Closed xrq-phys closed 2 years ago
@xrq-phys thank you for your great efforts! I merged PR.
@xrq-phys I now checked and found that an error occurs in using cmake (not using PFAFFIAN_BLOCKED).
CMake Error at CMakeLists.txt:74 (add_subdirectory):
add_subdirectory given source "src/pfapack/fortran" which is not an
existing directory.
This may be caused by the incorrect link for pfapack. Could you check the origin of this error?
I made Pfapack a submodule.
Sometimes git submodule update -r -i might fail to load due to clutters left inside the old directory. Could you try cleaning?
Best,
On Fri, Sep 24, 2021 at 9:09 Takahiro Misawa @.***> wrote:
@xrq-phys https://github.com/xrq-phys I now checked and found that an error occurs in using cmake (not using PFAFFIAN_BLOCKED).
CMake Error at CMakeLists.txt:74 (add_subdirectory): add_subdirectory given source "src/pfapack/fortran" which is not an existing directory.
This may be caused by the incorrect link for pfapack. Could you check the origin of this error?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/issp-center-dev/mVMC/pull/26#issuecomment-926251286, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB4GUGQIJAH3P5S2I25RS7LUDO6UZANCNFSM5ETPKUMQ .
--
Xu RuQing 許 ルーキン 東大理学系研究科 物理学専攻藤堂研 所属研究室Eメール: @. 東京大学生Eメール: @.
ok. Using git submodule update -r -i
, cmake now works.
thank you!
Btw with this CMake+prebuilt BLIS, Fugaku compilation no longer needs submitting compile.sh.
Cheers. 😜
mVMCconfig.sh
which builds BLIS locally, the CMake strategy downloads a system-specific prebuilt artifact from xrq-phys/blis@v0.8.1+arm.PFAFFIAN_BLOCKED
is created. If the option is set toOFF
, no blocked-Pfaffian would be built. No BLIS artifact would be downloaded, either./config
setPFAFFIAN_BLOCKED
toON
except forapple.cmake
(due to GitHub CI unavailability) andgcc.cmake
(since it's more or less system-agnostic and the auto-detection is not guaranteed to always work).One more thing done in this PR: User is able to use the environment variable
VMC_BLOCK_UPDATE_SIZE
to override the default block size of Pfaffian update. As this does not affect computational results, I didn't add it toreaddef.c
.