[REVIEW][FEA] Port nvtx.pyx to use non-legacy libcudf APIs

CHANGELOG.md

@@ -3,20 +3,18 @@
 ## New Features
 
 - PR #3577 Add initial dictionary support to column classes
-- PR #3917 Add dictionary add_keys function
 - PR #3777 Add support for dictionary column in gather
 - PR #3693 add string support, skipna to scan operation
 - PR #3662 Define and implement `shift`.
-- PR #3842 ORC writer: add support for column statistics
 - PR #3861 Added Series.sum feature for String
 - PR #4069 Added cast of numeric columns from/to String
 - PR #3681 Add cudf::experimental::boolean_mask_scatter
-- PR #4088 Added asString() on ColumnVector in Java that takes a format string
 - PR #4040 Add support for n-way merge of sorted tables
 - PR #4053 Multi-column quantiles.
 - PR #4100 Add set_keys function for dictionary columns
 - PR #3894 Add remove_keys functions for dictionary columns
 - PR #4107 Add groupby nunique aggregation
+- PR #4235 Port nvtx.pyx to use non-legacy libcudf APIs
 - PR #4153 Support Dask serialization protocol on cuDF objects
 - PR #4127 Add python API for n-way sorted merge (merge_sorted)
 - PR #4164 Add Buffer "constructor-kwargs" header
@@ -61,19 +59,15 @@
 - PR #3911 Adding null boolean handling for copy_if_else
 - PR #4003 Drop old `to_device` utility wrapper function
 - PR #4002 Adding to_frame and fix for categorical column issue
-- PR #4035 Port NVText tokenize function to libcudf++
 - PR #4009 build script update to enable cudf build without installing
 - PR #3897 Port cuIO JSON reader to cudf::column types
 - PR #4008 Eliminate extra copy in column constructor
 - PR #4013 Add cython definition for io readers cudf/io/io_types.hpp
 - PR #4028 Port json.pyx to use new libcudf APIs
 - PR #4014 ORC/Parquet: add count parameter to stripe/rowgroup-based reader API
-- PR #4042 Port cudf/io/functions.hpp to Cython for use in IO bindings
-- PR #3880 Add aggregation infrastructure support for reduction
 - PR #3880 Add aggregation infrastructure support for cudf::reduce
-- PR #4059 Add aggregation infrastructure support for cudf::scan
+- PR #4059 Add aggregation infrastructure support for cudf::scan 
 - PR #4021 Change quantiles signature for clarity.
-- PR #4058 Port hash.pyx to use libcudf++ APIs
 - PR #4057 Handle offsets in cython Column class
 - PR #4045 Reorganize `libxx` directory
 - PR #4029 Port stream_compaction.pyx to use libcudf++ APIs
@@ -99,7 +93,6 @@
 - PR #4098 Remove legacy calls from libcudf strings column code
 - PR #4044 Port join.pyx to use libcudf++ APIs
 - PR #4111 Use `Buffer`'s to serialize `StringColumn`
-- PR #4133 Mask cleanup and fixes: use `int32` dtype, ensure 64 byte padding, handle offsets
 - PR #4113 Get `len` of `StringColumn`s without `nvstrings`
 - PR #4147 Remove workaround for UNKNOWN_NULL_COUNT in contiguous_split.
 - PR #4130 Renames in-place `cudf::experimental::fill` to `cudf::experimental::fill_in_place`
@@ -186,13 +179,9 @@
 - PR #4089 Fix dask groupby mutliindex test case issues in join
 - PR #4097 Fix strings concatenate logic with column offsets
 - PR #4076 All null string entries should have null data buffer
-- PR #4145 Support empty index case in DataFrame._from_table
 - PR #4109 Use rmm::device_vector instead of thrust::device_vector
 - PR #4113 Use `.nvstrings` in `StringColumn.sum(...)`
 - PR #4116 Fix a bug in contiguous_split() where tables with mixed column types could corrupt string output
-- PR #4108 Fix dtype bugs in dask_cudf metadata (metadata_nonempty overhaul)
-- PR #4138 Really fix strings concatenate logic with column offsets
-- PR #4119 Fix binary ops slowdown using jitify -remove-unused-globals
 - PR #4125 Fix type enum to account for added Dictionary type in `types.hpp`
 - PR #4132 Fix `hash_partition` null mask allocation
 - PR #4137 Update Java for mutating fill and rolling window changes

python/cudf/cudf/_lib/avro.pyx

@@ -14,7 +14,10 @@ from libcpp.vector cimport vector
 from libcpp.memory cimport unique_ptr
 
 from cudf.utils import ioutils
-from cudf._lib.nvtx import nvtx_range_push, nvtx_range_pop
+from cudf._libxx.nvtx import (
+    range_push as nvtx_range_push,
+    range_pop as nvtx_range_pop
+)
 
 from io import BytesIO
 import errno

python/cudf/cudf/_lib/csv.pyx

@@ -6,7 +6,10 @@ from cudf._lib.cudf cimport *
 from cudf._lib.cudf import *
 from cudf._lib.utils cimport *
 from cudf._lib.utils import *
-from cudf._lib.nvtx import nvtx_range_push, nvtx_range_pop
+from cudf._libxx.nvtx import (
+    range_push as nvtx_range_push,
+    range_pop as nvtx_range_pop
+)
 from cudf._lib.includes.csv cimport (
     reader as csv_reader,
     reader_options as csv_reader_options
@@ -87,7 +90,7 @@ cpdef read_csv(
     if delimiter is None:
         delimiter = sep
 
-    nvtx_range_push("CUDF_READ_CSV", "purple")
+    nvtx_range_push("CUDF_READ_CSV", "PURPLE")
 
     # Setup reader options
     cdef csv_reader_options args = csv_reader_options()
@@ -262,7 +265,7 @@ cpdef write_csv(
     cudf.io.csv.write_csv
     """
 
-    nvtx_range_push("CUDF_WRITE_CSV", "purple")
+    nvtx_range_push("CUDF_WRITE_CSV", "PURPLE")
 
     from cudf.core.series import Series
 

python/cudf/cudf/_lib/cudf.pxd

@@ -317,19 +317,6 @@ cdef extern from "cudf/cudf.h" nogil:
         size_type* out_indices
     ) except +
 
-    cdef gdf_error gdf_nvtx_range_push(
-        const char* const name,
-        gdf_color color
-    ) except +
-
-    cdef gdf_error gdf_nvtx_range_push_hex(
-        const char* const name,
-        unsigned int color
-    ) except +
-
-    cdef gdf_error gdf_nvtx_range_pop() except +
-
-
 cdef extern from "cudf/legacy/bitmask.hpp" nogil:
 
     cdef gdf_error gdf_count_nonzero_mask(

python/cudf/cudf/_libxx/__init__.py

@@ -11,6 +11,7 @@
     join,
     merge,
     null_mask,
+    nvtx,
     orc,
     quantiles,
     reduce,

python/cudf/cudf/_libxx/cpp/nvtx.pxd

@@ -0,0 +1,33 @@
+# Copyright (c) 2019, NVIDIA CORPORATION.
+from libc.stdint cimport uint32_t
+
+cdef extern from "cudf/utilities/nvtx_utils.hpp" namespace "cudf::nvtx" nogil:
+    ctypedef enum color:
+        GREEN 'cudf::nvtx::color::GREEN'
+        BLUE 'cudf::nvtx::color::BLUE'
+        YELLOW 'cudf::nvtx::color::YELLOW'
+        PURPLE 'cudf::nvtx::color::PURPLE'
+        CYAN 'cudf::nvtx::color::CYAN'
+        RED 'cudf::nvtx::color::RED'
+        WHITE 'cudf::nvtx::color::WHITE'
+        DARK_GREEN 'cudf::nvtx::color::DARK_GREEN'
+        ORANGE 'cudf::nvtx::color::ORANGE'
+
+    cdef color JOIN_COLOR 'cudf::nvtx::JOIN_COLOR'
+    cdef color GROUP_COLOR 'cudf::nvtx::GROUP_COLOR'
+    cdef color BINARY_OP_COLOR 'cudf::nvtx::BINARY_OP_COLOR'
+    cdef color PARTITION_COLOR 'cudf::nvtx::PARTITION_COLOR'
+    cdef color READ_CSV_COLOR 'cudf::nvtx::READ_CSV_COLOR'
+
+cdef extern from "cudf/utilities/nvtx_utils.hpp" namespace "cudf::nvtx" nogil:
+    cdef void range_push(
+        const char* const name,
+        color color
+    ) except +
+
+    cdef void range_push_hex(
+        const char* const name,
+        uint32_t color
+    ) except +
+
+    cdef void range_pop() except +

python/cudf/cudf/_libxx/nvtx.pxd

@@ -0,0 +1,5 @@
+# Copyright (c) 2020, NVIDIA CORPORATION.
+
+from libc.stdint cimport uint32_t
+
+ctypedef uint32_t underlying_type_t_color

python/cudf/cudf/_libxx/nvtx.pyx

@@ -0,0 +1,57 @@
+# Copyright (c) 2020, NVIDIA CORPORATION.
+
+from enum import IntEnum
+from libc.stdint cimport uint32_t
+from libcpp.string cimport string
+from cudf._libxx.cpp.nvtx cimport (
+    range_push as cpp_range_push,
+    range_push_hex as cpp_range_push_hex,
+    range_pop as cpp_range_pop,
+    color as color_types
+)
+
+cimport cudf._libxx.nvtx as underlying_type_t_color
+
+
+class Color(IntEnum):
+    GREEN = <underlying_type_t_color> color_types.GREEN
+    BLUE = <underlying_type_t_color> color_types.BLUE
+    YELLOW = <underlying_type_t_color> color_types.YELLOW
+    PURPLE = <underlying_type_t_color> color_types.PURPLE
+    CYAN = <underlying_type_t_color> color_types.CYAN
+    RED = <underlying_type_t_color> color_types.RED
+    WHITE = <underlying_type_t_color> color_types.WHITE
+    DARK_GREEN = <underlying_type_t_color> color_types.DARK_GREEN
+    ORANGE = <underlying_type_t_color> color_types.ORANGE
+
+
+def range_push(object name, object color='GREEN'):
+    """
+    Demarcate the beginning of a user-defined NVTX range.
+
+    Parameters
+    ----------
+    name : str
+        The name of the NVTX range
+    color : str
+        The color to use for the range.
+        Can be named color or hex RGB string.
+    """
+    try:
+        color = int(color, 16)
+    except ValueError:
+        color = int(Color[color])
+
+    cdef const char* _name = <bytes> name
+    cdef uint32_t _color = color
+
+    with nogil:
+        cpp_range_push_hex(_name, _color)
+
+
+def range_pop():
+    """
+    Demarcate the end of a user-defined NVTX range.
+    """
+    with nogil:
+        cpp_range_pop()

python/cudf/cudf/core/column/datetime.py

@@ -264,9 +264,9 @@ def is_unique(self):
 
 
 def binop(lhs, rhs, op, out_dtype):
-    libcudf.nvtx.nvtx_range_push("CUDF_BINARY_OP", "orange")
+    libcudfxx.nvtx.range_push("CUDF_BINARY_OP", "ORANGE")
     masked = lhs.nullable or rhs.nullable
     out = column.column_empty_like(lhs, dtype=out_dtype, masked=masked)
     _ = libcudf.binops.apply_op(lhs, rhs, out, op)
-    libcudf.nvtx.nvtx_range_pop()
+    libcudfxx.nvtx.range_pop()
     return out

python/cudf/cudf/core/column/numerical.py

@@ -415,7 +415,7 @@ def can_cast_safely(self, to_dtype):
 def _numeric_column_binop(lhs, rhs, op, out_dtype, reflect=False):
     if reflect:
         lhs, rhs = rhs, lhs
-    libcudf.nvtx.nvtx_range_push("CUDF_BINARY_OP", "orange")
+    libcudfxx.nvtx.range_push("CUDF_BINARY_OP", "ORANGE")
     # Allocate output
     masked = False
     if np.isscalar(lhs):
@@ -443,7 +443,7 @@ def _numeric_column_binop(lhs, rhs, op, out_dtype, reflect=False):
     if is_op_comparison:
         out = out.fillna(op == "ne")
 
-    libcudf.nvtx.nvtx_range_pop()
+    libcudfxx.nvtx.range_pop()
     return out
 
 

python/cudf/cudf/core/column/string.py

@@ -14,7 +14,10 @@
 import cudf._lib as libcudf
 import cudf._libxx as libcudfxx
 import cudf._libxx.string_casting as str_cast
-from cudf._lib.nvtx import nvtx_range_pop, nvtx_range_push
+from cudf._libxx.nvtx import (
+    range_pop as nvtx_range_pop,
+    range_push as nvtx_range_push,
+)
 from cudf._libxx.strings.capitalize import (
     capitalize as cpp_capitalize,
     title as cpp_title,
@@ -2181,7 +2184,7 @@ def _mimic_inplace(self, other_col, inplace=False):
 
 
 def _string_column_binop(lhs, rhs, op):
-    nvtx_range_push("CUDF_BINARY_OP", "orange")
+    nvtx_range_push("CUDF_BINARY_OP", "ORANGE")
     # Allocate output
     masked = lhs.nullable or rhs.nullable
     out = column.column_empty_like(lhs, dtype="bool", masked=masked)

python/cudf/cudf/core/dataframe.py

@@ -1854,7 +1854,7 @@ def nans_to_nulls(self):
     @classmethod
     def _concat(cls, objs, axis=0, ignore_index=False):
 
-        libcudf.nvtx.nvtx_range_push("CUDF_CONCAT", "orange")
+        libcudfxx.nvtx.range_push("CUDF_CONCAT", "ORANGE")
 
         if ignore_index:
             index = RangeIndex(sum(map(len, objs)))
@@ -1892,7 +1892,7 @@ def _concat(cls, objs, axis=0, ignore_index=False):
         else:
             out.columns = unique_columns_ordered_ls
 
-        libcudf.nvtx.nvtx_range_pop()
+        libcudfxx.nvtx.range_pop()
         return out
 
     def as_gpu_matrix(self, columns=None, order="F"):
@@ -2295,7 +2295,7 @@ def merge(
         4    3    13.0
         2    4    14.0    12.0
         """
-        libcudf.nvtx.nvtx_range_push("CUDF_JOIN", "blue")
+        libcudfxx.nvtx.range_push("CUDF_JOIN", "BLUE")
         if indicator:
             raise NotImplementedError(
                 "Only indicator=False is currently supported"
@@ -2375,7 +2375,7 @@ def join(
         - *on* is not supported yet due to lack of multi-index support.
         """
 
-        libcudf.nvtx.nvtx_range_push("CUDF_JOIN", "blue")
+        libcudfxx.nvtx.range_push("CUDF_JOIN", "BLUE")
 
         # Outer joins still use the old implementation
         if type != "":
@@ -2576,7 +2576,7 @@ def groupby(
 
             # The corresponding pop() is in
             # DataFrameGroupBy._apply_aggregation()
-            libcudf.nvtx.nvtx_range_push("CUDF_GROUPBY", "purple")
+            libcudfxx.nvtx.range_push("CUDF_GROUPBY", "PURPLE")
 
             result = DataFrameGroupBy(
                 self,
@@ -2674,7 +2674,7 @@ def query(self, expr, local_dict={}):
                 )
             )
 
-        libcudf.nvtx.nvtx_range_push("CUDF_QUERY", "purple")
+        libcudfxx.nvtx.range_push("CUDF_QUERY", "PURPLE")
         # Get calling environment
         callframe = inspect.currentframe().f_back
         callenv = {
@@ -2691,7 +2691,7 @@ def query(self, expr, local_dict={}):
             newseries = self[col][selected]
             newdf[col] = newseries
         result = newdf
-        libcudf.nvtx.nvtx_range_pop()
+        libcudfxx.nvtx.range_pop()
         return result
 
     @applyutils.doc_apply()

python/cudf/cudf/core/groupby/groupby.py

@@ -6,6 +6,7 @@
 
 import cudf
 import cudf._lib as libcudf
+import cudf._libxx as libcudfxx
 from cudf import MultiIndex
 from cudf.core.column import deserialize_columns, serialize_columns
 from cudf.utils.dtypes import is_scalar
@@ -130,7 +131,7 @@ def _apply_aggregation(self, agg):
         Applies the aggregation function(s) ``agg`` on all columns
         """
         result = self._groupby.compute_result(agg)
-        libcudf.nvtx.nvtx_range_pop()
+        libcudfxx.nvtx.range_pop()
         return result
 
     def __getitem__(self, arg):