build: upgrade CUDA toolkit to 12.6.1 (PROOF-917)

README.md

@@ -44,7 +44,7 @@
   </a>
 
   <a href="https://developer.nvidia.com/cuda-downloads">
-    <img alt="CUDA" src="https://img.shields.io/badge/CUDA-12.4.1-green?style=flat&logo=nvidia">
+    <img alt="CUDA" src="https://img.shields.io/badge/CUDA-12.6-green?style=flat&logo=nvidia">
     </a>
   </a>
 
@@ -182,7 +182,7 @@ See the [example](./example) folder for some examples.
 Prerequisites:
 * `x86_64` Linux instance.
 * Nix with flake support (check out [The Determinate Nix Installer](https://github.com/DeterminateSystems/nix-installer))
-* Nvidia GPU capable of running CUDA 12.4.1 code.
+* Nvidia GPU capable of running CUDA 12.6 code.
 
 From your terminal, run the following command in the root of the source directory to set
 up a build environment. 

nix/cuda.nix

@@ -7,8 +7,8 @@ with pkgs;
 pkgs.stdenvNoCC.mkDerivation {
   name = "cudatoolkit";
   src = fetchurl {
-    url = "https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda_12.4.1_550.54.15_linux.run";
-    sha256 = "367d2299b3a4588ab487a6d27276ca5d9ead6e394904f18bccb9e12433b9c4fb";
+    url = "https://developer.download.nvidia.com/compute/cuda/12.6.1/local_installers/cuda_12.6.1_560.35.03_linux.run";
+    sha256 = "179fx44j0f6fm4m8afx5vhivv3ywyv7mv7shb7r2b5ji8drcxb3k";
   };
   patches = [
     # patch host_defines.h to work with libc++

sxt/algorithm/block/BUILD

@@ -10,8 +10,8 @@ sxt_cc_component(
         "//sxt/memory/resource:managed_device_resource",
     ],
     deps = [
-        "//sxt/base/device:cub",
         "//sxt/base/device:synchronization",
         "//sxt/base/macro:cuda_callable",
+        "@local_cuda//:cub",
     ],
 )

sxt/algorithm/block/runlength_count.h

@@ -16,7 +16,7 @@
  */
 #pragma once
 
-#include "sxt/base/device/cub.h"
+#include "cub/cub.cuh"
 #include "sxt/base/macro/cuda_callable.h"
 
 namespace sxt::algbk {

sxt/base/device/BUILD

@@ -14,14 +14,6 @@ sxt_cc_component(
     ],
 )
 
-sxt_cc_component(
-    name = "cub",
-    with_test = False,
-    deps = [
-        "@local_cuda//:cub",
-    ],
-)
-
 sxt_cc_component(
     name = "pointer_attributes",
     with_test = False,

sxt/multiexp/base/BUILD

@@ -46,8 +46,8 @@ sxt_cc_component(
 sxt_cc_component(
     name = "scalar_array",
     impl_deps = [
+        "@local_cuda//:cub",
         "//sxt/base/container:span_utility",
-        "//sxt/base/device:cub",
         "//sxt/base/device:memory_utility",
         "//sxt/base/device:stream",
         "//sxt/base/error:assert",

sxt/multiexp/base/scalar_array.cc

@@ -18,8 +18,8 @@
 
 #include <vector>
 
+#include "cub/cub.cuh"
 #include "sxt/base/container/span_utility.h"
-#include "sxt/base/device/cub.h"
 #include "sxt/base/device/memory_utility.h"
 #include "sxt/base/device/stream.h"
 #include "sxt/base/num/ceil_log2.h"

sxt/multiexp/bucket_method2/BUILD

@@ -93,8 +93,8 @@ sxt_cc_component(
     deps = [
         ":constants",
         "//sxt/algorithm/block:runlength_count",
-        "//sxt/base/device:cub",
         "//sxt/base/device:stream",
+        "@local_cuda//:cub",
     ],
 )
 

sxt/multiexp/bucket_method2/multiproduct_table.cc

@@ -16,8 +16,8 @@
  */
 #include "sxt/multiexp/bucket_method2/multiproduct_table.h"
 
+#include "cub/cub.cuh"
 #include "sxt/algorithm/iteration/for_each.h"
-#include "sxt/base/device/cub.h"
 #include "sxt/base/device/memory_utility.h"
 #include "sxt/base/device/stream.h"
 #include "sxt/base/log/log.h"

sxt/multiexp/bucket_method2/multiproduct_table_kernel.h

@@ -18,8 +18,8 @@
 
 #include <utility>
 
+#include "cub/cub.cuh"
 #include "sxt/algorithm/block/runlength_count.h"
-#include "sxt/base/device/cub.h"
 #include "sxt/base/device/stream.h"
 #include "sxt/base/num/constexpr_switch.h"
 #include "sxt/base/num/divide_up.h"