fix CI and updated readme

Author: muhos

.github/workflows/test-build.yml

@@ -12,16 +12,13 @@ on:
 
 jobs:
   build:
-
-    runs-on: ubuntu-24.04
-
+    runs-on: ubuntu-latest
+    container: nvidia/cuda:12.6.0-devel-ubuntu24.04
     steps:
-    - uses: actions/checkout@v3
-    - name: configure
-      run: |
-          wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
-          sudo dpkg -i cuda-keyring_1.1-1_all.deb
-          sudo apt-get update
-          sudo apt-get -y install cuda-toolkit-12-8
-    - name: make check
+      - uses: actions/checkout@v4
+      - name: Install dependencies
+        run: apt-get update && apt-get install -y cmake git build-essential
+      - name: Clone cuarena
+        run: git clone https://github.com/muhos/cuArena.git $HOME/cuarena
+      - name: Build
       run: bash install.sh -c -g

README.md

@@ -30,9 +30,10 @@ To install either the CPU or the GPU solvers, use the `install.sh` script which
        -a or --all           enable all above flags except 'assert'
        --ncolors             disable colors in all solver outputs
        --clean=<target>      remove old installation of <cpu | gpu | all> solvers
-       --standard=<n>        compile with <11 | 14 | 17> c++ standard
+       --standard=<n>        compile with <14 | 17 | 20> c++ standard
        --gextra="flags"      pass extra "flags" to the GPU compiler (nvcc)
        --cextra="flags"      pass extra "flags" to the CPU compiler (g++)
+       --cuarena=<path>      path to cuArena root (default: dep/cuarena; NONE to disable)
 
 
 ### GPU solver
@@ -51,6 +52,34 @@ installation guide in https://docs.nvidia.com/cuda/.
 Now the GPU solver is ready to install by running the install script via the command `./install.sh -g`. 
 The `parafrost` binary, the library `libparafrost.a`, and the main include file `solver.hpp` will be created by default in the build directory.<br>
 
+#### cuArena device pool backend
+
+The GPU solver can optionally use [cuArena](https://github.com/muhos/cuArena) as its GPU memory backend. cuArena pre-allocates a single contiguous GPU pool at startup and services all subsequent allocations entirely in host-side C++ with no `cudaMalloc` call per allocation. For ParaFROST this eliminates driver overhead during inprocessing, where the simplifier repeatedly allocates and frees buffers across many simplifying rounds.
+
+The pool is split into two regions:
+- **Stable**: long-term data that persists across simplification rounds; never moved by compaction.
+- **Dynamic**: the temporary buffers, which grow and shrink each phase. Defragmentation is supported and performed automatically when needed.
+
+To enable it, clone cuArena into `dep/cuarena` (the default search path) before building:
+
+```bash
+mkdir -p dep
+git clone https://github.com/muhos/cuArena.git dep/cuarena
+./install.sh -g
+```
+
+If cuArena is found and CMake is available, the install script automatically builds it and links it in (`-DUSE_CUARENA`). To point to a different location:
+
+```bash
+./install.sh -g --cuarena=/path/to/cuarena
+```
+
+To build without cuArena even if the directory exists:
+
+```bash
+./install.sh -g --cuarena=NONE
+```
+
 ### CPU solver
 To build a CPU-only version of the solver, run `./install.sh -c`.<br>