You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "alkis (via GitHub)" <gi...@apache.org> on 2023/03/06 16:09:19 UTC
[GitHub] [spark] alkis opened a new pull request, #40301: [SPARK-42685] optimize Utils.byteToString routines
alkis opened a new pull request, #40301:
URL: https://github.com/apache/spark/pull/40301
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a faster review.
7. If you want to add a new configuration, please read the guideline first for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
Optimize `Utils.byteToString`. Arithmetic ops on `BigInt` and `BigDecimal` are order(s) of magnitude slower than the ops on primitive types. Division is an especially slow operation and it is used en masse here.
To avoid heating up the Earth while formatting byte counts for human consumption we observe that most formatting operations are not in the 10s of EiBs but on counts that fit in 64-bits and use (fastpath) 64-bit operations to format them.
### Why are the changes needed?
Use of `Utils.byteToString` is prevalent through the codebase and they are mainly used in logs. If the logs are emitted then this becomes a heavyweight operation.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen closed pull request #40301: [SPARK-42685][CORE] Optimize Utils.bytesToString routines
Posted by "srowen (via GitHub)" <gi...@apache.org>.
srowen closed pull request #40301: [SPARK-42685][CORE] Optimize Utils.bytesToString routines
URL: https://github.com/apache/spark/pull/40301
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen commented on pull request #40301: [SPARK-42685][CORE] Optimize Utils.bytesToString routines
Posted by "srowen (via GitHub)" <gi...@apache.org>.
srowen commented on PR #40301:
URL: https://github.com/apache/spark/pull/40301#issuecomment-1464132393
Merged to master
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] alkis commented on a diff in pull request #40301: [SPARK-42685][CORE] Optimize Utils.bytesToString routines
Posted by "alkis (via GitHub)" <gi...@apache.org>.
alkis commented on code in PR #40301:
URL: https://github.com/apache/spark/pull/40301#discussion_r1130795171
##########
core/src/main/scala/org/apache/spark/util/Utils.scala:
##########
@@ -1305,41 +1305,30 @@ private[spark] object Utils extends Logging {
(JavaUtils.byteStringAsBytes(str) / 1024 / 1024).toInt
}
+ private[this] val siByteSizes =
+ Array(1L << 60, 1L << 50, 1L << 40, 1L << 30, 1L << 20, 1L << 10, 1)
+ private[this] val siByteSuffixes =
+ Array("EiB", "PiB", "TiB", "GiB", "MiB", "KiB", "B")
/**
* Convert a quantity in bytes to a human-readable string such as "4.0 MiB".
*/
- def bytesToString(size: Long): String = bytesToString(BigInt(size))
+ def bytesToString(size: Long): String = {
+ var i = 0
+ while (i < siByteSizes.length - 1 && size < 2 * siByteSizes(i)) i += 1
+ "%.1f %s".formatLocal(Locale.US, size.toDouble / siByteSizes(i), siByteSuffixes(i))
+ }
def bytesToString(size: BigInt): String = {
val EiB = 1L << 60
- val PiB = 1L << 50
- val TiB = 1L << 40
- val GiB = 1L << 30
- val MiB = 1L << 20
- val KiB = 1L << 10
-
- if (size >= BigInt(1L << 11) * EiB) {
+ if (size.isValidLong) {
+ // Common case, most sizes fit in 64 bits and all ops on BigInt are order(s) of magnitude
+ // slower than Long/Double.
+ bytesToString(size.toLong)
+ } else if (size < BigInt(2000) * EiB) {
Review Comment:
Changed it back to match.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] cloud-fan commented on a diff in pull request #40301: [SPARK-42685][CORE] Optimize Utils.bytesToString routines
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40301:
URL: https://github.com/apache/spark/pull/40301#discussion_r1130695992
##########
core/src/main/scala/org/apache/spark/util/Utils.scala:
##########
@@ -1305,41 +1305,30 @@ private[spark] object Utils extends Logging {
(JavaUtils.byteStringAsBytes(str) / 1024 / 1024).toInt
}
+ private[this] val siByteSizes =
+ Array(1L << 60, 1L << 50, 1L << 40, 1L << 30, 1L << 20, 1L << 10, 1)
+ private[this] val siByteSuffixes =
+ Array("EiB", "PiB", "TiB", "GiB", "MiB", "KiB", "B")
/**
* Convert a quantity in bytes to a human-readable string such as "4.0 MiB".
*/
- def bytesToString(size: Long): String = bytesToString(BigInt(size))
+ def bytesToString(size: Long): String = {
+ var i = 0
+ while (i < siByteSizes.length - 1 && size < 2 * siByteSizes(i)) i += 1
+ "%.1f %s".formatLocal(Locale.US, size.toDouble / siByteSizes(i), siByteSuffixes(i))
+ }
def bytesToString(size: BigInt): String = {
val EiB = 1L << 60
- val PiB = 1L << 50
- val TiB = 1L << 40
- val GiB = 1L << 30
- val MiB = 1L << 20
- val KiB = 1L << 10
-
- if (size >= BigInt(1L << 11) * EiB) {
+ if (size.isValidLong) {
+ // Common case, most sizes fit in 64 bits and all ops on BigInt are order(s) of magnitude
+ // slower than Long/Double.
+ bytesToString(size.toLong)
+ } else if (size < BigInt(2000) * EiB) {
Review Comment:
the previous code use `1L << 11`, why do we change it to 2000?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org