Show Harden Runner enabled runs only
When `MxFp4Codec::QuantizeBlock` encounters an all-zero input block, the early-out path zeroes `kBlockSize` (32) bytes instead of the encoded block size `sizeof(MxFp4Stream)` (17 bytes: 1 scale byte plus 16 nibble bytes). Writing 15 bytes past the end of the block corrupts the adjacent block when blocks are encoded out of order (such as across parallel batch boundaries in `Compress` or via `packed_ofs` sub-ranges) and writes past `PackedEnd` on the final block of a tensor.
#1045
build #8212build.yml10 days ago
5Destinations
7Actions
Insights
When `MxFp4Codec::QuantizeBlock` encounters an all-zero input block, the early-out path zeroes `kBlockSize` (32) bytes instead of the encoded block size `sizeof(MxFp4Stream)` (17 bytes: 1 scale byte plus 16 nibble bytes). Writing 15 bytes past the end of the block corrupts the adjacent block when blocks are encoded out of order (such as across parallel batch boundaries in `Compress` or via `packed_ofs` sub-ranges) and writes past `PackedEnd` on the final block of a tensor.
#1045
build #8211build.yml10 days ago
5Destinations
7Actions
Insights