[build] Blackwell Support LLVM20 + ARM Support #8735

johnnynunez · 2025-06-12T07:51:27Z

I open this PR if anyone want to use my fork
thanks to sonicflux fork also

454040244-f87ae1b3-b55f-4b6e-b385-8163a0009008.mp4

…es (#7)

…ies (#5)

…forms (#4)

…tures (#3)

* [update] Upgrade Vulkan SDK to version 1.4.309.0 and add support for ARM64 on Linux * [update] Refactor ImGui Vulkan integration for compatibility with version ≥ 1.90 * [logging] Include ranges header from spdlog for enhanced formatting capabilities

* [update] Upgrade Vulkan SDK to version 1.4.309.0 and add support for ARM64 on Linux * [update] Refactor ImGui Vulkan integration for compatibility with version ≥ 1.90 * [logging] Include ranges header from spdlog for enhanced formatting capabilities * Update vulkan.py

…us files

…nt changes

…TX handling, and implement new pass manager setup

…d missing include in window_base.cpp

… environment variable settings in llvm.py

…ity; preserve bool handling and adjust metadata usage

…bility; streamline bool handling and update metadata usage

…caching; switch to nvvm_ldu instructions for better compatibility

…with invariant metadata; remove nvvm.ldg intrinsics for improved compatibility

…ldg intrinsics with standard loads; update metadata usage for invariant loads

johnnynunez · 2025-06-25T23:47:26Z

@feisuzhu @yuanming-hu you can copy my files from here to merge for blackwell

…ple files

Paulo-21 · 2025-07-21T19:13:31Z

Hello, thank you for your work !
i m trying to use your branch but i have struggle to compile your branch with CUDA TOOLKIT flag enable .
I tried to compile it with cuda 12.9 but failed.
Do you think taichi is compatible with recent cuda version ?
TY

johnnynunez · 2025-07-21T19:39:32Z

Hello, thank you for your work ! i m trying to use your branch but i have struggle to compile your branch with CUDA TOOLKIT flag enable . I tried to compile it with cuda 12.9 but failed. Do you think taichi is compatible with recent cuda version ? TY

yes it is, i have it running in jetson and gh200

Paulo-21 · 2025-07-21T20:10:31Z

Ho so cool !
Would it be possible for you to explain me how did you buid or use cuda 12.9 for build taichi kernel for you recent nvidia gpu ?
I have a RTX 6000 ada, and i feel it's not used to it full potential because taichi use cuda 10.0.
I was able to build taichy with llvm 20.1.8 and when i print debug it show llvm : 20.1.8 succesfully.

I have see that maybe i have to replace slim_libdevice.10.bc with my 12.9 version. but i have struggle to find it.

import taichi as ti
ti.init(arch=ti.cuda)
print("CUDA version :", ti._lib.core._cuda_version())

This show cuda 10
Is it the right way ?
TY

Paulo-21 · 2025-07-21T20:53:35Z

And every time i tried to compile with TI_WITH_CUDA_TOOLKIT flag ON and with cuda toolkit 12.9
I have this :

The error Trace

FAILED: taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_codegen.cpp.o
      /usr/bin/clang++-20  -I/tmp/pip-build-env-p4ho7jse/overlay/lib/python3.12/site-packages/numpy/_core/include -I/home/paul/taichi-new -I/home/paul/taichi-new/external/eigen -I/home/paul/taichi-new/external/include -I/home/paul/taichi-new/external/SPIRV-Headers/include -I/home/paul/taichi-new/external/SPIRV-Reflect -I/home/paul/taichi-new/external/FP16/include -I/usr/lib/llvm-20/include -I/home/paul/taichi-new/external/spdlog/include -I/home/paul/taichi-new/external/SPIRV-Tools/include -ffunction-sections -fdata-sections -DTI_ISE_NONE -std=c++17 -fsized-deallocation -Wno-deprecated-declarations -Wno-shorten-64-to-32 -Wall  -Werror  -Wno-ignored-attributes  -Wno-nullability-completeness  -Wno-unused-private-field  -Wno-unneeded-internal-declaration  -Wno-unqualified-std-cast-call  -Wno-unused-but-set-variable  -DTI_ARCH_x64 -march=nehalem -DTI_PASS_EXCEPTION_TO_PYTHON -DTI_INCLUDED -DTI_WITH_LLVM -DTI_WITH_CUDA -DTI_WITH_OPENGL -DTI_WITH_CUDA_TOOLKIT -O3 -DNDEBUG -std=gnu++17 -fPIC -fvisibility=hidden -fvisibility-inlines-hidden -fcolor-diagnostics   -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -MD -MT taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_codegen.cpp.o -MF taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_codegen.cpp.o.d -o taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_codegen.cpp.o -c /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:8:53: error: use of undeclared identifier 'cusolverSpHandle_t'
          8 | PER_CUSOLVER_FUNCTION(csSpCreate, cusolverSpCreate, cusolverSpHandle_t * );
            |                                                     ^
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:8:1: error: expected expression
          8 | PER_CUSOLVER_FUNCTION(csSpCreate, cusolverSpCreate, cusolverSpHandle_t * );
            | ^
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:33: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:9:55: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
          9 | PER_CUSOLVER_FUNCTION(csSpDestory, cusolverSpDestroy, cusolverSpHandle_t );
            |                                                       ^~~~~~~~~~~~~~~~~~
            |                                                       cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:10:67: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         10 | PER_CUSOLVER_FUNCTION(csSpXcsrissymHost, cusolverSpXcsrissymHost, cusolverSpHandle_t,int ,int ,const cusparseMatDescr_t ,const void *,const void *,const void *,void *);
            |                                                                   ^~~~~~~~~~~~~~~~~~
            |                                                                   cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:11:69: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         11 | PER_CUSOLVER_FUNCTION(csSpXcsrsymrcmHost, cusolverSpXcsrsymrcmHost, cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,const void *,const void *,void *);
            |                                                                     ^~~~~~~~~~~~~~~~~~
            |                                                                     cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:12:69: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         12 | PER_CUSOLVER_FUNCTION(csSpXcsrsymamdHost, cusolverSpXcsrsymamdHost, cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,const void *,const void *,void *);
            |                                                                     ^~~~~~~~~~~~~~~~~~
            |                                                                     cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:13:87: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         13 | PER_CUSOLVER_FUNCTION(csSpXcsrperm_bufferSizeHost, cusolverSpXcsrperm_bufferSizeHost, cusolverSpHandle_t ,int ,int ,int ,const cusparseMatDescr_t ,void *,void *,const void *,const void *,size_t *);
            |                                                                                       ^~~~~~~~~~~~~~~~~~
            |                                                                                       cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:14:65: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         14 | PER_CUSOLVER_FUNCTION(csSpXcsrpermHost, cusolverSpXcsrpermHost, cusolverSpHandle_t ,int ,int ,int ,const cusparseMatDescr_t ,void *,void *,const void *,const void *,void *,void *);
            |                                                                 ^~~~~~~~~~~~~~~~~~
            |                                                                 cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:15:71: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         15 | PER_CUSOLVER_FUNCTION(csSpScsrlsvcholHost, cusolverSpScsrlsvcholHost, cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,const void *,const void *,const void *,const void *,float ,int ,void *,void *);
            |                                                                       ^~~~~~~~~~~~~~~~~~
            |                                                                       cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:18:75: error: use of undeclared identifier 'csrcholInfo_t'
         18 | PER_CUSOLVER_FUNCTION(csSpCreateCsrcholInfo, cusolverSpCreateCsrcholInfo, csrcholInfo_t*);
            |                                                                           ^
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:18:1: error: expected expression
         18 | PER_CUSOLVER_FUNCTION(csSpCreateCsrcholInfo, cusolverSpCreateCsrcholInfo, csrcholInfo_t*);
            | ^
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:33: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:19:77: error: unknown type name 'csrcholInfo_t'; did you mean 'csric02Info_t'?
         19 | PER_CUSOLVER_FUNCTION(csSpDestroyCsrcholInfo, cusolverSpDestroyCsrcholInfo, csrcholInfo_t);
            |                                                                             ^~~~~~~~~~~~~
            |                                                                             csric02Info_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:230:21: note: 'csric02Info_t' declared here
        230 | struct csric02Info* csric02Info_t CUSPARSE_DEPRECATED_TYPE;
            |                     ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:20:74: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         20 | PER_CUSOLVER_FUNCTION(csSpXcsrcholAnalysis, cusolverSpXcsrcholAnalysis,  cusolverSpHandle_t,int ,int ,const cusparseMatDescr_t , void *, void *,csrcholInfo_t );
            |                                                                          ^~~~~~~~~~~~~~~~~~
            |                                                                          cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:20:145: error: unknown type name 'csrcholInfo_t'; did you mean 'csric02Info_t'?
         20 | PER_CUSOLVER_FUNCTION(csSpXcsrcholAnalysis, cusolverSpXcsrcholAnalysis,  cusolverSpHandle_t,int ,int ,const cusparseMatDescr_t , void *, void *,csrcholInfo_t );
            |                                                                                                                                                 ^~~~~~~~~~~~~
            |                                                                                                                                                 csric02Info_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:230:21: note: 'csric02Info_t' declared here
        230 | struct csric02Info* csric02Info_t CUSPARSE_DEPRECATED_TYPE;
            |                     ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:21:78: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         21 | PER_CUSOLVER_FUNCTION(csSpScsrcholBufferInfo, cusolverSpScsrcholBufferInfo,  cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,void *,void *,void *,csrcholInfo_t ,size_t *,size_t *);
            |                                                                              ^~~~~~~~~~~~~~~~~~
            |                                                                              cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:21:155: error: unknown type name 'csrcholInfo_t'; did you mean 'csric02Info_t'?
         21 | PER_CUSOLVER_FUNCTION(csSpScsrcholBufferInfo, cusolverSpScsrcholBufferInfo,  cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,void *,void *,void *,csrcholInfo_t ,size_t *,size_t *);
            |                                                                                                                                                           ^~~~~~~~~~~~~
            |                                                                                                                                                           csric02Info_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:230:21: note: 'csric02Info_t' declared here
        230 | struct csric02Info* csric02Info_t CUSPARSE_DEPRECATED_TYPE;
            |                     ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:22:69: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         22 | PER_CUSOLVER_FUNCTION(csSpScsrcholFactor, cusolverSpScsrcholFactor, cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,void *,void *,void *,csrcholInfo_t ,void *);
            |                                                                     ^~~~~~~~~~~~~~~~~~
            |                                                                     cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:22:146: error: unknown type name 'csrcholInfo_t'; did you mean 'csric02Info_t'?
         22 | PER_CUSOLVER_FUNCTION(csSpScsrcholFactor, cusolverSpScsrcholFactor, cusolverSpHandle_t ,int ,int ,const cusparseMatDescr_t ,void *,void *,void *,csrcholInfo_t ,void *);
            |                                                                                                                                                  ^~~~~~~~~~~~~
            |                                                                                                                                                  csric02Info_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:230:21: note: 'csric02Info_t' declared here
        230 | struct csric02Info* csric02Info_t CUSPARSE_DEPRECATED_TYPE;
            |                     ^
      In file included from /home/paul/taichi-new/taichi/codegen/spirv/spirv_codegen.cpp:8:
      In file included from /home/paul/taichi-new/taichi/codegen/codegen_utils.h:2:
      In file included from /home/paul/taichi-new/taichi/program/program.h:29:
      In file included from /home/paul/taichi-new/taichi/program/sparse_matrix.h:8:
      In file included from /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:192:
      /home/paul/taichi-new/taichi/rhi/cuda/cusolver_functions.inc.h:23:75: error: unknown type name 'cusolverSpHandle_t'; did you mean 'cusparseHandle_t'?
         23 | PER_CUSOLVER_FUNCTION(csSpScsrcholZeroPivot, cusolverSpScsrcholZeroPivot, cusolverSpHandle_t, csrcholInfo_t ,float ,void *);
            |                                                                           ^~~~~~~~~~~~~~~~~~
            |                                                                           cusparseHandle_t
      /home/paul/taichi-new/taichi/rhi/cuda/cuda_driver.h:191:22: note: expanded from macro 'PER_CUSOLVER_FUNCTION'
        191 |   CUDADriverFunction<__VA_ARGS__> name;
            |                      ^
      /usr/local/cuda-12.9/include/cusparse.h:215:33: note: 'cusparseHandle_t' declared here
        215 | typedef struct cusparseContext* cusparseHandle_t;
            |                                 ^
      fatal error: too many errors emitted, stopping now [-ferror-limit=]
      20 errors generated.
      [440/647] Building CXX object taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/kernel_utils.cpp.o
      [441/647] Building CXX object taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_types.cpp.o
      [442/647] Building CXX object taichi/codegen/spirv/CMakeFiles/spirv_codegen.dir/spirv_ir_builder.cpp.o
      ninja: build stopped: subcommand failed.
      Traceback (most recent call last):
        File "/tmp/pip-build-env-p4ho7jse/overlay/lib/python3.12/site-packages/skbuild/setuptools_wrap.py", line 668, in setup
          cmkr.make(make_args, install_target=cmake_install_target, env=env)
        File "/tmp/pip-build-env-p4ho7jse/overlay/lib/python3.12/site-packages/skbuild/cmaker.py", line 696, in make
          self.make_impl(clargs=clargs, config=config, source_dir=source_dir, install_target=install_target, env=env)
        File "/tmp/pip-build-env-p4ho7jse/overlay/lib/python3.12/site-packages/skbuild/cmaker.py", line 741, in make_impl
          raise SKBuildError(msg)
      
      An error occurred while building with CMake.
        Command:
          /tmp/pip-build-env-p4ho7jse/overlay/lib/python3.12/site-packages/cmake/data/bin/cmake --build . --target install --config Release -- -j4
        Install target:
          install
        Source directory:
          /home/paul/taichi-new
        Working directory:
          /home/paul/taichi-new/_skbuild/linux-x86_64-3.12/cmake-build
      Please check the install target is valid and see CMake's output for more information.
      
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.

johnnynunez and others added 30 commits May 8, 2025 14:33

[feature] Add support for extracting .tar.xz archives in dep.py (#8)

2cae070

[llvm] Enhance LLVM installation logic for specific Linux architectur…

c3e8aa2

…es (#7)

[fix] Update setup function to use Miniforge and correct download URLs (

323026b

#6)

[build] Replace libtinfo-dev with libncurses-dev in package dependenc…

fb220b7

…ies (#5)

[fix] Update sccache download URLs to version 0.10.0 for various plat…

bc504cb

…forms (#4)

[fix] Update platform detection for manylinux support on ARM architec…

fde8884

…tures (#3)

Vulkan (#2)

998d76c

* [update] Upgrade Vulkan SDK to version 1.4.309.0 and add support for ARM64 on Linux * [update] Refactor ImGui Vulkan integration for compatibility with version ≥ 1.90 * [logging] Include ranges header from spdlog for enhanced formatting capabilities

[update] Update hash values for various dependencies (#1)

2efbff4

Vulkan (#9)

c54bbd1

* [update] Upgrade Vulkan SDK to version 1.4.309.0 and add support for ARM64 on Linux * [update] Refactor ImGui Vulkan integration for compatibility with version ≥ 1.90 * [logging] Include ranges header from spdlog for enhanced formatting capabilities * Update vulkan.py

Merge branch 'taichi-dev:master' into master

f4abba1

fix

c79a45d

fix

309f586

Update compiler.py

9301aba

Update llvm.py

abf20a1

Update llvm.py

8a476be

Update llvm.py

a1cb73b

Update check_clang_tidy.sh

88eb2f0

Update llvm.py

1b3a244

Add IR dumping and loading functionality in LLVM codegen

091a8b7

Add constructor for ASTBuilder to initialize state

c8efe96

Refactor ASTBuilder constructor declaration for improved clarity

cf76d97

Remove unnecessary print statement from IRNode handling

bf18da4

Refactor CMakeLists.txt and improve IR dumping functionality in vario…

ae44a22

…us files

LLVM-20

764f6bb

Update LLVM API calls in codegen_cuda.cpp for compatibility with rece…

c24757a

…nt changes

Add CHANGELOG.md to document recent updates and improvements

11f240e

Update Vulkan setup to check for x86_64 architecture on Linux

638564e

Fix include directive for IR analysis header in codegen_cuda.cpp

68249bd

Refactor JIT compilation in CUDA: update function pointers, enhance P…

38625e8

…TX handling, and implement new pass manager setup

Fix header include for PassBuilder in jit_cuda.cpp

a8805e2

johnnynunez and others added 10 commits June 11, 2025 18:25

Refactor CMakeLists.txt for improved readability and organization; ad…

eec70ab

…d missing include in window_base.cpp

Update CUDA intrinsic load handling for LLVM 20 compatibility; adjust…

8ca16de

… environment variable settings in llvm.py

Update intrinsic load handling in TaskCodeGenCUDA for LLVM compatibil…

403d73c

…ity; preserve bool handling and adjust metadata usage

Refactor intrinsic load handling in codegen_cuda.cpp for LLVM compati…

5caa5db

…bility; streamline bool handling and update metadata usage

Update intrinsic load handling in codegen_cuda.cpp for improved data …

8dd3ff3

…caching; switch to nvvm_ldu instructions for better compatibility

Update intrinsic load handling in codegen_cuda.cpp to use CreateLoad …

62fec58

…with invariant metadata; remove nvvm.ldg intrinsics for improved compatibility

Refactor intrinsic load handling in codegen_cuda.cpp to replace nvvm.…

074cf13

…ldg intrinsics with standard loads; update metadata usage for invariant loads

Fix unused variable warning for Vulkan command buffer in gui.cpp

60112ac

Remove commented-out code related to opaque pointers in codegen_cuda.cpp

add2df3

Update entry.py

649c58d

johnnynunez added 2 commits June 26, 2025 23:57

Merge branch 'taichi-dev:master' into master

fc11f2e

Merge branch 'taichi-dev:master' into master

aafbcff

johnnynunez mentioned this pull request Jul 1, 2025

LLVM 15 IS OLD #8674

Open

johnnynunez added 2 commits July 2, 2025 00:44

Update Miniforge download URLs to version 25.3.0-3 for all platforms

48fe16c

Clean up code formatting and remove unnecessary comments across multi…

6a6d508

…ple files

johnnynunez changed the title ~~Blackwell Support LLVM20 + ARM Support~~ [build] Blackwell Support LLVM20 + ARM Support Jul 2, 2025

johnnynunez added 5 commits July 2, 2025 23:58

Update ospkg.py

b21d29f

Update ospkg.py

24f6084

Update ospkg.py

963e0a6

Update PythonNumpyPybind11.cmake

ae1275e

Update ospkg.py

93c190f

This was referenced Jul 16, 2025

Taichi and 50-Series GPUs: The Vulkan 1.4 Challenge #8743

Open

[Bug]: CUDA_ERROR_ILLEGAL_ADDRESS when using RTX 5080 GPU #8728

Open

Vulkan-1.4.314-Released #8710

Closed

Paulo-21 mentioned this pull request Jul 24, 2025

[Feature]: Make Taichi support Cuda 12.X Genesis-Embodied-AI/gstaichi#166

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[build] Blackwell Support LLVM20 + ARM Support #8735

[build] Blackwell Support LLVM20 + ARM Support #8735

Uh oh!

johnnynunez commented Jun 12, 2025

Uh oh!

johnnynunez commented Jun 25, 2025

Uh oh!

Paulo-21 commented Jul 21, 2025

Uh oh!

johnnynunez commented Jul 21, 2025

Uh oh!

Paulo-21 commented Jul 21, 2025 •

edited

Loading

Uh oh!

Paulo-21 commented Jul 21, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

[build] Blackwell Support LLVM20 + ARM Support #8735

Are you sure you want to change the base?

[build] Blackwell Support LLVM20 + ARM Support #8735

Uh oh!

Conversation

johnnynunez commented Jun 12, 2025

Uh oh!

johnnynunez commented Jun 25, 2025

Uh oh!

Paulo-21 commented Jul 21, 2025

Uh oh!

johnnynunez commented Jul 21, 2025

Uh oh!

Paulo-21 commented Jul 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Paulo-21 commented Jul 21, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Paulo-21 commented Jul 21, 2025 •

edited

Loading