[DOC] add Databricks support matrix [skip ci]

docs/databricks-support.md

@@ -0,0 +1,72 @@
+---
+layout: page
+title: Databricks Support Matrix
+nav_order: 4
+---
+
+# Databricks Support Matrix
+
+This page summarizes the Databricks runtime combinations supported by the
+current RAPIDS Accelerator release. Use it together with the
+[Download](./download.md) page to select the correct plugin artifact before
+deploying to a Databricks cluster.
+
+## Runtime Compatibility
+
+The following matrix applies to RAPIDS Accelerator for Apache Spark v26.06.0.
+Databricks runtime images provide the JVM and system libraries for each row; use
+the runtime-provided JDK unless a Databricks support note explicitly instructs
+otherwise.
+
+| RAPIDS Accelerator | Databricks Runtime | Apache Spark | Scala | JDK runtime | CUDA jar variants | Minimum NVIDIA driver | Notes |
+|--------------------|--------------------|--------------|-------|-------------|-------------------|-----------------------|-------|
+| v26.06.0 | 13.3 ML LTS GPU | 3.4.1 | 2.12 | Databricks runtime default | CUDA 12, CUDA 13 | R525+ | Supported Databricks 13.3 runtime line. |
+| v26.06.0 | 14.3 ML LTS GPU | 3.5.0 | 2.12 | Databricks runtime default | CUDA 12, CUDA 13 | R525+ | Supported Databricks 14.3 runtime line. |
+| v26.06.0 | 17.3 ML LTS GPU | 4.0.0 | 2.13 | Databricks runtime default | CUDA 12, CUDA 13 | R525+ | Spark 4 / Scala 2.13 Databricks runtime line. |
+
+The v26.06.0 download page publishes Scala 2.12 and Scala 2.13 artifacts for
+both CUDA 12 and CUDA 13. Use the Scala artifact that matches the Databricks
+runtime's Spark/Scala line. The CUDA classifier controls which RAPIDS native
+libraries are bundled in the plugin jar; it does not change the Spark or Scala
+compatibility of the artifact.
+
+## Delta Lake GPU Support on Databricks
+
+Databricks runtimes use Databricks-specific Delta Lake implementations. The
+following table summarizes Delta Lake GPU support for the Databricks runtime
+lines with feature-specific coverage in the current release. `GPU` means the
+operation is expected to run on the GPU when the rest of the query plan is also
+GPU-compatible. `CPU fallback` means the RAPIDS Accelerator leaves that Delta
+operation on the CPU for that runtime.
+
+| Delta feature | DBR 14.3 | DBR 17.3 |
+|---------------|----------|----------|
+| Reads without deletion vectors | GPU | GPU |
+| Deletion vector reads | CPU fallback | GPU only with metadata row index and RAPIDS deletion-vector predicate pushdown |
+| Delta writes | GPU for append, overwrite, CTAS, and RTAS | GPU for append and overwrite. CTAS and RTAS fall back to CPU. |
+| Delta writes with deletion vectors | CPU fallback | CPU fallback |
+| DELETE and UPDATE | GPU for copy-on-write. Operations that write deletion vectors fall back to CPU. | Same as DBR 14.3. Liquid-clustered paths also fall back to CPU. |
+| MERGE | GPU, including liquid clustering | GPU for non-liquid-clustered tables only. Liquid-clustered and persistent deletion-vector writes fall back to CPU. |
+| OPTIMIZE | CPU fallback | GPU for standard deletion-vector-free, non-liquid-clustered tables only |
+| Auto compaction | GPU when triggered by supported GPU writes | GPU for inline deletion-vector-free, non-liquid-clustered tables only |
+| Liquid clustering | GPU support | CPU fallback |
+
+## Compatibility Caveats
+
+- Databricks may patch existing runtime versions without changing the public
+  runtime line. Use a RAPIDS Accelerator release that explicitly lists the
+  Databricks runtime you plan to run.
+- Runtime patch changes can surface as binary compatibility failures such as
+  `NoSuchMethodError` against Spark or Databricks-internal classes. If this
+  happens, first verify the RAPIDS Accelerator release and Databricks runtime
+  combination against this page and the release notes.
+- Earlier v26.04.0 artifacts on Databricks 17.3 could encounter a
+  `NoSuchMethodError` on `CatalogTable.copy` in
+  `GpuCreateDataSourceTableAsSelectCommand`; use v26.04.1 or later for that
+  runtime line.
+- Delta feature support is operation-specific. Even when a runtime is listed as
+  supported, individual Delta reads, writes, DML, `OPTIMIZE`, auto compaction,
+  deletion-vector, or liquid-clustering paths may still fall back to CPU as
+  shown above.
+- `spark.rapids.sql.explain=NOT_ON_GPU` can be used to confirm whether a
+  particular query plan stayed on the GPU or fell back to CPU.

docs/download.md

@@ -65,6 +65,9 @@ The plugin is designed to work on NVIDIA Volta, Turing, Ampere, Ada Lovelace, Ho
 		Spark runtime 2.3 LTS
 		Spark runtime 3.0
 
+See the [Databricks support matrix](./databricks-support.md) for runtime-specific Spark, Scala,
+CUDA, and Delta Lake feature support details.
+
 *Some hardware may have a minimum driver version greater than R470. Check the GPU spec sheet
 for your hardware's minimum driver version.