[Bf-blender-cvs] [737bd549b6e] master: Cycles: Add support for native OptiX curve primitive
Patrick Mours
noreply at git.blender.org
Tue Jul 7 15:39:36 CEST 2020
Commit: 737bd549b6eeee81f0573ad8e305fb8d888d82ec
Author: Patrick Mours
Date: Mon Jul 6 12:25:54 2020 +0200
Branches: master
https://developer.blender.org/rB737bd549b6eeee81f0573ad8e305fb8d888d82ec
Cycles: Add support for native OptiX curve primitive
This patch adds support for the curve primitive from OptiX to Cycles. It's currently hidden
behind a debug option, since there can be some slight rendering differences still (because no
backface culling is performed and something seems off with endcaps). The curve primitive
was added with the OptiX 7.1 SDK and requires a r450 driver or newer, so this also updates
the codebase to be able to build with the new SDK.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D8223
===================================================================
M intern/cycles/blender/addon/properties.py
M intern/cycles/blender/addon/ui.py
M intern/cycles/blender/blender_python.cpp
M intern/cycles/device/device_optix.cpp
M intern/cycles/kernel/bvh/bvh.h
M intern/cycles/kernel/geom/geom_curve_intersect.h
M intern/cycles/kernel/kernels/optix/kernel_optix.cu
M intern/cycles/util/util_debug.cpp
M intern/cycles/util/util_debug.h
===================================================================
diff --git a/intern/cycles/blender/addon/properties.py b/intern/cycles/blender/addon/properties.py
index 0cc419c7025..def35573ec3 100644
--- a/intern/cycles/blender/addon/properties.py
+++ b/intern/cycles/blender/addon/properties.py
@@ -830,6 +830,7 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
debug_use_cuda_split_kernel: BoolProperty(name="Split Kernel", default=False)
debug_optix_cuda_streams: IntProperty(name="CUDA Streams", default=1, min=1)
+ debug_optix_curves_api: BoolProperty(name="Native OptiX Curve Primitive", default=False)
debug_opencl_kernel_type: EnumProperty(
name="OpenCL Kernel Type",
diff --git a/intern/cycles/blender/addon/ui.py b/intern/cycles/blender/addon/ui.py
index bf81a360020..cce3f805b39 100644
--- a/intern/cycles/blender/addon/ui.py
+++ b/intern/cycles/blender/addon/ui.py
@@ -2031,6 +2031,7 @@ class CYCLES_RENDER_PT_debug(CyclesButtonsPanel, Panel):
col = layout.column()
col.label(text="OptiX Flags:")
col.prop(cscene, "debug_optix_cuda_streams")
+ col.prop(cscene, "debug_optix_curves_api")
col.separator()
diff --git a/intern/cycles/blender/blender_python.cpp b/intern/cycles/blender/blender_python.cpp
index 3e595c3ee52..25c77b74ce3 100644
--- a/intern/cycles/blender/blender_python.cpp
+++ b/intern/cycles/blender/blender_python.cpp
@@ -92,6 +92,7 @@ bool debug_flags_sync_from_scene(BL::Scene b_scene)
flags.cuda.split_kernel = get_boolean(cscene, "debug_use_cuda_split_kernel");
/* Synchronize OptiX flags. */
flags.optix.cuda_streams = get_int(cscene, "debug_optix_cuda_streams");
+ flags.optix.curves_api = get_boolean(cscene, "debug_optix_curves_api");
/* Synchronize OpenCL device type. */
switch (get_enum(cscene, "debug_opencl_device_type")) {
case 0:
diff --git a/intern/cycles/device/device_optix.cpp b/intern/cycles/device/device_optix.cpp
index ececca3df53..266222c74c5 100644
--- a/intern/cycles/device/device_optix.cpp
+++ b/intern/cycles/device/device_optix.cpp
@@ -131,8 +131,12 @@ class OptiXDevice : public CUDADevice {
PG_RGEN,
PG_MISS,
PG_HITD, // Default hit group
- PG_HITL, // __BVH_LOCAL__ hit group
PG_HITS, // __SHADOW_RECORD_ALL__ hit group
+ PG_HITL, // __BVH_LOCAL__ hit group (only used for triangles)
+# if OPTIX_ABI_VERSION >= 36
+ PG_HITD_MOTION,
+ PG_HITS_MOTION,
+# endif
# ifdef WITH_CYCLES_DEBUG
PG_EXCP,
# endif
@@ -177,6 +181,7 @@ class OptiXDevice : public CUDADevice {
OptixDeviceContext context = NULL;
OptixModule optix_module = NULL; // All necessary OptiX kernels are in one module
+ OptixModule builtin_modules[2] = {};
OptixPipeline pipelines[NUM_PIPELINES] = {};
bool motion_blur = false;
@@ -264,6 +269,9 @@ class OptiXDevice : public CUDADevice {
// Unload modules
if (optix_module != NULL)
optixModuleDestroy(optix_module);
+ for (unsigned int i = 0; i < 2; ++i)
+ if (builtin_modules[i] != NULL)
+ optixModuleDestroy(builtin_modules[i]);
for (unsigned int i = 0; i < NUM_PIPELINES; ++i)
if (pipelines[i] != NULL)
optixPipelineDestroy(pipelines[i]);
@@ -338,6 +346,12 @@ class OptiXDevice : public CUDADevice {
optixModuleDestroy(optix_module);
optix_module = NULL;
}
+ for (unsigned int i = 0; i < 2; ++i) {
+ if (builtin_modules[i] != NULL) {
+ optixModuleDestroy(builtin_modules[i]);
+ builtin_modules[i] = NULL;
+ }
+ }
for (unsigned int i = 0; i < NUM_PIPELINES; ++i) {
if (pipelines[i] != NULL) {
optixPipelineDestroy(pipelines[i]);
@@ -369,6 +383,18 @@ class OptiXDevice : public CUDADevice {
# endif
pipeline_options.pipelineLaunchParamsVariableName = "__params"; // See kernel_globals.h
+# if OPTIX_ABI_VERSION >= 36
+ pipeline_options.usesPrimitiveTypeFlags = OPTIX_PRIMITIVE_TYPE_FLAGS_TRIANGLE;
+ if (requested_features.use_hair) {
+ if (DebugFlags().optix.curves_api && requested_features.use_hair_thick) {
+ pipeline_options.usesPrimitiveTypeFlags |= OPTIX_PRIMITIVE_TYPE_FLAGS_ROUND_CUBIC_BSPLINE;
+ }
+ else {
+ pipeline_options.usesPrimitiveTypeFlags |= OPTIX_PRIMITIVE_TYPE_FLAGS_CUSTOM;
+ }
+ }
+# endif
+
// Keep track of whether motion blur is enabled, so to enable/disable motion in BVH builds
// This is necessary since objects may be reported to have motion if the Vector pass is
// active, but may still need to be rendered without motion blur if that isn't active as well
@@ -442,6 +468,34 @@ class OptiXDevice : public CUDADevice {
group_descs[PG_HITD].hitgroup.entryFunctionNameIS = "__intersection__curve_ribbon";
group_descs[PG_HITS].hitgroup.entryFunctionNameIS = "__intersection__curve_ribbon";
}
+
+# if OPTIX_ABI_VERSION >= 36
+ if (DebugFlags().optix.curves_api && requested_features.use_hair_thick) {
+ OptixBuiltinISOptions builtin_options;
+ builtin_options.builtinISModuleType = OPTIX_PRIMITIVE_TYPE_ROUND_CUBIC_BSPLINE;
+ builtin_options.usesMotionBlur = false;
+
+ check_result_optix_ret(optixBuiltinISModuleGet(
+ context, &module_options, &pipeline_options, &builtin_options, &builtin_modules[0]));
+
+ group_descs[PG_HITD].hitgroup.moduleIS = builtin_modules[0];
+ group_descs[PG_HITD].hitgroup.entryFunctionNameIS = nullptr;
+ group_descs[PG_HITS].hitgroup.moduleIS = builtin_modules[0];
+ group_descs[PG_HITS].hitgroup.entryFunctionNameIS = nullptr;
+
+ if (motion_blur) {
+ builtin_options.usesMotionBlur = true;
+
+ check_result_optix_ret(optixBuiltinISModuleGet(
+ context, &module_options, &pipeline_options, &builtin_options, &builtin_modules[1]));
+
+ group_descs[PG_HITD_MOTION] = group_descs[PG_HITD];
+ group_descs[PG_HITD_MOTION].hitgroup.moduleIS = builtin_modules[1];
+ group_descs[PG_HITS_MOTION] = group_descs[PG_HITS];
+ group_descs[PG_HITS_MOTION].hitgroup.moduleIS = builtin_modules[1];
+ }
+ }
+# endif
}
if (requested_features.use_subsurface || requested_features.use_shader_raytrace) {
@@ -493,8 +547,14 @@ class OptiXDevice : public CUDADevice {
unsigned int trace_css = stack_size[PG_HITD].cssCH;
// This is based on the maximum of closest-hit and any-hit/intersection programs
trace_css = std::max(trace_css, stack_size[PG_HITD].cssIS + stack_size[PG_HITD].cssAH);
- trace_css = std::max(trace_css, stack_size[PG_HITL].cssIS + stack_size[PG_HITL].cssAH);
trace_css = std::max(trace_css, stack_size[PG_HITS].cssIS + stack_size[PG_HITS].cssAH);
+ trace_css = std::max(trace_css, stack_size[PG_HITL].cssIS + stack_size[PG_HITL].cssAH);
+# if OPTIX_ABI_VERSION >= 36
+ trace_css = std::max(trace_css,
+ stack_size[PG_HITD_MOTION].cssIS + stack_size[PG_HITD_MOTION].cssAH);
+ trace_css = std::max(trace_css,
+ stack_size[PG_HITS_MOTION].cssIS + stack_size[PG_HITS_MOTION].cssAH);
+# endif
OptixPipelineLinkOptions link_options;
link_options.maxTraceDepth = 1;
@@ -503,17 +563,23 @@ class OptiXDevice : public CUDADevice {
# else
link_options.debugLevel = OPTIX_COMPILE_DEBUG_LEVEL_LINEINFO;
# endif
- link_options.overrideUsesMotionBlur = pipeline_options.usesMotionBlur;
+# if OPTIX_ABI_VERSION < 24
+ link_options.overrideUsesMotionBlur = motion_blur;
+# endif
{ // Create path tracing pipeline
OptixProgramGroup pipeline_groups[] = {
- groups[PG_RGEN],
- groups[PG_MISS],
- groups[PG_HITD],
- groups[PG_HITS],
- groups[PG_HITL],
+ groups[PG_RGEN],
+ groups[PG_MISS],
+ groups[PG_HITD],
+ groups[PG_HITS],
+ groups[PG_HITL],
+# if OPTIX_ABI_VERSION >= 36
+ groups[PG_HITD_MOTION],
+ groups[PG_HITS_MOTION],
+# endif
# ifdef WITH_CYCLES_DEBUG
- groups[PG_EXCP],
+ groups[PG_EXCP],
# endif
};
check_result_optix_ret(
@@ -530,8 +596,8 @@ class OptiXDevice : public CUDADevice {
const unsigned int css = stack_size[PG_RGEN].cssRG + link_options.maxTraceDepth * trace_css;
// Set stack size depending on pipeline options
- check_result_optix_ret(optixPipelineSetStackSize(
- pipelines[PIP_PATH_TRACE], 0, 0, css, (pipeline_options.usesMotionBlur ? 3 : 2)));
+ check_result_optix_ret(
+ optixPipelineSetStackSize(pipelines[PIP_PATH_TRACE], 0, 0, css, (motion_blur ? 3 : 2)));
}
// Only need to create shader evaluation pipeline if one of these features is used:
@@ -541,15 +607,19 @@ class OptiXDevice : public CUDADevice {
if (use_shader_eval_pipeline) { // Create shader evaluation pipeline
OptixProgramGroup pipeline_groups[] = {
- groups[PG_BAKE],
- groups[PG_DISP],
- groups[PG_BACK],
- groups[PG_MISS],
- groups[PG_HITD],
- groups[PG_HITS],
- groups[PG_HITL],
+ groups[PG_BAKE],
+ groups[PG_DISP],
+ groups[PG_BACK],
+ groups[PG_MISS],
+ groups[PG_HITD],
+ groups[PG_HITS],
+ groups[PG_HITL],
+# if OPTIX_ABI_VERSION >= 36
+ groups[PG_HITD_MOTION],
+ groups[PG_HITS_MOTION],
+# endif
# ifdef WITH_CYCLES_DEBUG
- groups[PG_EXCP],
+ groups[PG_EXCP],
# endif
};
check_result_optix_ret(
@@ -672,7 +742,11 @@ class OptiXDevice : public CUDADevice {
sbt_params.missRecordCount = 1;
sbt_params.hitgroupRecordBase = sbt_data.device_pointer + PG_HITD * sizeof(SbtRecord);
sbt_params.hitgroupRecordStrideInBytes = sizeof(SbtRecord);
- sbt_params.hitgroupRecordCount = 3; // PG_HITD, PG_HITL, PG_HITS
+# if OPTIX_ABI_VERSION >= 36
+ sbt_params.hitgroupRecordCount = 5; // PG_HITD(_MOTION), PG_HITS(_MOTION), PG_HITL
+# else
+ sbt_params.hitgroupRecordCount = 3; // PG_HITD, PG_HITS, PG_HITL
+# endif
// Launch the ray generation program
check_result_optix(optixLaunch(pipelines[PIP_PATH_TRACE],
@@ -836,7 +910,9 @@ class OptiXDevice : public CUDADevice {
assert(task.den
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list