Public bug reported: # What was expected
I installed librocsparse1-tests on in a Resolute docker container, then I installed the AMD ROCm 7.1.0 packages for /opt/rocm-7.1.0 in a Noble docker container. I copied the contents of /opt/rocm-7.1.0 in the noble container into my Resolute container. This provided me with two copies of ROCm 7.1.0, so that I could mix-and-match libraries. I ran rocsparse-test using the libamd_comgr3.so from /opt/rocm through the use of LD_PRELOAD. This resulted in a passing test. The LD_PRELOAD should not be necessary. # LD_PRELOAD=/opt/rocm-7.1.0/lib/libamd_comgr.so /usr/libexec/rocm/librocsparse1-tests/rocsparse-test --gtest_filter='quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0' Query device success: there are 2 devices Device ID 0: AMD Radeon RX 9070 XT ------------------------------------------------------------------------- with 16304MB memory, clock rate 2520MHz @ computing capability 12.0 maxGridDimX 2147483647, sharedMemPerBlock 64KB, maxThreadsPerBlock 1024 wavefrontSize 32 ------------------------------------------------------------------------- Device ID 1: AMD Ryzen 9 7950X 16-Core Processor ------------------------------------------------------------------------- with 47800MB memory, clock rate 2200MHz @ computing capability 10.3 maxGridDimX 2147483647, sharedMemPerBlock 64KB, maxThreadsPerBlock 1024 wavefrontSize 32 ------------------------------------------------------------------------- Using device ID 0 (AMD Radeon RX 9070 XT) for rocSPARSE ------------------------------------------------------------------------- rocSPARSE version: 4.1.0- rocSPARSE data path: /usr/share/librocsparse1-tests/data/ rocsparse-test: debug arguments verbose is disabled for testings (use --force-debug-arguments-verbose to skip the disabling) rocsparse-test: warnings are disabled for testings (use --force-warnings to skip the disabling) Note: Google Test filter = quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0 [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from quick/axpby [ RUN ] quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0 [ OK ] quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0 (39 ms) [----------] 1 test from quick/axpby (39 ms total) [----------] Global test environment tear-down [==========] 1 test from 1 test suite ran. (46 ms total) [ PASSED ] 1 test. Failed to free ptr:0x7fd9d2e00000 size:2097152 The warning at the end of the output is https://bugs.launchpad.net/ubuntu/+source/rocr-runtime/+bug/2142805. # What happened Using the comgr provided by libamd_comgr3 7.1.0+dfsg-0ubuntu4, there is instead a fatal missing symbol error when the GPU kernel is invoked: # /usr/libexec/rocm/librocsparse1-tests/rocsparse-test --gtest_filter='quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0' Query device success: there are 2 devices Device ID 0: AMD Radeon RX 9070 XT ------------------------------------------------------------------------- with 16304MB memory, clock rate 2520MHz @ computing capability 12.0 maxGridDimX 2147483647, sharedMemPerBlock 64KB, maxThreadsPerBlock 1024 wavefrontSize 32 ------------------------------------------------------------------------- Device ID 1: AMD Ryzen 9 7950X 16-Core Processor ------------------------------------------------------------------------- with 47800MB memory, clock rate 2200MHz @ computing capability 10.3 maxGridDimX 2147483647, sharedMemPerBlock 64KB, maxThreadsPerBlock 1024 wavefrontSize 32 ------------------------------------------------------------------------- Using device ID 0 (AMD Radeon RX 9070 XT) for rocSPARSE ------------------------------------------------------------------------- rocSPARSE version: 4.1.0- rocSPARSE data path: /usr/share/librocsparse1-tests/data/ rocsparse-test: debug arguments verbose is disabled for testings (use --force-debug-arguments-verbose to skip the disabling) rocsparse-test: warnings are disabled for testings (use --force-warnings to skip the disabling) Note: Google Test filter = quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0 [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from quick/axpby [ RUN ] quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0 :0:./hipamd/src/hip_global.cpp:109 : 195188722870 us: Cannot find Symbol with name: _ZN9rocsparseL12axpyi_kernelILj256EfiffEEvT1_NS_24const_host_device_scalarIT0_EEPKT2_PKS1_PT3_21rocsparse_index_base_b Aborted (core dumped) /usr/libexec/rocm/librocsparse1-tests/rocsparse-test --gtest_filter='quick/axpby.level1/i32_f32_r_f32_r_f32_r_1200_5_1_0_0_0_0b_0' # ltrace info I ran the two with ltrace -e 'amd_comgr*' -e 'hip*' -e 'hsa*'. The logs are attached. I also prepared a diff of the logs after stripping function arguments from the ltrace output. # System Info ## lsb_release -rd Description: Ubuntu Resolute Raccoon (development branch) Release: 26.04 # apt-cache policy libamd-comgr3 libamd-comgr3: Installed: 7.1.0+dfsg-0ubuntu4 Candidate: 7.1.0+dfsg-0ubuntu4 Version table: *** 7.1.0+dfsg-0ubuntu4 500 500 http://archive.ubuntu.com/ubuntu resolute/universe amd64 Packages 100 /var/lib/dpkg/status ** Affects: rocm-llvm (Ubuntu) Importance: Undecided Status: New ** Attachment added: "diff of passing and failing ltrace logs" https://bugs.launchpad.net/bugs/2142813/+attachment/5948838/+files/rocsparse-test.diff -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2142813 Title: missing symbol error loading compressed bundles To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/rocm-llvm/+bug/2142813/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
