A technique for decoding an instruction in a variable-length instruction set. In one embodiment, an instruction encoding is described, in which legacy, present, and future instruction set extensions are supported, and increased functionality is provided, without expanding the code size and, in some cases, reducing the code size.

Systems, apparatuses, and methods related to a computer system having a processor and a main memory storing scrambled data are described. The processor may have a cache, a register, an execution unit, and an unscrambler. The processor can load the scrambled data into the cache; and the unscrambler may convert the scrambled data into unscrambled data just in time for the register or the execution unit during instruction execution. The unscrambled data can be an instruction, an address, or an operand of an instruction. Unscrambling can be performed just before loading the data item in a scrambled form from the cache into the register in an unscrambled form, or after the data item leaves the register in the scrambled form as input to the execution unit in the unscrambled form. The unscrambled data and the scrambled data may have the same set of bits arranged in different orders.

Systems, methods, and apparatuses relating to circuitry to implement an instruction to create and/or use data that is restricted in how it can be used are described. In one embodiment, a hardware processor comprises a decoder of a core to decode a single instruction into a decoded single instruction, the single instruction comprising a first input operand of a handle including a ciphertext of an encryption key (e.g., cryptographic key), an authentication tag, and additional authentication data, and a second input operand of data encrypted with the encryption key, and an execution unit of the core to execute the decoded single instruction to: perform a first check of the authentication tag against the ciphertext and the additional authentication data for any modification to the ciphertext or the additional authentication data, perform a second check of a current request of the core against one or more restrictions specified by the additional authentication data of the handle, decrypt the ciphertext to generate the encryption key only when the first check indicates no modification to the ciphertext or the additional authentication data, and the second check indicates the one or more restrictions are not violated, decrypt the data encrypted with the encryption key to generate unencrypted data, and provide the unencrypted data as a resultant of the single instruction.

An apparatus to facilitate partitioning of local memory is disclosed. The apparatus includes a plurality of execution units to execute a plurality of execution threads, a memory coupled to share access between the plurality of execution units and partitioning hardware to partition the memory to be used as a cache and as shared local memory (SLM), wherein the partitioning hardware partitions the memory based on a quantity of the plurality of execution threads executing on the execution units that are active.

A processor includes a register to store an encoded pointer to a memory location in memory and the encoded pointer is to include an encrypted portion. The processor further includes circuitry to determine a first data encryption factor based on a first data access instruction, decode the encoded pointer to obtain a memory address of the memory location, use the memory address to access an encrypted first data element, and decrypt the encrypted first data element using a cryptographic algorithm with first inputs to generate a decrypted first data element. The first inputs include the first data encryption factor based on the first data access instruction and a second data encryption factor from the encoded pointer.

An apparatus to facilitate matrix multiplication operations. The apparatus comprises multiplication hardware to operate in a dot product mode, wherein a multiplication stage included in the multiplication hardware is configured as a dot product of a number of bit vectors (N) to perform N×N multiplication operations on a plurality of multiplicands and perform addition operations on results of the N×N multiplication operations.

Techniques and mechanisms for processor circuitry to execute a load and expand instruction of an instruction set to generate decompressed matrix data. In an embodiment, the instruction comprises a source operand which indicates a location from which compressed matrix data, and corresponding metadata, are to be accessed. A destination operand of the instruction indicates a location which is to receive decompressed metadata, which is generated, during execution of the instruction, based on the compressed matrix data and the corresponding metadata. The metadata comprises compression mask information which identifies which elements of the matrix have been masked from the compressed matrix data. In another embodiment, the instruction further comprises a count operand which identifies a total number of the unmasked matrix elements which are represented in the compressed matrix data.

A code decompression engine reads compressed code from a memory containing a series of code parts and a dictionary part. The code parts each have a bit indicating compressed or uncompressed. When the code part is compressed, it has a value indicating the number of segments, followed by the segments, followed by an index into the dictionary part. The decompressed instruction is the dictionary value specified by the index, which is modified by the segments. Each segment describes the modification to the dictionary part specified by the index by a mask type, a mask offset, and a mask.

Aspects of the invention include receiving a first dictionary that includes key/value pairs for a first code listing. Key instances of the first dictionary include instruction addresses of the first code listing, and value instances of the first dictionary include hashes of data that uniquely identify instructions at the instruction addresses in the first code listing. A second dictionary that includes key/value pairs for a second code listing is received. Key instances of the second dictionary include hashes of data that uniquely identify instructions at instruction addresses in the second code listing, and value instances of the second dictionary include instruction addresses of the second code listing. A match between a value instance of the first dictionary and a key instance of the second dictionary is identified. Responsive to identifying the match, an instruction address is translated between the first code listing and the second code listing.

A memory circuit included in a computer system includes a memory array that stores multiple program instructions included in compressed program code. In response to receiving a fetch instruction from a processor circuit, the memory circuit may retrieve a particular instruction from the memory array. The memory circuit may, in response to a determination that the particular instruction is a particular type of instruction, retrieve additional program instructions from the memory array using an address included in the particular instruction, and send the particular program instruction and the additional program instructions to the processor circuit.