freedreno/ir3: Add ir3 intrinsics for tessellation

These provide the iovas for system memory buffers used for
tessellation as well as a new HW specific system value.

Signed-off-by: Kristian H. Kristensen <hoegsberg@google.com>
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>

diff --git a/src/compiler/nir/nir_intrinsics.py b/src/compiler/nir/nir_intrinsics.py
index 9c1418f2a4ec3328a36a8b1c09e8e654fd2f96c4..55050a9e0583d17597eb1405b2611bd9e7cc01cd 100644 (file)
--- a/src/compiler/nir/nir_intrinsics.py
+++ b/src/compiler/nir/nir_intrinsics.py
@@ -792,6 +792,12 @@ system_value("vs_vertex_stride_ir3", 1)
 system_value("gs_header_ir3", 1)
 system_value("primitive_location_ir3", 1, indices=[DRIVER_LOCATION])
 
+# System values for freedreno tessellation shaders.
+system_value("hs_patch_stride_ir3", 1)
+system_value("tess_factor_base_ir3", 2)
+system_value("tess_param_base_ir3", 2)
+system_value("tcs_header_ir3", 1)
+
 # IR3-specific load/store intrinsics. These access a buffer used to pass data
 # between geometry stages - perhaps it's explicit access to the vertex cache.
 

diff --git a/src/compiler/shader_enums.c b/src/compiler/shader_enums.c
index afaad50adf680eaa2db3a8805a9ae6eeaefc919f..7b908eab6f609af2324be7430a62f0552190b30e 100644 (file)
--- a/src/compiler/shader_enums.c
+++ b/src/compiler/shader_enums.c
@@ -255,6 +255,7 @@ gl_system_value_name(gl_system_value sysval)
      ENUM(SYSTEM_VALUE_BARYCENTRIC_CENTROID),
      ENUM(SYSTEM_VALUE_BARYCENTRIC_SIZE),
      ENUM(SYSTEM_VALUE_GS_HEADER_IR3),
+     ENUM(SYSTEM_VALUE_TCS_HEADER_IR3),
    };
    STATIC_ASSERT(ARRAY_SIZE(names) == SYSTEM_VALUE_MAX);
    return NAME(sysval);

diff --git a/src/compiler/shader_enums.h b/src/compiler/shader_enums.h
index f9b2b8c1d736646759ac463d8a27e54040672e32..07f96f3733cfb116ff1176df220f1cc1bd25e706 100644 (file)
--- a/src/compiler/shader_enums.h
+++ b/src/compiler/shader_enums.h
@@ -642,11 +642,12 @@ typedef enum
    SYSTEM_VALUE_BARYCENTRIC_SIZE,
 
    /**
-    * IR3 specific geometry shader system value that packs invocation id,
-    * thread id and vertex id.  Having this as a nir level system value lets
-    * us do the unpacking in nir.
+    * IR3 specific geometry shader and tesselation control shader system
+    * values that packs invocation id, thread id and vertex id.  Having this
+    * as a nir level system value lets us do the unpacking in nir.
     */
    SYSTEM_VALUE_GS_HEADER_IR3,
+   SYSTEM_VALUE_TCS_HEADER_IR3,
 
    SYSTEM_VALUE_MAX             /**< Number of values */
 } gl_system_value;

diff --git a/src/freedreno/ir3/ir3_compiler_nir.c b/src/freedreno/ir3/ir3_compiler_nir.c
index f7af73f8227c6d1b12316d1be1e899e3505d0cff..f4e94332a2a0179f20384bab362ee77aff5a3eef 100644 (file)
--- a/src/freedreno/ir3/ir3_compiler_nir.c
+++ b/src/freedreno/ir3/ir3_compiler_nir.c
@@ -1363,6 +1363,21 @@ emit_intrinsic(struct ir3_context *ctx, nir_intrinsic_instr *intr)
        case nir_intrinsic_load_vs_vertex_stride_ir3:
                dst[0] = create_uniform(b, primitive_param + 1);
                break;
+       case nir_intrinsic_load_hs_patch_stride_ir3:
+               dst[0] = create_uniform(b, primitive_param + 2);
+               break;
+       case nir_intrinsic_load_patch_vertices_in:
+               dst[0] = create_uniform(b, primitive_param + 3);
+               break;
+       case nir_intrinsic_load_tess_param_base_ir3:
+               dst[0] = create_uniform(b, primitive_param + 4);
+               dst[1] = create_uniform(b, primitive_param + 5);
+               break;
+       case nir_intrinsic_load_tess_factor_base_ir3:
+               dst[0] = create_uniform(b, primitive_param + 6);
+               dst[1] = create_uniform(b, primitive_param + 7);
+               break;
+
        case nir_intrinsic_load_primitive_location_ir3:
                idx = nir_intrinsic_driver_location(intr);
                dst[0] = create_uniform(b, primitive_map + idx);
@@ -1371,6 +1386,9 @@ emit_intrinsic(struct ir3_context *ctx, nir_intrinsic_instr *intr)
        case nir_intrinsic_load_gs_header_ir3:
                dst[0] = ctx->gs_header;
                break;
+       case nir_intrinsic_load_tcs_header_ir3:
+               dst[0] = ctx->tcs_header;
+               break;
 
        case nir_intrinsic_load_primitive_id:
                dst[0] = ctx->primitive_id;

diff --git a/src/freedreno/ir3/ir3_context.h b/src/freedreno/ir3/ir3_context.h
index 2a1f90711180373fe6dd62021c5a3ec4fbef676f..bb283a76326a7f6ff42e6485f4ef626beec63e36 100644 (file)
--- a/src/freedreno/ir3/ir3_context.h
+++ b/src/freedreno/ir3/ir3_context.h
@@ -80,6 +80,11 @@ struct ir3_context {
        struct ir3_instruction *primitive_id;
        struct ir3_instruction *gs_header;
 
+       /* For tessellation shaders: */
+       struct ir3_instruction *patch_vertices_in;
+       struct ir3_instruction *tcs_header;
+       struct ir3_instruction *tess_coord;
+
        /* Compute shader inputs: */
        struct ir3_instruction *local_invocation_id, *work_group_id;
 

diff --git a/src/freedreno/ir3/ir3_shader.c b/src/freedreno/ir3/ir3_shader.c
index 0cfb28e1c495ad276a9b8f93861eba1672975021..4260aee98b324eb4318d051a3c8cb9eadef16526 100644 (file)
--- a/src/freedreno/ir3/ir3_shader.c
+++ b/src/freedreno/ir3/ir3_shader.c
@@ -363,6 +363,8 @@ output_name(struct ir3_shader_variant *so, int i)
                        return "GS_HEADER";
                case VARYING_SLOT_GS_VERTEX_FLAGS_IR3:
                        return "GS_VERTEX_FLAGS";
+               case VARYING_SLOT_TCS_HEADER_IR3:
+                       return "TCS_HEADER";
                default:
                        return gl_varying_slot_name(so->outputs[i].slot);
                }

diff --git a/src/freedreno/ir3/ir3_shader.h b/src/freedreno/ir3/ir3_shader.h
index dd613f2b52cbb79ba9e2512b18456f4cb6f904d9..84a7807a0c897a8c6a1a2a4d4d554ad73bbcb4a3 100644 (file)
--- a/src/freedreno/ir3/ir3_shader.h
+++ b/src/freedreno/ir3/ir3_shader.h
@@ -738,6 +738,7 @@ ir3_find_output_regid(const struct ir3_shader_variant *so, unsigned slot)
 
 #define VARYING_SLOT_GS_HEADER_IR3                     (VARYING_SLOT_MAX + 0)
 #define VARYING_SLOT_GS_VERTEX_FLAGS_IR3       (VARYING_SLOT_MAX + 1)
+#define VARYING_SLOT_TCS_HEADER_IR3                    (VARYING_SLOT_MAX + 2)
 
 
 static inline uint32_t