[
https://issues.apache.org/jira/browse/ARROW-15571?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17487181#comment-17487181
]
Yaron Gvili edited comment on ARROW-15571 at 2/4/22, 4:52 PM:
--------------------------------------------------------------
OK regarding power working on floating point values. This
[SO|https://stackoverflow.com/questions/17417490/difference-between-sqrtx-and-powx-0-5]
suggests sqrt(x ) would be faster and a bit more accurate than pow(x, 0.5), so
there is room for a sqrt operation.
Regarding my goal, it is to apply the operations to a single column with best
performance in Arrow. Here's an example for how I'm currently invoking the code
in my PR:
{code:java}
ProjectNodeOptions(
{call("minimum",
{call("maximum",
{call("square_root",
{field_ref("column_0")}),
literal(low_threshold)}),
literal(high_threshold)})},
{"result_0"}){code}
Note that the above is `using namespace arrow::compute` and that "column_0" is
of type double. How should I achieve the same functionality with the existing
operations? If the solution involves aggregate operations, do they provide best
performance, as compared to (the missing) scalar operations?
was (Author: JIRAUSER284707):
OK regarding power working on floating point values. This
[SO|https://stackoverflow.com/questions/17417490/difference-between-sqrtx-and-powx-0-5]
suggests sqrt(x) would be faster and a bit more accurate than pow(x, 0.5), so
there is room for a sqrt operation.
Regarding my goal, it is to apply the operations to a single column with best
performance in Arrow. Here's an example for how I'm currently invoking the code
in my PR:
{code:java}
ProjectNodeOptions(
{call("minimum",
{call("maximum",
{call("square_root",
{field_ref("column_0")}),
literal(low_threshold)}),
literal(high_threshold)})},
{"result_0"}){code}
Note that the above is `using namespace arrow::compute` and that "column_0" is
of type double. How should I achieve the same functionality with the existing
operations? If the solution involves aggregate operations, do they provide best
performance, as compared to (the missing) scalar operations?
> [C++] Add min/max/sqrt scalar kernels to execution engine
> ---------------------------------------------------------
>
> Key: ARROW-15571
> URL: https://issues.apache.org/jira/browse/ARROW-15571
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Yaron Gvili
> Priority: Major
> Labels: kernel, pull-request-available
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> The list of execution engine's scalar kernels currently available in
> `cpp/src/arrow/compute/kernels/scalar_arithmetic.cc` does not cover the
> common minimum, maximum, and square-root functions.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
