NVIDIA · thirtiseven · Dec 19, 2025 · Dec 22, 2025 · Dec 23, 2025 · Dec 23, 2025
diff --git a/docs/additional-functionality/advanced_configs.md b/docs/additional-functionality/advanced_configs.md
@@ -123,6 +123,9 @@ Name | Description | Default Value | Applicable at
 <a name="sql.format.parquet.reader.type"></a>spark.rapids.sql.format.parquet.reader.type|Sets the Parquet reader type. We support different types that are optimized for different environments. The original Spark style reader can be selected by setting this to PERFILE which individually reads and copies files to the GPU. Loading many small files individually has high overhead, and using either COALESCING or MULTITHREADED is recommended instead. The COALESCING reader is good when using a local file system where the executors are on the same nodes or close to the nodes the data is being read on. This reader coalesces all the files assigned to a task into a single host buffer before sending it down to the GPU. It copies blocks from a single file into a host buffer in separate threads in parallel, see spark.rapids.sql.multiThreadedRead.numThreads. MULTITHREADED is good for cloud environments where you are reading from a blobstore that is totally separate and likely has a higher I/O read cost. Many times the cloud environments also get better throughput when you have multiple readers in parallel. This reader uses multiple threads to read each file in parallel and each file is sent to the GPU separately. This allows the CPU to keep reading while GPU is also doing work. See spark.rapids.sql.multiThreadedRead.numThreads and spark.rapids.sql.format.parquet.multiThreadedRead.maxNumFilesParallel to control the number of threads and amount of memory used. By default this is set to AUTO so we select the reader we think is best. This will either be the COALESCING or the MULTITHREADED based on whether we think the file is in the cloud. See spark.rapids.cloudSchemes.|AUTO|Runtime
 <a name="sql.format.parquet.write.enabled"></a>spark.rapids.sql.format.parquet.write.enabled|When set to false disables parquet output acceleration|true|Runtime
 <a name="sql.format.parquet.writer.int96.enabled"></a>spark.rapids.sql.format.parquet.writer.int96.enabled|When set to false, disables accelerated parquet write if the spark.sql.parquet.outputTimestampType is set to INT96|true|Runtime
+<a name="sql.format.sequencefile.multiThreadedRead.maxNumFilesParallel"></a>spark.rapids.sql.format.sequencefile.multiThreadedRead.maxNumFilesParallel|A limit on the maximum number of files per task processed in parallel on the CPU side before the file is sent to the GPU. This affects the amount of host memory used when reading the files in parallel. Used with MULTITHREADED reader, see spark.rapids.sql.format.sequencefile.reader.type.|2147483647|Runtime
+<a name="sql.format.sequencefile.rddScan.physicalReplace.enabled"></a>spark.rapids.sql.format.sequencefile.rddScan.physicalReplace.enabled|Enable physical-plan replacement for SequenceFile RDD scans (RDDScanExec) when the lineage can be safely identified as a simple SequenceFile scan with BinaryType key/value output. Unsupported or risky cases automatically remain on CPU.|true|Runtime
+<a name="sql.format.sequencefile.reader.type"></a>spark.rapids.sql.format.sequencefile.reader.type|Sets the SequenceFile reader type. Since SequenceFile decoding happens on the CPU (using Hadoop's SequenceFile.Reader), COALESCING mode is not supported and will throw an exception. MULTITHREADED uses multiple threads to read files in parallel, utilizing multiple CPU cores for I/O and decoding. MULTITHREADED is recommended when reading many files as it allows the CPU to keep reading while GPU is also doing work. See spark.rapids.sql.multiThreadedRead.numThreads and spark.rapids.sql.format.sequencefile.multiThreadedRead.maxNumFilesParallel to control the number of threads and amount of memory used. AUTO is kept for compatibility, but MULTITHREADED is the default for SequenceFile.|MULTITHREADED|Runtime
 <a name="sql.formatNumberFloat.enabled"></a>spark.rapids.sql.formatNumberFloat.enabled|format_number with floating point types on the GPU returns results that have a different precision than the default results of Spark.|true|Runtime
 <a name="sql.hasExtendedYearValues"></a>spark.rapids.sql.hasExtendedYearValues|Spark 3.2.0+ extended parsing of years in dates and timestamps to support the full range of possible values. Prior to this it was limited to a positive 4 digit year. The Accelerator does not support the extended range yet. This config indicates if your data includes this extended range or not, or if you don't care about getting the correct values on values with the extended range.|true|Runtime
 <a name="sql.hashOptimizeSort.enabled"></a>spark.rapids.sql.hashOptimizeSort.enabled|Whether sorts should be inserted after some hashed operations to improve output ordering. This can improve output file sizes when saving to columnar formats.|false|Runtime
@@ -473,6 +476,7 @@ Name | Description | Default Value | Notes
 <a name="sql.exec.ProjectExec"></a>spark.rapids.sql.exec.ProjectExec|The backend for most select, withColumn and dropColumn statements|true|None|
 <a name="sql.exec.RangeExec"></a>spark.rapids.sql.exec.RangeExec|The backend for range operator|true|None|
 <a name="sql.exec.SampleExec"></a>spark.rapids.sql.exec.SampleExec|The backend for the sample operator|true|None|
+<a name="sql.exec.SerializeFromObjectExec"></a>spark.rapids.sql.exec.SerializeFromObjectExec|Serialize object rows to binary columns for SequenceFile RDD scans|true|None|
 <a name="sql.exec.SortExec"></a>spark.rapids.sql.exec.SortExec|The backend for the sort operator|true|None|
 <a name="sql.exec.SubqueryBroadcastExec"></a>spark.rapids.sql.exec.SubqueryBroadcastExec|Plan to collect and transform the broadcast key values|true|None|
 <a name="sql.exec.TakeOrderedAndProjectExec"></a>spark.rapids.sql.exec.TakeOrderedAndProjectExec|Take the first limit elements as defined by the sortOrder, and do projection if needed|true|None|

diff --git a/docs/supported_ops.md b/docs/supported_ops.md
@@ -421,6 +421,32 @@ Accelerator supports are described below.
 <td>S</td>
 </tr>
 <tr>
+<td rowspan="1">SerializeFromObjectExec</td>
+<td rowspan="1">Serialize object rows to binary columns for SequenceFile RDD scans</td>
+<td rowspan="1">None</td>
+<td>Input/Output</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>S</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+<tr>
 <td rowspan="1">SortExec</td>
 <td rowspan="1">The backend for the sort operator</td>
 <td rowspan="1">None</td>
@@ -499,32 +525,6 @@ Accelerator supports are described below.
 <td><b>NS</b></td>
 </tr>
 <tr>
-<td rowspan="1">UnionExec</td>
-<td rowspan="1">The backend for the union operator</td>
-<td rowspan="1">None</td>
-<td>Input/Output</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td><em>PS<br/>UTC is only supported TZ for TIMESTAMP</em></td>
-<td>S</td>
-<td>S</td>
-<td>S</td>
-<td><b>NS</b></td>
-<td><b>NS</b></td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
-<td><em>PS<br/>unionByName will not optionally impute nulls for missing struct fields when the column is a struct and there are non-overlapping fields;<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
-<td><b>NS</b></td>
-<td><b>NS</b></td>
-<td><b>NS</b></td>
-</tr>
-<tr>
 <th>Executor</th>
 <th>Description</th>
 <th>Notes</th>
@@ -551,6 +551,32 @@ Accelerator supports are described below.
 <th>YEARMONTH</th>
 </tr>
 <tr>
+<td rowspan="1">UnionExec</td>
+<td rowspan="1">The backend for the union operator</td>
+<td rowspan="1">None</td>
+<td>Input/Output</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td><em>PS<br/>UTC is only supported TZ for TIMESTAMP</em></td>
+<td>S</td>
+<td>S</td>
+<td>S</td>
+<td><b>NS</b></td>
+<td><b>NS</b></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>unionByName will not optionally impute nulls for missing struct fields when the column is a struct and there are non-overlapping fields;<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><b>NS</b></td>
+<td><b>NS</b></td>
+<td><b>NS</b></td>
+</tr>
+<tr>
 <td rowspan="1">AQEShuffleReadExec</td>
 <td rowspan="1">A wrapper of shuffle query stage</td>
 <td rowspan="1">None</td>
@@ -915,6 +941,32 @@ Accelerator supports are described below.
 <td>S</td>
 </tr>
 <tr>
+<th>Executor</th>
+<th>Description</th>
+<th>Notes</th>
+<th>Param(s)</th>
+<th>BOOLEAN</th>
+<th>BYTE</th>
+<th>SHORT</th>
+<th>INT</th>
+<th>LONG</th>
+<th>FLOAT</th>
+<th>DOUBLE</th>
+<th>DATE</th>
+<th>TIMESTAMP</th>
+<th>STRING</th>
+<th>DECIMAL</th>
+<th>NULL</th>
+<th>BINARY</th>
+<th>CALENDAR</th>
+<th>ARRAY</th>
+<th>MAP</th>
+<th>STRUCT</th>
+<th>UDT</th>
+<th>DAYTIME</th>
+<th>YEARMONTH</th>
+</tr>
+<tr>
 <td rowspan="4">BroadcastHashJoinExec</td>
 <td rowspan="4">Implementation of join using broadcast data</td>
 <td rowspan="4">None</td>
@@ -1010,32 +1062,6 @@ Accelerator supports are described below.
 <td><b>NS</b></td>
 </tr>
 <tr>
-<th>Executor</th>
-<th>Description</th>
-<th>Notes</th>
-<th>Param(s)</th>
-<th>BOOLEAN</th>
-<th>BYTE</th>
-<th>SHORT</th>
-<th>INT</th>
-<th>LONG</th>
-<th>FLOAT</th>
-<th>DOUBLE</th>
-<th>DATE</th>
-<th>TIMESTAMP</th>
-<th>STRING</th>
-<th>DECIMAL</th>
-<th>NULL</th>
-<th>BINARY</th>
-<th>CALENDAR</th>
-<th>ARRAY</th>
-<th>MAP</th>
-<th>STRUCT</th>
-<th>UDT</th>
-<th>DAYTIME</th>
-<th>YEARMONTH</th>
-</tr>
-<tr>
 <td rowspan="2">BroadcastNestedLoopJoinExec</td>
 <td rowspan="2">Implementation of join using brute force. Full outer joins and joins where the broadcast side matches the join side (e.g.: LeftOuter with left broadcast) are not supported</td>
 <td rowspan="2">None</td>
@@ -1301,6 +1327,32 @@ Accelerator supports are described below.
 <td><b>NS</b></td>
 </tr>
 <tr>
+<th>Executor</th>
+<th>Description</th>
+<th>Notes</th>
+<th>Param(s)</th>
+<th>BOOLEAN</th>
+<th>BYTE</th>
+<th>SHORT</th>
+<th>INT</th>
+<th>LONG</th>
+<th>FLOAT</th>
+<th>DOUBLE</th>
+<th>DATE</th>
+<th>TIMESTAMP</th>
+<th>STRING</th>
+<th>DECIMAL</th>
+<th>NULL</th>
+<th>BINARY</th>
+<th>CALENDAR</th>
+<th>ARRAY</th>
+<th>MAP</th>
+<th>STRUCT</th>
+<th>UDT</th>
+<th>DAYTIME</th>
+<th>YEARMONTH</th>
+</tr>
+<tr>
 <td rowspan="1">AggregateInPandasExec</td>
 <td rowspan="1">The backend for an Aggregation Pandas UDF. This accelerates the data transfer between the Java process and the Python process. It also supports scheduling GPU resources for the Python process when enabled.</td>
 <td rowspan="1">None</td>
@@ -1405,32 +1457,6 @@ Accelerator supports are described below.
 <td><b>NS</b></td>
 </tr>
 <tr>
-<th>Executor</th>
-<th>Description</th>
-<th>Notes</th>
-<th>Param(s)</th>
-<th>BOOLEAN</th>
-<th>BYTE</th>
-<th>SHORT</th>
-<th>INT</th>
-<th>LONG</th>
-<th>FLOAT</th>
-<th>DOUBLE</th>
-<th>DATE</th>
-<th>TIMESTAMP</th>
-<th>STRING</th>
-<th>DECIMAL</th>
-<th>NULL</th>
-<th>BINARY</th>
-<th>CALENDAR</th>
-<th>ARRAY</th>
-<th>MAP</th>
-<th>STRUCT</th>
-<th>UDT</th>
-<th>DAYTIME</th>
-<th>YEARMONTH</th>
-</tr>
-<tr>
 <td rowspan="1">MapInPandasExec</td>
 <td rowspan="1">The backend for Map Pandas Iterator UDF. Accelerates the data transfer between the Java process and the Python process. It also supports scheduling GPU resources for the Python process when enabled.</td>
 <td rowspan="1">None</td>
@@ -8267,7 +8293,7 @@ are limited.
 <td> </td>
 <td> </td>
 <td> </td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
 <td> </td>
 <td> </td>
 <td> </td>
@@ -8290,7 +8316,7 @@ are limited.
 <td> </td>
 <td> </td>
 <td> </td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
 <td> </td>
 <td> </td>
 <td> </td>