loop192: VPERM IN1, IN1, MASK, KEY // vperm 3,2,2,5 VSLDOI $12, ZERO, IN0, TMP // vsldoi 6,0,1,12 VCIPHERLAST KEY, RCON, KEY // vcipherlast 3,3,4 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $12, ZERO, TMP, TMP // vsldoi 6,0,6,12 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $12, ZERO, TMP, TMP // vsldoi 6,0,6,12 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $8, ZERO, IN1, STAGE // vsldoi 7,0,2,8 VSPLTW $3, IN0, TMP // vspltw 6,1,3 VXOR TMP, IN1, TMP // vxor 6,6,2 VSLDOI $12, ZERO, IN1, IN1 // vsldoi 2,0,2,12 VADDUWM RCON,...

loop192: VPERM IN1, IN1, MASK, KEY // vperm 3,2,2,5 VSLDOI $12, ZERO, IN0, TMP // vsldoi 6,0,1,12 VCIPHERLAST KEY, RCON, KEY // vcipherlast 3,3,4 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $12, ZERO, TMP, TMP // vsldoi 6,0,6,12 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $12, ZERO, TMP, TMP // vsldoi 6,0,6,12 VXOR IN0, TMP, IN0 // vxor 1,1,6 VSLDOI $8, ZERO, IN1, STAGE // vsldoi 7,0,2,8 VSPLTW $3, IN0, TMP // vspltw 6,1,3 VXOR TMP, IN1, TMP // vxor 6,6,2 VSLDOI $12, ZERO, IN1, IN1 // vsldoi 2,0,2,12 VADDUWM RCON,...

    large_constant]`) into 32-bit arithmetic in SASS.

    As a result, these versions of `ptxas` miscompile most XLA programs which
    use more than 4GB of temp memory. This results in garbage results and/or
    `CUDA_ERROR_ILLEGAL_ADDRESS` failures.

    A fix in CUDA 9.1.121 is expected in late February 2018. We do not expect a

* 📝 Add External Link: Instrument a FastAPI service adding tracing with OpenTelemetry and send/show traces in Grafana Tempo. PR [#9440](https://github.com/tiangolo/fastapi/pull/9440) by [@softwarebloat](https://github.com/softwarebloat).