[cherry-pick][release 2.8] Update OpenBLAS commit (pytorch#151547) (pytorch#158243)

* Update OpenBLAS commit  (pytorch#151547)

Motivation: Update OpenBLAS and change build script to enable SBGEMM kernels . Update pytorch `jammy` builds for aarch64 to use `install_openblas.sh` instead of `conda_install`

Link to full [TorchInductor Performance Dashboard AArch64](https://hud.pytorch.org/benchmark/compilers?dashboard=torchinductor&startTime=Fri%2C%2006%20Jun%202025%2009%3A46%3A35%20GMT&stopTime=Fri%2C%2013%20Jun%202025%2009%3A46%3A35%20GMT&granularity=hour&mode=inference&dtype=bfloat16&deviceName=cpu%20(aarch64)&lBranch=adi/update_openblas&lCommit=0218b65bcf61971c1861cfe8bc586168b73aeb5f&rBranch=main&rCommit=9d59b516e9b3026948918e3ff8c2ef55a33d13ad)

1. This shows a promising speedup across most of the HF models in benchmark, specifically giving a significant boost to SDPA layers.
2. Overall torch-bench pass-rate (cpp_wrapper mode) increased `[87%, 65/75 → 96%, 72/75]`

<img width="676" alt="Screenshot 2025-06-20 at 17 05 15" src="https://github.com/user-attachments/assets/2ca9c1bc-80c6-464a-8db6-b758f2476582" />

Pull Request resolved: pytorch#151547
Approved by: https://github.com/malfet, https://github.com/snadampal, https://github.com/fadara01

Co-authored-by: Christopher Sidebottom <[email protected]>
Co-authored-by: Ryo Suzuki <[email protected]>
Co-authored-by: Ye Tao <[email protected]>
Co-authored-by: Nikita Shulga <[email protected]>

* Update .ci/docker/common/install_conda.sh

Co-authored-by: Andrey Talman <[email protected]>

* Update .ci/docker/common/install_conda.sh

Co-authored-by: Andrey Talman <[email protected]>

* Update .ci/docker/common/install_conda.sh

Co-authored-by: Andrey Talman <[email protected]>

* try reverting conda install

---------

Co-authored-by: Aditya Tewari <[email protected]>
Co-authored-by: Christopher Sidebottom <[email protected]>
Co-authored-by: Ryo Suzuki <[email protected]>
Co-authored-by: Ye Tao <[email protected]>
Co-authored-by: Nikita Shulga <[email protected]>
Co-authored-by: Andrey Talman <[email protected]>

Author: 7 people

.ci/docker/build.sh

@@ -333,6 +333,8 @@ case "$tag" in
     GCC_VERSION=11
     ACL=yes
     VISION=yes
+    CONDA_CMAKE=yes
+    OPENBLAS=yes
     # snadampal: skipping llvm src build install because the current version
     # from pytorch/llvm:9.0.1 is x86 specific
     SKIP_LLVM_SRC_BUILD_INSTALL=yes
@@ -342,6 +344,8 @@ case "$tag" in
     GCC_VERSION=11
     ACL=yes
     VISION=yes
+    CONDA_CMAKE=yes
+    OPENBLAS=yes
     # snadampal: skipping llvm src build install because the current version
     # from pytorch/llvm:9.0.1 is x86 specific
     SKIP_LLVM_SRC_BUILD_INSTALL=yes
@@ -428,6 +432,7 @@ docker build \
        --build-arg "XPU_VERSION=${XPU_VERSION}" \
        --build-arg "UNINSTALL_DILL=${UNINSTALL_DILL}" \
        --build-arg "ACL=${ACL:-}" \
+       --build-arg "OPENBLAS=${OPENBLAS:-}" \
        --build-arg "SKIP_SCCACHE_INSTALL=${SKIP_SCCACHE_INSTALL:-}" \
        --build-arg "SKIP_LLVM_SRC_BUILD_INSTALL=${SKIP_LLVM_SRC_BUILD_INSTALL:-}" \
        -f $(dirname ${DOCKERFILE})/Dockerfile \

.ci/docker/common/install_openblas.sh

@@ -4,18 +4,18 @@
 set -ex
 
 cd /
-git clone https://github.com/OpenMathLib/OpenBLAS.git -b "${OPENBLAS_VERSION:-v0.3.29}" --depth 1 --shallow-submodules
+git clone https://github.com/OpenMathLib/OpenBLAS.git -b "${OPENBLAS_VERSION:-v0.3.30}" --depth 1 --shallow-submodules
 
+OPENBLAS_CHECKOUT_DIR="OpenBLAS"
 OPENBLAS_BUILD_FLAGS="
 NUM_THREADS=128
 USE_OPENMP=1
 NO_SHARED=0
 DYNAMIC_ARCH=1
 TARGET=ARMV8
 CFLAGS=-O3
+BUILD_BFLOAT16=1
 "
 
-OPENBLAS_CHECKOUT_DIR="OpenBLAS"
-
 make -j8 ${OPENBLAS_BUILD_FLAGS} -C ${OPENBLAS_CHECKOUT_DIR}
 make -j8 ${OPENBLAS_BUILD_FLAGS} install -C ${OPENBLAS_CHECKOUT_DIR}

.ci/docker/manywheel/build.sh

@@ -41,7 +41,7 @@ case ${image} in
         GPU_IMAGE=arm64v8/almalinux:8
         DOCKER_GPU_BUILD_ARG=" --build-arg DEVTOOLSET_VERSION=13 --build-arg NINJA_VERSION=1.12.1"
         MANY_LINUX_VERSION="2_28_aarch64"
-        OPENBLAS_VERSION="v0.3.29"
+        OPENBLAS_VERSION="v0.3.30"
         ;;
     manylinuxcxx11-abi-builder:cpu-cxx11-abi)
         TARGET=final

.ci/docker/ubuntu/Dockerfile

@@ -147,6 +147,12 @@ RUN if [ -n "${ACL}" ]; then bash ./install_acl.sh; fi
 RUN rm install_acl.sh
 ENV INSTALLED_ACL ${ACL}
 
+ARG OPENBLAS
+COPY ./common/install_openblas.sh install_openblas.sh
+RUN if [ -n "${OPENBLAS}" ]; then bash ./install_openblas.sh; fi
+RUN rm install_openblas.sh
+ENV INSTALLED_OPENBLAS ${OPENBLAS}
+
 # Install ccache/sccache (do this last, so we get priority in PATH)
 ARG SKIP_SCCACHE_INSTALL
 COPY ./common/install_cache.sh install_cache.sh