i965: Rework vertex uploads for Broadwell.
authorKenneth Graunke <kenneth@whitecape.org>
Tue, 4 Dec 2012 02:28:29 +0000 (18:28 -0800)
committerKenneth Graunke <kenneth@whitecape.org>
Sat, 1 Feb 2014 01:50:07 +0000 (17:50 -0800)
v2: Emit a dummy 3DSTATE_VF_SGVS packet when not needed.

v3: Add WARN_ONCE and perf_debugs requested by Eric Anholt.

v4: Program 3DSTATE_SGVS even in the no-elements case so gl_VertexID
    continues working.  Fix 3DSTATE_VF_INSTANCING to not use an
    element index to access the buffers array.  Some ARB_draw_indirect
    prep work.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
src/mesa/drivers/dri/i965/Makefile.sources
src/mesa/drivers/dri/i965/brw_context.h
src/mesa/drivers/dri/i965/brw_defines.h
src/mesa/drivers/dri/i965/brw_draw_upload.c
src/mesa/drivers/dri/i965/brw_state.h
src/mesa/drivers/dri/i965/brw_state_upload.c
src/mesa/drivers/dri/i965/gen8_draw_upload.c [new file with mode: 0644]

index 57afc9a822df3f3bd30d562d1620111525fd96a8..87962d92fd818a917434cb6190cc86da1b6886e7 100644 (file)
@@ -143,6 +143,7 @@ i965_FILES = \
        gen7_wm_surface_state.c \
        gen8_blend_state.c \
        gen8_disasm.c \
+       gen8_draw_upload.c \
        gen8_fs_generator.cpp \
        gen8_generator.cpp \
        gen8_instruction.c \
index 8d098e6c3b7f326d423db8f0b37961b7d2a90dc5..b6a1363a6a1a6e3fde9bac96556d2eb61d50efc2 100644 (file)
@@ -1617,6 +1617,7 @@ gl_clip_plane *brw_select_clip_planes(struct gl_context *ctx);
 unsigned brw_get_vertex_surface_type(struct brw_context *brw,
                                      const struct gl_client_array *glarray);
 unsigned brw_get_index_type(GLenum type);
+void brw_prepare_vertices(struct brw_context *brw);
 
 /* brw_wm_surface_state.c */
 void brw_init_surface_formats(struct brw_context *brw);
index a7f5d1a5f2feb251f89d5119e00c0d39acbd1c08..4496e57311e8c18060118969ad3b9bd6168d4338 100644 (file)
@@ -1409,6 +1409,19 @@ enum brw_message_target {
 #define _3DSTATE_VF                             0x780c /* GEN7.5+ */
 #define HSW_CUT_INDEX_ENABLE                            (1 << 8)
 
+#define _3DSTATE_VF_INSTANCING                  0x7849 /* GEN8+ */
+# define GEN8_VF_INSTANCING_ENABLE                      (1 << 8)
+
+#define _3DSTATE_VF_SGVS                        0x784a /* GEN8+ */
+# define GEN8_SGVS_ENABLE_INSTANCE_ID                   (1 << 31)
+# define GEN8_SGVS_INSTANCE_ID_COMPONENT_SHIFT          29
+# define GEN8_SGVS_INSTANCE_ID_ELEMENT_OFFSET_SHIFT     16
+# define GEN8_SGVS_ENABLE_VERTEX_ID                     (1 << 15)
+# define GEN8_SGVS_VERTEX_ID_COMPONENT_SHIFT            13
+# define GEN8_SGVS_VERTEX_ID_ELEMENT_OFFSET_SHIFT       0
+
+#define _3DSTATE_VF_TOPOLOGY                    0x784b /* GEN8+ */
+
 #define _3DSTATE_URB_VS                         0x7830 /* GEN7+ */
 #define _3DSTATE_URB_HS                         0x7831 /* GEN7+ */
 #define _3DSTATE_URB_DS                         0x7832 /* GEN7+ */
index 5418398afdf6efb5b63e918c12a1908788c23f7d..52d38478d8da9edccd5cc3e501b8065c6fd7f9e2 100644 (file)
@@ -400,7 +400,8 @@ copy_array_to_vbo_array(struct brw_context *brw,
    buffer->stride = dst_stride;
 }
 
-static void brw_prepare_vertices(struct brw_context *brw)
+void
+brw_prepare_vertices(struct brw_context *brw)
 {
    struct gl_context *ctx = &brw->ctx;
    /* CACHE_NEW_VS_PROG */
index 619f4c8d29a4f13e7f8bb3303ca031e946e976f8..dbad4661d3e12f12730e5eab8e152e6641edd6ce 100644 (file)
@@ -131,6 +131,7 @@ extern const struct brw_tracked_state gen7_urb;
 extern const struct brw_tracked_state gen7_vs_state;
 extern const struct brw_tracked_state gen7_wm_state;
 extern const struct brw_tracked_state haswell_cut_index;
+extern const struct brw_tracked_state gen8_index_buffer;
 extern const struct brw_tracked_state gen8_ps_blend;
 extern const struct brw_tracked_state gen8_ps_extra;
 extern const struct brw_tracked_state gen8_ps_state;
@@ -140,6 +141,8 @@ extern const struct brw_tracked_state gen8_raster_state;
 extern const struct brw_tracked_state gen8_sbe_state;
 extern const struct brw_tracked_state gen8_sf_state;
 extern const struct brw_tracked_state gen8_state_base_address;
+extern const struct brw_tracked_state gen8_vertices;
+extern const struct brw_tracked_state gen8_vf_topology;
 extern const struct brw_tracked_state gen8_vs_state;
 
 /* brw_misc_state.c */
index e4f8104bfaba13e65eeca3b72fe1db2be2f5c41f..76698dac3a7c2dd9e50c37f5a51a531a1f207f3b 100644 (file)
@@ -322,9 +322,11 @@ static const struct brw_tracked_state *gen8_atoms[] =
 
    &brw_drawing_rect,
 
+   &gen8_vf_topology,
+
    &brw_indices,
-   &brw_index_buffer,
-   &brw_vertices,
+   &gen8_index_buffer,
+   &gen8_vertices,
 
    &haswell_cut_index,
 };
diff --git a/src/mesa/drivers/dri/i965/gen8_draw_upload.c b/src/mesa/drivers/dri/i965/gen8_draw_upload.c
new file mode 100644 (file)
index 0000000..f927c13
--- /dev/null
@@ -0,0 +1,250 @@
+/*
+ * Copyright © 2012 Intel Corporation
+ *
+ * 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 without limitation
+ * the rights to use, copy, modify, merge, publish, 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 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 NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS 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/glheader.h"
+#include "main/bufferobj.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/macros.h"
+
+#include "brw_draw.h"
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_state.h"
+
+#include "intel_batchbuffer.h"
+#include "intel_buffer_objects.h"
+
+static void
+gen8_emit_vertices(struct brw_context *brw)
+{
+   struct gl_context *ctx = &brw->ctx;
+
+   brw_prepare_vertices(brw);
+
+   if (brw->vs.prog_data->uses_vertexid) {
+      unsigned vue = brw->vb.nr_enabled;
+
+      WARN_ONCE(brw->vs.prog_data->inputs_read & VERT_BIT_EDGEFLAG,
+                "Using VID/IID with edgeflags, need to reorder the "
+                "vertex attributes");
+      WARN_ONCE(vue >= 33,
+                "Trying to insert VID/IID past 33rd vertex element, "
+                "need to reorder the vertex attrbutes.");
+
+      BEGIN_BATCH(2);
+      OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
+      OUT_BATCH(GEN8_SGVS_ENABLE_VERTEX_ID |
+                (0 << GEN8_SGVS_VERTEX_ID_COMPONENT_SHIFT) |   /* .x channel */
+                (vue << GEN8_SGVS_VERTEX_ID_ELEMENT_OFFSET_SHIFT) |
+                GEN8_SGVS_ENABLE_INSTANCE_ID |
+                (1 << GEN8_SGVS_INSTANCE_ID_COMPONENT_SHIFT) | /* .y channel */
+                (vue << GEN8_SGVS_INSTANCE_ID_ELEMENT_OFFSET_SHIFT));
+      ADVANCE_BATCH();
+   } else {
+      BEGIN_BATCH(2);
+      OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
+      OUT_BATCH(0);
+      ADVANCE_BATCH();
+   }
+
+   /* If the VS doesn't read any inputs (calculating vertex position from
+    * a state variable for some reason, for example), emit a single pad
+    * VERTEX_ELEMENT struct and bail.
+    *
+    * The stale VB state stays in place, but they don't do anything unless
+    * a VE loads from them.
+    */
+   if (brw->vb.nr_enabled == 0) {
+      BEGIN_BATCH(3);
+      OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (3 - 2));
+      OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
+                GEN6_VE0_VALID |
+                (BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
+                (0 << BRW_VE0_SRC_OFFSET_SHIFT));
+      OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
+      ADVANCE_BATCH();
+      return;
+   }
+
+   /* Now emit 3DSTATE_VERTEX_BUFFERS and 3DSTATE_VERTEX_ELEMENTS packets. */
+   if (brw->vb.nr_buffers) {
+      assert(brw->vb.nr_buffers <= 33);
+
+      perf_debug("Missing MOCS setup for 3DSTATE_VERTEX_BUFFERS.");
+
+      BEGIN_BATCH(1 + 4*brw->vb.nr_buffers);
+      OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4*brw->vb.nr_buffers - 1));
+      for (unsigned i = 0; i < brw->vb.nr_buffers; i++) {
+         struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
+         uint32_t dw0 = 0;
+
+         dw0 |= i << GEN6_VB0_INDEX_SHIFT;
+         dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
+         dw0 |= buffer->stride << BRW_VB0_PITCH_SHIFT;
+
+         OUT_BATCH(dw0);
+         OUT_RELOC64(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
+         OUT_BATCH(buffer->bo->size);
+      }
+      ADVANCE_BATCH();
+   }
+
+   unsigned nr_elements = brw->vb.nr_enabled;
+
+   /* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS,
+    * presumably for VertexID/InstanceID.
+    */
+   assert(nr_elements <= 34);
+
+   struct brw_vertex_element *gen6_edgeflag_input = NULL;
+
+   BEGIN_BATCH(1 + nr_elements * 2);
+   OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
+   for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
+      struct brw_vertex_element *input = brw->vb.enabled[i];
+      uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
+      uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
+
+      /* The gen4 driver expects edgeflag to come in as a float, and passes
+       * that float on to the tests in the clipper.  Mesa's current vertex
+       * attribute value for EdgeFlag is stored as a float, which works out.
+       * glEdgeFlagPointer, on the other hand, gives us an unnormalized
+       * integer ubyte.  Just rewrite that to convert to a float.
+       */
+      if (input->attrib == VERT_ATTRIB_EDGEFLAG) {
+         /* Gen6+ passes edgeflag as sideband along with the vertex, instead
+          * of in the VUE.  We have to upload it sideband as the last vertex
+          * element according to the B-Spec.
+          */
+         gen6_edgeflag_input = input;
+         continue;
+      }
+
+      switch (input->glarray->Size) {
+      case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
+      case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
+      case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
+      case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
+                                              : BRW_VE1_COMPONENT_STORE_1_FLT;
+         break;
+      }
+
+      OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
+                GEN6_VE0_VALID |
+                (format << BRW_VE0_FORMAT_SHIFT) |
+                (input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
+
+      OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
+                (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
+                (comp3 << BRW_VE1_COMPONENT_3_SHIFT));
+   }
+
+   if (gen6_edgeflag_input) {
+      uint32_t format =
+         brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
+
+      OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
+                GEN6_VE0_VALID |
+                GEN6_VE0_EDGE_FLAG_ENABLE |
+                (format << BRW_VE0_FORMAT_SHIFT) |
+                (gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
+      OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
+                (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
+   }
+   ADVANCE_BATCH();
+
+   for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
+      const struct brw_vertex_element *input = brw->vb.enabled[i];
+      const struct brw_vertex_buffer *buffer = &brw->vb.buffers[input->buffer];
+
+      BEGIN_BATCH(3);
+      OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
+      OUT_BATCH(i | (buffer->step_rate ? GEN8_VF_INSTANCING_ENABLE : 0));
+      OUT_BATCH(buffer->step_rate);
+      ADVANCE_BATCH();
+   }
+}
+
+const struct brw_tracked_state gen8_vertices = {
+   .dirty = {
+      .mesa = _NEW_POLYGON,
+      .brw = BRW_NEW_BATCH | BRW_NEW_VERTICES,
+      .cache = CACHE_NEW_VS_PROG,
+   },
+   .emit = gen8_emit_vertices,
+};
+
+static void
+gen8_emit_index_buffer(struct brw_context *brw)
+{
+   const struct _mesa_index_buffer *index_buffer = brw->ib.ib;
+
+   if (index_buffer == NULL)
+      return;
+
+   perf_debug("Missing MOCS setup for 3DSTATE_INDEX_BUFFER.");
+
+   BEGIN_BATCH(5);
+   OUT_BATCH(CMD_INDEX_BUFFER << 16 | (5 - 2));
+   OUT_BATCH(brw_get_index_type(index_buffer->type) << 8);
+   OUT_RELOC64(brw->ib.bo, I915_GEM_DOMAIN_VERTEX, 0, 0);
+   OUT_BATCH(brw->ib.bo->size);
+   ADVANCE_BATCH();
+}
+
+const struct brw_tracked_state gen8_index_buffer = {
+   .dirty = {
+      .mesa = 0,
+      .brw = BRW_NEW_BATCH | BRW_NEW_INDEX_BUFFER,
+      .cache = 0,
+   },
+   .emit = gen8_emit_index_buffer,
+};
+
+static void
+gen8_emit_vf_topology(struct brw_context *brw)
+{
+   BEGIN_BATCH(2);
+   OUT_BATCH(_3DSTATE_VF_TOPOLOGY << 16 | (2 - 2));
+   OUT_BATCH(brw->primitive);
+   ADVANCE_BATCH();
+}
+
+const struct brw_tracked_state gen8_vf_topology = {
+   .dirty = {
+      .mesa = 0,
+      .brw = BRW_NEW_PRIMITIVE,
+      .cache = 0,
+   },
+   .emit = gen8_emit_vf_topology,
+};