Disable cl_intel_pvc_rt_validation extension

Keep only `cl_intel_rt_production` extension enabled

Author: aratajew

IGC/BiFModule/CMakeLists.txt

@@ -1,6 +1,6 @@
 #=========================== begin_copyright_notice ============================
 #
-# Copyright (C) 2017-2021 Intel Corporation
+# Copyright (C) 2017-2023 Intel Corporation
 #
 # SPDX-License-Identifier: MIT
 #
@@ -481,7 +481,8 @@ set(KHR_DEFINES "cl_khr_f16" "cl_khr_fp64" "cl_khr_gl_msaa_sharing" "cl_khr_mipm
                 "cl_khr_subgroup_extended_types" "cl_khr_subgroup_non_uniform_vote" "cl_khr_subgroup_ballot" "cl_khr_subgroup_shuffle"
                 "cl_khr_subgroup_shuffle_relative" "cl_khr_subgroup_non_uniform_arithmetic" "cl_khr_subgroup_clustered_reduce"
                 "cl_khr_extended_bit_ops" "cl_intel_bit_instructions" "cl_intel_global_float_atomics")
-set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_subgroup_matrix_multiply_accumulate" "cl_intel_subgroup_split_matrix_multiply_accumulate" "cl_intel_pvc_rt_validation")
+set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_subgroup_matrix_multiply_accumulate" "cl_intel_subgroup_split_matrix_multiply_accumulate")
+set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_rt_production")
 set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_subgroup_matrix_multiply_accumulate_tf32")
 set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_subgroup_extended_block_read")
 set(KHR_DEFINES ${KHR_DEFINES} "cl_intel_pvc_lsc_validation")

IGC/BiFModule/Implementation/IGCBiF_Intrinsics.cl

@@ -872,7 +872,11 @@ void  __builtin_IB_system_memfence(bool fence_typed_memory);
 // i64 CAS SLM atomic (PVC+)
 long  __builtin_IB_atomic_cmpxchg_local_i64(__local long*, long, long);
 
-// -----------------helpers for intel_rt_production--------------------
+#if defined(cl_intel_pvc_rt_validation) || defined(cl_intel_rt_production)
+struct rtglobals_t;
+typedef __global struct rtglobals_t *rtglobals_t;
+struct rtfence_t;
+typedef __private struct rtfence_t *rtfence_t;
 
 rtfence_t __builtin_IB_intel_query_rt_fence(intel_ray_query_t rayquery);
 rtglobals_t __builtin_IB_intel_query_rt_globals(intel_ray_query_t rayquery);
@@ -895,25 +899,14 @@ void __builtin_IB_intel_update_ray_query(
     uint ctrl,
     uint bvhLevel);
 
-// --------------------------------------------------------------------
-
-#ifdef cl_intel_pvc_rt_validation
-// PVC raytracing extensions
-
-struct rtglobals_t;
-typedef __global struct rtglobals_t *rtglobals_t;
-struct rtfence_t;
-typedef __private struct rtfence_t *rtfence_t;
-
 void* __builtin_IB_intel_get_rt_stack(rtglobals_t rt_dispatch_globals);
 void* __builtin_IB_intel_get_thread_btd_stack(rtglobals_t rt_dispatch_globals);
 void* __builtin_IB_intel_get_global_btd_stack(rtglobals_t rt_dispatch_globals);
-void* __builtin_IB_intel_get_global_btd_stack(rtglobals_t rt_dispatch_globals);
 rtfence_t __builtin_IB_intel_dispatch_trace_ray_query(
   rtglobals_t rt_dispatch_globals, uint bvh_level, uint traceTayCtrl);
 void __builtin_IB_intel_rt_sync(rtfence_t fence);
 global void* __builtin_IB_intel_get_implicit_dispatch_globals();
-#endif // cl_intel_pvc_rt_validation
+#endif // defined(cl_intel_pvc_rt_validation) || defined(cl_intel_rt_production)
 
 void    __builtin_IB_hdc_uncompressed_write_uchar(__global uchar *buf, uchar val);
 

IGC/BiFModule/Languages/OpenCL/IBiF_Impl.cl

@@ -1,6 +1,6 @@
 /*========================== begin_copyright_notice ============================
 
-Copyright (C) 2017-2021 Intel Corporation
+Copyright (C) 2017-2023 Intel Corporation
 
 SPDX-License-Identifier: MIT
 
@@ -73,7 +73,7 @@ SPDX-License-Identifier: MIT
 #include "IBiF_Atomics.cl"
 
 //*****************************************************************************/
-// Intel raytracing extension
+// Intel production raytracing extension
 //*****************************************************************************/
 #include "IBiF_intel_rt_production.cl"
 

IGC/BiFModule/Languages/OpenCL/PreRelease/IBIF_PreRelease_Impl.cl

@@ -1,6 +1,6 @@
 /*========================== begin_copyright_notice ============================
 
-Copyright (C) 2017-2021 Intel Corporation
+Copyright (C) 2017-2023 Intel Corporation
 
 SPDX-License-Identifier: MIT
 
@@ -46,3 +46,8 @@ SPDX-License-Identifier: MIT
 //*****************************************************************************/
 #include "IBiF_Dot_Product.cl"
 
+//*****************************************************************************/
+// Intel validation raytracing extension
+//*****************************************************************************/
+#include "IBiF_intel_rt_validation.cl"
+

IGC/BiFModule/Languages/OpenCL/PreRelease/IBiF_intel_rt_validation.cl

@@ -0,0 +1,43 @@
+/*========================== begin_copyright_notice ============================
+
+Copyright (C) 2023 Intel Corporation
+
+SPDX-License-Identifier: MIT
+
+============================= end_copyright_notice ===========================*/
+
+#if defined(cl_intel_pvc_rt_validation)
+
+void* intel_get_rt_stack(rtglobals_t rt_dispatch_globals)
+{
+    return __builtin_IB_intel_get_rt_stack(rt_dispatch_globals);
+}
+
+void* intel_get_thread_btd_stack(rtglobals_t rt_dispatch_globals)
+{
+    return __builtin_IB_intel_get_thread_btd_stack(rt_dispatch_globals);
+}
+
+void* intel_get_global_btd_stack(rtglobals_t rt_dispatch_globals)
+{
+    return __builtin_IB_intel_get_global_btd_stack(rt_dispatch_globals);
+}
+
+rtfence_t intel_dispatch_trace_ray_query(
+    rtglobals_t rt_dispatch_globals, uint bvh_level, uint traceRayCtrl)
+{
+    return __builtin_IB_intel_dispatch_trace_ray_query(
+        rt_dispatch_globals, bvh_level, traceRayCtrl);
+}
+
+void intel_rt_sync(rtfence_t fence)
+{
+    return __builtin_IB_intel_rt_sync(fence);
+}
+
+global void* intel_get_implicit_dispatch_globals()
+{
+    return __builtin_IB_intel_get_implicit_dispatch_globals();
+}
+
+#endif // defined(cl_intel_pvc_rt_validation)

IGC/BiFModule/Languages/OpenCL/PreRelease/opencl_cth_pre_release.h

@@ -1,6 +1,6 @@
 /*========================== begin_copyright_notice ============================
 
-Copyright (C) 2017-2021 Intel Corporation
+Copyright (C) 2017-2023 Intel Corporation
 
 SPDX-License-Identifier: MIT
 
@@ -2589,6 +2589,7 @@ void __attribute__((overloadable)) work_group_named_barrier(local NamedBarrier_t
 void __attribute__((overloadable)) work_group_named_barrier(local NamedBarrier_t *barrier, cl_mem_fence_flags flags, memory_scope scope);
 #endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
 
+#if defined(cl_intel_pvc_rt_validation)
 
 struct rtglobals_t;
 typedef __global struct rtglobals_t *rtglobals_t;
@@ -2608,149 +2609,4 @@ void intel_rt_sync(rtfence_t fence);
 
 global void *intel_get_implicit_dispatch_globals();
 
-// ----------- Raytracing production API code -----------
-
-// --- Opaque types ---
-typedef private struct intel_ray_query_opaque_t *intel_ray_query_t;
-typedef global struct intel_raytracing_acceleration_structure_opaque_t *intel_raytracing_acceleration_structure_t;
-
-// --- Enum and struct definitions ---
-
-typedef enum // intel_ray_flags_t
-{
-    intel_ray_flags_none = 0x00,
-    intel_ray_flags_force_opaque =
-        0x01, // forces geometry to be opaque (no anyhit shader invokation)
-    intel_ray_flags_force_non_opaque =
-        0x02, // forces geometry to be non-opqaue (invoke anyhit shader)
-    intel_ray_flags_accept_first_hit_and_end_search =
-        0x04, // terminates traversal on the first hit found (shadow rays)
-    intel_ray_flags_skip_closest_hit_shader =
-        0x08, // skip execution of the closest hit shader
-    intel_ray_flags_cull_back_facing_triangles =
-        0x10, // back facing triangles to not produce a hit
-    intel_ray_flags_cull_front_facing_triangles =
-        0x20,                               // front facing triangles do not produce a hit
-    intel_ray_flags_cull_opaque     = 0x40, // opaque geometry does not produce a hit
-    intel_ray_flags_cull_non_opaque = 0x80, // non-opaque geometry does not produce a hit
-    intel_ray_flags_skip_triangles = 0x100, // treat all triangle intersections as misses.
-    intel_ray_flags_skip_procedural_primitives =
-        0x200, // skip execution of intersection shaders
-} intel_ray_flags_t;
-
-typedef enum intel_hit_type_t
-{
-    intel_hit_type_committed_hit = 0,
-    intel_hit_type_potential_hit = 1,
-} intel_hit_type_t;
-
-typedef enum
-{
-    intel_raytracing_ext_flag_ray_query   = 1 << 0, // true if ray queries are supported
-} intel_raytracing_ext_flag_t;
-
-typedef struct // intel_float2
-{
-    float x, y;
-} intel_float2;
-
-typedef struct // intel_float3
-{
-    float x, y, z;
-} intel_float3;
-
-typedef struct // intel_float4x3
-{
-    intel_float3 vx, vy, vz, p;
-} intel_float4x3;
-
-typedef struct // intel_ray_desc_t
-{
-    intel_float3      origin;
-    intel_float3      direction;
-    float             tmin;
-    float             tmax;
-    uint              mask;
-    intel_ray_flags_t flags;
-} intel_ray_desc_t;
-
-// if traversal returns one can test if a triangle or procedural is hit
-typedef enum // intel_candidate_type_t
-{
-    intel_candidate_type_triangle,
-    intel_candidate_type_procedural
-} intel_candidate_type_t;
-
-// --- API functions ---
-
-// check supported ray tracing features
-intel_raytracing_ext_flag_t intel_get_raytracing_ext_flag();
-
-// initialize a ray query
-intel_ray_query_t intel_ray_query_init(
-    intel_ray_desc_t ray, intel_raytracing_acceleration_structure_t accel);
-
-// setup for instance traversal using a transformed ray and bottom-level AS
-void intel_ray_query_forward_ray(
-    intel_ray_query_t                         query,
-    intel_ray_desc_t                          ray,
-    intel_raytracing_acceleration_structure_t accel);
-
-// commit the potential hit
-void intel_ray_query_commit_potential_hit(intel_ray_query_t query);
-
-// commit the potential hit and override hit distance and UVs
-void intel_ray_query_commit_potential_hit_override(
-    intel_ray_query_t query, float override_hit_distance, intel_float2 override_uv);
-
-// start traversal of a ray query
-void intel_ray_query_start_traversal(intel_ray_query_t query);
-
-// Synchronize ray_query execution. If a ray was traversed,
-// this must be called prior to accessing the ray query.
-void intel_ray_query_sync(intel_ray_query_t query);
-
-// Signal that a ray query will not be used further. This is the moral
-// equivalent of a delete. This function does an implicit sync.
-void intel_ray_query_abandon(intel_ray_query_t query);
-
-// read hit information during shader execution
-uint intel_get_hit_bvh_level(intel_ray_query_t query, intel_hit_type_t hit_type);
-float intel_get_hit_distance(intel_ray_query_t query, intel_hit_type_t hit_type);
-intel_float2 intel_get_hit_barycentrics(intel_ray_query_t query, intel_hit_type_t hit_type);
-bool intel_get_hit_front_face(intel_ray_query_t query, intel_hit_type_t hit_type);
-uint intel_get_hit_geometry_id(intel_ray_query_t query, intel_hit_type_t hit_type);
-uint intel_get_hit_primitive_id(intel_ray_query_t query, intel_hit_type_t hit_type);
-uint intel_get_hit_triangle_primitive_id(
-    intel_ray_query_t query,
-    intel_hit_type_t  hit_type); // fast path for triangles
-uint intel_get_hit_procedural_primitive_id(
-    intel_ray_query_t query,
-    intel_hit_type_t  hit_type); // fast path for procedurals
-uint intel_get_hit_instance_id(intel_ray_query_t query, intel_hit_type_t hit_type);
-uint intel_get_hit_instance_user_id(intel_ray_query_t query, intel_hit_type_t hit_type);
-intel_float4x3 intel_get_hit_world_to_object(intel_ray_query_t query, intel_hit_type_t hit_type);
-intel_float4x3 intel_get_hit_object_to_world(intel_ray_query_t query, intel_hit_type_t hit_type);
-
-intel_candidate_type_t intel_get_hit_candidate(intel_ray_query_t query, intel_hit_type_t hit_type);
-
-// fetch triangle vertices for a hit
-void intel_get_hit_triangle_vertices(
-    intel_ray_query_t query, intel_float3 vertices_out[3], intel_hit_type_t hit_type);
-
-// Read ray-data. This is used to read transformed rays produced by HW
-// instancing pipeline during any-hit or intersection shader execution.
-intel_float3 intel_get_ray_origin(intel_ray_query_t query, uint bvh_level);
-intel_float3 intel_get_ray_direction(intel_ray_query_t query, uint bvh_level);
-float intel_get_ray_tmin(intel_ray_query_t query, uint bvh_level);
-intel_ray_flags_t intel_get_ray_flags(intel_ray_query_t query, uint bvh_level);
-int intel_get_ray_mask(intel_ray_query_t query, uint bvh_level);
-
-// Test whether traversal has terminated.  If false, the ray has reached
-// a procedural leaf or a non-opaque triangle leaf, and requires shader
-// processing.
-bool intel_is_traversal_done(intel_ray_query_t query);
-
-// if traversal is done one can test for the presence of a committed hit to
-// either invoke miss or closest hit shader
-bool intel_has_committed_hit(intel_ray_query_t query);
+#endif // defined(cl_intel_pvc_rt_validation)

IGC/BiFModule/Languages/OpenCL/Raytracing/IBiF_intel_rt_production.cl

@@ -1,45 +1,14 @@
 /*========================== begin_copyright_notice ============================
 
-Copyright (C) 2022 Intel Corporation
+Copyright (C) 2022-2023 Intel Corporation
 
 SPDX-License-Identifier: MIT
 
 ============================= end_copyright_notice ===========================*/
 
 #include "IBiF_intel_rt_struct_defs.cl"
 
-
-void* intel_get_rt_stack(rtglobals_t rt_dispatch_globals)
-{
-    return __builtin_IB_intel_get_rt_stack(rt_dispatch_globals);
-}
-
-void* intel_get_thread_btd_stack(rtglobals_t rt_dispatch_globals)
-{
-    return __builtin_IB_intel_get_thread_btd_stack(rt_dispatch_globals);
-}
-
-void* intel_get_global_btd_stack(rtglobals_t rt_dispatch_globals)
-{
-    return __builtin_IB_intel_get_global_btd_stack(rt_dispatch_globals);
-}
-
-rtfence_t intel_dispatch_trace_ray_query(
-    rtglobals_t rt_dispatch_globals, uint bvh_level, uint traceRayCtrl)
-{
-    return __builtin_IB_intel_dispatch_trace_ray_query(
-        rt_dispatch_globals, bvh_level, traceRayCtrl);
-}
-
-void intel_rt_sync(rtfence_t fence)
-{
-    return __builtin_IB_intel_rt_sync(fence);
-}
-
-global void* intel_get_implicit_dispatch_globals()
-{
-    return __builtin_IB_intel_get_implicit_dispatch_globals();
-}
+#if defined(cl_intel_rt_production)
 
 intel_raytracing_ext_flag_t intel_get_raytracing_ext_flag()
 {
@@ -52,9 +21,9 @@ intel_ray_query_t intel_ray_query_init(
     global HWAccel* hwaccel   = to_global((HWAccel*)accel);
     unsigned int    bvh_level = 0;
 
-    rtglobals_t     dispatchGlobalsPtr = (rtglobals_t)intel_get_implicit_dispatch_globals();
+    rtglobals_t     dispatchGlobalsPtr = (rtglobals_t)__builtin_IB_intel_get_implicit_dispatch_globals();
     global RTStack* rtStack =
-        to_global((RTStack*)intel_get_rt_stack((rtglobals_t)dispatchGlobalsPtr));
+        to_global((RTStack*)__builtin_IB_intel_get_rt_stack(dispatchGlobalsPtr));
 
     /* init ray */
     rtStack->ray[bvh_level].org[0] = ray.origin.x;
@@ -197,7 +166,7 @@ void intel_ray_query_start_traversal(intel_ray_query_t rayquery)
 
     uint bvh_level = __builtin_IB_intel_query_bvh_level(rayquery);
 
-    rtfence_t fence = intel_dispatch_trace_ray_query(
+    rtfence_t fence = __builtin_IB_intel_dispatch_trace_ray_query(
         dispatchGlobalsPtr, bvh_level, ctrl);
 
     __builtin_IB_intel_update_ray_query(
@@ -213,7 +182,7 @@ void intel_ray_query_start_traversal(intel_ray_query_t rayquery)
 void intel_ray_query_sync(intel_ray_query_t rayquery)
 {
     rtfence_t fence = __builtin_IB_intel_query_rt_fence(rayquery);
-    intel_rt_sync(fence);
+    __builtin_IB_intel_rt_sync(fence);
 
     global RTStack* rtStack = __builtin_IB_intel_query_rt_stack(rayquery);
 
@@ -445,3 +414,5 @@ bool intel_has_committed_hit(intel_ray_query_t rayquery)
 {
     return MemHit_getValid(get_query_hit(rayquery, intel_hit_type_committed_hit));
 }
+
+#endif // defined(cl_intel_rt_production)