`tf.distribute.experimental.MultiWorkerMirroredStrategy`
    *   Update NVIDIA `NCCL` to `2.5.7-1` for better performance and performance
        tuning. Please see
        [nccl developer guide](https://docs.nvidia.com/deeplearning/sdk/nccl-developer-guide/docs/env.html)
        for more information on this.
    *   Support gradient `allreduce` in `float16`. See this

src+24(FP), SRC MOVD ivlo+32(FP), IV_LOW_LE MOVD ivhi+40(FP), IV_HIGH_LE {{/* Prepare plain from IV and blockIndex. */}} {{/* Copy to plaintext registers. */}} {{ range $i := xrange $N }} REV IV_LOW_LE, IV_LOW_BE REV IV_HIGH_LE, IV_HIGH_BE {{- /* https://developer.arm.com/documentation/dui0801/g/A64-SIMD-Vector-Instructions/MOV--vector--from-general- */}} VMOV IV_LOW_BE, V{{ block_reg $i }}.D[1] VMOV IV_HIGH_BE, V{{ block_reg $i }}.D[0] {{- if ne (add $i 1) $N }} ADDS $1, IV_LOW_LE ADC $0, IV_HIGH_LE {{ end...