X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;ds=sidebyside;f=src%2Fgallium%2Fdrivers%2Fllvmpipe%2Flp_rast.h;h=c209f47f0f5e2e82841c42fa10aac8ccc136e3d7;hb=6f7371619c5edcde647ae8f4822d1b5bae050ef6;hp=3c8dae6b01f3d1a067e229dc90bd9c65dfd016d1;hpb=f25836d7b2c21e046a725cf13c8649d3981693b7;p=mesa.git

diff --git a/src/gallium/drivers/llvmpipe/lp_rast.h b/src/gallium/drivers/llvmpipe/lp_rast.h
index 3c8dae6b01f..c209f47f0f5 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast.h
@@ -38,6 +38,7 @@
 #define LP_RAST_H
 
 #include "pipe/p_compiler.h"
+#include "util/u_pack_color.h"
 #include "lp_jit.h"
 
 
@@ -46,10 +47,28 @@ struct lp_scene;
 struct lp_fence;
 struct cmd_bin;
 
+#define FIXED_TYPE_WIDTH 64
 /** For sub-pixel positioning */
-#define FIXED_ORDER 4
+#define FIXED_ORDER 8
 #define FIXED_ONE (1<<FIXED_ORDER)
+#define FIXED_SHIFT (FIXED_TYPE_WIDTH - 1)
+/** Maximum length of an edge in a primitive in pixels.
+ *  If the framebuffer is large we have to think about fixed-point
+ *  integer overflow. Coordinates need ((FIXED_TYPE_WIDTH/2) - 1) bits
+ *  to be able to fit product of two such coordinates inside
+ *  FIXED_TYPE_WIDTH, any larger and we could overflow a
+ *  FIXED_TYPE_WIDTH_-bit int.
+ */
+#define MAX_FIXED_LENGTH (1 << (((FIXED_TYPE_WIDTH/2) - 1) - FIXED_ORDER))
+
+#define MAX_FIXED_LENGTH32 (1 << (((32/2) - 1) - FIXED_ORDER))
+
+/* Rasterizer output size going to jit fs, width/height */
+#define LP_RASTER_BLOCK_SIZE 4
 
+#define LP_MAX_ACTIVE_BINNED_QUERIES 64
+
+#define IMUL64(a, b) (((int64_t)(a)) * ((int64_t)(b)))
 
 struct lp_rasterizer_task;
 
@@ -78,30 +97,25 @@ struct lp_rast_state {
  * These pointers point into the bin data buffer.
  */
 struct lp_rast_shader_inputs {
-   float facing;     /** Positive for front-facing, negative for back-facing */
-   boolean disable:1;  /** Partially binned, disable this command */
-   boolean opaque:1;   /** Is opaque */
-
-   float (*a0)[4];
-   float (*dadx)[4];
-   float (*dady)[4];
-
-   const struct lp_rast_state *state;
+   unsigned frontfacing:1;      /** True for front-facing */
+   unsigned disable:1;          /** Partially binned, disable this command */
+   unsigned opaque:1;           /** Is opaque */
+   unsigned pad0:29;            /* wasted space */
+   unsigned stride;             /* how much to advance data between a0, dadx, dady */
+   unsigned layer;              /* the layer to render to (from gs, already clamped) */
+   unsigned viewport_index;     /* the active viewport index (from gs, already clamped) */
+   /* followed by a0, dadx, dady and planes[] */
 };
 
-
 struct lp_rast_plane {
-   /* one-pixel sized trivial accept offsets for each plane */
-   int ei;
-
-   /* one-pixel sized trivial reject offsets for each plane */
-   int eo;
-
    /* edge function values at minx,miny ?? */
-   int c;
+   int64_t c;
 
-   int dcdx;
-   int dcdy;
+   int32_t dcdx;
+   int32_t dcdy;
+
+   /* one-pixel sized trivial reject offsets for each plane */
+   int64_t eo;
 };
 
 /**
@@ -111,17 +125,30 @@ struct lp_rast_plane {
  * Objects of this type are put into the lp_setup_context::data buffer.
  */
 struct lp_rast_triangle {
-   /* inputs for the shader */
-   struct lp_rast_shader_inputs inputs;
-
 #ifdef DEBUG
    float v[3][2];
+   float pad0;
+   float pad1;
 #endif
 
-   struct lp_rast_plane plane[8]; /* NOTE: may allocate fewer planes */
+   /* inputs for the shader */
+   struct lp_rast_shader_inputs inputs;
+   /* planes are also allocated here */
 };
 
 
+struct lp_rast_clear_rb {
+   union util_color color_val;
+   unsigned cbuf;
+};
+
+
+#define GET_A0(inputs) ((float (*)[4])((inputs)+1))
+#define GET_DADX(inputs) ((float (*)[4])((char *)((inputs) + 1) + (inputs)->stride))
+#define GET_DADY(inputs) ((float (*)[4])((char *)((inputs) + 1) + 2 * (inputs)->stride))
+#define GET_PLANES(tri) ((struct lp_rast_plane *)((char *)(&(tri)->inputs + 1) + 3 * (tri)->inputs.stride))
+
+
 
 struct lp_rasterizer *
 lp_rast_create( unsigned num_threads );
@@ -129,9 +156,6 @@ lp_rast_create( unsigned num_threads );
 void
 lp_rast_destroy( struct lp_rasterizer * );
 
-unsigned
-lp_rast_get_num_threads( struct lp_rasterizer * );
-
 void 
 lp_rast_queue_scene( struct lp_rasterizer *rast,
                      struct lp_scene *scene );
@@ -147,11 +171,12 @@ union lp_rast_cmd_arg {
       unsigned plane_mask;
    } triangle;
    const struct lp_rast_state *set_state;
-   uint8_t clear_color[4];
+   const struct lp_rast_clear_rb *clear_rb;
    struct {
-      unsigned value;
-      unsigned mask;
+      uint64_t value;
+      uint64_t mask;
    } clear_zstencil;
+   const struct lp_rast_state *state;
    struct lp_fence *fence;
    struct llvmpipe_query *query_obj;
 };
@@ -177,6 +202,22 @@ lp_rast_arg_triangle( const struct lp_rast_triangle *triangle,
    return arg;
 }
 
+/**
+ * Build argument for a contained triangle.
+ *
+ * All planes are enabled, so instead of the plane mask we pass the upper
+ * left coordinates of the a block that fully encloses the triangle.
+ */
+static INLINE union lp_rast_cmd_arg
+lp_rast_arg_triangle_contained( const struct lp_rast_triangle *triangle,
+                                unsigned x, unsigned y)
+{
+   union lp_rast_cmd_arg arg;
+   arg.triangle.tri = triangle;
+   arg.triangle.plane_mask = x | (y << 8);
+   return arg;
+}
+
 static INLINE union lp_rast_cmd_arg
 lp_rast_arg_state( const struct lp_rast_state *state )
 {
@@ -195,7 +236,7 @@ lp_rast_arg_fence( struct lp_fence *fence )
 
 
 static INLINE union lp_rast_cmd_arg
-lp_rast_arg_clearzs( unsigned value, unsigned mask )
+lp_rast_arg_clearzs( uint64_t value, uint64_t mask )
 {
    union lp_rast_cmd_arg arg;
    arg.clear_zstencil.value = value;
@@ -236,13 +277,27 @@ lp_rast_arg_null( void )
 #define LP_RAST_OP_TRIANGLE_6        0x7
 #define LP_RAST_OP_TRIANGLE_7        0x8
 #define LP_RAST_OP_TRIANGLE_8        0x9
-#define LP_RAST_OP_TRIANGLE_3_16     0xa
-#define LP_RAST_OP_SHADE_TILE        0xb
-#define LP_RAST_OP_SHADE_TILE_OPAQUE 0xc
-#define LP_RAST_OP_BEGIN_QUERY       0xd
-#define LP_RAST_OP_END_QUERY         0xe
-
-#define LP_RAST_OP_MAX               0xf
+#define LP_RAST_OP_TRIANGLE_3_4      0xa
+#define LP_RAST_OP_TRIANGLE_3_16     0xb
+#define LP_RAST_OP_TRIANGLE_4_16     0xc
+#define LP_RAST_OP_SHADE_TILE        0xd
+#define LP_RAST_OP_SHADE_TILE_OPAQUE 0xe
+#define LP_RAST_OP_BEGIN_QUERY       0xf
+#define LP_RAST_OP_END_QUERY         0x10
+#define LP_RAST_OP_SET_STATE         0x11
+#define LP_RAST_OP_TRIANGLE_32_1     0x12
+#define LP_RAST_OP_TRIANGLE_32_2     0x13
+#define LP_RAST_OP_TRIANGLE_32_3     0x14
+#define LP_RAST_OP_TRIANGLE_32_4     0x15
+#define LP_RAST_OP_TRIANGLE_32_5     0x16
+#define LP_RAST_OP_TRIANGLE_32_6     0x17
+#define LP_RAST_OP_TRIANGLE_32_7     0x18
+#define LP_RAST_OP_TRIANGLE_32_8     0x19
+#define LP_RAST_OP_TRIANGLE_32_3_4   0x1a
+#define LP_RAST_OP_TRIANGLE_32_3_16  0x1b
+#define LP_RAST_OP_TRIANGLE_32_4_16  0x1c
+
+#define LP_RAST_OP_MAX               0x1d
 #define LP_RAST_OP_MASK              0xff
 
 void
@@ -252,4 +307,18 @@ lp_debug_draw_bins_by_cmd_length( struct lp_scene *scene );
 void
 lp_debug_draw_bins_by_coverage( struct lp_scene *scene );
 
+
+#ifdef PIPE_ARCH_SSE
+#include <emmintrin.h>
+#include "util/u_sse.h"
+
+static INLINE __m128i
+lp_plane_to_m128i(const struct lp_rast_plane *plane)
+{
+   return _mm_setr_epi32((int32_t)plane->c, (int32_t)plane->dcdx,
+                         (int32_t)plane->dcdy, (int32_t)plane->eo);
+}
+
+#endif
+
 #endif