intel: Add a batch flush between front-buffer downsample and X protocol.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_cc.c
index c37376ef0de6a91ea4fa1dd7d6a5fdfedc1f9582..6f85f6355cbd7dc25d62c30759dbec944afe0010 100644 (file)
 #include "brw_defines.h"
 #include "brw_util.h"
 #include "main/macros.h"
+#include "main/stencil.h"
 #include "intel_batchbuffer.h"
 
-void
-brw_update_cc_vp(struct brw_context *brw)
+static void
+brw_upload_cc_vp(struct brw_context *brw)
 {
-   struct gl_context *ctx = &brw->intel.ctx;
-   struct brw_cc_viewport ccv;
+   struct gl_context *ctx = &brw->ctx;
+   struct brw_cc_viewport *ccv;
 
-   memset(&ccv, 0, sizeof(ccv));
+   ccv = brw_state_batch(brw, AUB_TRACE_CC_VP_STATE,
+                        sizeof(*ccv), 32, &brw->cc.vp_offset);
 
-   /* _NEW_TRANSOFORM */
+   /* _NEW_TRANSFORM */
    if (ctx->Transform.DepthClamp) {
       /* _NEW_VIEWPORT */
-      ccv.min_depth = MIN2(ctx->Viewport.Near, ctx->Viewport.Far);
-      ccv.max_depth = MAX2(ctx->Viewport.Near, ctx->Viewport.Far);
+      ccv->min_depth = MIN2(ctx->Viewport.Near, ctx->Viewport.Far);
+      ccv->max_depth = MAX2(ctx->Viewport.Near, ctx->Viewport.Far);
    } else {
-      ccv.min_depth = 0.0;
-      ccv.max_depth = 1.0;
+      ccv->min_depth = 0.0;
+      ccv->max_depth = 1.0;
    }
 
-   drm_intel_bo_unreference(brw->cc.vp_bo);
-   brw->cc.vp_bo = brw_cache_data(&brw->cache, BRW_CC_VP, &ccv, sizeof(ccv));
+   brw->state.dirty.cache |= CACHE_NEW_CC_VP;
 }
 
+const struct brw_tracked_state brw_cc_vp = {
+   .dirty = {
+      .mesa = _NEW_VIEWPORT | _NEW_TRANSFORM,
+      .brw = BRW_NEW_BATCH,
+      .cache = 0
+   },
+   .emit = brw_upload_cc_vp
+};
+
 /**
  * Modify blend function to force destination alpha to 1.0
  *
@@ -66,8 +76,8 @@ brw_update_cc_vp(struct brw_context *brw)
  * replace it with a function that hard-wires destination alpha to 1.0.  This
  * is used when rendering to xRGB targets.
  */
-static GLenum
-fix_xRGB_alpha(GLenum function)
+GLenum
+brw_fix_xRGB_alpha(GLenum function)
 {
    switch (function) {
    case GL_DST_ALPHA:
@@ -81,24 +91,19 @@ fix_xRGB_alpha(GLenum function)
    return function;
 }
 
-static void prepare_cc_unit(struct brw_context *brw)
-{
-   brw_add_validated_bo(brw, brw->cc.vp_bo);
-}
-
 /**
  * Creates the state cache entry for the given CC unit key.
  */
 static void upload_cc_unit(struct brw_context *brw)
 {
-   struct intel_context *intel = &brw->intel;
-   struct gl_context *ctx = &brw->intel.ctx;
+   struct gl_context *ctx = &brw->ctx;
    struct brw_cc_unit_state *cc;
 
-   cc = brw_state_batch(brw, sizeof(*cc), 64, &brw->cc.state_offset);
+   cc = brw_state_batch(brw, AUB_TRACE_CC_STATE,
+                       sizeof(*cc), 64, &brw->cc.state_offset);
    memset(cc, 0, sizeof(*cc));
 
-   /* _NEW_STENCIL */
+   /* _NEW_STENCIL | _NEW_BUFFERS */
    if (ctx->Stencil._Enabled) {
       const unsigned back = ctx->Stencil._BackFace;
 
@@ -111,7 +116,7 @@ static void upload_cc_unit(struct brw_context *brw)
         intel_translate_stencil_op(ctx->Stencil.ZFailFunc[0]);
       cc->cc0.stencil_pass_depth_pass_op =
         intel_translate_stencil_op(ctx->Stencil.ZPassFunc[0]);
-      cc->cc1.stencil_ref = ctx->Stencil.Ref[0];
+      cc->cc1.stencil_ref = _mesa_get_stencil_ref(ctx, 0);
       cc->cc1.stencil_write_mask = ctx->Stencil.WriteMask[0];
       cc->cc1.stencil_test_mask = ctx->Stencil.ValueMask[0];
 
@@ -125,7 +130,7 @@ static void upload_cc_unit(struct brw_context *brw)
            intel_translate_stencil_op(ctx->Stencil.ZFailFunc[back]);
         cc->cc0.bf_stencil_pass_depth_pass_op =
            intel_translate_stencil_op(ctx->Stencil.ZPassFunc[back]);
-        cc->cc1.bf_stencil_ref = ctx->Stencil.Ref[back];
+        cc->cc1.bf_stencil_ref = _mesa_get_stencil_ref(ctx, back);
         cc->cc2.bf_stencil_write_mask = ctx->Stencil.WriteMask[back];
         cc->cc2.bf_stencil_test_mask = ctx->Stencil.ValueMask[back];
       }
@@ -138,7 +143,7 @@ static void upload_cc_unit(struct brw_context *brw)
    }
 
    /* _NEW_COLOR */
-   if (ctx->Color._LogicOpEnabled && ctx->Color.LogicOp != GL_COPY) {
+   if (ctx->Color.ColorLogicOpEnabled && ctx->Color.LogicOp != GL_COPY) {
       cc->cc2.logicop_enable = 1;
       cc->cc5.logicop_func = intel_translate_logic_op(ctx->Color.LogicOp);
    } else if (ctx->Color.BlendEnabled) {
@@ -154,10 +159,10 @@ static void upload_cc_unit(struct brw_context *brw)
        * with GL_ONE and GL_ONE_MINUS_DST_ALPHA with GL_ZERO.
        */
       if (ctx->DrawBuffer->Visual.alphaBits == 0) {
-        srcRGB = fix_xRGB_alpha(srcRGB);
-        srcA   = fix_xRGB_alpha(srcA);
-        dstRGB = fix_xRGB_alpha(dstRGB);
-        dstA   = fix_xRGB_alpha(dstA);
+        srcRGB = brw_fix_xRGB_alpha(srcRGB);
+        srcA   = brw_fix_xRGB_alpha(srcA);
+        dstRGB = brw_fix_xRGB_alpha(dstRGB);
+        dstA   = brw_fix_xRGB_alpha(dstA);
       }
 
       if (eqRGB == GL_MIN || eqRGB == GL_MAX) {
@@ -205,46 +210,43 @@ static void upload_cc_unit(struct brw_context *brw)
       cc->cc2.depth_write_enable = ctx->Depth.Mask;
    }
 
-   if (intel->stats_wm || unlikely(INTEL_DEBUG & DEBUG_STATS))
+   if (brw->stats_wm || unlikely(INTEL_DEBUG & DEBUG_STATS))
       cc->cc5.statistics_enable = 1;
 
    /* CACHE_NEW_CC_VP */
-   cc->cc4.cc_viewport_state_offset = brw->cc.vp_bo->offset >> 5; /* reloc */
+   cc->cc4.cc_viewport_state_offset = (brw->batch.bo->offset +
+                                      brw->cc.vp_offset) >> 5; /* reloc */
 
    brw->state.dirty.cache |= CACHE_NEW_CC_UNIT;
 
    /* Emit CC viewport relocation */
-   drm_intel_bo_emit_reloc(brw->intel.batch.bo,
+   drm_intel_bo_emit_reloc(brw->batch.bo,
                           (brw->cc.state_offset +
                            offsetof(struct brw_cc_unit_state, cc4)),
-                          brw->cc.vp_bo, 0,
+                          brw->batch.bo, brw->cc.vp_offset,
                           I915_GEM_DOMAIN_INSTRUCTION, 0);
 }
 
 const struct brw_tracked_state brw_cc_unit = {
    .dirty = {
-      .mesa = _NEW_STENCIL | _NEW_COLOR | _NEW_DEPTH,
-      .brw = BRW_NEW_BATCH,
+      .mesa = _NEW_STENCIL | _NEW_COLOR | _NEW_DEPTH | _NEW_BUFFERS,
+      .brw = BRW_NEW_BATCH | BRW_NEW_STATS_WM,
       .cache = CACHE_NEW_CC_VP
    },
-   .prepare = prepare_cc_unit,
    .emit = upload_cc_unit,
 };
 
 static void upload_blend_constant_color(struct brw_context *brw)
 {
-   struct gl_context *ctx = &brw->intel.ctx;
-   struct brw_blend_constant_color bcc;
-
-   memset(&bcc, 0, sizeof(bcc));
-   bcc.header.opcode = _3DSTATE_BLEND_CONSTANT_COLOR;
-   bcc.header.length = sizeof(bcc)/4-2;
-   bcc.blend_constant_color[0] = ctx->Color.BlendColor[0];
-   bcc.blend_constant_color[1] = ctx->Color.BlendColor[1];
-   bcc.blend_constant_color[2] = ctx->Color.BlendColor[2];
-   bcc.blend_constant_color[3] = ctx->Color.BlendColor[3];
-
-   BRW_CACHED_BATCH_STRUCT(brw, &bcc);
+   struct gl_context *ctx = &brw->ctx;
+
+   BEGIN_BATCH(5);
+   OUT_BATCH(_3DSTATE_BLEND_CONSTANT_COLOR << 16 | (5-2));
+   OUT_BATCH_F(ctx->Color.BlendColorUnclamped[0]);
+   OUT_BATCH_F(ctx->Color.BlendColorUnclamped[1]);
+   OUT_BATCH_F(ctx->Color.BlendColorUnclamped[2]);
+   OUT_BATCH_F(ctx->Color.BlendColorUnclamped[3]);
+   CACHED_BATCH();
 }
 
 const struct brw_tracked_state brw_blend_constant_color = {