diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index 40517d542b3a0e942f4b78b90a9ed4b2e9360244..0f522beb1ac1f33bf977bfe1dd008ba62214e896 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -17,7 +17,12 @@
 #    reports:
 #      junit: test/pytest.xml
 
+stages:
+  - test
+  - deploy
+
 Pytest POCL:
+  stage: test
   script:
   - export PY_EXE=python3
   - export PYOPENCL_TEST=portable:pthread
@@ -34,6 +39,7 @@ Pytest POCL:
       junit: test/pytest.xml
 
 Pytest Titan V:
+  stage: test
   script:
   - py_version=3
   - export PYOPENCL_TEST=nvi:titan
@@ -51,6 +57,7 @@ Pytest Titan V:
       junit: test/pytest.xml
 
 Pytest Conda:
+  stage: test
   script:
   # Disable caching to ensure SymEngine code generation is exercised.
   - export SUMPY_NO_CACHE=1
@@ -68,6 +75,7 @@ Pytest Conda:
       junit: test/pytest.xml
 
 Pytest POCL Titan V:
+  stage: test
   script:
   # Disable caching to ensure SymEngine code generation is exercised.
   - export SUMPY_NO_CACHE=1
@@ -85,6 +93,7 @@ Pytest POCL Titan V:
       junit: test/pytest.xml
 
 Examples Conda:
+  stage: test
   script: |
     grep -v symengine .test-conda-env-py3.yml > .test-conda-env.yml
     CONDA_ENVIRONMENT=.test-conda-env.yml
@@ -99,14 +108,26 @@ Examples Conda:
   - tags
 
 Documentation:
-  script:
-  - EXTRA_INSTALL="pybind11 numpy mako"
-  - curl -L -O -k https://gitlab.tiker.net/inducer/ci-support/raw/main/build-docs.sh
-  - ". ./build-docs.sh"
+  stage: deploy
+  script: |
+    EXTRA_INSTALL="pybind11 numpy mako"
+    curl -L -O -k https://gitlab.tiker.net/inducer/ci-support/raw/main/ci-support.sh
+    . ci-support.sh
+    build_py_project_in_venv
+    pushd $PWD
+    build_docs
+    popd
+    if [[ -d ~/.scicomp-benchmarks/asv/$PROJECT ]]; then
+      cp -rf ~/.scicomp-benchmarks/asv/$PROJECT doc/_build/html/benchmarks
+    fi
+    maybe_upload_docs
+
   tags:
-  - python3
+  - linux
+  - benchmark
 
 Flake8:
+  stage: test
   script:
   - curl -L -O -k https://gitlab.tiker.net/inducer/ci-support/raw/main/prepare-and-run-flake8.sh
   - . ./prepare-and-run-flake8.sh "$CI_PROJECT_NAME" test examples benchmarks
@@ -116,6 +137,7 @@ Flake8:
   - tags
 
 Benchmarks:
+  stage: test
   script:
   - CONDA_ENVIRONMENT=.test-conda-env-py3.yml
   - PROJECT=sumpy
diff --git a/README.rst b/README.rst
index 7d4041be88d22e6f3102e914b543494863cbb804..3c5bf836f21956861106c5aac58e57652849e3c1 100644
--- a/README.rst
+++ b/README.rst
@@ -35,6 +35,4 @@ Resources:
 
 * `documentation <https://documen.tician.de/sumpy>`_
 * `source code via git <https://github.com/inducer/sumpy>`_
-
-If you can see inside the UIUC firewall, you may browse
-`benchmark results <https://koelsch.d.tiker.net/benchmarks/asv/sumpy/>`_.
+* `benchmarks <https://documen.tician.de/sumpy/benchmarks>`_
diff --git a/asv.conf.json b/asv.conf.json
index 4c34933d134fb2279ff05e11498e3f16a34beed6..800817d68e353523789be7bae34551d51f9fb636 100644
--- a/asv.conf.json
+++ b/asv.conf.json
@@ -20,7 +20,7 @@
 
     // List of branches to benchmark. If not provided, defaults to "master"
     // (for git) or "default" (for mercurial).
-    // "branches": ["master"], // for git
+    "branches": ["main"], // for git
     // "branches": ["default"],    // for mercurial
 
     // The DVCS being used.  If not set, it will be automatically