WO1999056202A1 - Method and apparatus for floating point operations and format conversion operations - Google Patents
Method and apparatus for floating point operations and format conversion operations Download PDFInfo
- Publication number
- WO1999056202A1 WO1999056202A1 PCT/US1999/009156 US9909156W WO9956202A1 WO 1999056202 A1 WO1999056202 A1 WO 1999056202A1 US 9909156 W US9909156 W US 9909156W WO 9956202 A1 WO9956202 A1 WO 9956202A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- floating point
- exponent
- circuit
- shift
- conversion
- Prior art date
Links
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims description 10
- 238000010586 diagram Methods 0.000 description 11
- 230000001419 dependent effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
- H03M7/14—Conversion to or from non-weighted codes
- H03M7/24—Conversion to or from floating-point codes
Definitions
- the present invention is in the field of computer operations on numbers in binary formats, including conversions between formats.
- the required range of numbers is very large. While it is possible to represent such numbers as multibyte integers or multibyte fractions, the memory required for storage is excessive. Also, when the number of significant bits required is small, the use of a multibyte representation is wasteful of memory. In addition, most very large or very small numbers do not require the precision of a multibyte representation.
- a more efficient representation of very large or very small decimal numbers is floating point notation or format. In digital systems, floating point format is the counterpart of scientific notation. Floating point is useful for performing operations that require many precise calculations, such as operations in a graphics application.
- Figure 1 shows base 2 representations of the number 6 in both floating point and integer formats.
- Number 106 is a base 2 floating point representation of the number 6.
- Number 108 is an integer base 2 representation of the number 6. Numbers in floating point format are always aligned on the left, that is, they are always normalized so that only one, non-zero number appears to the left of the decimal point as in number
- Integer representations are aligned on the right, that is, the decimal point appears as far to the right as possible to represent
- Steps 110 and 112 show how floating point number 106 is converted to the base 10 integer "6".
- Steps 114 and 116 show how integer 108 is converted to the base 10 integer "6".
- processors that perform floating point operations typically include special floating point circuitry to perform operations such as addition, subtraction, etc. Because it is not necessary or efficient for floating point numbers to be used for every application that may be executed on a processor, processors have the capability of performing operations using either floating point numbers or integer numbers. Conversions between the two formats are therefore often required. As seen by a comparison of number 106 with number 108, conversion between floating point and integer formats involves shifting the decimal point to the right or left. Floating point arithmetic units in typical processors include shift circuits. For example, a multiplication operation is equal to a shift of one bit position for every multiplication by 2.
- existing floating point arithmetic circuitry could be used for both performing floating point arithmetic operations and converting operands between floating point and integer formats.
- This is not a straightforward process, however, because the number of significant bits for a floating point format is not the same as the number of significant bits for a corresponding integer format. For instance, an integer format that includes 32 significant bits converts to a floating point format with a 23-bit mantissa.
- the data path provided in the floating point arithmetic unit to process floating point numbers during processing is not adequate to perform conversions between floating point and integer formats.
- a floating point arithmetic apparatus for converting numbers between an integer format and a floating point format, wherein a conversion operation requires a greater data path width than a arithmetic operation.
- the apparatus comprises right shift circuitry that receives a number in the floating point format, wherein the right shift circuitry includes additional register positions to accommodate a shift beyond a data path width required by an arithmetic operation.
- Figure 1 shows differences between a floating point format and an integer format.
- Figure 2 is a block diagram of a floating point arithmetic and conversion unit according to one embodiment of the present invention.
- Figure 3a is a diagram of a floating point format to integer format conversion operation when the floating point exponent is less than 23, according to one embodiment.
- Figure 3b is a diagram of a floating point format to integer format conversion operation when the floating point exponent is greater than or equal to 23, according to one embodiment.
- Figure 4 is a diagram of an integer format to floating point format conversion operation according to one embodiment.
- FIG. 5 is a diagram of a selection circuit according to one embodiment.
- Figure 6 is a block diagram of a selection circuit according to one embodiment.
- a modified floating point arithmetic unit allows for performance of conversion operations that would not be possible in an unmodified unit.
- the floating point arithmetic unit operates on single precision floating point numbers.
- a data width of an unmodified single precision floating point unit is inadequate to perform conversion operations between single precision floating point and 32-bit integer formats because a maximum possible shift left or right exceeds the unmodified data path width.
- additional bit positions are added to operational units and data paths of the floating point arithmetic unit to accommodate the maximum possible shift required by a conversion operation.
- a circuit is provided to latch numbers coming into the floating point unit for conversion in a particular manner.
- a circuit is also provided to more quickly determine a number of bits to be shifted and a direction of shift in order to perform a conversion operation in fewer clock cycles.
- FIG. 2 is a block diagram of a floating point arithmetic unit 300 according to one embodiment of the present invention.
- floating point unit 300 performs addition and subtraction on floating point numbers in a single precision format.
- Floating point unit 300 includes duplicate sets of functional units to perform parallel operations on two sets of floating point operands. Duplicate, parallel functional units are designated in Figure 2 with like numbers, for example 304a and 304b.
- Floating point unit 300 can therefore perform parallel operations on packed data formats, for example, as required by single instruction multiple data (SIMD) operations. Because functional units "a” are equivalent to functional units "b", only functional units "a" will be described.
- SIMD single instruction multiple data
- Control unit 302 includes circuitry for controlling the operation of functional units within floating point unit 300. For example, control unit 302 receives instructions and determines which functional units must be used, and in which manner, to carry out an instruction.
- Functional unit 308a is a mantissa comparison, multiplexing, and right shifting unit. The comparison capability of unit 308a is only needed in addition and subtraction operations. Specifically, when exponents of two floating point operands are equal, it is necessary to compare mantissa portions in order to determine the smaller mantissa and arrange operands so as to avoid a negative result.
- Exponent compare unit 304 compares incoming exponents in order to determine which is the larger exponent.
- exponent compare unit receives an exponent 1 and an exponent 2 from an operand 1 and an operand 2, respectively.
- an exponent of an incoming floating point number to be converted to an integer format is compared to a predetermined number for the purpose of determining whether a conversion operation will require a shift beyond data path space normally allotted for single precision floating point numbers in floating point unit 300.
- floating point unit 300 includes additional data path space in each of its functional units to accommodate shifts beyond a normal single precision floating point capacity.
- one of two conversion constants is used in a floating point to integer conversion operation.
- One conversion constant is used for the case in which normal single precision floating point data path capacity is adequate to perform the conversion.
- Another constant is used for the case in which the conversion operation will require a shift beyond the normal data path capacity required for single precision floating point numbers.
- Exponent subtract unit 306a in a conversion operation, performs a subtraction between an exponent of an incoming floating point number to be converted and a constant.
- constants used for conversion between a floating point format and integer format contain an exponent field that contains a predetermined number and a mantissa field that is filled with zeros.
- the constant takes the place of a second floating point operand in a conversion operation in floating point unit 300.
- the constant is a predetermined number whose value is dependent on a specific
- the conversion constant also includes a bias.
- processors commonly represent exponents in a biased form. This means that a constant value is added to an actual exponent so that the biased exponent is always a positive number.
- the value of a bias depends on the number of bits available for representing exponents in the floating point format being used. The bias and constant are chosen so that the smallest normalized number can be reciprocated without overflow. For example, in a common external format, -126 10 is the maximum negative value representable. In this case, a bias of +127 10 is added to the exponent.
- processor internal formats the actual bias is larger because the exponent field is larger than that of an external representation.
- exponent subtract unit 306a is used to subtract an exponent of an incoming floating point number from a constant (where the constant includes a constant plus a constant bias) in order to determine how many positions to the right the floating point must be shifted in order to right align the number for integer format.
- Mantissa addition unit 312a is used for adding floating point numbers in an addition operation.
- Mantissa unit 312a can also be used in a conversion operation for complementing a negative number.
- floating point numbers are always represented by positive fractions in the mantissa with a sign bit indicating the sign of the number. Incoming negative integer numbers may need to be converted to positive numbers for floating point representation.
- Mantissa leading zero anticipation (LZA)/left shifting unit 314a is used for left shifting in the case of a conversion from integer to floating
- Unit 314a is also used to anticipate leading 0s. When a left shift has been performed in a conversion from integer to floating point, leading zero anticipation circuitry of unit 314a determines the number of zeros to the left of a leftmost 1 if a resultant floating point number is not normalized. The number of leading zeros is transmitted to exponent adjust unit 310a. Exponent adjust unit 310a receives the number of leading zeros and shifts the number as necessary to normalize the floating point number.
- Figure 3a is a diagram illustrating an operation to convert a floating point number having an exponent that is smaller than a certain number to an integer number according to one embodiment.
- floating pint unit 300 is designed to operate on single precision floating point numbers with 23-bit mantissas. Therefore, in this embodiment, the certain number is 23. If an exponent of an incoming floating point number to be converted has an exponent that is greater than or equal to 23, it is possible that conversion will require a shift right beyond the normal data path width of a floating point unit designed to perform floating point arithmetic.
- Floating point number 602 is a number to be converted that is latched into right shifter 604 of floating point unit 300 as shown.
- Floating point number 602 is a number in an internal floating point format according to one embodiment.
- the internal format of floating point number 602 is a format on which floating point unit 300 operates, and differs from an external format that is output from or input to a processor including floating point unit 300.
- Several external formats are known. Commonly external format conform to an Institute of Electronics and Electrical Engineers (IEEE) specification. The present invention can be used with different external and internal formats.
- the leftmost field contains a sign bit indicating whether number 602 is positive or negative.
- Number 602 also includes an exponent field that, in this embodiment, includes 10 bits that indicate a power to which the mantissa portion of number 602 is to be raised.
- the mantissa portion includes the "J" bit and the fraction field.
- the J bit is a one-bit binary integer immediately to the left of an implied decimal point (or floating point) as shown, for example, in Figure 1, item 106. In other embodiments, the one-bit binary integer is implied and not explicitly shown as it is in this embodiment.
- the fraction field contains a binary fraction of 27 bits.
- the "O" bit is a single bit that indicates whether a mantissa overflow exists in the number represented.
- some numbers cannot be represented with just the exponent field, the J bit, and the fraction field. For example, in a floating point operation adding the numbers 1 and 1.5, the resultant number 2.5 requires that the O bit and the J bit contain a 1 and a 0, respectively.
- Numbers that have an overflow condition are not in the "normalized” floating point format. Normalized floating point format includes a value of zero in the O bit, a value of 1 in the J bit, and a binary value in the fraction field that indicates a power of two to which the number must be raised.
- the "GRS" field includes 3 bits that are used for rounding control as is known in the art.
- the GRS bits are the guard, round, and sticky bits.
- the value stored in the GRS bits is used by a rounding circuit in floating point arithmetic operations to round a resultant number according to some predetermined method.
- floating point unit 300 performs operations on single precision floating point numbers. Without additional storage capacity, floating point unit 300 could not accurately perform conversion operations on numbers that required a shift to the left or right over a certain number of bit positions.
- Figure 3b is a diagram illustrating an operation to convert a floating point number having an exponent that is greater smaller than a certain number to an integer number according to one embodiment.
- incoming floating point number 702 is a single precision floating point number that, in one case, may require the floating point to be shifted 31 bit positions in order to convert floating point number 702 to an integer format.
- a maximum shift of 24 bit positions is permitted in order to avoid overflowing into the GRS field. This is because the GRS field must be preserved for rounding.
- it is determined whether an operation to be performed is an arithmetic operation or a conversion operation (as explained more fully below). If it is determined that the operation to be performed is a conversion operation that requires a shift into and beyond the GRS field, "virtual shifting" and “virtual rounding" are enabled. Virtual shifting uses a special, larger
- the control unit of floating point unit 300 generates a signal to disable the rounding circuitry (this may be referred to as virtual rounding). It is not necessary to perform rounding or use actual GRS bits because the initial GRS bits(before the start of a conversion operation) are known to be zero.
- bits potentially shifted beyond the rightmost bit of right shifter 704 are known to be zeros and do not have to be accounted for by bits in a GRS field.
- the result of the conversion operation is a final 32-bit integer that is available at the outputs of floating point unit 300, for example, in a register such as register 706.
- Figure 4 is an illustration of a conversion from 32-bit integer format to single precision floating point format according to one embodiment of the present invention using floating point unit 300.
- Integer number 402 is a number in 32-bit integer format coming into an input of floating point unit 300.
- Floating point adder left shifter 404 is part of mantissa /left zero anticipation/left shifting unit 314a.
- Left shifter 404 includes additional bit positions to the right of the GRS field. In this embodiment, four additional bit positions are provided to accommodate a maximum shift to the left of 31 bits. The additional bit positions prevent a possible shift of significant bits into the exponent field of the resultant floating point number, which would cause a meaningless number to be created.
- floating point unit 300 determines how to latch an incoming
- floating point unit 300 latches incoming number 402 so as to align the rightmost bit of number 402 with the rightmost bit of the additional bit added to the right of the GRS field in left shifter 404. After shifting is performed by left shifter 404, the result is transferred to rounder 406.
- rounder 406 is a separate unit from floating point unit 300. In other embodiments, rounder 406 could be in a same unit as left shifter 404. Resultant floating point number 408 is output from rounder 406.
- Figure 4 does not show every intermediate operation that may be required in a conversion from integer format to floating point format.
- mantissa addition unit 312a may be required to convert a negative integer number from its 2's compliment representation to absolute value and sign representation required for floating point.
- Figure 5 is a diagram of a selection circuit used to direct floating point unit 300 to latch an incoming number in a particular way.
- the selection circuit of Figure 5 includes a multiplexor 500 controlled by integer convert signal 502. If an instruction received by floating point unit 300 is an integer convert instruction, integer number 506 is selected to be latched in the manner shown in Figure 4. In this case, latch integer signal 506 is output on output 510 of multiplexor 500. In the case where a floating point to integer conversion is required by an instruction or the case where a floating point arithmetic operation is to be performed, integer convert signal 502 is not active and latch floating point signal 504 is output.
- This embodiment includes the advantage of gracefully handling the case of the maximum negative integer number as input to a conversion operation.
- the maximum negative integer comes into floating point arithmetic unit 300 with a "1" in the O bit of left shifter 404 and a zero in the J bit of left shifter 404.
- the maximum negative integer does not change when complimented. Therefore, when the number arrives at rounder 406, it appears as if an overflow condition exists. Rounder 406 will therefore shift the number right and adjust the floating point exponent accordingly so that a correct floating point representation is produced.
- Figure 6 is a block diagram of a selection circuit that determines whether an incoming floating point number to be converted will require a shift of more than 23 bit positions (in this embodiment using single precision floating point numbers) and enables virtual shifting accordingly.
- the selection circuit of Figure 6 speeds execution of a conversion operation by simultaneously calculating both of two possible numbers of bit positions to be shifted. In this way, a data dependency is removed. Specifically, it is not necessary to wait for the result of a subtraction operation performed on two incoming operand exponents, determine whether the result is negative or positive, and then compliment as necessary to obtain a correct shift value.
- the selection circuit of Figure 6 is also used in addition operations.
- floating point addition operations typically align the exponents of the two operands by shifting the number having the larger exponent to match the smaller exponent before addition takes place.
- addition as well as that of conversion it is necessary to determine a correct number of bits to shift by performing a subtraction operation.
- Exponent 2 represents an exponent of an arithmetic operand when an operation to be performed is an arithmetic operation.
- K represents a conversion constant comprised of a constant bias plus 23 for this embodiment. 23 for a single precision floating point number is the maximum number of bit positions that can be shifted without shifting into the GRS field.
- Multiplexor 704 has inputs exponent 2 and K v . Exponent 2 is the same exponent 2 as is input to multiplexor 702.
- K v is an alternative conversion constant that allows virtual shifting into the GRS field and into bits to the right of the GRS field. In this embodiment, K v is 31 plus the constant bias.
- K v Because there is a difference of 4 between K v and K, a shift of 7 additional bit positions (virtual shifting) is allowed when K v is used. In this embodiment, use of K v allows bits to be shifted through the GRS field and the additional 4 bit positions to the right of the GRS field.
- Both multiplexors 702 and 704 are controlled by a signal that indicates whether the instruction is a conversion instruction or an arithmetic instruction. If the instruction is a conversion instruction, multiplexor 702 outputs K and multiplexor 704 outputs K v . Selector circuit
- Subtractor 700 is also comprised of two subtraction circuits that perform two subtraction operations simultaneously.
- Subtractor 706 has exponent 1 on one input.
- Exponent 1 is the exponent of a second arithmetic operand when the operation is an arithmetic operation and exponent 1 is the exponent of the floating point number to be converted when the operation is a conversion operation.
- subtract circuit 706 performs either a subtraction of exponent 1 from K or a subtraction of exponent 2 from exponent 1.
- Subtract circuit 708 has one input that receives exponent 1 where exponent 1 is the same exponent received by subtract circuit 706. Subtract circuit 708 also receives the output of multiplexor 704. Dependent upon the output of multiplexor 704, subtract circuit 708 performs a subtraction of exponent 1 from K v or a subtraction of exponent 1 from exponent 2.
- Multiplexor 710 receives the output of subtract circuit 706 and the output of subtract circuit 708. Multiplexor 710 is controlled by a signal that indicates, in the case of an arithmetic operation, whether exponent 1 is greater than or equal to exponent 2. In the case of a conversion operation, 710 is controlled by a signal that indicates whether exponent 1 is greater than or equal to K. In the case of a conversion operation, if exponent 1 is greater than or equal to K multiplexor 710 will output the difference of exponent 1 and K v as a right shift control. If exponent 1 is not greater than or equal to
- multiplexor 710 will output the difference of exponent 1 and K as a right shift control.
- constant values K and K v are stored in a read only memory (ROM).
Abstract
Description
Claims
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB0026500A GB2353882B (en) | 1998-04-30 | 1999-04-27 | Method and apparatus for floating point operations and format conversion operations |
DE19983175T DE19983175B4 (en) | 1998-04-30 | 1999-04-27 | Method, processor and device for floating point operations and format conversion operations |
AU37673/99A AU3767399A (en) | 1998-04-30 | 1999-04-27 | Method and apparatus for floating point operations and format conversion operations |
HK01102411A HK1032124A1 (en) | 1998-04-30 | 2001-04-04 | Method and apparatus for floating point operations and format conversion operations |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US09/071,466 US6282554B1 (en) | 1998-04-30 | 1998-04-30 | Method and apparatus for floating point operations and format conversion operations |
US09/071,466 | 1998-04-30 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1999056202A1 true WO1999056202A1 (en) | 1999-11-04 |
Family
ID=22101514
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US1999/009156 WO1999056202A1 (en) | 1998-04-30 | 1999-04-27 | Method and apparatus for floating point operations and format conversion operations |
Country Status (6)
Country | Link |
---|---|
US (2) | US6282554B1 (en) |
AU (1) | AU3767399A (en) |
DE (1) | DE19983175B4 (en) |
GB (1) | GB2353882B (en) |
HK (1) | HK1032124A1 (en) |
WO (1) | WO1999056202A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2549153A (en) * | 2016-04-08 | 2017-10-11 | Advanced Risc Mach Ltd | Apparatus and method for supporting a conversion instruction |
Families Citing this family (91)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6263426B1 (en) * | 1998-04-30 | 2001-07-17 | Intel Corporation | Conversion from packed floating point data to packed 8-bit integer data in different architectural registers |
US6266769B1 (en) * | 1998-04-30 | 2001-07-24 | Intel Corporation | Conversion between packed floating point data and packed 32-bit integer data in different architectural registers |
US6650327B1 (en) * | 1998-06-16 | 2003-11-18 | Silicon Graphics, Inc. | Display system having floating point rasterization and floating point framebuffering |
US6523050B1 (en) * | 1999-08-19 | 2003-02-18 | National Semiconductor Corporation | Integer to floating point conversion using one's complement with subsequent correction to eliminate two's complement in critical path |
US6369725B1 (en) * | 2000-09-26 | 2002-04-09 | International Business Machines Corporation | Method for binary to decimal conversion |
KR100465371B1 (en) * | 2001-01-26 | 2005-01-13 | 학교법인연세대학교 | apparatus and method for design of the floating point ALU performing addition and round operations in parallel |
JP3779602B2 (en) * | 2001-11-28 | 2006-05-31 | 松下電器産業株式会社 | SIMD operation method and SIMD operation device |
US7219336B2 (en) * | 2002-01-03 | 2007-05-15 | Intel Corporation | Tracking format of registers having multiple content formats in binary translation |
US6990505B2 (en) * | 2002-05-09 | 2006-01-24 | Sun Microsystems, Inc. | Method/apparatus for conversion of higher order bits of 64-bit integer to floating point using 53-bit adder hardware |
US20030231660A1 (en) * | 2002-06-14 | 2003-12-18 | Bapiraju Vinnakota | Bit-manipulation instructions for packet processing |
US20040103086A1 (en) * | 2002-11-26 | 2004-05-27 | Bapiraju Vinnakota | Data structure traversal instructions for packet processing |
JP4228706B2 (en) * | 2003-01-27 | 2009-02-25 | 株式会社デンソー | Electronic control device and storage device of the electronic control device |
US7424501B2 (en) | 2003-06-30 | 2008-09-09 | Intel Corporation | Nonlinear filtering and deblocking applications utilizing SIMD sign and absolute value operations |
US7899855B2 (en) | 2003-09-08 | 2011-03-01 | Intel Corporation | Method, apparatus and instructions for parallel data conversions |
TWI252430B (en) * | 2004-01-19 | 2006-04-01 | Mediatek Inc | Joint adaptive fixed-point representation and related arithmetic and processor thereof |
US7426630B1 (en) | 2004-06-30 | 2008-09-16 | Sun Microsystems, Inc. | Arbitration of window swap operations |
US8095778B1 (en) | 2004-06-30 | 2012-01-10 | Open Computing Trust I & II | Method and system for sharing functional units of a multithreaded processor |
US7178005B1 (en) | 2004-06-30 | 2007-02-13 | Sun Microsystems, Inc. | Efficient implementation of timers in a multithreaded processor |
US7941642B1 (en) | 2004-06-30 | 2011-05-10 | Oracle America, Inc. | Method for selecting between divide instructions associated with respective threads in a multi-threaded processor |
US7861063B1 (en) | 2004-06-30 | 2010-12-28 | Oracle America, Inc. | Delay slot handling in a processor |
US7185178B1 (en) | 2004-06-30 | 2007-02-27 | Sun Microsystems, Inc. | Fetch speculation in a multithreaded processor |
US7774393B1 (en) | 2004-06-30 | 2010-08-10 | Oracle America, Inc. | Apparatus and method for integer to floating-point format conversion |
US7533248B1 (en) | 2004-06-30 | 2009-05-12 | Sun Microsystems, Inc. | Multithreaded processor including a functional unit shared between multiple requestors and arbitration therefor |
US7437538B1 (en) | 2004-06-30 | 2008-10-14 | Sun Microsystems, Inc. | Apparatus and method for reducing execution latency of floating point operations having special case operands |
US7747771B1 (en) | 2004-06-30 | 2010-06-29 | Oracle America, Inc. | Register access protocol in a multihreaded multi-core processor |
US7353364B1 (en) | 2004-06-30 | 2008-04-01 | Sun Microsystems, Inc. | Apparatus and method for sharing a functional unit execution resource among a plurality of functional units |
US7216216B1 (en) | 2004-06-30 | 2007-05-08 | Sun Microsystems, Inc. | Register window management using first pipeline to change current window and second pipeline to read operand from old window and write operand to new window |
US7478225B1 (en) | 2004-06-30 | 2009-01-13 | Sun Microsystems, Inc. | Apparatus and method to support pipelining of differing-latency instructions in a multithreaded processor |
US7401206B2 (en) * | 2004-06-30 | 2008-07-15 | Sun Microsystems, Inc. | Apparatus and method for fine-grained multithreading in a multipipelined processor core |
US7383403B1 (en) | 2004-06-30 | 2008-06-03 | Sun Microsystems, Inc. | Concurrent bypass to instruction buffers in a fine grain multithreaded processor |
US7434000B1 (en) | 2004-06-30 | 2008-10-07 | Sun Microsystems, Inc. | Handling duplicate cache misses in a multithreaded/multi-core processor |
US7370243B1 (en) | 2004-06-30 | 2008-05-06 | Sun Microsystems, Inc. | Precise error handling in a fine grain multithreaded multicore processor |
US7890734B2 (en) * | 2004-06-30 | 2011-02-15 | Open Computing Trust I & II | Mechanism for selecting instructions for execution in a multithreaded processor |
US7330988B2 (en) * | 2004-06-30 | 2008-02-12 | Sun Microsystems, Inc. | Method and apparatus for power throttling in a multi-thread processor |
US7373489B1 (en) | 2004-06-30 | 2008-05-13 | Sun Microsystems, Inc. | Apparatus and method for floating-point exception prediction and recovery |
US7343474B1 (en) | 2004-06-30 | 2008-03-11 | Sun Microsystems, Inc. | Minimal address state in a fine grain multithreaded processor |
US7676655B2 (en) * | 2004-06-30 | 2010-03-09 | Sun Microsystems, Inc. | Single bit control of threads in a multithreaded multicore processor |
US7523330B2 (en) * | 2004-06-30 | 2009-04-21 | Sun Microsystems, Inc. | Thread-based clock enabling in a multi-threaded processor |
US8225034B1 (en) | 2004-06-30 | 2012-07-17 | Oracle America, Inc. | Hybrid instruction buffer |
US7702887B1 (en) | 2004-06-30 | 2010-04-20 | Sun Microsystems, Inc. | Performance instrumentation in a fine grain multithreaded multicore processor |
US7277990B2 (en) | 2004-09-30 | 2007-10-02 | Sanjeev Jain | Method and apparatus providing efficient queue descriptor memory access |
US8037250B1 (en) | 2004-12-09 | 2011-10-11 | Oracle America, Inc. | Arbitrating cache misses in a multithreaded/multi-core processor |
US7418543B2 (en) * | 2004-12-21 | 2008-08-26 | Intel Corporation | Processor having content addressable memory with command ordering |
US7555630B2 (en) | 2004-12-21 | 2009-06-30 | Intel Corporation | Method and apparatus to provide efficient communication between multi-threaded processing elements in a processor unit |
US7467256B2 (en) * | 2004-12-28 | 2008-12-16 | Intel Corporation | Processor having content addressable memory for block-based queue structures |
US7685213B2 (en) * | 2005-03-31 | 2010-03-23 | Siebel Systems, Inc. | Conversion of floating-point numbers from binary into string format |
US20070162475A1 (en) * | 2005-12-30 | 2007-07-12 | Intel Corporation | Method and apparatus for hardware-based dynamic escape detection in managed run-time environments |
US8732226B2 (en) * | 2006-06-06 | 2014-05-20 | Intel Corporation | Integer rounding operation |
US7865882B2 (en) * | 2006-08-18 | 2011-01-04 | International Business Machines Corporation | Fast correctly rounding floating point conversion and identifying exceptional conversion |
US7565492B2 (en) * | 2006-08-31 | 2009-07-21 | Intel Corporation | Method and apparatus for preventing software side channel attacks |
US7610448B2 (en) * | 2006-12-27 | 2009-10-27 | Intel Corporation | Obscuring memory access patterns |
US7991811B2 (en) * | 2007-04-17 | 2011-08-02 | International Business Machines Corporation | Method and system for optimizing floating point conversion between different bases |
US8327120B2 (en) | 2007-12-29 | 2012-12-04 | Intel Corporation | Instructions with floating point control override |
US7877582B2 (en) * | 2008-01-31 | 2011-01-25 | International Business Machines Corporation | Multi-addressable register file |
US7849294B2 (en) | 2008-01-31 | 2010-12-07 | International Business Machines Corporation | Sharing data in internal and memory representations with dynamic data-driven conversion |
US8392489B2 (en) * | 2008-02-15 | 2013-03-05 | International Business Machines Corporation | ASCII to binary floating point conversion of decimal real numbers on a vector processor |
US20100115494A1 (en) * | 2008-11-03 | 2010-05-06 | Gorton Jr Richard C | System for dynamic program profiling |
US8024719B2 (en) | 2008-11-03 | 2011-09-20 | Advanced Micro Devices, Inc. | Bounded hash table sorting in a dynamic program profiling system |
US8478948B2 (en) * | 2008-12-04 | 2013-07-02 | Oracle America, Inc. | Method and system for efficient tracing and profiling of memory accesses during program execution |
US8407271B2 (en) * | 2009-08-28 | 2013-03-26 | Advanced Micro Devices, Inc. | Method for floating point round to integer operation |
US8417733B2 (en) * | 2010-03-04 | 2013-04-09 | Oracle International Corporation | Dynamic atomic bitsets |
US10089278B2 (en) * | 2011-01-21 | 2018-10-02 | Nxp Usa, Inc. | Device and method for computing a function value of a function |
GB2488322B (en) * | 2011-02-22 | 2017-09-20 | Advanced Risc Mach Ltd | Number format pre-conversion instructions |
US9411585B2 (en) | 2011-09-16 | 2016-08-09 | International Business Machines Corporation | Multi-addressable register files and format conversions associated therewith |
US9727336B2 (en) | 2011-09-16 | 2017-08-08 | International Business Machines Corporation | Fine-grained instruction enablement at sub-function granularity based on an indicated subrange of registers |
WO2013095603A1 (en) * | 2011-12-23 | 2013-06-27 | Intel Corporation | Apparatus and method for down conversion of data types |
WO2013100926A1 (en) * | 2011-12-28 | 2013-07-04 | Intel Corporation | Memory cell array with dedicated nanoprocessors |
US9335993B2 (en) | 2011-12-29 | 2016-05-10 | International Business Machines Corporation | Convert from zoned format to decimal floating point format |
US9329861B2 (en) | 2011-12-29 | 2016-05-03 | International Business Machines Corporation | Convert to zoned format from decimal floating point format |
US9025494B1 (en) * | 2012-03-27 | 2015-05-05 | Infoblox Inc. | IPv6 network device discovery |
US9710227B2 (en) | 2012-09-15 | 2017-07-18 | John W. Ogilvie | Formatting floating point numbers |
US9606803B2 (en) * | 2013-07-15 | 2017-03-28 | Texas Instruments Incorporated | Highly integrated scalable, flexible DSP megamodule architecture |
US9264066B2 (en) | 2013-07-30 | 2016-02-16 | Apple Inc. | Type conversion using floating-point unit |
US9729167B2 (en) * | 2013-10-01 | 2017-08-08 | Oracle International Corporation | System and method for conversion of numeric values between different number base formats, for use with software applications |
US9619214B2 (en) | 2014-08-13 | 2017-04-11 | International Business Machines Corporation | Compiler optimizations for vector instructions |
US9588746B2 (en) | 2014-12-19 | 2017-03-07 | International Business Machines Corporation | Compiler method for generating instructions for vector operations on a multi-endian processor |
US10169014B2 (en) | 2014-12-19 | 2019-01-01 | International Business Machines Corporation | Compiler method for generating instructions for vector operations in a multi-endian instruction set |
US9722629B2 (en) * | 2015-01-15 | 2017-08-01 | Huawei Technologies Co., Ltd. | Method and apparatus for converting from floating point to integer representation |
US9880821B2 (en) | 2015-08-17 | 2018-01-30 | International Business Machines Corporation | Compiler optimizations for vector operations that are reformatting-resistant |
US9594668B1 (en) | 2015-09-04 | 2017-03-14 | International Business Machines Corporation | Debugger display of vector register contents after compiler optimizations for vector instructions |
US10706101B2 (en) | 2016-04-14 | 2020-07-07 | Advanced Micro Devices, Inc. | Bucketized hash tables with remap entries |
CN106502626A (en) * | 2016-11-03 | 2017-03-15 | 北京百度网讯科技有限公司 | Data processing method and device |
US10223114B1 (en) | 2017-09-29 | 2019-03-05 | Intel Corporation | Fixed point to floating point conversion |
US10224954B1 (en) | 2017-09-29 | 2019-03-05 | Intel Corporation | Floating point to fixed point conversion |
US10970078B2 (en) | 2018-04-05 | 2021-04-06 | Apple Inc. | Computation engine with upsize/interleave and downsize/deinterleave options |
US10831488B1 (en) | 2018-08-20 | 2020-11-10 | Apple Inc. | Computation engine with extract instructions to minimize memory access |
US11372643B2 (en) | 2018-11-09 | 2022-06-28 | Intel Corporation | Systems and methods for performing instructions to convert to 16-bit floating-point format |
CN111008230B (en) * | 2019-11-22 | 2023-08-04 | 远景智能国际私人投资有限公司 | Data storage method, device, computer equipment and storage medium |
CN111290790B (en) * | 2020-01-22 | 2023-03-24 | 安徽大学 | Conversion device for converting fixed point into floating point |
CN113703717B (en) * | 2021-08-31 | 2024-01-26 | 南京英锐创电子科技有限公司 | Binary floating-point number multiplication circuit, control method thereof and computing device |
US20230126672A1 (en) * | 2021-10-27 | 2023-04-27 | Jpmorgan Chase Bank, N.A. | Systems and methods for mixed precision machine learning with fully homomorphic encryption |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5257215A (en) * | 1992-03-31 | 1993-10-26 | Intel Corporation | Floating point and integer number conversions in a floating point adder |
US5627773A (en) * | 1995-06-30 | 1997-05-06 | Digital Equipment Corporation | Floating point unit data path alignment |
Family Cites Families (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3872442A (en) * | 1972-12-14 | 1975-03-18 | Sperry Rand Corp | System for conversion between coded byte and floating point format |
US4949291A (en) | 1988-12-22 | 1990-08-14 | Intel Corporation | Apparatus and method for converting floating point data formats in a microprocessor |
US5157388A (en) | 1989-02-14 | 1992-10-20 | Intel Corporation | Method and apparatus for graphics data interpolation |
US5303174A (en) * | 1991-10-07 | 1994-04-12 | Nec Corporation | Floating pointing arithmetic operation system |
US5404469A (en) | 1992-02-25 | 1995-04-04 | Industrial Technology Research Institute | Multi-threaded microprocessor architecture utilizing static interleaving |
EP0578950A3 (en) | 1992-07-15 | 1995-11-22 | Ibm | Method and apparatus for converting floating-point pixel values to byte pixel values by table lookup |
US5463484A (en) | 1993-10-29 | 1995-10-31 | Brody; Thomas P. | Method for manufacturing laminated U V-sensitive color filters for liquid crystal displays |
GB2287111B (en) | 1994-03-01 | 1998-08-05 | Intel Corp | Method for pipeline processing of instructions by controlling access to a reorder buffer using a register file outside the reorder buffer |
US5673407A (en) | 1994-03-08 | 1997-09-30 | Texas Instruments Incorporated | Data processor having capability to perform both floating point operations and memory access in response to a single instruction |
US5761103A (en) | 1995-03-08 | 1998-06-02 | Texas Instruments Incorporated | Left and right justification of single precision mantissa in a double precision rounding unit |
US5696709A (en) | 1995-03-31 | 1997-12-09 | International Business Machines Corporation | Program controlled rounding modes |
US5764548A (en) | 1995-09-29 | 1998-06-09 | Intel Corporation | Fast floating-point to integer conversion |
US5933160A (en) | 1995-11-27 | 1999-08-03 | Sun Microsystems | High-performance band combine function |
US5802219A (en) | 1995-11-27 | 1998-09-01 | Sun Microsystems, Inc. | Methods and apparatus for table lookup transformation of digital images |
US5805486A (en) | 1995-11-28 | 1998-09-08 | Intel Corporation | Moderately coupled floating point and integer units |
US5848284A (en) | 1995-11-28 | 1998-12-08 | Intel Corporation | Method of transferring data between moderately coupled integer and floating point units |
US5852726A (en) | 1995-12-19 | 1998-12-22 | Intel Corporation | Method and apparatus for executing two types of instructions that specify registers of a shared logical register file in a stack and a non-stack referenced manner |
US5729724A (en) | 1995-12-20 | 1998-03-17 | Intel Corporation | Adaptive 128-bit floating point load and store operations for quadruple precision compatibility |
US5740093A (en) | 1995-12-20 | 1998-04-14 | Intel Corporation | 128-bit register file and 128-bit floating point load and store for quadruple precision compatibility |
US5968116A (en) | 1996-03-27 | 1999-10-19 | Intel Corporation | Method and apparatus for facilitating the management of networked devices |
US5874969A (en) | 1996-07-01 | 1999-02-23 | Sun Microsystems, Inc. | Three-dimensional graphics accelerator which implements multiple logical buses using common data lines for improved bus communication |
US5768170A (en) | 1996-07-25 | 1998-06-16 | Motorola Inc. | Method and apparatus for performing microprocessor integer division operations using floating point hardware |
US6058465A (en) | 1996-08-19 | 2000-05-02 | Nguyen; Le Trong | Single-instruction-multiple-data processing in a multimedia signal processor |
US5978901A (en) | 1997-08-21 | 1999-11-02 | Advanced Micro Devices, Inc. | Floating point and multimedia unit with data type reclassification capability |
DE69801678T2 (en) | 1997-10-23 | 2003-01-16 | Advanced Micro Devices Inc | MULTIFUNCTIONAL TWO-PIECE LOOKUP TABLE |
US6061782A (en) | 1997-11-29 | 2000-05-09 | Ip First Llc | Mechanism for floating point to integer conversion with RGB bias multiply |
US5873407A (en) * | 1998-03-27 | 1999-02-23 | Inventec Corporation | Windblown-type heat-dissipating device for computer mother board |
US5995122A (en) | 1998-04-30 | 1999-11-30 | Intel Corporation | Method and apparatus for parallel conversion of color values from a single precision floating point format to an integer format |
-
1998
- 1998-04-30 US US09/071,466 patent/US6282554B1/en not_active Expired - Lifetime
-
1999
- 1999-04-27 AU AU37673/99A patent/AU3767399A/en not_active Abandoned
- 1999-04-27 WO PCT/US1999/009156 patent/WO1999056202A1/en active Application Filing
- 1999-04-27 GB GB0026500A patent/GB2353882B/en not_active Expired - Fee Related
- 1999-04-27 DE DE19983175T patent/DE19983175B4/en not_active Expired - Fee Related
-
2001
- 2001-02-14 US US09/783,875 patent/US7216138B2/en not_active Expired - Fee Related
- 2001-04-04 HK HK01102411A patent/HK1032124A1/en not_active IP Right Cessation
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5257215A (en) * | 1992-03-31 | 1993-10-26 | Intel Corporation | Floating point and integer number conversions in a floating point adder |
US5627773A (en) * | 1995-06-30 | 1997-05-06 | Digital Equipment Corporation | Floating point unit data path alignment |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2549153A (en) * | 2016-04-08 | 2017-10-11 | Advanced Risc Mach Ltd | Apparatus and method for supporting a conversion instruction |
US10310809B2 (en) | 2016-04-08 | 2019-06-04 | Arm Limited | Apparatus and method for supporting a conversion instruction |
GB2549153B (en) * | 2016-04-08 | 2019-09-11 | Advanced Risc Mach Ltd | Apparatus and method for supporting a conversion instruction |
Also Published As
Publication number | Publication date |
---|---|
US7216138B2 (en) | 2007-05-08 |
GB2353882B (en) | 2003-02-19 |
US20040268094A1 (en) | 2004-12-30 |
GB2353882A (en) | 2001-03-07 |
GB0026500D0 (en) | 2000-12-13 |
DE19983175T1 (en) | 2001-05-10 |
US6282554B1 (en) | 2001-08-28 |
HK1032124A1 (en) | 2001-07-06 |
AU3767399A (en) | 1999-11-16 |
DE19983175B4 (en) | 2006-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6282554B1 (en) | Method and apparatus for floating point operations and format conversion operations | |
US4841467A (en) | Architecture to implement floating point multiply/accumulate operations | |
US5963461A (en) | Multiplication apparatus and methods which generate a shift amount by which the product of the significands is shifted for normalization or denormalization | |
US6529928B1 (en) | Floating-point adder performing floating-point and integer operations | |
US5161117A (en) | Floating point conversion device and method | |
US5027308A (en) | Circuit for adding/subtracting two floating point operands | |
US4941120A (en) | Floating point normalization and rounding prediction circuit | |
EP0472139B1 (en) | A floating-point processor | |
US8489663B2 (en) | Decimal floating-point adder with leading zero anticipation | |
EP1857925B1 (en) | Method and apparatus for decimal number multiplication using hardware for binary number operations | |
US5010508A (en) | Prenormalization for a floating-point adder | |
US5317526A (en) | Format conversion method of floating point number and device employing the same | |
KR100203468B1 (en) | Arithmetic apparatus for floating point numbers | |
US5687106A (en) | Implementation of binary floating point using hexadecimal floating point unit | |
US20030028572A1 (en) | Fast single precision floating point accumulator using base 32 system | |
US8326908B2 (en) | Residue number system arithmetic operating system, scaling operator, scaling operation method and program and recording medium of the same | |
US5954790A (en) | Method and apparatus for parallel prediction and computation of massive cancellation in floating point subtraction | |
US4110831A (en) | Method and means for tracking digit significance in arithmetic operations executed on decimal computers | |
KR19980082906A (en) | How to Convert Floating-Point Numbers to Integer Types | |
EP0332215B1 (en) | Operation circuit based on floating-point representation | |
KR100223997B1 (en) | Normalization circuit device of floating point computation device | |
KR19980064405A (en) | Arithmetic operation system | |
US5867722A (en) | Sticky bit detector for a floating-point processor | |
JPH05150948A (en) | Rounding circuit in floating point multiplier | |
US6044391A (en) | Method of generating the sticky-bit from the input operands |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG US UZ VN YU ZA ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH GM KE LS MW SD SL SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
ENP | Entry into the national phase |
Ref document number: 200026500 Country of ref document: GB Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: KR |
|
RET | De translation (de og part 6b) |
Ref document number: 19983175 Country of ref document: DE Date of ref document: 20010510 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 19983175 Country of ref document: DE |
|
122 | Ep: pct application non-entry in european phase | ||
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8607 |