[Bf-blender-cvs] [f5ad9fd] cycles_split_kernel: Cycles: Add names to buffer allocations
Mai Lavelle
noreply at git.blender.org
Fri Dec 16 11:07:41 CET 2016
Commit: f5ad9fd13b7138b583e260d643233145c02d9c92
Author: Mai Lavelle
Date: Tue Dec 13 20:45:09 2016 -0500
Branches: cycles_split_kernel
https://developer.blender.org/rBf5ad9fd13b7138b583e260d643233145c02d9c92
Cycles: Add names to buffer allocations
This is to help debug and track memory usage for generic buffers. We
have similar for textures already since those require a name, but for
buffers the name is only for debugging proposes.
===================================================================
M intern/cycles/device/device.cpp
M intern/cycles/device/device.h
M intern/cycles/device/device_cpu.cpp
M intern/cycles/device/device_cuda.cpp
M intern/cycles/device/device_multi.cpp
M intern/cycles/device/device_split_kernel.cpp
M intern/cycles/device/opencl/opencl.h
M intern/cycles/device/opencl/opencl_base.cpp
M intern/cycles/device/opencl/opencl_split.cpp
M intern/cycles/render/bake.cpp
M intern/cycles/render/buffers.cpp
M intern/cycles/render/light.cpp
M intern/cycles/render/mesh_displace.cpp
===================================================================
diff --git a/intern/cycles/device/device.cpp b/intern/cycles/device/device.cpp
index 31c99f4..6b07b9d 100644
--- a/intern/cycles/device/device.cpp
+++ b/intern/cycles/device/device.cpp
@@ -80,7 +80,7 @@ Device::~Device()
void Device::pixels_alloc(device_memory& mem)
{
- mem_alloc(mem, MEM_READ_WRITE);
+ mem_alloc("pixels", mem, MEM_READ_WRITE);
}
void Device::pixels_copy_from(device_memory& mem, int y, int w, int h)
diff --git a/intern/cycles/device/device.h b/intern/cycles/device/device.h
index 4d270e9..1564d9f 100644
--- a/intern/cycles/device/device.h
+++ b/intern/cycles/device/device.h
@@ -256,7 +256,7 @@ public:
Stats &stats;
/* regular memory */
- virtual void mem_alloc(device_memory& mem, MemoryType type) = 0;
+ virtual void mem_alloc(const char *name, device_memory& mem, MemoryType type) = 0;
virtual void mem_copy_to(device_memory& mem) = 0;
virtual void mem_copy_from(device_memory& mem,
int y, int w, int h, int elem) = 0;
diff --git a/intern/cycles/device/device_cpu.cpp b/intern/cycles/device/device_cpu.cpp
index 07cbaad..3511195 100644
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -237,8 +237,14 @@ public:
return (TaskScheduler::num_threads() == 1);
}
- void mem_alloc(device_memory& mem, MemoryType /*type*/)
+ void mem_alloc(const char *name, device_memory& mem, MemoryType /*type*/)
{
+ if(name) {
+ VLOG(1) << "Buffer allocate: " << name << ", "
+ << string_human_readable_number(mem.memory_size()) << " bytes. ("
+ << string_human_readable_size(mem.memory_size()) << ")";
+ }
+
mem.device_pointer = mem.data_pointer;
if(!mem.device_pointer) {
@@ -824,7 +830,7 @@ protected:
virtual void alloc_kernel_globals(device_memory& mem)
{
mem.resize(sizeof(KernelGlobals));
- mem_alloc(mem, MEM_READ_WRITE);
+ mem_alloc("kernel_globals", mem, MEM_READ_WRITE);
KernelGlobals *kg = (KernelGlobals*)mem.device_pointer;
*kg = thread_kernel_globals_init();
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index 233f94b..1316a3e 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -466,8 +466,14 @@ public:
}
}
- void mem_alloc(device_memory& mem, MemoryType /*type*/)
+ void mem_alloc(const char *name, device_memory& mem, MemoryType /*type*/)
{
+ if(name) {
+ VLOG(1) << "Buffer allocate: " << name << ", "
+ << string_human_readable_number(mem.memory_size()) << " bytes. ("
+ << string_human_readable_size(mem.memory_size()) << ")";
+ }
+
cuda_push_context();
CUdeviceptr device_pointer;
size_t size = mem.memory_size();
@@ -617,7 +623,7 @@ public:
/* Data Storage */
if(interpolation == INTERPOLATION_NONE) {
if(has_bindless_textures) {
- mem_alloc(mem, MEM_READ_ONLY);
+ mem_alloc(NULL, mem, MEM_READ_ONLY);
mem_copy_to(mem);
cuda_push_context();
@@ -641,7 +647,7 @@ public:
cuda_pop_context();
}
else {
- mem_alloc(mem, MEM_READ_ONLY);
+ mem_alloc(NULL, mem, MEM_READ_ONLY);
mem_copy_to(mem);
cuda_push_context();
diff --git a/intern/cycles/device/device_multi.cpp b/intern/cycles/device/device_multi.cpp
index 31b8006..7d658df 100644
--- a/intern/cycles/device/device_multi.cpp
+++ b/intern/cycles/device/device_multi.cpp
@@ -106,11 +106,11 @@ public:
return true;
}
- void mem_alloc(device_memory& mem, MemoryType type)
+ void mem_alloc(const char *name, device_memory& mem, MemoryType type)
{
foreach(SubDevice& sub, devices) {
mem.device_pointer = 0;
- sub.device->mem_alloc(mem, type);
+ sub.device->mem_alloc(name, mem, type);
sub.ptr_map[unique_ptr] = mem.device_pointer;
}
diff --git a/intern/cycles/device/device_split_kernel.cpp b/intern/cycles/device/device_split_kernel.cpp
index 170ffc4..d400bfb 100644
--- a/intern/cycles/device/device_split_kernel.cpp
+++ b/intern/cycles/device/device_split_kernel.cpp
@@ -191,24 +191,24 @@ bool DeviceSplitKernel::path_trace(DeviceTask *task,
/* Allocate work_pool_wgs memory. */
work_pool_wgs.resize(max_work_groups * sizeof(unsigned int));
- device->mem_alloc(work_pool_wgs, MEM_READ_WRITE);
+ device->mem_alloc("work_pool_wgs", work_pool_wgs, MEM_READ_WRITE);
#endif /* __WORK_STEALING__ */
queue_index.resize(NUM_QUEUES * sizeof(int));
- device->mem_alloc(queue_index, MEM_READ_WRITE);
+ device->mem_alloc("queue_index", queue_index, MEM_READ_WRITE);
use_queues_flag.resize(sizeof(char));
- device->mem_alloc(use_queues_flag, MEM_READ_WRITE);
+ device->mem_alloc("use_queues_flag", use_queues_flag, MEM_READ_WRITE);
device->alloc_kernel_globals(kgbuffer);
ray_state.resize(num_global_elements);
- device->mem_alloc(ray_state, MEM_READ_WRITE);
+ device->mem_alloc("ray_state", ray_state, MEM_READ_WRITE);
split_data.resize(split_data_buffer_size(num_global_elements,
current_max_closure,
per_thread_output_buffer_size));
- device->mem_alloc(split_data, MEM_READ_WRITE);
+ device->mem_alloc("split_data", split_data, MEM_READ_WRITE);
}
if(device->have_error()) {
diff --git a/intern/cycles/device/opencl/opencl.h b/intern/cycles/device/opencl/opencl.h
index d42de47..efde79d 100644
--- a/intern/cycles/device/opencl/opencl.h
+++ b/intern/cycles/device/opencl/opencl.h
@@ -248,7 +248,7 @@ public:
virtual bool load_kernels(const DeviceRequestedFeatures& requested_features,
vector<OpenCLProgram*> &programs) = 0;
- void mem_alloc(device_memory& mem, MemoryType type);
+ void mem_alloc(const char *name, device_memory& mem, MemoryType type);
void mem_copy_to(device_memory& mem);
void mem_copy_from(device_memory& mem, int y, int w, int h, int elem);
void mem_zero(device_memory& mem);
diff --git a/intern/cycles/device/opencl/opencl_base.cpp b/intern/cycles/device/opencl/opencl_base.cpp
index 8350f2f..2d42fd0 100644
--- a/intern/cycles/device/opencl/opencl_base.cpp
+++ b/intern/cycles/device/opencl/opencl_base.cpp
@@ -244,8 +244,14 @@ bool OpenCLDeviceBase::load_kernels(const DeviceRequestedFeatures& requested_fea
return true;
}
-void OpenCLDeviceBase::mem_alloc(device_memory& mem, MemoryType type)
+void OpenCLDeviceBase::mem_alloc(const char *name, device_memory& mem, MemoryType type)
{
+ if(name) {
+ VLOG(1) << "Buffer allocate: " << name << ", "
+ << string_human_readable_number(mem.memory_size()) << " bytes. ("
+ << string_human_readable_size(mem.memory_size()) << ")";
+ }
+
size_t size = mem.memory_size();
cl_mem_flags mem_flag;
@@ -339,7 +345,7 @@ void OpenCLDeviceBase::const_copy_to(const char *name, void *host, size_t size)
device_vector<uchar> *data = new device_vector<uchar>();
data->copy((uchar*)host, size);
- mem_alloc(*data, MEM_READ_ONLY);
+ mem_alloc(name, *data, MEM_READ_ONLY);
i = const_mem_map.insert(ConstMemMap::value_type(name, data)).first;
}
else {
@@ -358,7 +364,7 @@ void OpenCLDeviceBase::tex_alloc(const char *name,
VLOG(1) << "Texture allocate: " << name << ", "
<< string_human_readable_number(mem.memory_size()) << " bytes. ("
<< string_human_readable_size(mem.memory_size()) << ")";
- mem_alloc(mem, MEM_READ_ONLY);
+ mem_alloc(NULL, mem, MEM_READ_ONLY);
mem_copy_to(mem);
assert(mem_map.find(name) == mem_map.end());
mem_map.insert(MemMap::value_type(name, mem.device_pointer));
diff --git a/intern/cycles/device/opencl/opencl_split.cpp b/intern/cycles/device/opencl/opencl_split.cpp
index 6bbf1b5..bc85ba3 100644
--- a/intern/cycles/device/opencl/opencl_split.cpp
+++ b/intern/cycles/device/opencl/opencl_split.cpp
@@ -100,7 +100,7 @@ public:
} KernelGlobals;
mem.resize(sizeof(KernelGlobals));
- mem_alloc(mem, MEM_READ_WRITE);
+ mem_alloc("kernel_globals", mem, MEM_READ_WRITE);
}
virtual void free_kernel_globals(device_memory& mem)
diff --git a/intern/cycles/render/bake.cpp b/intern/cycles/render/bake.cpp
index d9a2970..c2f6293 100644
--- a/intern/cycles/render/bake.cpp
+++ b/intern/cycles/render/bake.cpp
@@ -171,9 +171,9 @@ bool BakeManager::bake(Device *device, DeviceScene *dscene, Scene *scene, Progre
/* needs to be up to data for attribute access */
device->const_copy_to("__data", &dscene->data, sizeof(dscene->data));
- device->mem_alloc(d_input, MEM_READ_ONLY);
+ device->mem_alloc("bake_input", d_input, MEM_READ_ONLY);
device->mem_copy_to(d_input);
- device->mem_alloc(d_output, MEM_READ_WRITE);
+ device->mem_alloc("bake_output", d_output, MEM_READ_WRITE);
DeviceTask task(DeviceTask::SHADER);
task.shader_input = d_input.device_pointer;
diff --git a/intern/cycles/render/buffers.cpp b/intern/cycles/render/buffers.cpp
index 8a38643..aea14aa 100644
--- a/intern/cycles/render/buffers.cpp
+++ b/intern/cycles/render/buffers.cpp
@@ -135,13 +135,13 @@ void RenderBuffers::reset(Device *device, BufferParams& params_)
/* allocate buffer */
buffer.resize(params.width*params.height*params.get_passes_size());
- device->mem_alloc(buffer, MEM_READ_WRITE);
+ device->mem_alloc("render_buffer", buffer, MEM_READ_WRITE);
device->mem_zero(buffer);
/* allocate rng state */
rng_state.resize(params.width, params.height);
- device->mem_alloc(rng_state, MEM_READ_WRITE);
+ device->mem_alloc("rng_state", rng_state, MEM_READ_WRITE);
}
bool RenderBuffers::copy_from_device()
diff --git a/intern/cycles/render/light.cpp b/intern/cycles/render/light.cpp
index 2245c86..473bf77 100644
--- a/intern/cycles/render/light.cpp
+++ b/intern/cycles/render/light.cpp
@@ -57,9 +57,9 @@ static void shade_background_pixels(Device *device, DeviceScene *dscene, int res
device->const_copy_to("__data", &dscene->data, sizeof(dscene->data));
- device->mem_alloc(d_input, MEM_READ_ONLY);
+ device->mem_alloc("shade_background_pixels_input", d_input, MEM_READ_ONLY);
device->mem_copy_to(d_input);
- device->mem_alloc(d_output, MEM_WRITE_ONLY);
+ device->mem_alloc("shade_background_pixels_output", d_output, MEM_WRITE_ONLY);
DeviceTask main_task(DeviceTask::SHADER);
main_task.shader_input = d_input.device_pointer;
diff --git a/intern/cycles/render/mesh_displace.cpp b/intern/cycles/
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list