pytorch - 💡(How to fix) Fix ROCm `F.layer_norm` silently produces wrong output for large inputs (even below `M*N > 2^32` threshold)

"""ROCm F.layer_norm correctness repro.

Compares F.layer_norm(x) against a chunked reference (same op, slice-by-slice
along the leading dim) and counts how many rows differ by more than 1e-2.

Two shapes are tested:
  A. (1,1,2048,5247,267) — M*N = 2.87 B (BELOW 2^32)
  B. (1,1,5247,5247,267) — M*N = 7.35 B (ABOVE 2^32 — matches #181555)
"""
import torch
import torch.nn.functional as F


def check(shape, C, chunk=256):
    numel = 1
    for s in shape:
        numel *= s
    print(f"shape={shape} M*N={numel:,} ({numel / 2**32:.2f}x 2^32)")
    torch.manual_seed(0)
    x = torch.randn(*shape, device="cuda", dtype=torch.float32)
    w = torch.ones(C, device="cuda", dtype=torch.float32)

    out_full = F.layer_norm(x, (C,), weight=w, bias=None, eps=1e-5)
    torch.cuda.synchronize()

    bad = 0
    first_bad = -1
    for i in range(0, shape[2], chunk):
        ci = x[:, :, i:i + chunk].contiguous()
        cr = F.layer_norm(ci, (C,), weight=w, bias=None, eps=1e-5)
        diff = (cr - out_full[:, :, i:i + chunk]).abs()
        per_row = diff.amax(dim=(0, 1, 3, 4))
        for r in (per_row > 0.01).nonzero().flatten().tolist():
            bad += 1
            if first_bad < 0:
                first_bad = i + r
        del ci, cr, diff, per_row
    del x, w, out_full
    torch.cuda.empty_cache()
    print(f"  bad_rows={bad}/{shape[2]} first_bad={first_bad}")


if __name__ == "__main__":
    print(f"PyTorch: {torch.__version__}")
    print(f"HIP: {torch.version.hip}")
    print(f"Device: {torch.cuda.get_device_name(0)}\n")
    check((1, 1, 2048, 5247, 267), 267)
    check((1, 1, 5247, 5247, 267), 267)

---

shape=(1, 1, 2048, 5247, 267) M*N=2,869,143,552 (0.67x 2^32)
  bad_rows=1599/2048 first_bad=449
shape=(1, 1, 5247, 5247, 267) M*N=7,350,779,403 (1.71x 2^32)
  bad_rows=4797/5247 first_bad=450

---

[101877579776.0, 101877596160.0, 101877612544.0, 101877628928.0, ...]

---

Collecting environment information...
PyTorch version: 2.13.0.dev20260521+rocm7.2
Is debug build: False
CUDA used to build PyTorch: N/A
ROCM used to build PyTorch: 7.2.53211

OS: Ubuntu 22.04.5 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.10.12 (main, Jan 26 2026, 14:55:28) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.15.0-1074-oracle-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to:
GPU models and configuration: AMD Instinct MI300X (gfx942:sramecc+:xnack-)
Nvidia driver version: Could not collect
cuDNN version: Could not collect
Is XPU available: False
HIP runtime version: 7.2.53211
MIOpen runtime version: 3.5.1
Is XNNPACK available: True
Caching allocator config: N/A

CPU:
Architecture:                         x86_64
CPU op-mode(s):                       32-bit, 64-bit
Address sizes:                        46 bits physical, 57 bits virtual
Byte Order:                           Little Endian
CPU(s):                               224
On-line CPU(s) list:                  0-223
Vendor ID:                            GenuineIntel
Model name:                           Intel(R) Xeon(R) Platinum 8480+
CPU family:                           6
Model:                                143
Thread(s) per core:                   2
Core(s) per socket:                   56
Socket(s):                            2
Stepping:                             8
CPU max MHz:                          3800.0000
CPU min MHz:                          800.0000
BogoMIPS:                             4000.00
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx
fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology
nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid
dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb
cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid
ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb
intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local
split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku
ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect
cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8
flush_l1d arch_capabilities
Virtualization:                       VT-x
L1d cache:                            5.3 MiB (112 instances)
L1i cache:                            3.5 MiB (112 instances)
L2 cache:                             224 MiB (112 instances)
L3 cache:                             210 MiB (2 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-55,112-167
NUMA node1 CPU(s):                    56-111,168-223
Vulnerability Gather data sampling:   Not affected
Vulnerability Itlb multihit:          Not affected
Vulnerability L1tf:                   Not affected
Vulnerability Mds:                    Not affected
Vulnerability Meltdown:               Not affected
Vulnerability Mmio stale data:        Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed:               Not affected
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Vulnerable
Vulnerability Spectre v1:             Vulnerable: __user pointer sanitization and usercopy barriers only; no swapgs barriers
Vulnerability Spectre v2:             Vulnerable; IBPB: disabled; STIBP: disabled; PBRSB-eIBRS: Vulnerable; BHI: Vulnerable
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected

Versions of relevant libraries:
[pip3] torch==2.13.0.dev20260521+rocm7.2
[pip3] triton-rocm==3.7.0+git88b227e2
[conda] Could not collect