i965: only try print GLSL IR once when using INTEL_DEBUG to dump ir
[mesa.git] / src / mesa / drivers / dri / i965 / gen8_sf_state.c
index 8d8b00dd3ea300224ce9a475add617ccff8ccbf3..5d77b39819d3ec076b5fbcf1d62284096ca819ff 100644 (file)
@@ -21,6 +21,7 @@
  * IN THE SOFTWARE.
  */
 
+#include "compiler/nir/nir.h"
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
@@ -33,16 +34,21 @@ static void
 upload_sbe(struct brw_context *brw)
 {
    struct gl_context *ctx = &brw->ctx;
-   /* CACHE_NEW_WM_PROG */
-   uint32_t num_outputs = brw->wm.prog_data->num_varying_inputs;
+   /* BRW_NEW_FS_PROG_DATA */
+   const struct brw_wm_prog_data *wm_prog_data =
+      brw_wm_prog_data(brw->wm.base.prog_data);
+   uint32_t num_outputs = wm_prog_data->num_varying_inputs;
    uint16_t attr_overrides[VARYING_SLOT_MAX];
    uint32_t urb_entry_read_length;
+   uint32_t urb_entry_read_offset;
    uint32_t point_sprite_enables;
-   uint32_t flat_enables;
+   int sbe_cmd_length;
 
    uint32_t dw1 =
       GEN7_SBE_SWIZZLE_ENABLE |
       num_outputs << GEN7_SBE_NUM_OUTPUTS_SHIFT;
+   uint32_t dw4 = 0;
+   uint32_t dw5 = 0;
 
    /* _NEW_BUFFERS */
    bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
@@ -57,13 +63,15 @@ upload_sbe(struct brw_context *brw)
    else
       dw1 |= GEN6_SF_POINT_SPRITE_UPPERLEFT;
 
-   /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM |
-    * CACHE_NEW_WM_PROG
+   /* _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM,
+    * BRW_NEW_FS_PROG_DATA | BRW_NEW_FRAGMENT_PROGRAM |
+    * BRW_NEW_GS_PROG_DATA | BRW_NEW_PRIMITIVE | BRW_NEW_TES_PROG_DATA |
+    * BRW_NEW_VUE_MAP_GEOM_OUT
     */
    calculate_attr_overrides(brw, attr_overrides,
                             &point_sprite_enables,
-                            &flat_enables,
-                            &urb_entry_read_length);
+                            &urb_entry_read_length,
+                            &urb_entry_read_offset);
 
    /* Typically, the URB entry read length and offset should be programmed in
     * 3DSTATE_VS and 3DSTATE_GS; SBE inherits it from the last active stage
@@ -75,15 +83,42 @@ upload_sbe(struct brw_context *brw)
     */
    dw1 |=
       urb_entry_read_length << GEN7_SBE_URB_ENTRY_READ_LENGTH_SHIFT |
-      BRW_SF_URB_ENTRY_READ_OFFSET << GEN8_SBE_URB_ENTRY_READ_OFFSET_SHIFT |
+      urb_entry_read_offset << GEN8_SBE_URB_ENTRY_READ_OFFSET_SHIFT |
       GEN8_SBE_FORCE_URB_ENTRY_READ_LENGTH |
       GEN8_SBE_FORCE_URB_ENTRY_READ_OFFSET;
 
-   BEGIN_BATCH(4);
-   OUT_BATCH(_3DSTATE_SBE << 16 | (4 - 2));
+   if (brw->gen == 8) {
+      sbe_cmd_length = 4;
+   } else {
+      sbe_cmd_length = 6;
+
+      /* prepare the active component dwords */
+      int input_index = 0;
+      for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) {
+         if (!(brw->fragment_program->info.inputs_read &
+               BITFIELD64_BIT(attr))) {
+            continue;
+         }
+
+         assert(input_index < 32);
+
+         if (input_index < 16)
+            dw4 |= (GEN9_SBE_ACTIVE_COMPONENT_XYZW << (input_index << 1));
+         else
+            dw5 |= (GEN9_SBE_ACTIVE_COMPONENT_XYZW << ((input_index - 16) << 1));
+
+         ++input_index;
+      }
+   }
+   BEGIN_BATCH(sbe_cmd_length);
+   OUT_BATCH(_3DSTATE_SBE << 16 | (sbe_cmd_length - 2));
    OUT_BATCH(dw1);
    OUT_BATCH(point_sprite_enables);
-   OUT_BATCH(flat_enables);
+   OUT_BATCH(wm_prog_data->flat_inputs);
+   if (sbe_cmd_length >= 6) {
+      OUT_BATCH(dw4);
+      OUT_BATCH(dw5);
+   }
    ADVANCE_BATCH();
 
    BEGIN_BATCH(11);
@@ -101,11 +136,18 @@ upload_sbe(struct brw_context *brw)
 
 const struct brw_tracked_state gen8_sbe_state = {
    .dirty = {
-      .mesa  = _NEW_BUFFERS | _NEW_LIGHT | _NEW_POINT | _NEW_PROGRAM,
-      .brw   = BRW_NEW_CONTEXT |
+      .mesa  = _NEW_BUFFERS |
+               _NEW_LIGHT |
+               _NEW_POINT |
+               _NEW_POLYGON |
+               _NEW_PROGRAM,
+      .brw   = BRW_NEW_BLORP |
+               BRW_NEW_CONTEXT |
                BRW_NEW_FRAGMENT_PROGRAM |
+               BRW_NEW_FS_PROG_DATA |
+               BRW_NEW_GS_PROG_DATA |
+               BRW_NEW_TES_PROG_DATA |
                BRW_NEW_VUE_MAP_GEOM_OUT,
-      .cache = CACHE_NEW_WM_PROG,
    },
    .emit = upload_sbe,
 };
@@ -117,30 +159,35 @@ upload_sf(struct brw_context *brw)
    uint32_t dw1 = 0, dw2 = 0, dw3 = 0;
    float point_size;
 
-   dw1 = GEN6_SF_STATISTICS_ENABLE | GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
+   dw1 = GEN6_SF_STATISTICS_ENABLE;
+
+   if (brw->sf.viewport_transform_enable)
+       dw1 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
 
    /* _NEW_LINE */
-   uint32_t line_width_u3_7 = U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7);
-   if (line_width_u3_7 == 0)
-      line_width_u3_7 = 1;
-   dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
+   uint32_t line_width_u3_7 = brw_get_line_width(brw);
+   if (brw->gen >= 9 || brw->is_cherryview) {
+      dw1 |= line_width_u3_7 << GEN9_SF_LINE_WIDTH_SHIFT;
+   } else {
+      dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
+   }
 
    if (ctx->Line.SmoothFlag) {
       dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
    }
 
-   /* Clamp to ARB_point_parameters user limits */
+   /* _NEW_POINT - Clamp to ARB_point_parameters user limits */
    point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
 
    /* Clamp to the hardware limits and convert to fixed point */
-   dw3 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
+   dw3 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3);
 
-   /* _NEW_PROGRAM | _NEW_POINT */
-   if (!(ctx->VertexProgram.PointSizeEnabled || ctx->Point._Attenuated))
+   /* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */
+   if (use_state_point_size(brw))
       dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH;
 
    /* _NEW_POINT | _NEW_MULTISAMPLE */
-   if ((ctx->Point.SmoothFlag || ctx->Multisample._Enabled) &&
+   if ((ctx->Point.SmoothFlag || _mesa_is_multisample_enabled(ctx)) &&
        !ctx->Point.PointSprite) {
       dw3 |= GEN8_SF_SMOOTH_POINT_ENABLE;
    }
@@ -171,8 +218,9 @@ const struct brw_tracked_state gen8_sf_state = {
                _NEW_LINE |
                _NEW_MULTISAMPLE |
                _NEW_POINT,
-      .brw   = BRW_NEW_CONTEXT,
-      .cache = 0,
+      .brw   = BRW_NEW_BLORP |
+               BRW_NEW_CONTEXT |
+               BRW_NEW_VUE_MAP_GEOM_OUT,
    },
    .emit = upload_sf,
 };
@@ -187,7 +235,7 @@ upload_raster(struct brw_context *brw)
    bool render_to_fbo = _mesa_is_user_fbo(brw->ctx.DrawBuffer);
 
    /* _NEW_POLYGON */
-   if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
+   if (ctx->Polygon._FrontBit == render_to_fbo)
       dw1 |= GEN8_RASTER_FRONT_WINDING_CCW;
 
    if (ctx->Polygon.CullFlag) {
@@ -212,7 +260,7 @@ upload_raster(struct brw_context *brw)
    if (ctx->Point.SmoothFlag)
       dw1 |= GEN8_RASTER_SMOOTH_POINT_ENABLE;
 
-   if (ctx->Multisample._Enabled)
+   if (_mesa_is_multisample_enabled(ctx))
       dw1 |= GEN8_RASTER_API_MULTISAMPLE_ENABLE;
 
    if (ctx->Polygon.OffsetFill)
@@ -262,15 +310,21 @@ upload_raster(struct brw_context *brw)
       dw1 |= GEN8_RASTER_SCISSOR_ENABLE;
 
    /* _NEW_TRANSFORM */
-   if (!ctx->Transform.DepthClamp)
-      dw1 |= GEN8_RASTER_VIEWPORT_Z_CLIP_TEST_ENABLE;
+   if (!ctx->Transform.DepthClamp) {
+      if (brw->gen >= 9) {
+         dw1 |= GEN9_RASTER_VIEWPORT_Z_NEAR_CLIP_TEST_ENABLE |
+                GEN9_RASTER_VIEWPORT_Z_FAR_CLIP_TEST_ENABLE;
+      } else {
+         dw1 |= GEN8_RASTER_VIEWPORT_Z_CLIP_TEST_ENABLE;
+      }
+   }
 
    BEGIN_BATCH(5);
    OUT_BATCH(_3DSTATE_RASTER << 16 | (5 - 2));
    OUT_BATCH(dw1);
    OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant.  copied from gen4 */
    OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
-   OUT_BATCH_F(0.0);
+   OUT_BATCH_F(ctx->Polygon.OffsetClamp); /* global depth offset clamp */
    ADVANCE_BATCH();
 }
 
@@ -283,8 +337,8 @@ const struct brw_tracked_state gen8_raster_state = {
                _NEW_POLYGON |
                _NEW_SCISSOR |
                _NEW_TRANSFORM,
-      .brw   = BRW_NEW_CONTEXT,
-      .cache = 0,
+      .brw   = BRW_NEW_BLORP |
+               BRW_NEW_CONTEXT,
    },
    .emit = upload_raster,
 };