leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

Fix typos 2 (#842)

Browse Source 
Co-authored-by: Haicheng Wu <57973641+hwu36@users.noreply.github.com>
This commit is contained in:
Alexander Pivovarov
2023-03-09 23:22:56 -05:00
committed by GitHub
co-authored by Haicheng Wu
parent c4f6b8c6bc
commit 7e370c9637
161 changed files with 310 additions and 309 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/profiler/src/conv2d_operation_profiler.cu
+13 -13
View File
@@ -270,7 +270,7 @@ Status Conv2dOperationProfiler::initialize_configuration(
}
//////////////////////// Convolution output dimensions p and q ////////////////////////
// Cutlass convolutions support arbitrary output sizes and not constriant by //
// Cutlass convolutions support arbitrary output sizes and not constrained by //
// input, filter, padding, striding, dilation sizes. //
// cuDNN sets the output dimensions (p, q) using following equations: //
// //
@@ -502,7 +502,7 @@ void Conv2dOperationProfiler::initialize_result_(
// Bytes of activation, filter, and output tensors
result.bytes = problem_.bytes(operation_desc);
// Theoritical flops required for the computation
// Theoretical flops required for the computation
result.flops = problem_.flops(operation_desc);
// Measured runtime
@@ -510,7 +510,7 @@ void Conv2dOperationProfiler::initialize_result_(
}
/// Initialize reduction problem dimenstions and library::Operation
/// Initialize reduction problem dimensions and library::Operation
bool Conv2dOperationProfiler::initialize_reduction_configuration_(
Options const &options,
PerformanceReport &report,
@@ -535,7 +535,7 @@ bool Conv2dOperationProfiler::initialize_reduction_configuration_(
/// This chooses the appropriate stride element of the row-major C tensor.
int const & tensor_c_stride_idx = (conv_kind == library::ConvKind::kWgrad ? 2 : 0);
/// intialize library::ReductionConfiguration
/// initialize library::ReductionConfiguration
conv_workspace_.reduction_configuration.problem_size = problem_.eq_gemm_size(conv_kind).mn();
conv_workspace_.reduction_configuration.partitions = int(problem_.split_k_slices);
conv_workspace_.reduction_configuration.partition_stride = problem_.eq_gemm_size(conv_kind).mn().product();
@@ -773,7 +773,7 @@ bool Conv2dOperationProfiler::verify_cutlass(
conv_workspace_.arguments.alpha = problem_.alpha_one.data();
conv_workspace_.arguments.beta = problem_.beta_zero.data();
/// intialize library::ReductionArguments
/// initialize library::ReductionArguments
conv_workspace_.reduction_arguments.workspace = conv_workspace_.device_workspace.data();
conv_workspace_.reduction_arguments.source = conv_workspace_.C->data();
conv_workspace_.reduction_arguments.destination = conv_workspace_.Computed->data();
@@ -961,7 +961,7 @@ bool Conv2dOperationProfiler::verify_with_host_reference_(
conv_desc.tile_description.math_instruction.element_accumulator,
conv_desc.element_epilogue);
#if 0 // debug print to check which host refererence instance is selected
#if 0 // debug print to check which host reference instance is selected
std::cout << conv2d_key << "\n";
#endif
@@ -982,7 +982,7 @@ bool Conv2dOperationProfiler::verify_with_host_reference_(
return true;
}
// host refernce has only one instances in Conv2dOperationVectorMap
// host reference has only one instances in Conv2dOperationVectorMap
library::Operation const *reference_op = cc_it->second[0];
//
@@ -1009,7 +1009,7 @@ bool Conv2dOperationProfiler::verify_with_host_reference_(
conv_workspace_.arguments.pointer_mode = library::ScalarPointerMode::kHost;
//
// Intialize host reference operation
// Initialize host reference operation
//
std::vector<uint8_t> host_workspace_reference_op;
@@ -1114,11 +1114,11 @@ bool Conv2dOperationProfiler::verify_with_device_reference_(
return true;
}
// device refernce has only one instances in Conv2dOperationVectorMap
// device reference has only one instances in Conv2dOperationVectorMap
library::Operation const *reference_op = cc_it->second[0];
//
// Intialize device reference operation
// Initialize device reference operation
//
std::vector<uint8_t> host_workspace_reference_op;
@@ -1205,7 +1205,7 @@ bool Conv2dOperationProfiler::profile(
conv_workspace_.arguments.alpha = problem_.alpha_one.data();
conv_workspace_.arguments.beta = problem_.beta_zero.data();
/// intialize library::ReductionArguments
/// initialize library::ReductionArguments
conv_workspace_.reduction_arguments.workspace = conv_workspace_.device_workspace.data();
conv_workspace_.reduction_arguments.source = conv_workspace_.C->data();
conv_workspace_.reduction_arguments.destination = conv_workspace_.Computed->data();
@@ -1276,7 +1276,7 @@ Status Conv2dOperationProfiler::profile_cutlass_(
// update library::ConvArguments for parallel split-k reduction
conv_arguments->D = conv_workspace_.device_workspace.data();
/// intialize library::ReductionArguments
/// initialize library::ReductionArguments
conv_workspace_.reduction_arguments.workspace = conv_workspace_.device_workspace.data();
conv_workspace_.reduction_arguments.source = conv_workspace_.C->batch_data(problem_idx);
conv_workspace_.reduction_arguments.destination = conv_workspace_.Computed->batch_data(problem_idx);
@@ -1329,7 +1329,7 @@ Status Conv2dOperationProfiler::profile_cutlass_(
// update library::ConvArguments for parallel split-k reduction
conv_arguments->D = conv_workspace_.device_workspace.data();
/// intialize library::ReductionArguments
/// initialize library::ReductionArguments
conv_workspace_.reduction_arguments.workspace = conv_workspace_.device_workspace.data();
conv_workspace_.reduction_arguments.source = conv_workspace_.C->batch_data(problem_idx);
conv_workspace_.reduction_arguments.destination = conv_workspace_.Computed->batch_data(problem_idx);