Skip to content

[CUB] Cleanup some dispatch arguments #9603

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
miscco wants to merge 1 commit into from
+150 −141
Draft

[CUB] Cleanup some dispatch arguments #9603

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
291 changes: 150 additions & 141 deletions cub/cub/device/dispatch/dispatch_histogram.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -518,29 +518,28 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t __dispatch_even_device
constexpr int PRIVATIZED_SMEM_BINS = 0;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
true, // IsDeviceInit
true, // IsEven
false // IsByteSample
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ true,
/* IsEven = */ true,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand All @@ -551,29 +550,28 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t __dispatch_even_device
constexpr int PRIVATIZED_SMEM_BINS = detail::histogram::max_privatized_smem_bins;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
true, // IsDeviceInit
true, // IsEven
false // IsByteSample
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ true,
/* IsEven = */ true,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand Down Expand Up @@ -692,29 +690,28 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t __dispatch_even_device
constexpr int PRIVATIZED_SMEM_BINS = 256;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
true, // IsDeviceInit
true, // IsEven
true // IsByteSample
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_privatized_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ true,
/* IsEven = */ true,
/* IsByteSample = */ true>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_privatized_levels,
num_output_levels,
upper_level,
lower_level,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand Down Expand Up @@ -844,29 +841,28 @@ CUB_RUNTIME_FUNCTION static cudaError_t dispatch_range(
constexpr int PRIVATIZED_SMEM_BINS = 256;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
false, // IsDeviceInit
false, // IsEven (unused for host-init)
false // IsByteSample (unused for host-init)
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_privatized_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_privatized_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand Down Expand Up @@ -902,29 +898,28 @@ CUB_RUNTIME_FUNCTION static cudaError_t dispatch_range(
constexpr int PRIVATIZED_SMEM_BINS = 0;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
false, // IsDeviceInit
false, // IsEven (unused for host-init)
false // IsByteSample (unused for host-init)
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand All @@ -935,29 +930,28 @@ CUB_RUNTIME_FUNCTION static cudaError_t dispatch_range(
constexpr int PRIVATIZED_SMEM_BINS = max_privatized_smem_bins;

if (const auto error = CubDebug(
(detail::histogram::dispatch<
NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
false, // IsDeviceInit
false, // IsEven (unused for host-init)
false // IsByteSample (unused for host-init)
>(d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
d_output_histograms,
num_output_levels,
num_output_levels,
output_decode_op,
privatized_decode_op,
max_num_output_bins,
num_row_pixels,
num_rows,
row_stride_samples,
stream,
policy_selector,
kernel_source,
launcher_factory))))
{
return error;
}
Expand Down Expand Up @@ -1040,7 +1034,12 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t dispatch_even(
constexpr int PRIVATIZED_SMEM_BINS = 256;

if (const auto error = CubDebug(
(detail::histogram::dispatch<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, PRIVATIZED_SMEM_BINS, false, false, false>(
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
Expand Down Expand Up @@ -1103,7 +1102,12 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t dispatch_even(
constexpr int PRIVATIZED_SMEM_BINS = 0;

if (const auto error = CubDebug(
(detail::histogram::dispatch<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, PRIVATIZED_SMEM_BINS, false, false, false>(
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
Expand All @@ -1129,7 +1133,12 @@ CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t dispatch_even(
constexpr int PRIVATIZED_SMEM_BINS = max_privatized_smem_bins;

if (const auto error = CubDebug(
(detail::histogram::dispatch<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, PRIVATIZED_SMEM_BINS, false, false, false>(
(detail::histogram::dispatch<NUM_CHANNELS,
NUM_ACTIVE_CHANNELS,
PRIVATIZED_SMEM_BINS,
/* IsDeviceInit = */ false,
/* IsEven = */ false,
/* IsByteSample = */ false>(
d_temp_storage,
temp_storage_bytes,
d_samples,
Expand Down
Loading