREE_LAYOUT_TILING_ANY) == 0);
+   assert((layout_flags & MIPTREE_LAYOUT_TILING_NONE) == 0);
+
    layout_flags |= MIPTREE_LAYOUT_FOR_BO;
    mt = intel_miptree_create_layout(brw, target, format,
                                     0, 0,
                                     width, height, depth, 0,
-                                    INTEL_MIPTREE_TILING_ANY,
                                     layout_flags);
    if (!mt)
       return NULL;
@@ -800,11 +847,13 @@ intel_miptree_create_for_renderbuffer(struct brw_context *brw,
    uint32_t depth = 1;
    bool ok;
    GLenum target = num_samples > 1 ? GL_TEXTURE_2D_MULTISAMPLE : GL_TEXTURE_2D;
+   const uint32_t layout_flags = MIPTREE_LAYOUT_ACCELERATED_UPLOAD |
+                                 MIPTREE_LAYOUT_TILING_ANY;
+
 
    mt = intel_miptree_create(brw, target, format, 0, 0,
                              width, height, depth, num_samples,
-                             INTEL_MIPTREE_TILING_ANY,
-                             MIPTREE_LAYOUT_ACCELERATED_UPLOAD);
+                             layout_flags);
    if (!mt)
       goto fail;
 
@@ -872,12 +921,18 @@ intel_miptree_release(struct intel_mipmap_tree **mt)
    *mt = NULL;
 }
 
+
 void
-intel_miptree_get_dimensions_for_image(struct gl_texture_image *image,
-                                       int *width, int *height, int *depth)
+intel_get_image_dims(struct gl_texture_image *image,
+                     int *width, int *height, int *depth)
 {
    switch (image->TexObject->Target) {
    case GL_TEXTURE_1D_ARRAY:
+      /* For a 1D Array texture the OpenGL API will treat the image height as
+       * the number of array slices. For Intel hardware, we treat the 1D array
+       * as a 2D Array with a height of 1. So, here we want to swap image
+       * height and depth.
+       */
       *width = image->Width;
       *height = 1;
       *depth = image->Height;
@@ -921,7 +976,7 @@ intel_miptree_match_image(struct intel_mipmap_tree *mt,
    if (image->TexFormat != mt_format)
       return false;
 
-   intel_miptree_get_dimensions_for_image(image, &width, &height, &depth);
+   intel_get_image_dims(image, &width, &height, &depth);
 
    if (mt->target == GL_TEXTURE_CUBE_MAP)
       depth = 6;
@@ -1006,37 +1061,82 @@ intel_miptree_get_image_offset(const struct intel_mipmap_tree *mt,
    *y = mt->level[level].slice[slice].y_offset;
 }
 
+
+/**
+ * This function computes the tile_w (in bytes) and tile_h (in rows) of
+ * different tiling patterns. If the BO is untiled, tile_w is set to cpp
+ * and tile_h is set to 1.
+ */
+void
+intel_get_tile_dims(uint32_t tiling, uint32_t tr_mode, uint32_t cpp,
+                    uint32_t *tile_w, uint32_t *tile_h)
+{
+   if (tr_mode == INTEL_MIPTREE_TRMODE_NONE) {
+      switch (tiling) {
+      case I915_TILING_X:
+         *tile_w = 512;
+         *tile_h = 8;
+         break;
+      case I915_TILING_Y:
+         *tile_w = 128;
+         *tile_h = 32;
+         break;
+      case I915_TILING_NONE:
+         *tile_w = cpp;
+         *tile_h = 1;
+         break;
+      default:
+         unreachable("not reached");
+      }
+   } else {
+      uint32_t aspect_ratio = 1;
+      assert(_mesa_is_pow_two(cpp));
+
+      switch (cpp) {
+      case 1:
+         *tile_h = 64;
+         break;
+      case 2:
+      case 4:
+         *tile_h = 32;
+         break;
+      case 8:
+      case 16:
+         *tile_h = 16;
+         break;
+      default:
+         unreachable("not reached");
+      }
+
+      if (cpp == 2 || cpp == 8)
+         aspect_ratio = 2;
+
+      if (tr_mode == INTEL_MIPTREE_TRMODE_YS)
+         *tile_h *= 4;
+
+      *tile_w = *tile_h * aspect_ratio * cpp;
+   }
+}
+
+
 /**
  * This function computes masks that may be used to select the bits of the X
  * and Y coordinates that indicate the offset within a tile.  If the BO is
  * untiled, the masks are set to 0.
  */
 void
-intel_miptree_get_tile_masks(const struct intel_mipmap_tree *mt,
-                             uint32_t *mask_x, uint32_t *mask_y,
-                             bool map_stencil_as_y_tiled)
+intel_get_tile_masks(uint32_t tiling, uint32_t tr_mode, uint32_t cpp,
+                     bool map_stencil_as_y_tiled,
+                     uint32_t *mask_x, uint32_t *mask_y)
 {
-   int cpp = mt->cpp;
-   uint32_t tiling = mt->tiling;
-
+   uint32_t tile_w_bytes, tile_h;
    if (map_stencil_as_y_tiled)
       tiling = I915_TILING_Y;
 
-   switch (tiling) {
-   default:
-      unreachable("not reached");
-   case I915_TILING_NONE:
-      *mask_x = *mask_y = 0;
-      break;
-   case I915_TILING_X:
-      *mask_x = 512 / cpp - 1;
-      *mask_y = 7;
-      break;
-   case I915_TILING_Y:
-      *mask_x = 128 / cpp - 1;
-      *mask_y = 31;
-      break;
-   }
+   intel_get_tile_dims(tiling, tr_mode, cpp, &tile_w_bytes, &tile_h);
+
+   *mask_x = tile_w_bytes / cpp - 1;
+   *mask_y = tile_h - 1;
 }
 
 /**
@@ -1101,7 +1201,7 @@ intel_miptree_get_tile_offsets(const struct intel_mipmap_tree *mt,
    uint32_t x, y;
    uint32_t mask_x, mask_y;
 
-   intel_miptree_get_tile_masks(mt, &mask_x, &mask_y, false);
+   intel_get_tile_masks(mt->tiling, mt->tr_mode, mt->cpp, false, &mask_x, &mask_y);
    intel_miptree_get_image_offset(mt, level, slice, &x, &y);
 
    *tile_x = x & mask_x;
@@ -1195,8 +1295,10 @@ intel_miptree_copy_slice(struct brw_context *brw,
    assert(src_mt->format == dst_mt->format);
 
    if (dst_mt->compressed) {
-      height = ALIGN(height, dst_mt->align_h) / dst_mt->align_h;
-      width = ALIGN(width, dst_mt->align_w);
+      unsigned int i, j;
+      _mesa_get_format_block_size(dst_mt->format, &i, &j);
+      height = ALIGN_NPOT(height, j) / j;
+      width = ALIGN_NPOT(width, i) / i;
    }
 
    /* If it's a packed depth/stencil buffer with separate stencil, the blit
@@ -1307,6 +1409,8 @@ intel_miptree_alloc_mcs(struct brw_context *brw,
     *
     *     "The MCS surface must be stored as Tile Y."
     */
+   const uint32_t mcs_flags = MIPTREE_LAYOUT_ACCELERATED_UPLOAD |
+                              MIPTREE_LAYOUT_TILING_Y;
    mt->mcs_mt = intel_miptree_create(brw,
                                      mt->target,
                                      format,
@@ -1316,8 +1420,7 @@ intel_miptree_alloc_mcs(struct brw_context *brw,
                                      mt->logical_height0,
                                      mt->logical_depth0,
                                      0 /* num_samples */,
-                                     INTEL_MIPTREE_TILING_Y,
-                                     MIPTREE_LAYOUT_ACCELERATED_UPLOAD);
+                                     mcs_flags);
 
    /* From the Ivy Bridge PRM, Vol 2 Part 1 p326:
     *
@@ -1331,7 +1434,7 @@ intel_miptree_alloc_mcs(struct brw_context *brw,
     */
    void *data = intel_miptree_map_raw(brw, mt->mcs_mt);
    memset(data, 0xff, mt->mcs_mt->total_height * mt->mcs_mt->pitch);
-   intel_miptree_unmap_raw(brw, mt->mcs_mt);
+   intel_miptree_unmap_raw(mt->mcs_mt);
    mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_CLEAR;
 
    return mt->mcs_mt;
@@ -1357,7 +1460,7 @@ intel_miptree_alloc_non_msrt_mcs(struct brw_context *brw,
    const mesa_format format = MESA_FORMAT_R_UINT32;
    unsigned block_width_px;
    unsigned block_height;
-   intel_get_non_msrt_mcs_alignment(brw, mt, &block_width_px, &block_height);
+   intel_get_non_msrt_mcs_alignment(mt, &block_width_px, &block_height);
    unsigned width_divisor = block_width_px * 4;
    unsigned height_divisor = block_height * 8;
    unsigned mcs_width =
@@ -1365,9 +1468,11 @@ intel_miptree_alloc_non_msrt_mcs(struct brw_context *brw,
    unsigned mcs_height =
       ALIGN(mt->logical_height0, height_divisor) / height_divisor;
    assert(mt->logical_depth0 == 1);
-   uint32_t layout_flags = MIPTREE_LAYOUT_ACCELERATED_UPLOAD;
-   if (brw->gen >= 8)
+   uint32_t layout_flags = MIPTREE_LAYOUT_ACCELERATED_UPLOAD |
+                           MIPTREE_LAYOUT_TILING_Y;
+   if (brw->gen >= 8) {
       layout_flags |= MIPTREE_LAYOUT_FORCE_HALIGN16;
+   }
    mt->mcs_mt = intel_miptree_create(brw,
                                      mt->target,
                                      format,
@@ -1377,7 +1482,6 @@ intel_miptree_alloc_non_msrt_mcs(struct brw_context *brw,
                                      mcs_height,
                                      mt->logical_depth0,
                                      0 /* num_samples */,
-                                     INTEL_MIPTREE_TILING_Y,
                                      layout_flags);
 
    return mt->mcs_mt;
@@ -1468,7 +1572,7 @@ intel_gen7_hiz_buf_create(struct brw_context *brw,
       unsigned H_i = H0;
       unsigned Z_i = Z0;
       hz_height = 0;
-      for (int level = mt->first_level; level <= mt->last_level; ++level) {
+      for (unsigned level = mt->first_level; level <= mt->last_level; ++level) {
          unsigned h_i = ALIGN(H_i, vertical_align);
          /* sum(i=0 to m; h_i * max(1, floor(Z_Depth/2**i))) */
          hz_height += h_i * Z_i;
@@ -1532,21 +1636,23 @@ intel_gen8_hiz_buf_create(struct brw_context *brw,
    /* Gen7 PRM Volume 2, Part 1, 11.5.3 "Hierarchical Depth Buffer" documents
     * adjustments required for Z_Height and Z_Width based on multisampling.
     */
-   switch (mt->num_samples) {
-   case 0:
-   case 1:
-      break;
-   case 2:
-   case 4:
-      z_width *= 2;
-      z_height *= 2;
-      break;
-   case 8:
-      z_width *= 4;
-      z_height *= 2;
-      break;
-   default:
-      unreachable("unsupported sample count");
+   if (brw->gen < 9) {
+      switch (mt->num_samples) {
+      case 0:
+      case 1:
+         break;
+      case 2:
+      case 4:
+         z_width *= 2;
+         z_height *= 2;
+         break;
+      case 8:
+         z_width *= 4;
+         z_height *= 2;
+         break;
+      default:
+         unreachable("unsupported sample count");
+      }
    }
 
    const unsigned vertical_align = 8; /* 'j' in the docs */
@@ -1562,7 +1668,7 @@ intel_gen8_hiz_buf_create(struct brw_context *brw,
    unsigned Z_i = Z0;
    unsigned sum_h_i = 0;
    unsigned hz_height_3d_sum = 0;
-   for (int level = mt->first_level; level <= mt->last_level; ++level) {
+   for (unsigned level = mt->first_level; level <= mt->last_level; ++level) {
       unsigned i = level - mt->first_level;
       unsigned h_i = ALIGN(H_i, vertical_align);
       /* sum(i=2 to m; h_i) */
@@ -1628,6 +1734,7 @@ intel_hiz_miptree_buf_create(struct brw_context *brw,
    if (!buf)
       return NULL;
 
+   layout_flags |= MIPTREE_LAYOUT_TILING_ANY;
    buf->mt = intel_miptree_create(brw,
                                   mt->target,
                                   mt->format,
@@ -1637,7 +1744,6 @@ intel_hiz_miptree_buf_create(struct brw_context *brw,
                                   mt->logical_height0,
                                   mt->logical_depth0,
                                   mt->num_samples,
-                                  INTEL_MIPTREE_TILING_ANY,
                                   layout_flags);
    if (!buf->mt) {
       free(buf);
@@ -1695,11 +1801,11 @@ intel_miptree_alloc_hiz(struct brw_context *brw,
       return false;
 
    /* Mark that all slices need a HiZ resolve. */
-   for (int level = mt->first_level; level <= mt->last_level; ++level) {
+   for (unsigned level = mt->first_level; level <= mt->last_level; ++level) {
       if (!intel_miptree_level_enable_hiz(brw, mt, level))
          continue;
 
-      for (int layer = 0; layer < mt->level[level].depth; ++layer) {
+      for (unsigned layer = 0; layer < mt->level[level].depth; ++layer) {
          struct intel_resolve_map *m = malloc(sizeof(struct intel_resolve_map));
          exec_node_init(&m->link);
          m->level = level;
@@ -2000,8 +2106,7 @@ intel_miptree_map_raw(struct brw_context *brw, struct intel_mipmap_tree *mt)
 }
 
 void
-intel_miptree_unmap_raw(struct brw_context *brw,
-                        struct intel_mipmap_tree *mt)
+intel_miptree_unmap_raw(struct intel_mipmap_tree *mt)
 {
    drm_intel_bo_unmap(mt->bo);
 }
@@ -2024,7 +2129,9 @@ intel_miptree_map_gtt(struct brw_context *brw,
     */
    _mesa_get_format_block_size(mt->format, &bw, &bh);
    assert(y % bh == 0);
+   assert(x % bw == 0);
    y /= bh;
+   x /= bw;
 
    base = intel_miptree_map_raw(brw, mt) + mt->offset;
 
@@ -2050,13 +2157,9 @@ intel_miptree_map_gtt(struct brw_context *brw,
 }
 
 static void
-intel_miptree_unmap_gtt(struct brw_context *brw,
-                       struct intel_mipmap_tree *mt,
-                       struct intel_miptree_map *map,
-                       unsigned int level,
-                       unsigned int slice)
+intel_miptree_unmap_gtt(struct intel_mipmap_tree *mt)
 {
-   intel_miptree_unmap_raw(brw, mt);
+   intel_miptree_unmap_raw(mt);
 }
 
 static void
@@ -2065,16 +2168,18 @@ intel_miptree_map_blit(struct brw_context *brw,
                       struct intel_miptree_map *map,
                       unsigned int level, unsigned int slice)
 {
-   map->mt = intel_miptree_create(brw, GL_TEXTURE_2D, mt->format,
-                                  0, 0,
-                                  map->w, map->h, 1,
-                                  0, INTEL_MIPTREE_TILING_NONE, 0);
+   map->linear_mt = intel_miptree_create(brw, GL_TEXTURE_2D, mt->format,
+                                         /* first_level */ 0,
+                                         /* last_level */ 0,
+                                         map->w, map->h, 1,
+                                         /* samples */ 0,
+                                         MIPTREE_LAYOUT_TILING_NONE);
 
-   if (!map->mt) {
+   if (!map->linear_mt) {
       fprintf(stderr, "Failed to allocate blit temporary\n");
       goto fail;
    }
-   map->stride = map->mt->pitch;
+   map->stride = map->linear_mt->pitch;
 
    /* One of either READ_BIT or WRITE_BIT or both is set.  READ_BIT implies no
     * INVALIDATE_RANGE_BIT.  WRITE_BIT needs the original values read in unless
@@ -2085,7 +2190,7 @@ intel_miptree_map_blit(struct brw_context *brw,
       if (!intel_miptree_blit(brw,
                               mt, level, slice,
                               map->x, map->y, false,
-                              map->mt, 0, 0,
+                              map->linear_mt, 0, 0,
                               0, 0, false,
                               map->w, map->h, GL_COPY)) {
          fprintf(stderr, "Failed to blit\n");
@@ -2093,7 +2198,7 @@ intel_miptree_map_blit(struct brw_context *brw,
       }
    }
 
-   map->ptr = intel_miptree_map_raw(brw, map->mt);
+   map->ptr = intel_miptree_map_raw(brw, map->linear_mt);
 
    DBG("%s: %d,%d %dx%d from mt %p (%s) %d,%d = %p/%d\n", __func__,
        map->x, map->y, map->w, map->h,
@@ -2103,7 +2208,7 @@ intel_miptree_map_blit(struct brw_context *brw,
    return;
 
 fail:
-   intel_miptree_release(&map->mt);
+   intel_miptree_release(&map->linear_mt);
    map->ptr = NULL;
    map->stride = 0;
 }
@@ -2117,11 +2222,11 @@ intel_miptree_unmap_blit(struct brw_context *brw,
 {
    struct gl_context *ctx = &brw->ctx;
 
-   intel_miptree_unmap_raw(brw, map->mt);
+   intel_miptree_unmap_raw(map->linear_mt);
 
    if (map->mode & GL_MAP_WRITE_BIT) {
       bool ok = intel_miptree_blit(brw,
-                                   map->mt, 0, 0,
+                                   map->linear_mt, 0, 0,
                                    0, 0, false,
                                    mt, level, slice,
                                    map->x, map->y, false,
@@ -2129,7 +2234,7 @@ intel_miptree_unmap_blit(struct brw_context *brw,
       WARN_ONCE(!ok, "Failed to blit from linear temporary mapping");
    }
 
-   intel_miptree_release(&map->mt);
+   intel_miptree_release(&map->linear_mt);
 }
 
 /**
@@ -2189,7 +2294,7 @@ intel_miptree_map_movntdqa(struct brw_context *brw,
       _mesa_streaming_load_memcpy(dst_ptr, src_ptr, width_bytes);
    }
 
-   intel_miptree_unmap_raw(brw, mt);
+   intel_miptree_unmap_raw(mt);
 }
 
 static void
@@ -2238,7 +2343,7 @@ intel_miptree_map_s8(struct brw_context *brw,
         }
       }
 
-      intel_miptree_unmap_raw(brw, mt);
+      intel_miptree_unmap_raw(mt);
 
       DBG("%s: %d,%d %dx%d from mt %p %d,%d = %p/%d\n", __func__,
          map->x, map->y, map->w, map->h,
@@ -2274,7 +2379,7 @@ intel_miptree_unmap_s8(struct brw_context *brw,
         }
       }
 
-      intel_miptree_unmap_raw(brw, mt);
+      intel_miptree_unmap_raw(mt);
    }
 
    free(map->buffer);
@@ -2328,7 +2433,7 @@ intel_miptree_unmap_etc(struct brw_context *brw,
                                map->ptr, map->stride,
                                map->w, map->h, mt->etc_format);
 
-   intel_miptree_unmap_raw(brw, mt);
+   intel_miptree_unmap_raw(mt);
    free(map->buffer);
 }
 
@@ -2398,8 +2503,8 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
         }
       }
 
-      intel_miptree_unmap_raw(brw, s_mt);
-      intel_miptree_unmap_raw(brw, z_mt);
+      intel_miptree_unmap_raw(s_mt);
+      intel_miptree_unmap_raw(z_mt);
 
       DBG("%s: %d,%d %dx%d from z mt %p %d,%d, s mt %p %d,%d = %p/%d\n",
          __func__,
@@ -2458,8 +2563,8 @@ intel_miptree_unmap_depthstencil(struct brw_context *brw,
         }
       }
 
-      intel_miptree_unmap_raw(brw, s_mt);
-      intel_miptree_unmap_raw(brw, z_mt);
+      intel_miptree_unmap_raw(s_mt);
+      intel_miptree_unmap_raw(z_mt);
 
       DBG("%s: %d,%d %dx%d from z mt %p (%s) %d,%d, s mt %p %d,%d = %p/%d\n",
          __func__,
@@ -2614,7 +2719,9 @@ intel_miptree_map(struct brw_context *brw,
    } else if (use_intel_mipree_map_blit(brw, mt, mode, level, slice)) {
       intel_miptree_map_blit(brw, mt, map, level, slice);
 #if defined(USE_SSE41)
-   } else if (!(mode & GL_MAP_WRITE_BIT) && !mt->compressed && cpu_has_sse4_1) {
+   } else if (!(mode & GL_MAP_WRITE_BIT) &&
+              !mt->compressed && cpu_has_sse4_1 &&
+              (mt->pitch % 16 == 0)) {
       intel_miptree_map_movntdqa(brw, mt, map, level, slice);
 #endif
    } else {
@@ -2651,14 +2758,14 @@ intel_miptree_unmap(struct brw_context *brw,
       intel_miptree_unmap_etc(brw, mt, map, level, slice);
    } else if (mt->stencil_mt && !(map->mode & BRW_MAP_DIRECT_BIT)) {
       intel_miptree_unmap_depthstencil(brw, mt, map, level, slice);
-   } else if (map->mt) {
+   } else if (map->linear_mt) {
       intel_miptree_unmap_blit(brw, mt, map, level, slice);
 #if defined(USE_SSE41)
    } else if (map->buffer && cpu_has_sse4_1) {
       intel_miptree_unmap_movntdqa(brw, mt, map, level, slice);
 #endif
    } else {
-      intel_miptree_unmap_gtt(brw, mt, map, level, slice);
+      intel_miptree_unmap_gtt(mt);
    }
 
    intel_miptree_release_map(mt, level, slice);