Support float case of format_number with format_float kernel #9790
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
thirtiseven
requested review from
jlowe,
revans2,
tgravescs,
GaryShen2008,
NvTimLiu and
pxLi
as code owners
Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
|
@revans2 please take a look on this too if you have time, thanks! |
|
build |
revans2
reviewed
Jan 3, 2024
docs/compatibility.md
Outdated
| This configuration is enabled by default. To disable this operation on the GPU set | ||
| [`spark.rapids.sql.castFloatToString.enabled`](additional-functionality/advanced_configs.md#sql.castFloatToString.enabled) to `false`. | ||
|
|
||
| The `format_number` function also use ryu as the solution when formatting floating-point data types to |
| 'format_number(a, 5)').collect(), conf = float_format_number_conf) | ||
| gpu = with_gpu_session(lambda spark: unary_op_df(spark, gen).selectExpr('*', | ||
| 'format_number(a, 5)').collect(), conf = float_format_number_conf) | ||
| mismatched = sum(x[0] != x[1] for x in zip(cpu, gpu)) |
There was a problem hiding this comment.
I preferred the version that checked that when we parsed them back to a float the numbers were within the error bounds instead of saying that we cannot be wrong more than some set percentage.
|
The test code is really not strong enough, I made a fix from the kernel side: NVIDIA/spark-rapids-jni#1676 |
Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
|
build |
revans2
approved these changes
Jan 9, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #9173
This PR adds better support for the float case of format_number, which made it good enough to be enabled by default.
It will have known compatibility issues from ryu, as the same way as float to string.
Depends on: NVIDIA/spark-rapids-jni#1572
performance test results
10000000 random number generated by BigDataGen:
test code:
I plan to move special case checking to the kernel next, but personally I think it is a nit for this pr.done and it runs much faster than before the change.results before move nan/inf replacement to kernel