Bug#1056173: librocsparse0-tests: CBSRMV test failure on gfx1030

[Cordell Bloor <cgmb@slerp.xyz>]

Package: librocsparse0-tests
Version: 5.5.1-2
Severity: minor
X-Debbugs-Cc: cgmb@slerp.xyz

Dear Maintainer,

There is a test failing with gfx1030 on the CI:

    [ RUN      ] quick/bsrmv.level2/f32_c_500_842_1_1_n1_n0p5_row_2_NT_0b_rand_t2
    clients/common/rocsparse_check.cpp:287: Failure
    The difference between std::imag(A[i + j * LDA]) and std::imag(B[i + j * LDB]) is 5.7220458984375e-06, which exceeds std::imag(compare_val), where
    std::imag(A[i + j * LDA]) evaluates to -0.00034707784652709961,
    std::imag(B[i + j * LDB]) evaluates to -0.00034135580062866211, and
    std::imag(compare_val) evaluates to 1.1920928955078125e-06.
131s
    clients/common/rocsparse_check.cpp:287: Failure
    The difference between std::imag(A[i + j * LDA]) and std::imag(B[i + j * LDB]) is 5.7220458984375e-06, which exceeds std::imag(compare_val), where
    std::imag(A[i + j * LDA]) evaluates to -0.00034707784652709961,
    std::imag(B[i + j * LDB]) evaluates to -0.00034135580062866211, and
    std::imag(compare_val) evaluates to 1.1920928955078125e-06.
131s
    [  FAILED  ] quick/bsrmv.level2/f32_c_500_842_1_1_n1_n0p5_row_2_NT_0b_rand_t2, where GetParam() = { function: "bsrmv", index_type_I: "i32", index_type_J: "i32", a_type: "f32_c", b_type: "f32_c", c_type: "f32_c", x_type: "f32_c", y_type: "f32_c", compute_type: "f32_c", transA: "NT", transB: "NT", baseA: "0b", baseB: "0b", baseC: "0b", baseD: "0b", M: 500, N: 842, K: -1, nnz: -1, block_dim: 2, row_block_dimA: 2, col_block_dimA: 2, row_block_dimB: 2, col_block_dimB: 2, dim_x: 1, dim_y: 1, dim_z: 1, ll: -2, l: -1, u: 1, uu: 2, alpha: 1.0, alphai: 1.0, beta: -1.0, betai: -0.5, threshold: 1.0, percentage: 0.0, action: "num", part: "auto", matrix_type: "general", diag: "ND", uplo: "L", storage: "unsorted", analysis_policy: "reuse", solve_policy: "auto", direction: "row", order: "col", format: "coo", itilu0_alg: "async_inplace", sddmm_alg: "default", spmv_alg: "default", spsv_alg: "default", spitsv_alg: "default", spsm_alg: "default", spmm_alg: "default", spgemm_alg: "default", sparse_to_dense_alg: "default", dense_to_sparse_alg: "default", gtsv_interleaved_alg: "default", gpsv_interleaved_alg: "default", matrix: "rand", matrix_init_kind: "default", file: "*", algo: 0, numeric_boost: 0, boost_tol: 0.0, boost_val: 1.0, boost_vali: 0.0, tolm: 1.0, graph_test: false, name: "bsrmv", category: "quick", hardware: "all", unit_check: 1, timing: 0, iters: 10, denseld: -1, batch_count: -1, batch_count_A: -1, batch_count_B: -1, batch_count_C: -1, batch_stride: -1 }
     (5 ms)

It is difficult to be sure whether this is an indication of a real
problem or merely a test that is not numerically robust enough.

Sincerely,
Cory Bloor

-- System Information:
Debian Release: trixie/sid
  APT prefers testing
  APT policy: (500, 'testing'), (1, 'experimental')
Architecture: amd64 (x86_64)

Kernel: Linux 6.5.0-2-amd64 (SMP w/32 CPU threads; PREEMPT)
Locale: LANG=C, LC_CTYPE=C.UTF-8 (charmap=UTF-8), LANGUAGE not set
Shell: /bin/sh linked to /usr/bin/dash
Init: unable to detect

Versions of packages librocsparse0-tests depends on:
ii  libamdhip64-5             5.2.3-13
ii  libc6                     2.37-12
ii  libgcc-s1                 13.2.0-5
ii  librocsparse0             5.5.1-2
ii  librocsparse0-tests-data  5.5.1-2
ii  libstdc++6                13.2.0-5

librocsparse0-tests recommends no packages.

librocsparse0-tests suggests no packages.

-- no debconf information