i965: Alphabetize brw_tracked_state flags and use a consistent style.
[mesa.git] / src / mesa / drivers / dri / i965 / intel_blit.c
index da56f55c4df323082c347c7b744eb709cabf3689..73ab488f00df919fda61e58425744a8ee12346b2 100644 (file)
@@ -1,8 +1,8 @@
 /**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ *
+ * Copyright 2003 VMware, Inc.
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
  * distribute, sub license, 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 above copyright notice and this permission notice (including the
  * next paragraph) shall be included in all copies or substantial portions
  * of the Software.
- * 
+ *
  * 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.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * 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 "main/colormac.h"
 #include "main/fbobject.h"
 
+#include "brw_context.h"
+#include "brw_defines.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
-#include "intel_context.h"
 #include "intel_fbo.h"
 #include "intel_reg.h"
-#include "intel_regions.h"
 #include "intel_batchbuffer.h"
 #include "intel_mipmap_tree.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
 static void
-intel_miptree_set_alpha_to_one(struct intel_context *intel,
+intel_miptree_set_alpha_to_one(struct brw_context *brw,
                                struct intel_mipmap_tree *mt,
                                int x, int y, int width, int height);
 
@@ -85,8 +85,7 @@ br13_for_cpp(int cpp)
       return BR13_8;
       break;
    default:
-      assert(0);
-      return 0;
+      unreachable("not reached");
    }
 }
 
@@ -101,10 +100,10 @@ br13_for_cpp(int cpp)
  * server).
  */
 static void
-set_blitter_tiling(struct intel_context *intel,
+set_blitter_tiling(struct brw_context *brw,
                    bool dst_y_tiled, bool src_y_tiled)
 {
-   assert(intel->gen >= 6);
+   assert(brw->gen >= 6);
 
    /* Idle the blitter before we update how tiling is interpreted. */
    OUT_BATCH(MI_FLUSH_DW);
@@ -122,12 +121,12 @@ set_blitter_tiling(struct intel_context *intel,
 #define BEGIN_BATCH_BLT_TILED(n, dst_y_tiled, src_y_tiled) do {         \
       BEGIN_BATCH_BLT(n + ((dst_y_tiled || src_y_tiled) ? 14 : 0));     \
       if (dst_y_tiled || src_y_tiled)                                   \
-         set_blitter_tiling(intel, dst_y_tiled, src_y_tiled);           \
+         set_blitter_tiling(brw, dst_y_tiled, src_y_tiled);             \
    } while (0)
 
 #define ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled) do {              \
       if (dst_y_tiled || src_y_tiled)                                   \
-         set_blitter_tiling(intel, false, false);                       \
+         set_blitter_tiling(brw, false, false);                         \
       ADVANCE_BATCH();                                                  \
    } while (0)
 
@@ -147,7 +146,7 @@ set_blitter_tiling(struct intel_context *intel,
  * renderbuffers/textures.
  */
 bool
-intel_miptree_blit(struct intel_context *intel,
+intel_miptree_blit(struct brw_context *brw,
                    struct intel_mipmap_tree *src_mt,
                    int src_level, int src_slice,
                    uint32_t src_x, uint32_t src_y, bool src_flip,
@@ -157,12 +156,16 @@ intel_miptree_blit(struct intel_context *intel,
                    uint32_t width, uint32_t height,
                    GLenum logicop)
 {
+   /* The blitter doesn't understand multisampling at all. */
+   if (src_mt->num_samples > 0 || dst_mt->num_samples > 0)
+      return false;
+
    /* No sRGB decode or encode is done by the hardware blitter, which is
     * consistent with what we want in the callers (glCopyTexSubImage(),
     * glBlitFramebuffer(), texture validation, etc.).
     */
-   gl_format src_format = _mesa_get_srgb_format_linear(src_mt->format);
-   gl_format dst_format = _mesa_get_srgb_format_linear(dst_mt->format);
+   mesa_format src_format = _mesa_get_srgb_format_linear(src_mt->format);
+   mesa_format dst_format = _mesa_get_srgb_format_linear(dst_mt->format);
 
    /* The blitter doesn't support doing any format conversions.  We do also
     * support blitting ARGB8888 to XRGB8888 (trivial, the values dropped into
@@ -170,10 +173,10 @@ intel_miptree_blit(struct intel_context *intel,
     * channel to 1.0 at the end.
     */
    if (src_format != dst_format &&
-      ((src_format != MESA_FORMAT_ARGB8888 &&
-        src_format != MESA_FORMAT_XRGB8888) ||
-       (dst_format != MESA_FORMAT_ARGB8888 &&
-        dst_format != MESA_FORMAT_XRGB8888))) {
+      ((src_format != MESA_FORMAT_B8G8R8A8_UNORM &&
+        src_format != MESA_FORMAT_B8G8R8X8_UNORM) ||
+       (dst_format != MESA_FORMAT_B8G8R8A8_UNORM &&
+        dst_format != MESA_FORMAT_B8G8R8X8_UNORM))) {
       perf_debug("%s: Can't use hardware blitter from %s to %s, "
                  "falling back.\n", __FUNCTION__,
                  _mesa_get_format_name(src_format),
@@ -197,29 +200,29 @@ intel_miptree_blit(struct intel_context *intel,
     * pitches < 32k.
     *
     * As a result of these two limitations, we can only use the blitter to do
-    * this copy when the region's pitch is less than 32k.
+    * this copy when the miptree's pitch is less than 32k.
     */
-   if (src_mt->region->pitch > 32768 ||
-       dst_mt->region->pitch > 32768) {
-      perf_debug("Falling back due to >32k pitch\n");
+   if (src_mt->pitch >= 32768 ||
+       dst_mt->pitch >= 32768) {
+      perf_debug("Falling back due to >=32k pitch\n");
       return false;
    }
 
    /* The blitter has no idea about HiZ or fast color clears, so we need to
     * resolve the miptrees before we do anything.
     */
-   intel_miptree_slice_resolve_depth(intel, src_mt, src_level, src_slice);
-   intel_miptree_slice_resolve_depth(intel, dst_mt, dst_level, dst_slice);
-   intel_miptree_resolve_color(intel, src_mt);
-   intel_miptree_resolve_color(intel, dst_mt);
+   intel_miptree_slice_resolve_depth(brw, src_mt, src_level, src_slice);
+   intel_miptree_slice_resolve_depth(brw, dst_mt, dst_level, dst_slice);
+   intel_miptree_resolve_color(brw, src_mt);
+   intel_miptree_resolve_color(brw, dst_mt);
 
    if (src_flip)
-      src_y = src_mt->level[src_level].height - src_y - height;
+      src_y = minify(src_mt->physical_height0, src_level - src_mt->first_level) - src_y - height;
 
    if (dst_flip)
-      dst_y = dst_mt->level[dst_level].height - dst_y - height;
+      dst_y = minify(dst_mt->physical_height0, dst_level - dst_mt->first_level) - dst_y - height;
 
-   int src_pitch = src_mt->region->pitch;
+   int src_pitch = src_mt->pitch;
    if (src_flip != dst_flip)
       src_pitch = -src_pitch;
 
@@ -229,20 +232,40 @@ intel_miptree_blit(struct intel_context *intel,
    src_x += src_image_x;
    src_y += src_image_y;
 
+   /* The blitter interprets the 16-bit src x/y as a signed 16-bit value,
+    * where negative values are invalid.  The values we're working with are
+    * unsigned, so make sure we don't overflow.
+    */
+   if (src_x >= 32768 || src_y >= 32768) {
+      perf_debug("Falling back due to >=32k src offset (%d, %d)\n",
+                 src_x, src_y);
+      return false;
+   }
+
    uint32_t dst_image_x, dst_image_y;
    intel_miptree_get_image_offset(dst_mt, dst_level, dst_slice,
                                   &dst_image_x, &dst_image_y);
    dst_x += dst_image_x;
    dst_y += dst_image_y;
 
-   if (!intelEmitCopyBlit(intel,
+   /* The blitter interprets the 16-bit destination x/y as a signed 16-bit
+    * value.  The values we're working with are unsigned, so make sure we
+    * don't overflow.
+    */
+   if (dst_x >= 32768 || dst_y >= 32768) {
+      perf_debug("Falling back due to >=32k dst offset (%d, %d)\n",
+                 dst_x, dst_y);
+      return false;
+   }
+
+   if (!intelEmitCopyBlit(brw,
                           src_mt->cpp,
                           src_pitch,
-                          src_mt->region->bo, src_mt->offset,
-                          src_mt->region->tiling,
-                          dst_mt->region->pitch,
-                          dst_mt->region->bo, dst_mt->offset,
-                          dst_mt->region->tiling,
+                          src_mt->bo, src_mt->offset,
+                          src_mt->tiling,
+                          dst_mt->pitch,
+                          dst_mt->bo, dst_mt->offset,
+                          dst_mt->tiling,
                           src_x, src_y,
                           dst_x, dst_y,
                           width, height,
@@ -250,9 +273,9 @@ intel_miptree_blit(struct intel_context *intel,
       return false;
    }
 
-   if (src_mt->format == MESA_FORMAT_XRGB8888 &&
-       dst_mt->format == MESA_FORMAT_ARGB8888) {
-      intel_miptree_set_alpha_to_one(intel, dst_mt,
+   if (src_mt->format == MESA_FORMAT_B8G8R8X8_UNORM &&
+       dst_mt->format == MESA_FORMAT_B8G8R8A8_UNORM) {
+      intel_miptree_set_alpha_to_one(brw, dst_mt,
                                      dst_x, dst_y,
                                      width, height);
    }
@@ -263,7 +286,7 @@ intel_miptree_blit(struct intel_context *intel,
 /* Copy BitBlt
  */
 bool
-intelEmitCopyBlit(struct intel_context *intel,
+intelEmitCopyBlit(struct brw_context *brw,
                  GLuint cpp,
                  GLshort src_pitch,
                  drm_intel_bo *src_buffer,
@@ -284,7 +307,6 @@ intelEmitCopyBlit(struct intel_context *intel,
    drm_intel_bo *aper_array[3];
    bool dst_y_tiled = dst_tiling == I915_TILING_Y;
    bool src_y_tiled = src_tiling == I915_TILING_Y;
-   BATCH_LOCALS;
 
    if (dst_tiling != I915_TILING_NONE) {
       if (dst_offset & 4095)
@@ -294,17 +316,17 @@ intelEmitCopyBlit(struct intel_context *intel,
       if (src_offset & 4095)
         return false;
    }
-   if ((dst_y_tiled || src_y_tiled) && intel->gen < 6)
+   if ((dst_y_tiled || src_y_tiled) && brw->gen < 6)
       return false;
 
    /* do space check before going any further */
    do {
-       aper_array[0] = intel->batch.bo;
+       aper_array[0] = brw->batch.bo;
        aper_array[1] = dst_buffer;
        aper_array[2] = src_buffer;
 
        if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
-           intel_batchbuffer_flush(intel);
+           intel_batchbuffer_flush(brw);
            pass++;
        } else
            break;
@@ -313,7 +335,7 @@ intelEmitCopyBlit(struct intel_context *intel,
    if (pass >= 2)
       return false;
 
-   intel_batchbuffer_require_space(intel, 8 * 4, true);
+   intel_batchbuffer_require_space(brw, 8 * 4, BLT_RING);
    DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
        __FUNCTION__,
        src_buffer, src_pitch, src_offset, src_x, src_y,
@@ -357,7 +379,6 @@ intelEmitCopyBlit(struct intel_context *intel,
       return false;
    }
 
-#ifndef I915
    if (dst_tiling != I915_TILING_NONE) {
       CMD |= XY_DST_TILED;
       dst_pitch /= 4;
@@ -366,7 +387,6 @@ intelEmitCopyBlit(struct intel_context *intel,
       CMD |= XY_SRC_TILED;
       src_pitch /= 4;
    }
-#endif
 
    if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
       return true;
@@ -374,211 +394,48 @@ intelEmitCopyBlit(struct intel_context *intel,
 
    assert(dst_x < dst_x2);
    assert(dst_y < dst_y2);
+   assert(src_offset + (src_y + h - 1) * abs(src_pitch) +
+          (w * cpp) <= src_buffer->size);
+   assert(dst_offset + (dst_y + h - 1) * abs(dst_pitch) +
+          (w * cpp) <= dst_buffer->size);
 
-   BEGIN_BATCH_BLT_TILED(8, dst_y_tiled, src_y_tiled);
+   unsigned length = brw->gen >= 8 ? 10 : 8;
 
-   OUT_BATCH(CMD | (8 - 2));
+   BEGIN_BATCH_BLT_TILED(length, dst_y_tiled, src_y_tiled);
+   OUT_BATCH(CMD | (length - 2));
    OUT_BATCH(BR13 | (uint16_t)dst_pitch);
-   OUT_BATCH((dst_y << 16) | dst_x);
-   OUT_BATCH((dst_y2 << 16) | dst_x2);
-   OUT_RELOC_FENCED(dst_buffer,
-                   I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-                   dst_offset);
-   OUT_BATCH((src_y << 16) | src_x);
+   OUT_BATCH(SET_FIELD(dst_y, BLT_Y) | SET_FIELD(dst_x, BLT_X));
+   OUT_BATCH(SET_FIELD(dst_y2, BLT_Y) | SET_FIELD(dst_x2, BLT_X));
+   if (brw->gen >= 8) {
+      OUT_RELOC64(dst_buffer,
+                  I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                  dst_offset);
+   } else {
+      OUT_RELOC(dst_buffer,
+                I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                dst_offset);
+   }
+   OUT_BATCH(SET_FIELD(src_y, BLT_Y) | SET_FIELD(src_x, BLT_X));
    OUT_BATCH((uint16_t)src_pitch);
-   OUT_RELOC_FENCED(src_buffer,
-                   I915_GEM_DOMAIN_RENDER, 0,
-                   src_offset);
+   if (brw->gen >= 8) {
+      OUT_RELOC64(src_buffer,
+                  I915_GEM_DOMAIN_RENDER, 0,
+                  src_offset);
+   } else {
+      OUT_RELOC(src_buffer,
+                I915_GEM_DOMAIN_RENDER, 0,
+                src_offset);
+   }
 
    ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled);
 
-   intel_batchbuffer_emit_mi_flush(intel);
+   intel_batchbuffer_emit_mi_flush(brw);
 
    return true;
 }
 
-
-/**
- * Use blitting to clear the renderbuffers named by 'flags'.
- * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field
- * since that might include software renderbuffers or renderbuffers
- * which we're clearing with triangles.
- * \param mask  bitmask of BUFFER_BIT_* values indicating buffers to clear
- */
-GLbitfield
-intelClearWithBlit(struct gl_context *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint clear_depth_value, clear_depth_mask;
-   GLint cx, cy, cw, ch;
-   GLbitfield fail_mask = 0;
-   BATCH_LOCALS;
-
-   /* Note: we don't use this function on Gen7+ hardware, so we can safely
-    * ignore fast color clear issues.
-    */
-   assert(intel->gen < 7);
-
-   /*
-    * Compute values for clearing the buffers.
-    */
-   clear_depth_value = 0;
-   clear_depth_mask = 0;
-   if (mask & BUFFER_BIT_DEPTH) {
-      clear_depth_value = (GLuint) (fb->_DepthMax * ctx->Depth.Clear);
-      clear_depth_mask = XY_BLT_WRITE_RGB;
-   }
-   if (mask & BUFFER_BIT_STENCIL) {
-      clear_depth_value |= (ctx->Stencil.Clear & 0xff) << 24;
-      clear_depth_mask |= XY_BLT_WRITE_ALPHA;
-   }
-
-   cx = fb->_Xmin;
-   if (_mesa_is_winsys_fbo(fb))
-      cy = ctx->DrawBuffer->Height - fb->_Ymax;
-   else
-      cy = fb->_Ymin;
-   cw = fb->_Xmax - fb->_Xmin;
-   ch = fb->_Ymax - fb->_Ymin;
-
-   if (cw == 0 || ch == 0)
-      return 0;
-
-   /* Loop over all renderbuffers */
-   mask &= (1 << BUFFER_COUNT) - 1;
-   while (mask) {
-      GLuint buf = ffs(mask) - 1;
-      bool is_depth_stencil = buf == BUFFER_DEPTH || buf == BUFFER_STENCIL;
-      struct intel_renderbuffer *irb;
-      int x1, y1, x2, y2;
-      uint32_t clear_val;
-      uint32_t BR13, CMD;
-      struct intel_region *region;
-      int pitch, cpp;
-      drm_intel_bo *aper_array[2];
-
-      mask &= ~(1 << buf);
-
-      irb = intel_get_renderbuffer(fb, buf);
-      if (irb && irb->mt) {
-        region = irb->mt->region;
-        assert(region);
-        assert(region->bo);
-      } else {
-         fail_mask |= 1 << buf;
-         continue;
-      }
-
-      /* OK, clear this renderbuffer */
-      x1 = cx + irb->draw_x;
-      y1 = cy + irb->draw_y;
-      x2 = cx + cw + irb->draw_x;
-      y2 = cy + ch + irb->draw_y;
-
-      pitch = region->pitch;
-      cpp = region->cpp;
-
-      DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
-         __FUNCTION__,
-         region->bo, pitch,
-         x1, y1, x2 - x1, y2 - y1);
-
-      BR13 = 0xf0 << 16;
-      CMD = XY_COLOR_BLT_CMD;
-
-      /* Setup the blit command */
-      if (cpp == 4) {
-        if (is_depth_stencil) {
-           CMD |= clear_depth_mask;
-        } else {
-           /* clearing RGBA */
-           CMD |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
-        }
-      }
-
-      assert(region->tiling != I915_TILING_Y);
-
-#ifndef I915
-      if (region->tiling != I915_TILING_NONE) {
-        CMD |= XY_DST_TILED;
-        pitch /= 4;
-      }
-#endif
-      BR13 |= pitch;
-
-      if (is_depth_stencil) {
-        clear_val = clear_depth_value;
-      } else {
-        uint8_t clear[4];
-        GLfloat *color = ctx->Color.ClearColor.f;
-
-        _mesa_unclamped_float_rgba_to_ubyte(clear, color);
-
-        switch (intel_rb_format(irb)) {
-        case MESA_FORMAT_ARGB8888:
-        case MESA_FORMAT_XRGB8888:
-           clear_val = PACK_COLOR_8888(clear[3], clear[0],
-                                       clear[1], clear[2]);
-           break;
-        case MESA_FORMAT_RGB565:
-           clear_val = PACK_COLOR_565(clear[0], clear[1], clear[2]);
-           break;
-        case MESA_FORMAT_ARGB4444:
-           clear_val = PACK_COLOR_4444(clear[3], clear[0],
-                                       clear[1], clear[2]);
-           break;
-        case MESA_FORMAT_ARGB1555:
-           clear_val = PACK_COLOR_1555(clear[3], clear[0],
-                                       clear[1], clear[2]);
-           break;
-        case MESA_FORMAT_A8:
-           clear_val = PACK_COLOR_8888(clear[3], clear[3],
-                                       clear[3], clear[3]);
-           break;
-        default:
-           fail_mask |= 1 << buf;
-           continue;
-        }
-      }
-
-      BR13 |= br13_for_cpp(cpp);
-
-      assert(x1 < x2);
-      assert(y1 < y2);
-
-      /* do space check before going any further */
-      aper_array[0] = intel->batch.bo;
-      aper_array[1] = region->bo;
-
-      if (drm_intel_bufmgr_check_aperture_space(aper_array,
-                                               ARRAY_SIZE(aper_array)) != 0) {
-        intel_batchbuffer_flush(intel);
-      }
-
-      BEGIN_BATCH_BLT(6);
-      OUT_BATCH(CMD | (6 - 2));
-      OUT_BATCH(BR13);
-      OUT_BATCH((y1 << 16) | x1);
-      OUT_BATCH((y2 << 16) | x2);
-      OUT_RELOC_FENCED(region->bo,
-                      I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-                      0);
-      OUT_BATCH(clear_val);
-      ADVANCE_BATCH();
-
-      if (intel->always_flush_cache)
-        intel_batchbuffer_emit_mi_flush(intel);
-
-      if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL)
-        mask &= ~(BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
-   }
-
-   return fail_mask;
-}
-
 bool
-intelEmitImmediateColorExpandBlit(struct intel_context *intel,
+intelEmitImmediateColorExpandBlit(struct brw_context *brw,
                                  GLuint cpp,
                                  GLubyte *src_bits, GLuint src_size,
                                  GLuint fg_color,
@@ -600,8 +457,7 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
         return false;
    }
 
-   assert( logic_op - GL_CLEAR >= 0 );
-   assert( logic_op - GL_CLEAR < 0x10 );
+   assert((logic_op >= GL_CLEAR) && (logic_op <= (GL_CLEAR + 0x0f)));
    assert(dst_pitch > 0);
 
    if (w < 0 || h < 0)
@@ -611,20 +467,15 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
        __FUNCTION__,
        dst_buffer, dst_pitch, dst_offset, x, y, w, h, src_size, dwords);
 
-   intel_batchbuffer_require_space(intel,
-                                  (8 * 4) +
-                                  (3 * 4) +
-                                  dwords * 4, true);
+   intel_batchbuffer_require_space(brw, (8 * 4) + (3 * 4) + dwords * 4, BLT_RING);
 
    opcode = XY_SETUP_BLT_CMD;
    if (cpp == 4)
       opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
-#ifndef I915
    if (dst_tiling != I915_TILING_NONE) {
       opcode |= XY_DST_TILED;
       dst_pitch /= 4;
    }
-#endif
 
    br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
    br13 |= br13_for_cpp(cpp);
@@ -633,26 +484,36 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
    if (dst_tiling != I915_TILING_NONE)
       blit_cmd |= XY_DST_TILED;
 
-   BEGIN_BATCH_BLT(8 + 3);
-   OUT_BATCH(opcode | (8 - 2));
+   unsigned xy_setup_blt_length = brw->gen >= 8 ? 10 : 8;
+
+   BEGIN_BATCH_BLT(xy_setup_blt_length + 3);
+   OUT_BATCH(opcode | (xy_setup_blt_length - 2));
    OUT_BATCH(br13);
    OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
    OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
-   OUT_RELOC_FENCED(dst_buffer,
-                   I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-                   dst_offset);
+   if (brw->gen >= 8) {
+      OUT_RELOC64(dst_buffer,
+                  I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                  dst_offset);
+   } else {
+      OUT_RELOC(dst_buffer,
+                I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                dst_offset);
+   }
    OUT_BATCH(0); /* bg */
    OUT_BATCH(fg_color); /* fg */
    OUT_BATCH(0); /* pattern base addr */
+   if (brw->gen >= 8)
+      OUT_BATCH(0);
 
    OUT_BATCH(blit_cmd | ((3 - 2) + dwords));
-   OUT_BATCH((y << 16) | x);
-   OUT_BATCH(((y + h) << 16) | (x + w));
+   OUT_BATCH(SET_FIELD(y, BLT_Y) | SET_FIELD(x, BLT_X));
+   OUT_BATCH(SET_FIELD(y + h, BLT_Y) | SET_FIELD(x + w, BLT_X));
    ADVANCE_BATCH();
 
-   intel_batchbuffer_data(intel, src_bits, dwords * 4, true);
+   intel_batchbuffer_data(brw, src_bits, dwords * 4, BLT_RING);
 
-   intel_batchbuffer_emit_mi_flush(intel);
+   intel_batchbuffer_emit_mi_flush(brw);
 
    return true;
 }
@@ -662,14 +523,14 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
  * end to cover the last if we need.
  */
 void
-intel_emit_linear_blit(struct intel_context *intel,
+intel_emit_linear_blit(struct brw_context *brw,
                       drm_intel_bo *dst_bo,
                       unsigned int dst_offset,
                       drm_intel_bo *src_bo,
                       unsigned int src_offset,
                       unsigned int size)
 {
-   struct gl_context *ctx = &intel->ctx;
+   struct gl_context *ctx = &brw->ctx;
    GLuint pitch, height;
    bool ok;
 
@@ -679,7 +540,7 @@ intel_emit_linear_blit(struct intel_context *intel,
     */
    pitch = ROUND_DOWN_TO(MIN2(size, (1 << 15) - 1), 4);
    height = (pitch == 0) ? 1 : size / pitch;
-   ok = intelEmitCopyBlit(intel, 1,
+   ok = intelEmitCopyBlit(brw, 1,
                          pitch, src_bo, src_offset, I915_TILING_NONE,
                          pitch, dst_bo, dst_offset, I915_TILING_NONE,
                          0, 0, /* src x/y */
@@ -695,7 +556,7 @@ intel_emit_linear_blit(struct intel_context *intel,
    assert (size < (1 << 15));
    pitch = ALIGN(size, 4);
    if (size != 0) {
-      ok = intelEmitCopyBlit(intel, 1,
+      ok = intelEmitCopyBlit(brw, 1,
                             pitch, src_bo, src_offset, I915_TILING_NONE,
                             pitch, dst_bo, dst_offset, I915_TILING_NONE,
                             0, 0, /* src x/y */
@@ -716,55 +577,58 @@ intel_emit_linear_blit(struct intel_context *intel,
  * miptree.
  */
 static void
-intel_miptree_set_alpha_to_one(struct intel_context *intel,
+intel_miptree_set_alpha_to_one(struct brw_context *brw,
                               struct intel_mipmap_tree *mt,
                               int x, int y, int width, int height)
 {
-   struct intel_region *region = mt->region;
    uint32_t BR13, CMD;
    int pitch, cpp;
    drm_intel_bo *aper_array[2];
-   BATCH_LOCALS;
 
-   pitch = region->pitch;
-   cpp = region->cpp;
+   pitch = mt->pitch;
+   cpp = mt->cpp;
 
    DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
-       __FUNCTION__, region->bo, pitch, x, y, width, height);
+       __FUNCTION__, mt->bo, pitch, x, y, width, height);
 
    BR13 = br13_for_cpp(cpp) | 0xf0 << 16;
    CMD = XY_COLOR_BLT_CMD;
    CMD |= XY_BLT_WRITE_ALPHA;
 
-#ifndef I915
-   if (region->tiling != I915_TILING_NONE) {
+   if (mt->tiling != I915_TILING_NONE) {
       CMD |= XY_DST_TILED;
       pitch /= 4;
    }
-#endif
    BR13 |= pitch;
 
    /* do space check before going any further */
-   aper_array[0] = intel->batch.bo;
-   aper_array[1] = region->bo;
+   aper_array[0] = brw->batch.bo;
+   aper_array[1] = mt->bo;
 
    if (drm_intel_bufmgr_check_aperture_space(aper_array,
                                             ARRAY_SIZE(aper_array)) != 0) {
-      intel_batchbuffer_flush(intel);
+      intel_batchbuffer_flush(brw);
    }
 
-   bool dst_y_tiled = region->tiling == I915_TILING_Y;
+   unsigned length = brw->gen >= 8 ? 7 : 6;
+   bool dst_y_tiled = mt->tiling == I915_TILING_Y;
 
-   BEGIN_BATCH_BLT_TILED(6, dst_y_tiled, false);
-   OUT_BATCH(CMD | (6 - 2));
+   BEGIN_BATCH_BLT_TILED(length, dst_y_tiled, false);
+   OUT_BATCH(CMD | (length - 2));
    OUT_BATCH(BR13);
-   OUT_BATCH((y << 16) | x);
-   OUT_BATCH(((y + height) << 16) | (x + width));
-   OUT_RELOC_FENCED(region->bo,
-                   I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-                   0);
+   OUT_BATCH(SET_FIELD(y, BLT_Y) | SET_FIELD(x, BLT_X));
+   OUT_BATCH(SET_FIELD(y + height, BLT_Y) | SET_FIELD(x + width, BLT_X));
+   if (brw->gen >= 8) {
+      OUT_RELOC64(mt->bo,
+                  I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                  0);
+   } else {
+      OUT_RELOC(mt->bo,
+                I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+                0);
+   }
    OUT_BATCH(0xffffffff); /* white, but only alpha gets written */
    ADVANCE_BATCH_TILED(dst_y_tiled, false);
 
-   intel_batchbuffer_emit_mi_flush(intel);
+   intel_batchbuffer_emit_mi_flush(brw);
 }