Misc cleanups.

NVIDIA · Nov 15, 2024 · d826b67 · d826b67
1 parent b09c887
commit d826b67
Show file tree

Hide file tree

Showing 2 changed files with 36 additions and 26 deletions.
diff --git a/lib/cmake/thrust/README.md b/lib/cmake/thrust/README.md
@@ -38,6 +38,22 @@ If using Thrust from the CCCL sources, this would be
 $ cmake . -DThrust_DIR=<CCCL git repo root>/thrust/thrust/cmake/
 ```
 
+#### Large Array (>2^32 element) Handling: `DISPATCH`
+
+The `DISPATCH` option allows users to select the tradeoff of compile-time / binary-size
+vs. performance vs. scalability when given large inputs with greater than 2^32 elements.
+This currently only applies when DEVICE=CUDA.
+
+- `Dynamic` Compiles each kernel twice, once for 32-bit offsets and again for 64-bit offsets,
+  and chooses dynamically using the input size at runtime.
+  This significantly increases compile-time and binary-size, but provides optimal performance
+  for small input sizes while also supporting 64-bit indexed workloads.
+- `Force32bit` forces Thrust to use a 32 bit offset type. This improves compile time and
+  binary size but limits the input size.
+- `Force64bit` forces Thrust to use a 64 bit offset type. This improves compile time and
+  binary size and allows large input sizes. However, it may degrade runtime performance
+  for 32-bit indexed workloads.
+
 #### TBB / OpenMP
 
 To explicitly specify host/device systems, `HOST` and `DEVICE` arguments can be
@@ -56,33 +72,40 @@ host system, but will find and use TBB or OpenMP for the device system.
 
 To allow a Thrust target to be configurable easily via `cmake-gui` or
 `ccmake`, pass the `FROM_OPTIONS` flag to `thrust_create_target`. This will add
-`THRUST_HOST_SYSTEM` and `THRUST_DEVICE_SYSTEM` options to the CMake cache that
-allow selection from the systems supported by this version of Thrust.
+`THRUST_HOST_SYSTEM`, `THRUST_DEVICE_SYSTEM`, and `THRUST_DISPATCH_TYPE` options
+to the CMake cache that allow selection from the systems supported by this version
+of Thrust.
 
 ```cmake
 thrust_create_target(Thrust FROM_OPTIONS
   [HOST_OPTION <option name>]
   [DEVICE_OPTION <option name>]
+  [DISPATCH_OPTION <option name>]
   [HOST_OPTION_DOC <doc string>]
   [DEVICE_OPTION_DOC <doc string>]
+  [DISPATCH_OPTION_DOC <doc string>]
   [HOST <default host system name>]
   [DEVICE <default device system name>]
+  [DISPATCH <default device system name>]
   [ADVANCED]
 )
 ```
 
 The optional arguments have sensible defaults, but may be configured per
 `thrust_create_target` call:
 
-| Argument            | Default                 | Description                     |
-|---------------------|-------------------------|---------------------------------|
-| `HOST_OPTION`       | `THRUST_HOST_SYSTEM`    | Name of cache option for host   |
-| `DEVICE_OPTION`     | `THRUST_DEVICE_SYSTEM`  | Name of cache option for device |
-| `HOST_OPTION_DOC`   | Thrust's host system.   | Docstring for host option       |
-| `DEVICE_OPTION_DOC` | Thrust's device system. | Docstring for device option     |
-| `HOST`              | `CPP`                   | Default host system             |
-| `DEVICE`            | `CUDA`                  | Default device system           |
-| `ADVANCED`          | *N/A*                   | Mark cache options advanced     |
+| Argument              | Default                 | Description                       |
+|-----------------------|-------------------------|-----------------------------------|
+| `HOST_OPTION`         | `THRUST_HOST_SYSTEM`    | Name of cache option for host     |
+| `DEVICE_OPTION`       | `THRUST_DEVICE_SYSTEM`  | Name of cache option for device   |
+| `DISPATCH_OPTION`     | `THRUST_DISPATCH_TYPE`  | Name of cache option for dispatch |
+| `HOST_OPTION_DOC`     | Thrust's host system.   | Docstring for host option         |
+| `DEVICE_OPTION_DOC`   | Thrust's device system. | Docstring for device option       |
+| `DISPATCH_OPTION_DOC` | Thrust's dispatch type. | Docstring for dispatch option     |
+| `HOST`                | `CPP`                   | Default host system               |
+| `DEVICE`              | `CUDA`                  | Default device system             |
+| `DISPATCH`            | `Dispatch`                  | Default dispatch type             |
+| `ADVANCED`            | *N/A*                   | Mark cache options advanced       |
 
 ### Specifying Thrust Version Requirements
 

diff --git a/lib/cmake/thrust/thrust-config.cmake b/lib/cmake/thrust/thrust-config.cmake
@@ -3,8 +3,6 @@
 #
 # Provided by NVIDIA under the same license as the associated Thrust library.
 #
-# Reply-To: Allison Piper <[email protected]>
-#
 # *****************************************************************************
 # **     The following is a short reference to using Thrust from CMake.      **
 # ** For more details, see the README.md in the same directory as this file. **
@@ -62,18 +60,7 @@
 #   IGNORE_CUB_VERSION            # Skip configure-time and compile-time CUB version checks
 # )
 #
-# # DISPATCH options [Advanced]:
-# #
-# # Controls how Thrust handles large (>2^32 element) inputs. The default is `Dynamic`.
-# # This only applies when DEVICE=CUDA.
-# #
-# # - `Dynamic` lets Thrust choose the index type based on input size, allowing
-# #   large inputs and optimal performance at the cost of increased compile time and binary size,
-# #   as Thrust will compile each kernel twice, once for 32 bit and once for 64 bit.
-# # - `Force32bit` forces Thrust to use a 32 bit offset type. This improves compile time and
-# #   binary size but limits the input size.
-# # - `Force64bit` forces Thrust to use a 64 bit offset type. This improves compile time and
-# #   binary size and allows large input sizes. However, it might degrade runtime performance.
+# # DISPATCH options (See README):
 # thrust_create_target(TargetName DISPATCH Dynamic)
 # thrust_create_target(TargetName DISPATCH Force32bit)
 # thrust_create_target(TargetName DISPATCH Force64bit)
@@ -188,7 +175,7 @@ function(thrust_create_target target_name)
   _thrust_set_if_undefined(TCT_DISPATCH Dynamic)
   _thrust_set_if_undefined(TCT_HOST_OPTION THRUST_HOST_SYSTEM)
   _thrust_set_if_undefined(TCT_DEVICE_OPTION THRUST_DEVICE_SYSTEM)
-  _thrust_set_if_undefined(TCT_DISPATCH_OPTION THRUST_DISPATCH_TYPE)
+  _thrust_set_if_undefined(TCT_DISPATCH_TYPE THRUST_DISPATCH_TYPE)
   _thrust_set_if_undefined(TCT_HOST_OPTION_DOC "Thrust host system: ${THRUST_HOST_SYSTEM_OPTIONS}")
   _thrust_set_if_undefined(TCT_DEVICE_OPTION_DOC "Thrust device system: ${THRUST_DEVICE_SYSTEM_OPTIONS}")
   _thrust_set_if_undefined(TCT_DISPATCH_OPTION_DOC "Thrust dispatch type: ${THRUST_DISPATCH_TYPE_OPTIONS}")