feat(cuda): VPMM v3 and async app prover (#2165)

Co-authored-by: Jonathan Wang <[email protected]>
Co-authored-by: Ayush Shukla <[email protected]>

Author: 3 people

.github/workflows/base-tests.cuda.yml

@@ -27,7 +27,7 @@ env:
 jobs:
   tests-cuda:
     runs-on:
-      - runs-on=${{ github.run_id }}-base-tests-cuda-${{ github.run_attempt }}/runner=test-gpu-nvidia
+      - runs-on=${{ github.run_id }}-base-tests-cuda-${{ github.run_attempt }}/runner=test-gpu-nvidia/family=g6+g5 # ensure 24G vram for async test
 
     steps:
       - uses: runs-on/action@v2
@@ -65,3 +65,8 @@ jobs:
         working-directory: extensions/native/recursion
         run: |
           cargo nextest run --features cuda
+
+      - name: Run async concurrency test
+        working-directory: benchmarks/prove
+        run: |
+          CUDA_OPT_LEVEL=3 MAX_CONCURRENCY=14 cargo run --bin async_regex --features cuda,async -- --max-segment-length $((1<<20)) --segment-max-cells 150000000

.github/workflows/sdk.yml

@@ -44,6 +44,8 @@ jobs:
 
       - name: Install Foundry
         uses: foundry-rs/foundry-toolchain@v1
+        with:
+          version: v1.3.5
 
       - name: Install architecture specific tools
         run: |