Patent Grant
11416248
The embodiments of the invention relate generally to the field of computer processors. More particularly, the embodiments relate to an apparatus and method for efficient floating point compression.
Data compression is becoming more important as emerging applications are more data-intensive and the compute throughput of processors and accelerators increases. Unfortunately, conventional data compression solutions do not provide a reasonable compression ratio for floating-point (FP) data used in some key emerging applications such as Deep Neural Networks (DNNs) and real time ray tracing.
A better understanding of the present invention can be obtained from the following detailed description in conjunction with the following drawings, in which:
In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the invention described below. It will be apparent, however, to one skilled in the art that the embodiments of the invention may be practiced without some of these specific details. In other instances, well-known structures and devices are shown in block diagram form to avoid obscuring the underlying principles of the embodiments of the invention.
An instruction set includes one or more instruction formats. A given instruction format defines various fields (number of bits, location of bits) to specify, among other things, the operation to be performed (opcode) and the operand(s) on which that operation is to be performed. Some instruction formats are further broken down though the definition of instruction templates (or subformats). For example, the instruction templates of a given instruction format may be defined to have different subsets of the instruction format's fields (the included fields are typically in the same order, but at least some have different bit positions because there are less fields included) and/or defined to have a given field interpreted differently. Thus, each instruction of an ISA is expressed using a given instruction format (and, if defined, in a given one of the instruction templates of that instruction format) and includes fields for specifying the operation and the operands. For example, an exemplary ADD instruction has a specific opcode and an instruction format that includes an opcode field to specify that opcode and operand fields to select operands (source1/destination and source2); and an occurrence of this ADD instruction in an instruction stream will have specific contents in the operand fields that select specific operands.
Embodiments of the instruction(s) described herein may be embodied in different formats. Additionally, exemplary systems, architectures, and pipelines are detailed below. Embodiments of the instruction(s) may be executed on such systems, architectures, and pipelines, but are not limited to those detailed.
Generic Vector Friendly Instruction Format
A vector friendly instruction format is an instruction format that is suited for vector instructions (e.g., there are certain fields specific to vector operations). While embodiments are described in which both vector and scalar operations are supported through the vector friendly instruction format, alternative embodiments use only vector operations the vector friendly instruction format.
While embodiments of the invention will be described in which the vector friendly instruction format supports the following: a 64 byte vector operand length (or size) with 32 bit (4 byte) or 64 bit (8 byte) data element widths (or sizes) (and thus, a 64 byte vector consists of either 16 doubleword-size elements or alternatively, 8 quadword-size elements); a 64 byte vector operand length (or size) with 16 bit (2 byte) or 8 bit (1 byte) data element widths (or sizes); a 32 byte vector operand length (or size) with 32 bit (4 byte), 64 bit (8 byte), 16 bit (2 byte), or 8 bit (1 byte) data element widths (or sizes); and a 16 byte vector operand length (or size) with 32 bit (4 byte), 64 bit (8 byte), 16 bit (2 byte), or 8 bit (1 byte) data element widths (or sizes); alternative embodiments may support more, less and/or different vector operand sizes (e.g., 256 byte vector operands) with more, less, or different data element widths (e.g., 128 bit (16 byte) data element widths).
The class A instruction templates in
The generic vector friendly instruction format 100 includes the following fields listed below in the order illustrated in
Format field 140—a specific value (an instruction format identifier value) in this field uniquely identifies the vector friendly instruction format, and thus occurrences of instructions in the vector friendly instruction format in instruction streams. As such, this field is optional in the sense that it is not needed for an instruction set that has only the generic vector friendly instruction format.
Base operation field 142—its content distinguishes different base operations.
Register index field 144—its content, directly or through address generation, specifies the locations of the source and destination operands, be they in registers or in memory. These include a sufficient number of bits to select N registers from a P×Q (e.g. 32×512, 16×128, 32×1024, 64×1024) register file. While in one embodiment N may be up to three sources and one destination register, alternative embodiments may support more or less sources and destination registers (e.g., may support up to two sources where one of these sources also acts as the destination, may support up to three sources where one of these sources also acts as the destination, may support up to two sources and one destination).
Modifier field 146—its content distinguishes occurrences of instructions in the generic vector instruction format that specify memory access from those that do not; that is, between no memory access 105 instruction templates and memory access 120 instruction templates. Memory access operations read and/or write to the memory hierarchy (in some cases specifying the source and/or destination addresses using values in registers), while non-memory access operations do not (e.g., the source and destinations are registers). While in one embodiment this field also selects between three different ways to perform memory address calculations, alternative embodiments may support more, less, or different ways to perform memory address calculations.
Augmentation operation field 150—its content distinguishes which one of a variety of different operations to be performed in addition to the base operation. This field is context specific. In one embodiment of the invention, this field is divided into a class field 168, an alpha field 152, and a beta field 154. The augmentation operation field 150 allows common groups of operations to be performed in a single instruction rather than 2, 3, or 4 instructions.
Scale field 160—its content allows for the scaling of the index field's content for memory address generation (e.g., for address generation that uses 2scale*index+base).
Displacement Field 162A—its content is used as part of memory address generation (e.g., for address generation that uses 2scale*index+base+displacement).
Displacement Factor Field 162B (note that the juxtaposition of displacement field 162A directly over displacement factor field 162B indicates one or the other is used)—its content is used as part of address generation; it specifies a displacement factor that is to be scaled by the size of a memory access (N)—where N is the number of bytes in the memory access (e.g., for address generation that uses 2scale*index+base+scaled displacement). Redundant low-order bits are ignored and hence, the displacement factor field's content is multiplied by the memory operands total size (N) in order to generate the final displacement to be used in calculating an effective address. The value of N is determined by the processor hardware at runtime based on the full opcode field 174 (described later herein) and the data manipulation field 154C. The displacement field 162A and the displacement factor field 162B are optional in the sense that they are not used for the no memory access 105 instruction templates and/or different embodiments may implement only one or none of the two.
Data element width field 164—its content distinguishes which one of a number of data element widths is to be used (in some embodiments for all instructions; in other embodiments for only some of the instructions). This field is optional in the sense that it is not needed if only one data element width is supported and/or data element widths are supported using some aspect of the opcodes.
Write mask field 170—its content controls, on a per data element position basis, whether that data element position in the destination vector operand reflects the result of the base operation and augmentation operation. Class A instruction templates support merging-writemasking, while class B instruction templates support both merging- and zeroing-writemasking. When merging, vector masks allow any set of elements in the destination to be protected from updates during the execution of any operation (specified by the base operation and the augmentation operation); in other one embodiment, preserving the old value of each element of the destination where the corresponding mask bit has a 0. In contrast, when zeroing vector masks allow any set of elements in the destination to be zeroed during the execution of any operation (specified by the base operation and the augmentation operation); in one embodiment, an element of the destination is set to 0 when the corresponding mask bit has a 0 value. A subset of this functionality is the ability to control the vector length of the operation being performed (that is, the span of elements being modified, from the first to the last one); however, it is not necessary that the elements that are modified be consecutive. Thus, the write mask field 170 allows for partial vector operations, including loads, stores, arithmetic, logical, etc. While embodiments of the invention are described in which the write mask field's 170 content selects one of a number of write mask registers that contains the write mask to be used (and thus the write mask field's 170 content indirectly identifies that masking to be performed), alternative embodiments instead or additional allow the mask write field's 170 content to directly specify the masking to be performed.
Immediate field 172—its content allows for the specification of an immediate. This field is optional in the sense that is it not present in an implementation of the generic vector friendly format that does not support immediate and it is not present in instructions that do not use an immediate.
Class field 168—its content distinguishes between different classes of instructions. With reference to
Instruction Templates of Class A
In the case of the non-memory access 105 instruction templates of class A, the alpha field 152 is interpreted as an RS field 152A, whose content distinguishes which one of the different augmentation operation types are to be performed (e.g., round 152A.1 and data transform 152A.2 are respectively specified for the no memory access, round type operation 110 and the no memory access, data transform type operation 115 instruction templates), while the beta field 154 distinguishes which of the operations of the specified type is to be performed. In the no memory access 105 instruction templates, the scale field 160, the displacement field 162A, and the displacement scale filed 162B are not present.
No-Memory Access Instruction Templates—Full Round Control Type Operation
In the no memory access full round control type operation 110 instruction template, the beta field 154 is interpreted as a round control field 154A, whose content(s) provide static rounding. While in the described embodiments of the invention the round control field 154A includes a suppress all floating point exceptions (SAE) field 156 and a round operation control field 158, alternative embodiments may support may encode both these concepts into the same field or only have one or the other of these concepts/fields (e.g., may have only the round operation control field 158).
SAE field 156—its content distinguishes whether or not to disable the exception event reporting; when the SAE field's 156 content indicates suppression is enabled, a given instruction does not report any kind of floating-point exception flag and does not raise any floating point exception handler.
Round operation control field 158—its content distinguishes which one of a group of rounding operations to perform (e.g., Round-up, Round-down, Round-towards-zero and Round-to-nearest). Thus, the round operation control field 158 allows for the changing of the rounding mode on a per instruction basis. In one embodiment of the invention where a processor includes a control register for specifying rounding modes, the round operation control field's 150 content overrides that register value.
No Memory Access Instruction Templates—Data Transform Type Operation
In the no memory access data transform type operation 115 instruction template, the beta field 154 is interpreted as a data transform field 1546, whose content distinguishes which one of a number of data transforms is to be performed (e.g., no data transform, swizzle, broadcast).
In the case of a memory access 120 instruction template of class A, the alpha field 152 is interpreted as an eviction hint field 1526, whose content distinguishes which one of the eviction hints is to be used (in
Vector memory instructions perform vector loads from and vector stores to memory, with conversion support. As with regular vector instructions, vector memory instructions transfer data from/to memory in a data element-wise fashion, with the elements that are actually transferred is dictated by the contents of the vector mask that is selected as the write mask.
Memory Access Instruction Templates—Temporal
Temporal data is data likely to be reused soon enough to benefit from caching. This is, however, a hint, and different processors may implement it in different ways, including ignoring the hint entirely.
Memory Access Instruction Templates—Non-Temporal
Non-temporal data is data unlikely to be reused soon enough to benefit from caching in the 1st-level cache and should be given priority for eviction. This is, however, a hint, and different processors may implement it in different ways, including ignoring the hint entirely.
Instruction Templates of Class B
In the case of the instruction templates of class B, the alpha field 152 is interpreted as a write mask control (Z) field 152C, whose content distinguishes whether the write masking controlled by the write mask field 170 should be a merging or a zeroing.
In the case of the non-memory access 105 instruction templates of class B, part of the beta field 154 is interpreted as an RL field 157A, whose content distinguishes which one of the different augmentation operation types are to be performed (e.g., round 157A.1 and vector length (VSIZE) 157A.2 are respectively specified for the no memory access, write mask control, partial round control type operation 112 instruction template and the no memory access, write mask control, VSIZE type operation 117 instruction template), while the rest of the beta field 154 distinguishes which of the operations of the specified type is to be performed. In the no memory access 105 instruction templates, the scale field 160, the displacement field 162A, and the displacement scale filed 162B are not present.
In the no memory access, write mask control, partial round control type operation 110 instruction template, the rest of the beta field 154 is interpreted as a round operation field 159A and exception event reporting is disabled (a given instruction does not report any kind of floating-point exception flag and does not raise any floating point exception handler).
Round operation control field 159A—just as round operation control field 158, its content distinguishes which one of a group of rounding operations to perform (e.g., Round-up, Round-down, Round-towards-zero and Round-to-nearest). Thus, the round operation control field 159A allows for the changing of the rounding mode on a per instruction basis. In one embodiment of the invention where a processor includes a control register for specifying rounding modes, the round operation control field's 150 content overrides that register value.
In the no memory access, write mask control, VSIZE type operation 117 instruction template, the rest of the beta field 154 is interpreted as a vector length field 159B, whose content distinguishes which one of a number of data vector lengths is to be performed on (e.g., 128, 256, or 512 byte).
In the case of a memory access 120 instruction template of class B, part of the beta field 154 is interpreted as a broadcast field 157B, whose content distinguishes whether or not the broadcast type data manipulation operation is to be performed, while the rest of the beta field 154 is interpreted the vector length field 159B. The memory access 120 instruction templates include the scale field 160, and optionally the displacement field 162A or the displacement scale field 162B.
With regard to the generic vector friendly instruction format 100, a full opcode field 174 is shown including the format field 140, the base operation field 142, and the data element width field 164. While one embodiment is shown where the full opcode field 174 includes all of these fields, the full opcode field 174 includes less than all of these fields in embodiments that do not support all of them. The full opcode field 174 provides the operation code (opcode).
The augmentation operation field 150, the data element width field 164, and the write mask field 170 allow these features to be specified on a per instruction basis in the generic vector friendly instruction format.
The combination of write mask field and data element width field create typed instructions in that they allow the mask to be applied based on different data element widths.
The various instruction templates found within class A and class B are beneficial in different situations. In some embodiments of the invention, different processors or different cores within a processor may support only class A, only class B, or both classes. For instance, a high performance general purpose out-of-order core intended for general-purpose computing may support only class B, a core intended primarily for graphics and/or scientific (throughput) computing may support only class A, and a core intended for both may support both (of course, a core that has some mix of templates and instructions from both classes but not all templates and instructions from both classes is within the purview of the invention). Also, a single processor may include multiple cores, all of which support the same class or in which different cores support different class. For instance, in a processor with separate graphics and general purpose cores, one of the graphics cores intended primarily for graphics and/or scientific computing may support only class A, while one or more of the general purpose cores may be high performance general purpose cores with out of order execution and register renaming intended for general-purpose computing that support only class B. Another processor that does not have a separate graphics core, may include one more general purpose in-order or out-of-order cores that support both class A and class B. Of course, features from one class may also be implement in the other class in different embodiments of the invention. Programs written in a high level language would be put (e.g., just in time compiled or statically compiled) into an variety of different executable forms, including: 1) a form having only instructions of the class(es) supported by the target processor for execution; or 2) a form having alternative routines written using different combinations of the instructions of all classes and having control flow code that selects the routines to execute based on the instructions supported by the processor which is currently executing the code.
VEX Instruction Format
VEX encoding allows instructions to have more than two operands, and allows SIMD vector registers to be longer than 28 bits. The use of a VEX prefix provides for three-operand (or more) syntax. For example, previous two-operand instructions performed operations such as A=A+B, which overwrites a source operand. The use of a VEX prefix enables operands to perform nondestructive operations such as A=B+C.
VEX Prefix (Bytes 0-2) 202 is encoded in a three-byte form. The first byte is the Format Field 290 (VEX Byte 0, bits [7:0]), which contains an explicit C4 byte value (the unique value used for distinguishing the C4 instruction format). The second-third bytes (VEX Bytes 1-2) include a number of bit fields providing specific capability. Specifically, REX field 205 (VEX Byte 1, bits [7-5]) consists of a VEX.R bit field (VEX Byte 1, bit [7]-R), VEX.X bit field (VEX byte 1, bit [6]-X), and VEX.B bit field (VEX byte 1, bit[5]-B). Other fields of the instructions encode the lower three bits of the register indexes as is known in the art (rrr, xxx, and bbb), so that Rrrr, Xxxx, and Bbbb may be formed by adding VEX.R, VEX.X, and VEX.B. Opcode map field 215 (VEX byte 1, bits [4:0]-mmmmm) includes content to encode an implied leading opcode byte. W Field 264 (VEX byte 2, bit [7]-W)—is represented by the notation VEX.W, and provides different functions depending on the instruction. The role of VEX.vvvv 220 (VEX Byte 2, bits [6:3]-vvvv) may include the following: 1) VEX.vvvv encodes the first source register operand, specified in inverted (1s complement) form and is valid for instructions with 2 or more source operands; 2) VEX.vvvv encodes the destination register operand, specified in 1s complement form for certain vector shifts; or 3) VEX.vvvv does not encode any operand, the field is reserved and should contain 1111b. If VEX.L 268 Size field (VEX byte 2, bit [2]-L)=0, it indicates 28 bit vector; if VEX.L=1, it indicates 256 bit vector. Prefix encoding field 225 (VEX byte 2, bits [1:0]-pp) provides additional bits for the base operation field 241.
Real Opcode Field 230 (Byte 3) is also known as the opcode byte. Part of the opcode is specified in this field.
MOD R/M Field 240 (Byte 4) includes MOD field 242 (bits [7-6]), Reg field 244 (bits [5-3]), and R/M field 246 (bits [2-0]). The role of Reg field 244 may include the following: encoding either the destination register operand or a source register operand (the rrr of Rrrr), or be treated as an opcode extension and not used to encode any instruction operand. The role of R/M field 246 may include the following: encoding the instruction operand that references a memory address, or encoding either the destination register operand or a source register operand.
Scale, Index, Base (SIB)—The content of Scale field 250 (Byte 5) includes SS252 (bits [7-6]), which is used for memory address generation. The contents of SIB.xxx 254 (bits [5-3]) and SIB.bbb 256 (bits [2-0]) have been previously referred to with regard to the register indexes Xxxx and Bbbb.
The Displacement Field 262 and the immediate field (IMM8) 272 contain data.
Exemplary Register Architecture
General-purpose registers 325—in the embodiment illustrated, there are sixteen 64-bit general-purpose registers that are used along with the existing x86 addressing modes to address memory operands. These registers are referenced by the names RAX, RBX, RCX, RDX, RBP, RSI, RDI, RSP, and R8 through R15.
Scalar floating point stack register file (x87 stack) 345, on which is aliased the MMX packed integer flat register file 350—in the embodiment illustrated, the x87 stack is an eight-element stack used to perform scalar floating-point operations on 32/64/80-bit floating point data using the x87 instruction set extension; while the MMX registers are used to perform operations on 64-bit packed integer data, as well as to hold operands for some operations performed between the MMX and XMM registers.
Alternative embodiments of the invention may use wider or narrower registers. Additionally, alternative embodiments of the invention may use more, less, or different register files and registers.
Exemplary Core Architectures, Processors, and Computer Architectures
Processor cores may be implemented in different ways, for different purposes, and in different processors. For instance, implementations of such cores may include: 1) a general purpose in-order core intended for general-purpose computing; 2) a high performance general purpose out-of-order core intended for general-purpose computing; 3) a special purpose core intended primarily for graphics and/or scientific (throughput) computing. Implementations of different processors may include: 1) a CPU including one or more general purpose in-order cores intended for general-purpose computing and/or one or more general purpose out-of-order cores intended for general-purpose computing; and 2) a coprocessor including one or more special purpose cores intended primarily for graphics and/or scientific (throughput). Such different processors lead to different computer system architectures, which may include: 1) the coprocessor on a separate chip from the CPU; 2) the coprocessor on a separate die in the same package as a CPU; 3) the coprocessor on the same die as a CPU (in which case, such a coprocessor is sometimes referred to as special purpose logic, such as integrated graphics and/or scientific (throughput) logic, or as special purpose cores); and 4) a system on a chip that may include on the same die the described CPU (sometimes referred to as the application core(s) or application processor(s)), the above described coprocessor, and additional functionality. Exemplary core architectures are described next, followed by descriptions of exemplary processors and computer architectures. Detailed herein are circuits (units) that comprise exemplary cores, processors, etc.
Exemplary Core Architectures
In
The front end unit 430 includes a branch prediction unit 432 coupled to an instruction cache unit 434, which is coupled to an instruction translation lookaside buffer (TLB) 436, which is coupled to an instruction fetch unit 438, which is coupled to a decode unit 440. The decode unit 440 (or decoder) may decode instructions, and generate as an output one or more micro-operations, micro-code entry points, microinstructions, other instructions, or other control signals, which are decoded from, or which otherwise reflect, or are derived from, the original instructions. The decode unit 440 may be implemented using various different mechanisms. Examples of suitable mechanisms include, but are not limited to, look-up tables, hardware implementations, programmable logic arrays (PLAs), microcode read only memories (ROMs), etc. In one embodiment, the core 490 includes a microcode ROM or other medium that stores microcode for certain macroinstructions (e.g., in decode unit 440 or otherwise within the front end unit 430). The decode unit 440 is coupled to a rename/allocator unit 452 in the execution engine unit 450.
The execution engine unit 450 includes the rename/allocator unit 452 coupled to a retirement unit 454 and a set of one or more scheduler unit(s) 456. The scheduler unit(s) 456 represents any number of different schedulers, including reservations stations, central instruction window, etc. The scheduler unit(s) 456 is coupled to the physical register file(s) unit(s) 458. Each of the physical register file(s) units 458 represents one or more physical register files, different ones of which store one or more different data types, such as scalar integer, scalar floating point, packed integer, packed floating point, vector integer, vector floating point, status (e.g., an instruction pointer that is the address of the next instruction to be executed), etc. In one embodiment, the physical register file(s) unit 458 comprises a vector registers unit and a scalar registers unit. These register units may provide architectural vector registers, vector mask registers, and general purpose registers. The physical register file(s) unit(s) 458 is overlapped by the retirement unit 454 to illustrate various ways in which register renaming and out-of-order execution may be implemented (e.g., using a reorder buffer(s) and a retirement register file(s); using a future file(s), a history buffer(s), and a retirement register file(s); using a register maps and a pool of registers; etc.). The retirement unit 454 and the physical register file(s) unit(s) 458 are coupled to the execution cluster(s) 460. The execution cluster(s) 460 includes a set of one or more execution units 462 and a set of one or more memory access units 464. The execution units 462 may perform various operations (e.g., shifts, addition, subtraction, multiplication) and on various types of data (e.g., scalar floating point, packed integer, packed floating point, vector integer, vector floating point). While some embodiments may include a number of execution units dedicated to specific functions or sets of functions, other embodiments may include only one execution unit or multiple execution units that all perform all functions. The scheduler unit(s) 456, physical register file(s) unit(s) 458, and execution cluster(s) 460 are shown as being possibly plural because certain embodiments create separate pipelines for certain types of data/operations (e.g., a scalar integer pipeline, a scalar floating point/packed integer/packed floating point/vector integer/vector floating point pipeline, and/or a memory access pipeline that each have their own scheduler unit, physical register file(s) unit, and/or execution cluster—and in the case of a separate memory access pipeline, certain embodiments are implemented in which only the execution cluster of this pipeline has the memory access unit(s) 464). It should also be understood that where separate pipelines are used, one or more of these pipelines may be out-of-order issue/execution and the rest in-order.
The set of memory access units 464 is coupled to the memory unit 470, which includes a data TLB unit 472 coupled to a data cache unit 474 coupled to a level 2 (L2) cache unit 476. In one exemplary embodiment, the memory access units 464 may include a load unit, a store address unit, and a store data unit, each of which is coupled to the data TLB unit 472 in the memory unit 470. The instruction cache unit 434 is further coupled to a level 2 (L2) cache unit 476 in the memory unit 470. The L2 cache unit 476 is coupled to one or more other levels of cache and eventually to a main memory.
By way of example, the exemplary register renaming, out-of-order issue/execution core architecture may implement the pipeline 400 as follows: 1) the instruction fetch 438 performs the fetch and length decoding stages 402 and 404; 2) the decode unit 440 performs the decode stage 406; 3) the rename/allocator unit 452 performs the allocation stage 408 and renaming stage 410; 4) the scheduler unit(s) 456 performs the schedule stage 412; 5) the physical register file(s) unit(s) 458 and the memory unit 470 perform the register read/memory read stage 414; the execution cluster 460 perform the execute stage 416; 6) the memory unit 470 and the physical register file(s) unit(s) 458 perform the write back/memory write stage 418; 7) various units may be involved in the exception handling stage 422; and 8) the retirement unit 454 and the physical register file(s) unit(s) 458 perform the commit stage 424.
The core 490 may support one or more instructions sets (e.g., the x86 instruction set (with some extensions that have been added with newer versions); the MIPS instruction set of MIPS Technologies of Sunnyvale, Calif.; the ARM instruction set (with optional additional extensions such as NEON) of ARM Holdings of Sunnyvale, Calif.), including the instruction(s) described herein. In one embodiment, the core 490 includes logic to support a packed data instruction set extension (e.g., AVX1, AVX2), thereby allowing the operations used by many multimedia applications to be performed using packed data.
It should be understood that the core may support multithreading (executing two or more parallel sets of operations or threads), and may do so in a variety of ways including time sliced multithreading, simultaneous multithreading (where a single physical core provides a logical core for each of the threads that physical core is simultaneously multithreading), or a combination thereof (e.g., time sliced fetching and decoding and simultaneous multithreading thereafter such as in the Intel® Hyperthreading technology).
While register renaming is described in the context of out-of-order execution, it should be understood that register renaming may be used in an in-order architecture. While the illustrated embodiment of the processor also includes separate instruction and data cache units 434/474 and a shared L2 cache unit 476, alternative embodiments may have a single internal cache for both instructions and data, such as, for example, a Level 1 (L1) internal cache, or multiple levels of internal cache. In some embodiments, the system may include a combination of an internal cache and an external cache that is external to the core and/or the processor. Alternatively, all of the cache may be external to the core and/or the processor.
Specific Exemplary In-Order Core Architecture
The local subset of the L2 cache 504 is part of a global L2 cache that is divided into separate local subsets, one per processor core. Each processor core has a direct access path to its own local subset of the L2 cache 504. Data read by a processor core is stored in its L2 cache subset 504 and can be accessed quickly, in parallel with other processor cores accessing their own local L2 cache subsets. Data written by a processor core is stored in its own L2 cache subset 504 and is flushed from other subsets, if necessary. The ring network ensures coherency for shared data. The ring network is bi-directional to allow agents such as processor cores, L2 caches and other logic blocks to communicate with each other within the chip. Each ring data-path is 1024-bits wide per direction in some embodiments.
Processor with Integrated Memory Controller and Graphics
Thus, different implementations of the processor 600 may include: 1) a CPU with the special purpose logic 608 being integrated graphics and/or scientific (throughput) logic (which may include one or more cores), and the cores 602A-N being one or more general purpose cores (e.g., general purpose in-order cores, general purpose out-of-order cores, a combination of the two); 2) a coprocessor with the cores 602A-N being a large number of special purpose cores intended primarily for graphics and/or scientific (throughput); and 3) a coprocessor with the cores 602A-N being a large number of general purpose in-order cores. Thus, the processor 600 may be a general-purpose processor, coprocessor or special-purpose processor, such as, for example, a network or communication processor, compression engine, graphics processor, GPGPU (general purpose graphics processing unit), a high-throughput many integrated core (MIC) coprocessor (including 30 or more cores), embedded processor, or the like. The processor may be implemented on one or more chips. The processor 600 may be a part of and/or may be implemented on one or more substrates using any of a number of process technologies, such as, for example, BiCMOS, CMOS, or NMOS.
The memory hierarchy includes one or more levels of cache within the cores 604A-N, a set or one or more shared cache units 606, and external memory (not shown) coupled to the set of integrated memory controller units 614. The set of shared cache units 606 may include one or more mid-level caches, such as level 2 (L2), level 3 (L3), level 4 (L4), or other levels of cache, a last level cache (LLC), and/or combinations thereof. While in one embodiment a ring based interconnect unit 612 interconnects the integrated graphics logic 608, the set of shared cache units 606, and the system agent unit 610/integrated memory controller unit(s) 614, alternative embodiments may use any number of well-known techniques for interconnecting such units. In one embodiment, coherency is maintained between one or more cache units 606 and cores 602-A-N.
In some embodiments, one or more of the cores 602A-N are capable of multi-threading. The system agent 610 includes those components coordinating and operating cores 602A-N. The system agent unit 610 may include for example a power control unit (PCU) and a display unit. The PCU may be or include logic and components needed for regulating the power state of the cores 602A-N and the integrated graphics logic 608. The display unit is for driving one or more externally connected displays.
The cores 602A-N may be homogenous or heterogeneous in terms of architecture instruction set; that is, two or more of the cores 602A-N may be capable of execution the same instruction set, while others may be capable of executing only a subset of that instruction set or a different instruction set.
Exemplary Computer Architectures
Referring now to
The optional nature of additional processors 715 is denoted in
The memory 740 may be, for example, dynamic random access memory (DRAM), phase change memory (PCM), or a combination of the two. For at least one embodiment, the controller hub 720 communicates with the processor(s) 710, 715 via a multi-drop bus, such as a frontside bus (FSB), point-to-point interface, or similar connection 795.
In one embodiment, the coprocessor 745 is a special-purpose processor, such as, for example, a high-throughput MIC processor, a network or communication processor, compression engine, graphics processor, GPGPU, embedded processor, or the like. In one embodiment, controller hub 720 may include an integrated graphics accelerator.
There can be a variety of differences between the physical resources 710, 7155 in terms of a spectrum of metrics of merit including architectural, microarchitectural, thermal, power consumption characteristics, and the like.
In one embodiment, the processor 710 executes instructions that control data processing operations of a general type. Embedded within the instructions may be coprocessor instructions. The processor 710 recognizes these coprocessor instructions as being of a type that should be executed by the attached coprocessor 745. Accordingly, the processor 710 issues these coprocessor instructions (or control signals representing coprocessor instructions) on a coprocessor bus or other interconnect, to coprocessor 745. Coprocessor(s) 745 accept and execute the received coprocessor instructions.
Referring now to
Processors 870 and 880 are shown including integrated memory controller (IMC) units 872 and 882, respectively. Processor 870 also includes as part of its bus controller units point-to-point (P-P) interfaces 876 and 878; similarly, second processor 880 includes P-P interfaces 886 and 888. Processors 870, 880 may exchange information via a point-to-point (P-P) interface 850 using P-P interface circuits 878, 888. As shown in
Processors 870, 880 may each exchange information with a chipset 890 via individual P-P interfaces 852, 854 using point to point interface circuits 876, 894, 886, 898. Chipset 890 may optionally exchange information with the coprocessor 838 via a high-performance interface 892. In one embodiment, the coprocessor 838 is a special-purpose processor, such as, for example, a high-throughput MIC processor, a network or communication processor, compression engine, graphics processor, GPGPU, embedded processor, or the like.
A shared cache (not shown) may be included in either processor or outside of both processors, yet connected with the processors via P-P interconnect, such that either or both processors' local cache information may be stored in the shared cache if a processor is placed into a low power mode.
Chipset 890 may be coupled to a first bus 816 via an interface 896. In one embodiment, first bus 816 may be a Peripheral Component Interconnect (PCI) bus, or a bus such as a PCI Express bus or another I/O interconnect bus, although the scope of the present invention is not so limited.
As shown in
Referring now to
Referring now to
Embodiments of the mechanisms disclosed herein may be implemented in hardware, software, firmware, or a combination of such implementation approaches. Embodiments of the invention may be implemented as computer programs or program code executing on programmable systems comprising at least one processor, a storage system (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device.
Program code, such as code 830 illustrated in
The program code may be implemented in a high level procedural or object oriented programming language to communicate with a processing system. The program code may also be implemented in assembly or machine language, if desired. In fact, the mechanisms described herein are not limited in scope to any particular programming language. In any case, the language may be a compiled or interpreted language.
One or more aspects of at least one embodiment may be implemented by representative instructions stored on a machine-readable medium which represents various logic within the processor, which when read by a machine causes the machine to fabricate logic to perform the techniques described herein. Such representations, known as “IP cores” may be stored on a tangible, machine readable medium and supplied to various customers or manufacturing facilities to load into the fabrication machines that actually make the logic or processor.
Such machine-readable storage media may include, without limitation, non-transitory, tangible arrangements of articles manufactured or formed by a machine or device, including storage media such as hard disks, any other type of disk including floppy disks, optical disks, compact disk read-only memories (CD-ROMs), compact disk rewritable's (CD-RWs), and magneto-optical disks, semiconductor devices such as read-only memories (ROMs), random access memories (RAMs) such as dynamic random access memories (DRAMs), static random access memories (SRAMs), erasable programmable read-only memories (EPROMs), flash memories, electrically erasable programmable read-only memories (EEPROMs), phase change memory (PCM), magnetic or optical cards, or any other type of media suitable for storing electronic instructions.
Accordingly, embodiments of the invention also include non-transitory, tangible machine-readable media containing instructions or containing design data, such as Hardware Description Language (HDL), which defines structures, circuits, apparatuses, processors and/or system features described herein. Such embodiments may also be referred to as program products.
Emulation (including binary translation, code morphing, etc.)
In some cases, an instruction converter may be used to convert an instruction from a source instruction set to a target instruction set. For example, the instruction converter may translate (e.g., using static binary translation, dynamic binary translation including dynamic compilation), morph, emulate, or otherwise convert an instruction to one or more other instructions to be processed by the core. The instruction converter may be implemented in software, hardware, firmware, or a combination thereof. The instruction converter may be on processor, off processor, or part on and part off processor.
Apparatus and Method for Floating Point Compression
As mentioned, floating-point (FP) data is not adequately compressed using conventional compression techniques. Thus, the FP data used in some key emerging applications such as Deep Neural Networks (DNNs) and real time ray tracing consumes an unreasonable amount of storage.
Embodiments of the invention include a new hardware-friendly floating-point encoding scheme which provides a substantially higher compression ratio compared to current implementations. An overview of previous floating point compression techniques will be provided, followed by a description of embodiments of the invention.
One existing compression mechanism which is the basis for the fpzip library is described in Peter Lindstrom and Martin Isenburg, “Fast and Efficient Compression of Floating-Point Data,” IEEE Transactions on Visualization and Computer Graphics, 12(5):1245-1250, September-October 2006. It provides an implementation for floating-point compression that traverses data sequentially and predicts each visited data value from a subset of previously seen and encoded data (transformed into an integer representation and potentially truncated if lossy compression is desired). Residuals are partitioned into entropy codes and raw bits. This mechanism provides relatively high compression ratios but adds significant complexity and latency on the critical path of a read operation.
Another mechanism described in P. Ratanaworabhan, Jian Ke, and M. Burtscher, “Fast Lossless Compression of Scientific Floating-Point Data” Data Compression Conference, March 2006, depends on value prediction algorithms in compressing FP64 data. Each value is predicted using a value prediction mechanism and is stored in a compressed form if the prediction is accurate. However, it is limited by the predictability of nearby floating-point values and could lead to data expansion for mostly random data.
Previous floating point compression techniques simply compress raw bit data in an oblivious way without distinguishing between the exponent and mantissa components. In general, floating point (FP) data is less compressible than integer data, and previous lossless compression solutions typically achieve a low compression ratio unless there is a large portion of zeros in a block.
One embodiment of the invention encodes the exponent portion of floating-point numbers using delta encoding. For example, instead of storing a full 8-bit exponent for each FP32 value in a data block, this embodiment stores one 8-bit base exponent for a group of FP32 numbers and an N-bit delta value (e.g., 2 bits) for each of FP32 numbers in a data block.
To execute a floating-point instruction, one or more compressed floating point input operands 1202 are loaded from the cache/memory subsystem 1260 and decompressed by FP decompression circuitry/logic 1240 to generate a decompressed FP value 1204, which is then stored in a vector/FP register 1205. In the illustrated embodiment, exponent decompression circuitry/logic 1210 decompresses the exponent value 1202A in accordance with the techniques described herein to generate a decompressed exponent value 1204A. Significand decompression circuitry 1211 decompresses the compressed significand 1202B to generate a decompressed significand value 1204B. The FP circuitry/logic 1223 then reads the decompressed FP value 1204 to perform the specified floating point operation (e.g., as indicated by the opcode and possibly an immediate of the instruction). The exponent processing circuitry 1220 processes the uncompressed exponent 1204A while the significand processing circuitry 1221 processes the significand 1204B. While
A floating-point result 1206 generated by the FP processing circuitry/logic 1223 is stored in the vector/FP registers 1205 and compressed by FP compression circuitry/logic 1241 to generate a compressed FP result 1208 which may be stored to the cache/memory subsystem 1260. Exponent compression circuitry/logic 1212 compresses the exponent 1206A in accordance with the techniques described herein to produce a compressed exponent 1208A and significand compression circuitry/logic 1213 compresses the significand 1206B to produce a compressed significand 1208B. In one embodiment, control circuitry 1215 adjusts operation of the FP decompression circuitry/logic 1240 and the FP compression circuitry/logic 1241 as described herein.
One embodiment of the invention provides a significantly improved compression ratio when the FP numbers in a compression block are within a small dynamic range. One type of application for which the embodiments of the invention may be particularly useful is deep neural networks (DNNs). For example, most of the FP numbers in DNN weights and activations have similar exponent values. Additionally, emerging numeric formats for DNNs such as bfloat16 (Google/Intel) and FP8/FP9 (MS) have more exponent bits than mantissa bits, making the embodiments of the invention particularly useful. Note, however, that the underlying principles of the invention are not limited to any particular applications.
As described with respect to
While the embodiments of the invention are described below in the context of DNNs, these embodiments are generally applicable to other domain-specific applications including, but not limited to, graphics processing, ray tracing, and approximate computing.
This encoding scheme is referred to herein as EDn encoding, where the exponent is encoded with n bits. Using ED2 encoding, for example, a single 64B cache line that contains 32 bfloat16 values (1-bit sign, 8-bit exponent, 7-bit mantissa) can be compressed to 360 bits in the best case, leading to a 1.42× compression ratio. In the above case where roughly 81% of the numbers (26 elements out of 32) can be compressed with ED2 encoding, the compression ratio is 1.29×. Decompression is a simple operation that adds the base value to the delta to generate the uncompressed exponent, which is a much faster operation compared to other floating-point mechanisms and has little impact on performance. The above embodiments result in a savings of 6 bits×26 elements=156 bits, with an overhead of 1-bit×32 elements+8-bits for base=40 bits, for a net savings of 116 bits.
As illustrated in
The base value 1431 is provided to a delta encoder module 1420 which responsively encodes a delta value 1430A-D for all or a subset of the exponent values 1406A-D. As mentioned, for exponent values within a small range of one another, the delta values 1430A-D can be encoded with a significantly reduced number of bits. The delta values 1430A-D and base value 1431 are then stored together in the cache/memory subsystem 1260. In addition, in one embodiment, a compression bit is associated with each FP data element to indicate whether the FP value is compressed (e.g., if set to 1) or uncompressed (e.g., if set to 0).
In one embodiment, separate instructions are used to load and store compression FP data. For example, a Store-Compress instruction may be executed to compress FP data as described herein prior to storage in the memory subsystem. Similarly, a Load-Decompress instruction may be executed to decompress the compressed FP data prior to loading the FP data to a register file.
As illustrated in
The embodiments of the invention may be applied to any type of floating point data format including, but not limited to, FP32, FP64, bfloat16, as well as custom FP formats such as MS-FPB, which has a 2-bit mantissa and 6-bit exponent. All that is required is a simple layout change for the final compressed bit arrangement. One embodiment selects the compression granularity to be a multiple of cache lines to amortize the overhead of the additional base exponent.
In one embodiment, EDn encoding is used in combination with mantissa compression that compresses the mantissa bits and/or zero values of the FP data elements. Referring to
In one embodiment, the compression bit that indicates whether the element is compressed may be encoded in one of the mantissa bits. In the above bfloat16 case, for example, this provides an approximate 1.41× compression ratio. This lossy compression could be useful in many approximate computing and/or DNN applications. It has been observed that many DNN inference and training implementations do not see a significant accuracy drop even if one or more mantissa bits in FP32 is dropped and even in bfloat16. For example, the table in
The EDn encoding described herein can be used to improve performance and save energy. The following architectural support is provided in one embodiment:
Energy Savings in LLC: Using EDn encoding can help transfer fewer bits between different cache levels, where fewer bits can be transferred (saving energy) and decompress on the cache fill path. This requires little or no additional latency depending on the pipelined flow for cache fills and would save energy (˜23%) due to transferring fewer bits.
Performance Improvement in LLC: One embodiment takes advantage of an LLC capacity increase due to compression by allocating more logical cache lines than physical lines. In this context, if the baseline is a 16-way set-associative cache, for example, 5 logical lines can be allocated to 4 physical lines. Each set will have four additional tags, where the last tag in each group of five point to the aggregate of leftover bits from compressing the other four lines in that group. This leads to a 25% capacity increase, reducing misses by ˜11% and therefore improving application performance. Further capacity and performance improvements may be achieved by sharing the base exponent value between all data blocks in a 5-block group.
Performance Improvements in Main Memory:
EDn encoding can increase memory capacity which leads to performance improvements in cases where the memory footprint of an application exceeds the physical memory capacity. In this implementation an on-die indirection table is used to track where each compressed block starts in memory. In one particular implementation, the EDn encoding described herein is used as part of an overall main memory compression (or deduplication) architecture that uses EDn encoding in combination with other compression or deduplication algorithms.
A method in accordance with one embodiment of the invention is illustrated in
At 1701 a request is received to load-decompress or store-compress floating point data. In one embodiment, separate instructions are used to specify each operation.
If the request is a store-compress operation, determined at 1702, then at 1703 a set of exponent values are evaluated to determine a base value. As mentioned, the base value may be selected from within the range of the exponent values. At 1704 a delta value is calculated for each exponent value, indicating a distance from the base value. At 1705, compressed FP values including the base value and delta values are stored (e.g., to memory or cache).
If the request is a load-decompress request, then at 1706 the base value is read from the data and added to the delta values to generate the corresponding exponent values. At 1707 the resulting decompressed FP values are loaded to a register file so that the can be processed in response to the execution of an instruction.
In the foregoing specification, the embodiments of invention have been described with reference to specific exemplary embodiments thereof. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the appended claims. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
The following are example implementations of different embodiments of the invention.
A processor comprising: instruction fetch circuitry to fetch instructions from a memory, the instructions including floating-point instructions; execution circuitry to execute the floating-point instructions, each floating-point instruction having one or more floating-point operands, each floating-point operand comprising an exponent value and a significand value; floating-point compression circuitry to compress a plurality of the exponent values associated with a corresponding plurality of the floating-point operands, the floating-point compression circuitry comprising: base generation circuitry to evaluate the plurality of the exponent values to generate a first base value; and delta generation circuitry to determine a difference between the plurality of exponent values and the first base value and to generate a corresponding first plurality of delta values, wherein the floating-point compression circuitry is to store the first base value and the corresponding first plurality of delta values as a plurality of compressed exponent values.
The processor of example 1 wherein the floating-point compression circuitry is to compress the plurality of exponent values independently from any compression of the corresponding plurality of the significand values.
The processor of example 1 wherein the floating-point compression circuitry is to compress the plurality of significand values using a different compression technique from that used to generate the plurality of compressed exponent values.
The processor of example 1 further comprising: floating-point decompression circuitry to receive a plurality of compressed exponent values of compressed floating-point values read from the memory, the floating point decompression circuitry to add a second base value associated with the plurality of compressed exponent values to a second plurality of delta values to generate a corresponding plurality of decompressed exponent values.
The processor of example 4 wherein the floating-point decompression circuitry is to load a first floating-point value in a source register of a floating point instruction, the first floating-point value comprising a first decompressed exponent value of the plurality of decompressed exponent values and a first significand value.
The processor of example 1 wherein the corresponding plurality of the floating-point operands comprise one or more of: half-precision floating-point operands, single-precision floating-point operands, double-precision floating-point operands, bfloat16 operands, fp8 operands, and fp9 operands.
The processor of example 1 wherein the first base value comprises an 8-bit value and the delta values comprise 2-bit or 3-bit values.
The processor of example 1 wherein the floating-point compression circuitry is to set a compression bit associated with each of the plurality of compressed exponent values to indicate that the exponent values are compressed, the compression bits to be stored with the plurality of compressed exponent values.
A method comprising: fetching instructions from a memory, the instructions including floating-point instructions; executing the floating-point instructions, each floating-point instruction having one or more floating-point operands, each floating-point operand comprising an exponent value and a significand value; compressing a plurality of the exponent values associated with a corresponding plurality of the floating-point operands, wherein compressing comprises: evaluating the plurality of the exponent values to generate a first base value; and determining a difference between the plurality of exponent values and the first base value to generate a corresponding first plurality of delta values, storing the first base value and the corresponding first plurality of delta values as a plurality of compressed exponent values.
The method of example 9 wherein the plurality of exponent values are to be compressed independently from any compression of the corresponding plurality of the significand values.
The method of example 9 wherein the plurality of significand values are to be compressed using a different compression technique from that used to generate the plurality of compressed exponent values.
The method of example 9 further comprising: receiving a plurality of compressed exponent values from the memory; adding a second base value associated with the plurality of compressed exponent values to a second plurality of delta values to generate a corresponding plurality of decompressed exponent values.
The method of example 12 further comprising: loading a first floating-point value in a source register of a floating point instruction, the first floating-point value comprising a first decompressed exponent value of the plurality of decompressed exponent values and a first significand value.
The method of example 9 wherein the corresponding plurality of the floating-point operands comprise one or more of: half-precision floating-point operands, single-precision floating-point operands, double-precision floating-point operands, bfloat16 operands, fp8 operands, and fp9 operands.
The method example claim 9 wherein the first base value comprises an 8-bit value and the delta values comprise 2-bit or 3-bit values.
The method of claim 9 further comprising: setting a compression bit associated with each of the plurality of compressed exponent values to indicate that the exponent values are compressed, the compression bits to be stored with the plurality of compressed exponent values.
A machine-readable medium having program code stored thereon which, when executed by a machine, causes the machine to perform the operations of: fetching instructions from a memory, the instructions including floating-point instructions; executing the floating-point instructions, each floating-point instruction having one or more floating-point operands, each floating-point operand comprising an exponent value and a significand value; compressing a plurality of the exponent values associated with a corresponding plurality of the floating-point operands, wherein compressing comprises: evaluating the plurality of the exponent values to generate a first base value; and determining a difference between the plurality of exponent values and the first base value to generate a corresponding first plurality of delta values, storing the first base value and the corresponding first plurality of delta values as a plurality of compressed exponent values.
The machine-readable medium of example 17 wherein the plurality of exponent values are to be compressed independently from any compression of the corresponding plurality of the significand values.
The machine-readable medium of claim 17 wherein the plurality of significand values are to be compressed using a different compression technique from that used to generate the plurality of compressed exponent values.
The machine-readable medium of example 17 further comprising program code to cause the machine to perform the operations of: receiving a plurality of compressed exponent values from the memory; adding a second base value associated with the plurality of compressed exponent values to a second plurality of delta values to generate a corresponding plurality of decompressed exponent values.
The machine-readable medium of example 20 further comprising program code to cause the machine to perform the operations of: loading a first floating-point value in a source register of a floating point instruction, the first floating-point value comprising a first decompressed exponent value of the plurality of decompressed exponent values and a first significand value.
The machine-readable medium of example 17 wherein the corresponding plurality of the floating-point operands comprise one or more of: half-precision floating-point operands, single-precision floating-point operands, double-precision floating-point operands, bfloat16 operands, fp8 operands, and fp9 operands.
The machine-readable medium of example 17 wherein the first base value comprises an 8-bit value and the delta values comprise 2-bit or 3-bit values.
The machine-readable medium of example 17 further comprising program code to cause the machine to perform the operations of: setting a compression bit associated with each of the plurality of compressed exponent values to indicate that the exponent values are compressed, the compression bits to be stored with the plurality of compressed exponent values.
Embodiments of the invention may include various steps, which have been described above. The steps may be embodied in machine-executable instructions which may be used to cause a general-purpose or special-purpose processor to perform the steps. Alternatively, these steps may be performed by specific hardware components that contain hardwired logic for performing the steps, or by any combination of programmed computer components and custom hardware components.
As described herein, instructions may refer to specific configurations of hardware such as application specific integrated circuits (ASICs) configured to perform certain operations or having a predetermined functionality or software instructions stored in memory embodied in a non-transitory computer readable medium. Thus, the techniques shown in the Figures can be implemented using code and data stored and executed on one or more electronic devices (e.g., an end station, a network element, etc.). Such electronic devices store and communicate (internally and/or with other electronic devices over a network) code and data using computer machine-readable media, such as non-transitory computer machine-readable storage media (e.g., magnetic disks; optical disks; random access memory; read only memory; flash memory devices; phase-change memory) and transitory computer machine-readable communication media (e.g., electrical, optical, acoustical or other form of propagated signals—such as carrier waves, infrared signals, digital signals, etc.). In addition, such electronic devices typically include a set of one or more processors coupled to one or more other components, such as one or more storage devices (non-transitory machine-readable storage media), user input/output devices (e.g., a keyboard, a touchscreen, and/or a display), and network connections. The coupling of the set of processors and other components is typically through one or more busses and bridges (also termed as bus controllers). The storage device and signals carrying the network traffic respectively represent one or more machine-readable storage media and machine-readable communication media. Thus, the storage device of a given electronic device typically stores code and/or data for execution on the set of one or more processors of that electronic device. Of course, one or more parts of an embodiment of the invention may be implemented using different combinations of software, firmware, and/or hardware. Throughout this detailed description, for the purposes of explanation, numerous specific details were set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without some of these specific details. In certain instances, well known structures and functions were not described in elaborate detail in order to avoid obscuring the subject matter of the present invention. Accordingly, the scope and spirit of the invention should be judged in terms of the claims which follow.
| Number | Name | Date | Kind |
|---|---|---|---|
| 8301803 | Wegener | Oct 2012 | B2 |
| 8874794 | Wegener | Oct 2014 | B2 |
| 8880734 | Wegener | Nov 2014 | B2 |
| 8959129 | Wegener | Feb 2015 | B2 |
| 8959130 | Wegener | Feb 2015 | B2 |
| 9047118 | Wegener | Jun 2015 | B2 |
| 9104473 | Wegener | Aug 2015 | B2 |
| 9143792 | Kaempfer | Sep 2015 | B2 |
| 9298457 | Wegener | Mar 2016 | B2 |
| 9628107 | Dickie | Apr 2017 | B2 |
| 9660666 | Ciarlini | May 2017 | B1 |
| 9935650 | Dickie | Apr 2018 | B2 |
| 10153779 | Bordignon | Dec 2018 | B1 |
| 10200060 | Ciarlini | Feb 2019 | B1 |
| 20110099295 | Wegener | Apr 2011 | A1 |
| 20130007076 | Wegener | Jan 2013 | A1 |
| 20130007077 | Wegener | Jan 2013 | A1 |
| 20130007078 | Wegener | Jan 2013 | A1 |
| 20130054661 | Wegener | Feb 2013 | A1 |
| 20130060827 | Wegener | Mar 2013 | A1 |
| 20130138925 | Hickey | May 2013 | A1 |
| 20130262539 | Wegener | Oct 2013 | A1 |
| 20130301890 | Kaempfer | Nov 2013 | A1 |
| 20150286465 | Dickie | Oct 2015 | A1 |
| 20150288381 | Dickie | Oct 2015 | A1 |
| 20150379682 | Golas | Dec 2015 | A1 |
| 20190042243 | Tang | Feb 2019 | A1 |
| Number | Date | Country |
|---|---|---|
| H07199996 | Aug 1995 | JP |
| Entry |
|---|
| ‘Pipelining’ from Stanford.edu, archived on Nov. 26, 2017. (Year: 2017). |
| ‘MIPS Pipeline’ by Hakim Weatherspoon, CS 3410, Spring 2012, Computer Science, Cornell University. (Year: 2012). |
| ‘Fast and Efficient Compression of Floating-Point Data’ by Peter Lindstrom, IEEE Transactions on Visualization and Computer Graphics, vol. 12, No. 5, Sep./Oct. 2006. (Year: 2006). |
| ‘Redundant-Digit Floating-Point Addition Scheme Based on a Stored Rounding Value’ by Ghassem Jaberipur et al., IEEE Transactions on Computers, vol. 59, No. 5, May 2010. (Year: 2010). |
| Computing, “zfp & fpzip: Floating Point Compression”, Available Online at <https://computing.llnl.gov/projects/floating-point-compression>, Jul. 12, 2019, pp. 1-4. |
| Ratanaworabhan et al., “Fast Lossless Compression of Scientific Floating-Point Data”, Proceedings of the Data Compression Conference (DCC'06), IEEE Computer Society, 2006, 10 pages. |
| Number | Date | Country | |
|---|---|---|---|
| 20200225948 A1 | Jul 2020 | US |
