US20190087197A1 - Method and apparatus for flushing instructions from reservation stations - Google Patents
Method and apparatus for flushing instructions from reservation stations Download PDFInfo
- Publication number
- US20190087197A1 US20190087197A1 US16/056,723 US201816056723A US2019087197A1 US 20190087197 A1 US20190087197 A1 US 20190087197A1 US 201816056723 A US201816056723 A US 201816056723A US 2019087197 A1 US2019087197 A1 US 2019087197A1
- Authority
- US
- United States
- Prior art keywords
- index
- allocated
- circular buffer
- entry
- pointer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
- G06F9/3861—Recovery, e.g. branch miss-prediction, exception handling
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/30098—Register arrangements
- G06F9/30101—Special purpose registers
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
- G06F9/3836—Instruction issuing, e.g. dynamic instruction scheduling or out of order instruction execution
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
- G06F9/3854—Instruction completion, e.g. retiring, committing or graduating
- G06F9/3858—Result writeback, i.e. updating the architectural state or memory
- G06F9/38585—Result writeback, i.e. updating the architectural state or memory with result invalidation, e.g. nullification
-
- G06F9/3859—
Definitions
- the present invention relates to microprocessors, and more particularly, to efficiently perform recovery from exceptions.
- Modern processor microarchitectures implement branch prediction and speculative fetch of instructions from the predicted branch. On misprediction, speculatively fetched instructions need to be flushed from the processor.
- a processor may include one or more reservation stations configured to host instructions waiting to be issued to the execution units. On exception event, reservation stations may be configured to flush instructions that are younger, in program order, than the instruction executed with exception. Older instructions may remain in the reservation stations waiting to be executed.
- Flushing instructions from the reservation stations may be implemented by assigning flush masks to the hosted instructions. Each bit in the flush mask indicates dependence with prior branch instruction that may initiate flush. This approach leads to complex design solutions that do not scale well with the size of the flush mask. Keeping the flush mask as small as possible, on the other hand, introduces stalls at runtime.
- a processor may include reorder buffer, one or more reservation stations, and execution units.
- Reorder buffer assigns indexes to instructions, before they enter the reservation stations.
- Reservation stations host instructions with the assigned reorder buffer indexes, while waiting to be issued to the execution units.
- Execution units execute instructions and provide back reorder buffer index of the executed instructions.
- the reorder buffer is a circular buffer that implements a head pointer and a tail pointer to maintain the beginning and the end of the assigned indexes.
- reservation stations may be configured to flush instructions that are younger than the instruction executed with exception.
- Reservation stations may be configured to receive a first index A and a second index B, and to flush (remove) instructions with assigned indexes in the wrapped-around increasing interval from the first index A to the second index B.
- Execution units may provide the reorder buffer index EX of the instruction executed with exception as the first index A.
- the reorder buffer may provide the reorder buffer index TP stored in the tail pointer as the second index B.
- Reservation stations may be configured to flush instructions with reorder buffer indexes in the wrapped-around increasing interval from the index EX to the index TP (or HP).
- FIG. 1 shows an embodiment of a processor core
- FIGS. 2 A and B shows an embodiment of a reorder buffer
- FIG. 3 shows an embodiment of a reservation station
- FIG. 4 shows an embodiment of a flush node
- FIG. 5 shows a method for flushing instructions from reservation stations
- FIG. 6 shows an embodiment of central processing unit in accordance with the embodiments of the present invention
- FIG. 1 shows an embodiment of a processor core.
- the core 100 may include fetch and decode unit 102 , renaming unit 104 , reorder buffer 106 , reservation stations 108 , execution units 110 , and physical register file 112 .
- the core 100 may include other components and interfaces not shown on FIG. 1 to emphasize the embodiments described herein.
- the fetch and decode unit 102 is configured to fetch instructions from memory or L2 cache and to output, in parallel, one or more decoded instructions or instruction micro-operations.
- the fetch and decode unit 102 may be configured to fetch instructions from any instruction set architecture (e.g. PowerPCTM, ARMTM, SPARCTM, x86TM, etc.) and convert them into instruction (micro-)operations that can be executed in the execution units 110 .
- the fetch and decode unit 102 unit may be split into two or more units, for example fetch unit, decode unit, branch predictor, L1 cache, etc.
- the renaming unit 104 is configured to rename source and destination operands to the physical registers of the core 100 .
- the core 100 may include one or more physical register files 112 configured to store speculative results and architecturally visible results.
- the reorder buffer 106 is a circular buffer configured to assign indexes to instructions entering the reservation stations 108 . For each instruction entering the reservation stations 108 , the reorder buffer 106 provides an index in wrapped-around incrementing fashion.
- the reorder buffer 106 may include a head pointer, to store the index of the oldest in-flight instruction, and a tail pointer, to store the next-in-line for assignment index.
- Other embodiments may use additional circular structures that assign unique indexes to groups of one or more in-flight instructions.
- Reservation stations 108 may comprise of one or more reservation stations, wherein each reservation station may be coupled to one or more execution units 110 .
- Each reservation station may include a plurality of entries configured to host instructions with the assigned indexes. Instructions enter reservation stations 108 in program order, but may be issued to the execution units 110 in any order.
- Reservation stations 108 may be coupled to the reorder buffer 106 receive the index HP stored in the head pointer or the index TP stored in the tail pointer. The HP index or the TP index may be used to track relative age between instructions in the reservation stations 108 .
- Execution units 112 may include any number and type of execution units, e.g. branch units, floating point units, load/store units, integer units, etc.
- Execution units 110 are configured to execute instructions and to return the reorder buffer index of the executed instructions.
- An instruction may be executed with exception, which may require flushing the core 100 .
- Reservation stations 108 may be configured to flush (remove) instructions that are younger, in program order, than the instruction executed with exception. Older instructions may remain in the reservation stations 108 waiting to be issued.
- Execution units 110 may be coupled to the reservation stations 108 to provide the reorder buffer index EX of the instruction executed with exception.
- the reorder buffer 106 may be coupled to the reservation stations 108 to provide the index TP stored in the tail pointer or the index HP stored in the head pointer. Reservation stations 108 may be configured to flush instructions with reorder buffer indexes in the wrapped-around increasing interval from the reorder buffer index EX to the reorder buffer index TP (or HP).
- the reorder buffer (ROB) 200 is a circular buffer with a plurality of entries denoted with indexes from 0 to N 202 , where the index 0 is considered successor of the index N. For each instruction entering the reservation stations 108 in program order, the reorder buffer 200 allocates an entry and returns the index 202 of the allocated entry.
- the reorder buffer 200 may include a head pointer 204 and a tail pointer 206 to maintain the beginning and the end of the consecutive interval of allocated entries.
- the head pointer 204 points to the entry allocated to the oldest in-flight instruction.
- the tail pointer 206 points to the next-in-line for allocation entry.
- the tail pointer 206 is incremented with each allocation in wrapped-around incrementing order.
- the head pointer 204 is incremented with retirement of the instructions.
- the head pointer 204 or the tail pointer 206 may be used as reference point in determining the relative age between instructions in the reservation stations 108 , i.e. the closer, in wrapped-around incrementing order, the assigned index of an instruction to the tail pointer 206 is, the younger the instruction is.
- An instruction e.g. with reorder buffer index I
- Reservation stations 108 may be configured flush instructions that are younger, in program order, than the instruction executed with exception.
- Reorder buffer indexes of instructions that are younger than the instruction executed with exception are in the increasing interval from the index I to the index T, which is stored in the tail pointer 206 .
- Reservation stations 108 may be configured to flush instructions in the increasing interval from the index I to the index T. If the index I is larger than the index T ( FIG. 2 B), the increasing interval wraps around from the index N to the index 0.
- the flush interval may be defined from the index I to the index H, which is stored in the head pointer 204 .
- Reservation stations 108 may be configured to compare the index I with the index T. If I ⁇ T ( FIG. 2 A), reservation stations 108 may be configured to flush instructions with reorder buffer indexes that are larger than the index I and smaller than the index T. If I ⁇ T ( FIG. 2 B), reservation stations 108 may be configured to flush instructions with reorder buffer indexes that are larger than the index I or smaller than the index T.
- the reservation station 300 may include a plurality of entries 302 a - n , configured to host instructions, and a plurality of flush nodes 306 a - n , configured to flush instructions hosted in the entries 302 a - n .
- Each instruction may be hosted with an index, assigned from an interval of indexes from 0 to N, in a circular fashion.
- Reservation station entries 302 a - n may include Index field 304 a - n to store the index assigned to the hosted instruction.
- Flush nodes 306 a - n are coupled to the entries 302 a - n to receive the value in the Index field 304 a - n.
- Flush nodes 306 a - n are configured to receive a first index A and a second index B and to assert flush signal if the value in the coupled Index field 304 a - n is in the wrapped-around increasing interval from the index A to the index B. Asserted flush signal may initiate flush operation for the instruction hosted in the coupled entry 302 a - n.
- Flush nodes 306 a - n are configured to compare the first index A with the second index B. If A ⁇ B, each flush node, e.g. 306 i , is configured to compare the value in the Index field 304 i with the first index A and with the second index B, and to assert flush signal for the entry 302 i if the value in the Index field 304 i is larger than the index A and smaller than the index B. If A ⁇ B, each flush node, e.g.
- 306 i is configured to compare the value in the Index field 304 i with the first index A and with the second index B, and to assert flush signal for the entry 302 i if the value in the Index field 304 i is larger than the index A or smaller than the index B.
- reorder buffer index assigned to the hosted instruction may be stored in the Index field 304 a - n . Responsive to exception event that requires selectively flushing instructions from the reservation station 300 , reorder buffer 106 and execution units 110 may be coupled to the flush nodes 306 a - n . Execution units 110 may provide the reorder buffer index EX of the instruction executed with exception as the first index A. The reorder buffer 106 may provide reorder the buffer index TP stored in the tail pointer 206 or the reorder buffer index HP stored in the head pointer 204 as the second index B. Flush nodes 306 a - n are configured to flush entries 302 a - n with values in the Index field 304 a - n in the increasing interval from the index EX to the index TP (or HP).
- the flush node 400 may include comparators 402 - 406 , AND gates 408 - 412 , coupled to the comparators 402 - 406 , and an OR gate 414 , coupled to the AND gates 408 - 412 .
- a reservation station entry e.g. 302 i
- the flush node 400 is configured to receive a first index A and a second index B and to output a flush signal, which may flush the instruction hosted in the reservation station entry 302 i.
- the comparator 402 may be configured to output 1 if the Index is larger than A, and 0 otherwise.
- the comparator 404 may be configured to output 1 if A ⁇ B, and 0 otherwise.
- the comparator 406 may be configured to output 1 if the Index is smaller than B, and 0 otherwise.
- the output from the comparator 404 may be shared among all flush nodes 306 a - n.
- the AND gate 408 is configured to output 1 if A ⁇ B and the Index is larger than the index A.
- the AND gate 410 is configured to output 1 if A ⁇ B and the Index is larger than A and smaller than B.
- the AND gate 412 is configured to output 1 if A ⁇ B and the Index is smaller than B.
- the OR gate 414 is configured to assert the flush signal if at least one of the AND gates 408 - 412 outputs 1, which indicates that the Index is in a flush interval.
- Reservation stations are configured to host instructions with assigned indexes.
- the method 500 begins when a first index A and a second index B are received by the reservation stations (block 502 ).
- the index A is compared with the index B (conditional block 504 ). If the index A is smaller than the index B (Y branch of 504 ), assigned index of each instruction is compared with the index A and with the index B to determine if the assigned index is larger than A and smaller than B (conditional block 506 ). If the assigned index is larger than A and smaller than B (Y branch of 506 ), the instruction is flushed from the reservation stations (block 510 ).
- index A is larger than or equal to the index B (N branch of 504 )
- assigned index of each instruction is compared with the index A and with the index B to determine if the assigned index is smaller than B or larger than A (conditional block 508 ). If the assigned index is smaller than B or larger than A (Y branch of 508 ), the instruction is flushed from reservation station entry (block 510 ).
- the central processing unit 600 may be embodied as a hardware, software, combination of hardware and software, or computer program product, stored on a non-transitory storage media and later used to fabricate hardware comprising the embodiments described herein.
- the central processing unit 600 may be part of a desktop computer, server, laptop computer, tablet computer, cell or mobile phone, wearable device, special purpose computer, etc.
- the central processing unit 600 may be included within a system on a chip or integrated circuit, coupled to external memory 606 and peripheral units 608 .
- the CPU 600 may include one or more instances of core processors 602 a - n , shared cache 604 , interface units, power supply unit, etc. At least one of the core processors 602 a - n may include the embodiments described herein.
- External memory 606 may be any type of memory, such as dynamic random-access memory (DRAM), static random-access memory (SRAM), etc. In some systems, more than one instance of central processing units 600 and/or external memory 608 may be used on one or more integrated circuits.
- the peripheral unit 608 may include various types of communication interfaces, display, keyboard, etc.
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Advance Control (AREA)
Abstract
A processor may include a reorder buffer, reservation stations, and execution units. The reorder buffer may be a circular buffer with a head pointer and a tail pointer, configured to assign indexes to instructions. Reservation stations may be configured to host instructions with the assigned indexes, while waiting to be issued to the execution units. Responsive to exception event, reservation stations may be configured to flush instructions that are younger, in program order, than the instruction executed with exception. Execution units may provide the reorder buffer index EX of the instruction executed with exception. The reorder buffer may provide the reorder buffer index TP stored in the tail pointer. Reservation stations may be configured to flush instructions with assigned indexes in the wrapped-around increasing interval from the index EX to the index TP.
Description
- This application is a continuation application of U.S. patent application Ser. No. 15/821,884, filed on Nov. 24, 2017, which claims priority to U.S. provisional patent application No. 62/561,215, filed on Sep. 21, 2017.
- The present invention relates to microprocessors, and more particularly, to efficiently perform recovery from exceptions.
- Modern processor microarchitectures implement branch prediction and speculative fetch of instructions from the predicted branch. On misprediction, speculatively fetched instructions need to be flushed from the processor.
- A processor may include one or more reservation stations configured to host instructions waiting to be issued to the execution units. On exception event, reservation stations may be configured to flush instructions that are younger, in program order, than the instruction executed with exception. Older instructions may remain in the reservation stations waiting to be executed.
- Flushing instructions from the reservation stations may be implemented by assigning flush masks to the hosted instructions. Each bit in the flush mask indicates dependence with prior branch instruction that may initiate flush. This approach leads to complex design solutions that do not scale well with the size of the flush mask. Keeping the flush mask as small as possible, on the other hand, introduces stalls at runtime.
- Method and apparatus for flushing instructions from the reservation stations is contemplated. The embodiments described herein use the tail pointer or the head pointer of a circular buffer to determine the relative age between instructions in the reservation stations and to flush instructions that are younger than the instruction executed with exception.
- In one embodiment, a processor may include reorder buffer, one or more reservation stations, and execution units. Reorder buffer assigns indexes to instructions, before they enter the reservation stations. Reservation stations host instructions with the assigned reorder buffer indexes, while waiting to be issued to the execution units. Execution units execute instructions and provide back reorder buffer index of the executed instructions. The reorder buffer is a circular buffer that implements a head pointer and a tail pointer to maintain the beginning and the end of the assigned indexes.
- Responsive to exception event that requires flushing instructions, reservation stations may be configured to flush instructions that are younger than the instruction executed with exception. Reservation stations may be configured to receive a first index A and a second index B, and to flush (remove) instructions with assigned indexes in the wrapped-around increasing interval from the first index A to the second index B. Execution units may provide the reorder buffer index EX of the instruction executed with exception as the first index A. The reorder buffer may provide the reorder buffer index TP stored in the tail pointer as the second index B. Reservation stations may be configured to flush instructions with reorder buffer indexes in the wrapped-around increasing interval from the index EX to the index TP (or HP).
-
FIG. 1 shows an embodiment of a processor core; -
FIGS. 2 A and B shows an embodiment of a reorder buffer; -
FIG. 3 shows an embodiment of a reservation station; -
FIG. 4 shows an embodiment of a flush node; -
FIG. 5 shows a method for flushing instructions from reservation stations; -
FIG. 6 shows an embodiment of central processing unit in accordance with the embodiments of the present invention; -
FIG. 1 shows an embodiment of a processor core. Thecore 100 may include fetch anddecode unit 102, renamingunit 104,reorder buffer 106,reservation stations 108,execution units 110, andphysical register file 112. Thecore 100 may include other components and interfaces not shown onFIG. 1 to emphasize the embodiments described herein. - The fetch and
decode unit 102 is configured to fetch instructions from memory or L2 cache and to output, in parallel, one or more decoded instructions or instruction micro-operations. The fetch anddecode unit 102 may be configured to fetch instructions from any instruction set architecture (e.g. PowerPC™, ARM™, SPARC™, x86™, etc.) and convert them into instruction (micro-)operations that can be executed in theexecution units 110. In other embodiments, the fetch anddecode unit 102 unit may be split into two or more units, for example fetch unit, decode unit, branch predictor, L1 cache, etc. - The renaming
unit 104 is configured to rename source and destination operands to the physical registers of thecore 100. Thecore 100 may include one or morephysical register files 112 configured to store speculative results and architecturally visible results. - The
reorder buffer 106 is a circular buffer configured to assign indexes to instructions entering thereservation stations 108. For each instruction entering thereservation stations 108, thereorder buffer 106 provides an index in wrapped-around incrementing fashion. Thereorder buffer 106 may include a head pointer, to store the index of the oldest in-flight instruction, and a tail pointer, to store the next-in-line for assignment index. Other embodiments may use additional circular structures that assign unique indexes to groups of one or more in-flight instructions. -
Reservation stations 108 may comprise of one or more reservation stations, wherein each reservation station may be coupled to one ormore execution units 110. Each reservation station may include a plurality of entries configured to host instructions with the assigned indexes. Instructions enterreservation stations 108 in program order, but may be issued to theexecution units 110 in any order.Reservation stations 108 may be coupled to thereorder buffer 106 receive the index HP stored in the head pointer or the index TP stored in the tail pointer. The HP index or the TP index may be used to track relative age between instructions in thereservation stations 108. -
Execution units 112 may include any number and type of execution units, e.g. branch units, floating point units, load/store units, integer units, etc.Execution units 110 are configured to execute instructions and to return the reorder buffer index of the executed instructions. An instruction may be executed with exception, which may require flushing thecore 100.Reservation stations 108 may be configured to flush (remove) instructions that are younger, in program order, than the instruction executed with exception. Older instructions may remain in thereservation stations 108 waiting to be issued.Execution units 110 may be coupled to thereservation stations 108 to provide the reorder buffer index EX of the instruction executed with exception. Thereorder buffer 106 may be coupled to thereservation stations 108 to provide the index TP stored in the tail pointer or the index HP stored in the head pointer.Reservation stations 108 may be configured to flush instructions with reorder buffer indexes in the wrapped-around increasing interval from the reorder buffer index EX to the reorder buffer index TP (or HP). - Turning now to
FIGS. 2 A and B, an embodiment of a reorder buffer is shown. The reorder buffer (ROB) 200 is a circular buffer with a plurality of entries denoted with indexes from 0 toN 202, where theindex 0 is considered successor of the index N. For each instruction entering thereservation stations 108 in program order, thereorder buffer 200 allocates an entry and returns theindex 202 of the allocated entry. Thereorder buffer 200 may include ahead pointer 204 and atail pointer 206 to maintain the beginning and the end of the consecutive interval of allocated entries. Thehead pointer 204 points to the entry allocated to the oldest in-flight instruction. Thetail pointer 206 points to the next-in-line for allocation entry. Thetail pointer 206 is incremented with each allocation in wrapped-around incrementing order. Thehead pointer 204 is incremented with retirement of the instructions. Thehead pointer 204 or thetail pointer 206 may be used as reference point in determining the relative age between instructions in thereservation stations 108, i.e. the closer, in wrapped-around incrementing order, the assigned index of an instruction to thetail pointer 206 is, the younger the instruction is. - An instruction, e.g. with reorder buffer index I, may be executed with exception.
Reservation stations 108 may be configured flush instructions that are younger, in program order, than the instruction executed with exception. Reorder buffer indexes of instructions that are younger than the instruction executed with exception are in the increasing interval from the index I to the index T, which is stored in thetail pointer 206.Reservation stations 108 may be configured to flush instructions in the increasing interval from the index I to the index T. If the index I is larger than the index T (FIG. 2 B), the increasing interval wraps around from the index N to theindex 0. Since reorder buffer entries with indexes in the increasing interval from thetail pointer 206 to thehead pointer 204 are not associated with in-flight instructions, the flush interval may be defined from the index I to the index H, which is stored in thehead pointer 204. -
Reservation stations 108 may be configured to compare the index I with the index T. If I<T (FIG. 2 A),reservation stations 108 may be configured to flush instructions with reorder buffer indexes that are larger than the index I and smaller than the index T. If I≥T (FIG. 2 B),reservation stations 108 may be configured to flush instructions with reorder buffer indexes that are larger than the index I or smaller than the index T. - Turning now to
FIG. 3 , an embodiment of a reservation station is shown. Thereservation station 300 may include a plurality of entries 302 a-n, configured to host instructions, and a plurality of flush nodes 306 a-n, configured to flush instructions hosted in the entries 302 a-n. Each instruction may be hosted with an index, assigned from an interval of indexes from 0 to N, in a circular fashion. Reservation station entries 302 a-n may include Index field 304 a-n to store the index assigned to the hosted instruction. Flush nodes 306 a-n are coupled to the entries 302 a-n to receive the value in the Index field 304 a-n. - Flush nodes 306 a-n are configured to receive a first index A and a second index B and to assert flush signal if the value in the coupled Index field 304 a-n is in the wrapped-around increasing interval from the index A to the index B. Asserted flush signal may initiate flush operation for the instruction hosted in the coupled entry 302 a-n.
- Flush nodes 306 a-n are configured to compare the first index A with the second index B. If A<B, each flush node, e.g. 306 i, is configured to compare the value in the
Index field 304 i with the first index A and with the second index B, and to assert flush signal for theentry 302 i if the value in theIndex field 304 i is larger than the index A and smaller than the index B. If A≥B, each flush node, e.g. 306 i, is configured to compare the value in theIndex field 304 i with the first index A and with the second index B, and to assert flush signal for theentry 302 i if the value in theIndex field 304 i is larger than the index A or smaller than the index B. - In one embodiment, reorder buffer index assigned to the hosted instruction may be stored in the Index field 304 a-n. Responsive to exception event that requires selectively flushing instructions from the
reservation station 300,reorder buffer 106 andexecution units 110 may be coupled to the flush nodes 306 a-n.Execution units 110 may provide the reorder buffer index EX of the instruction executed with exception as the first index A. Thereorder buffer 106 may provide reorder the buffer index TP stored in thetail pointer 206 or the reorder buffer index HP stored in thehead pointer 204 as the second index B. Flush nodes 306 a-n are configured to flush entries 302 a-n with values in the Index field 304 a-n in the increasing interval from the index EX to the index TP (or HP). - Turning now to
FIG. 4 , an embodiment of a flush node is shown. The flush node 400 may include comparators 402-406, AND gates 408-412, coupled to the comparators 402-406, and anOR gate 414, coupled to the AND gates 408-412. Those of ordinary skill in the art will appreciate that the hardware may vary depending on the implementation. A reservation station entry, e.g. 302 i, is coupled to the flush node 400 to provide the value stored in theIndex field 304 i. The flush node 400 is configured to receive a first index A and a second index B and to output a flush signal, which may flush the instruction hosted in thereservation station entry 302 i. - The
comparator 402 may be configured tooutput 1 if the Index is larger than A, and 0 otherwise. Thecomparator 404 may be configured tooutput 1 if A<B, and 0 otherwise. Thecomparator 406 may be configured tooutput 1 if the Index is smaller than B, and 0 otherwise. The output from thecomparator 404 may be shared among all flush nodes 306 a-n. - The AND
gate 408 is configured tooutput 1 if A≥B and the Index is larger than the index A. The ANDgate 410 is configured tooutput 1 if A≤B and the Index is larger than A and smaller than B. The ANDgate 412 is configured tooutput 1 if A≥B and the Index is smaller than B. The ORgate 414 is configured to assert the flush signal if at least one of the AND gates 408-412outputs 1, which indicates that the Index is in a flush interval. - Turning now to
FIG. 5 , a method for flushing instructions from the reservation stations is shown. Reservation stations are configured to host instructions with assigned indexes. Themethod 500 begins when a first index A and a second index B are received by the reservation stations (block 502). - The index A is compared with the index B (conditional block 504). If the index A is smaller than the index B (Y branch of 504), assigned index of each instruction is compared with the index A and with the index B to determine if the assigned index is larger than A and smaller than B (conditional block 506). If the assigned index is larger than A and smaller than B (Y branch of 506), the instruction is flushed from the reservation stations (block 510).
- If the index A is larger than or equal to the index B (N branch of 504), assigned index of each instruction is compared with the index A and with the index B to determine if the assigned index is smaller than B or larger than A (conditional block 508). If the assigned index is smaller than B or larger than A (Y branch of 508), the instruction is flushed from reservation station entry (block 510).
- Referring now to
FIG. 6 , an embodiment of a central processing unit in accordance with the embodiments of the present invention is shown. It should be obvious to those skilled in the art that thecentral processing unit 600 may be embodied as a hardware, software, combination of hardware and software, or computer program product, stored on a non-transitory storage media and later used to fabricate hardware comprising the embodiments described herein. Thecentral processing unit 600 may be part of a desktop computer, server, laptop computer, tablet computer, cell or mobile phone, wearable device, special purpose computer, etc. Thecentral processing unit 600 may be included within a system on a chip or integrated circuit, coupled toexternal memory 606 andperipheral units 608. TheCPU 600 may include one or more instances of core processors 602 a-n, sharedcache 604, interface units, power supply unit, etc. At least one of the core processors 602 a-n may include the embodiments described herein.External memory 606 may be any type of memory, such as dynamic random-access memory (DRAM), static random-access memory (SRAM), etc. In some systems, more than one instance ofcentral processing units 600 and/orexternal memory 608 may be used on one or more integrated circuits. Theperipheral unit 608 may include various types of communication interfaces, display, keyboard, etc.
Claims (14)
1. A method comprising:
allocating entries of a circular buffer, wherein each circular buffer entry is identified with an index from an interval of consecutive integers;
maintaining a pointer, wherein the pointer may point to a not allocated circular buffer entry, or to a circular buffer entry allocated before other allocated circular buffer entries, or to a circular buffer entry allocated after other allocated circular buffer entries;
comparing a first index of a first allocated circular buffer entry with a second index of a second allocated circular buffer entry;
comparing the pointer with the first index;
comparing the pointer with the second index;
if the first index is smaller than the second index and the pointer is smaller than or equal to both the first index and the second index, or the first index is smaller than the second index and the pointer is larger than or equal to both the first index and the second index, or the first index is larger than the second index and the pointer is smaller than or equal to the first index and the pointer is larger than or equal to the second index
determining that the first allocated entry was allocated before the second allocated entry,
otherwise
determining that the second allocated entry was allocated before the first allocated entry.
2. The method in claim 1 , further comprising
maintaining a table with plurality of table entries, wherein each table entry comprises an index field configured to store an index of an allocated circular buffer entry.
3. The method in claim 2 , further comprising
storing the second index of the second allocated circular buffer entry in an index field of a second table entry.
4. The method in claim 3 , further comprising
flushing the second table entry in response to determining that the first allocated circular buffer entry was allocated before the second allocated circular buffer entry.
5. The method in claim 4 , wherein the circular buffer stores not committed instructions in a processor.
6. The method in claim 5 , wherein the table stores not executed instructions in the processor.
7. The method in claim 6 , wherein the first allocated circular buffer entry corresponds to an instruction executed with exception.
8. An apparatus comprising
a circular buffer comprising plurality of entries, wherein each circular buffer entry is identified with an index from an interval of consecutive integers;
a pointer configured to store an index of a not allocated circular buffer entry, or an index of a circular buffer entry allocated before other allocated circular buffer entries, or an index of a circular buffer entry allocated after other allocated circular buffer entries;
a circuitry configured to
compare a first index of a first circular allocated buffer entry with a second index of a second allocated circular buffer entry;
compare the first index with the pointer;
compare the second index with the pointer;
if the first index is smaller than the second index and the pointer is smaller than or equal to both the first index and the second index, or the first index is smaller than the second index and the pointer is larger than or equal to both the first index and the second index, or the first index is larger than the second index and the pointer is smaller than or equal to the first index and the pointer is larger than or equal to the second index
assert a signal to indicate that the first allocated circular buffer entry was allocated before the second allocated circular buffer entry
otherwise,
deassert the signal to indicate that the second allocated circular buffer entry was allocated before the first allocated circular buffer entry.
9. The apparatus in claim 8 , further comprising
a table with plurality of table entries, wherein each table entry comprises an index field configured to store an index of an allocated circular buffer entry.
10. The apparatus in claim 9 , wherein the second index of the second allocated circular buffer entry is stored in an index field of a second table entry.
11. The apparatus in claim 10 , wherein the circuitry is further configured to flush the second table entry in response to determining that the signal is asserted.
12. The method in claim 11 , wherein the circular buffer stores not committed instructions in a processor.
13. The method in claim 12 , wherein the table stores not executed instructions in the processor.
14. The method in claim 13 , wherein the first allocated circular buffer entry corresponds to an instruction executed with exception.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US16/056,723 US20190087197A1 (en) | 2017-09-21 | 2018-08-07 | Method and apparatus for flushing instructions from reservation stations |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US201762561215P | 2017-09-21 | 2017-09-21 | |
| US15/821,884 US10095525B2 (en) | 2017-09-21 | 2017-11-24 | Method and apparatus for flushing instructions from reservation stations |
| US16/056,723 US20190087197A1 (en) | 2017-09-21 | 2018-08-07 | Method and apparatus for flushing instructions from reservation stations |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US15/821,884 Continuation US10095525B2 (en) | 2017-09-21 | 2017-11-24 | Method and apparatus for flushing instructions from reservation stations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| US20190087197A1 true US20190087197A1 (en) | 2019-03-21 |
Family
ID=61904542
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US15/821,884 Active 2037-12-12 US10095525B2 (en) | 2017-09-21 | 2017-11-24 | Method and apparatus for flushing instructions from reservation stations |
| US16/056,723 Abandoned US20190087197A1 (en) | 2017-09-21 | 2018-08-07 | Method and apparatus for flushing instructions from reservation stations |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US15/821,884 Active 2037-12-12 US10095525B2 (en) | 2017-09-21 | 2017-11-24 | Method and apparatus for flushing instructions from reservation stations |
Country Status (1)
| Country | Link |
|---|---|
| US (2) | US10095525B2 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10782976B2 (en) | 2018-05-07 | 2020-09-22 | Dejan Spasov | Issuing and flushing instructions from reservation stations using wrap bits and indexes |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US12436714B2 (en) * | 2022-08-30 | 2025-10-07 | Micron Technology, Inc. | Access request reordering for memory-based communication queues |
| US12436712B2 (en) * | 2022-08-30 | 2025-10-07 | Micron Technology, Inc. | Access request reordering across a multiple-channel interface for memory-based communication queues |
| US20240192961A1 (en) * | 2022-12-07 | 2024-06-13 | Akeana, Inc. | Processor instruction exception handling |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100169622A1 (en) * | 2008-12-31 | 2010-07-01 | International Business Machines Corporation | Processor register recovery after flush operation |
| US20160357554A1 (en) * | 2015-06-05 | 2016-12-08 | Arm Limited | Controlling execution of instructions for a processing pipeline having first and second execution circuitry |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7711935B2 (en) | 2007-04-30 | 2010-05-04 | Netlogic Microsystems, Inc. | Universal branch identifier for invalidation of speculative instructions |
| US8099586B2 (en) | 2008-12-30 | 2012-01-17 | Oracle America, Inc. | Branch misprediction recovery mechanism for microprocessors |
| US9489207B2 (en) | 2009-04-14 | 2016-11-08 | International Business Machines Corporation | Processor and method for partially flushing a dispatched instruction group including a mispredicted branch |
-
2017
- 2017-11-24 US US15/821,884 patent/US10095525B2/en active Active
-
2018
- 2018-08-07 US US16/056,723 patent/US20190087197A1/en not_active Abandoned
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100169622A1 (en) * | 2008-12-31 | 2010-07-01 | International Business Machines Corporation | Processor register recovery after flush operation |
| US20160357554A1 (en) * | 2015-06-05 | 2016-12-08 | Arm Limited | Controlling execution of instructions for a processing pipeline having first and second execution circuitry |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10782976B2 (en) | 2018-05-07 | 2020-09-22 | Dejan Spasov | Issuing and flushing instructions from reservation stations using wrap bits and indexes |
Also Published As
| Publication number | Publication date |
|---|---|
| US20180107487A1 (en) | 2018-04-19 |
| US10095525B2 (en) | 2018-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9996348B2 (en) | Zero cycle load | |
| US9524164B2 (en) | Specialized memory disambiguation mechanisms for different memory read access types | |
| US6073217A (en) | Method for detecting updates to instructions which are within an instruction processing pipeline of a microprocessor | |
| US10838729B1 (en) | System and method for predicting memory dependence when a source register of a push instruction matches the destination register of a pop instruction | |
| US9575754B2 (en) | Zero cycle move | |
| US9501284B2 (en) | Mechanism for allowing speculative execution of loads beyond a wait for event instruction | |
| US11068271B2 (en) | Zero cycle move using free list counts | |
| US20190087197A1 (en) | Method and apparatus for flushing instructions from reservation stations | |
| US8239661B2 (en) | System and method for double-issue instructions using a dependency matrix | |
| US12204911B2 (en) | Retire queue compression | |
| CN109564510B (en) | Systems and methods for allocating load and store queues at address generation time | |
| CN111095203B (en) | Inter-cluster communication of real-time register values | |
| US20190361703A1 (en) | Method and apparatus for renaming source operands of instructions | |
| US11995445B2 (en) | Assignment of microprocessor register tags at issue time | |
| US10782976B2 (en) | Issuing and flushing instructions from reservation stations using wrap bits and indexes | |
| US10846095B2 (en) | System and method for processing a load micro-operation by allocating an address generation scheduler queue entry without allocating a load queue entry | |
| US12112169B2 (en) | Register freeing latency | |
| US11520586B2 (en) | Method and apparatus for renaming source operands of instructions | |
| US20250383875A1 (en) | Processor, information processing apparatus, and method for controlling processor | |
| US10628164B1 (en) | Branch resolve pointer optimization | |
| CN114761919A (en) | Register write suppression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |
|
| STPP | Information on status: patent application and granting procedure in general |
Free format text: FINAL REJECTION MAILED |
|
| STPP | Information on status: patent application and granting procedure in general |
Free format text: RESPONSE AFTER FINAL ACTION FORWARDED TO EXAMINER |
|
| STPP | Information on status: patent application and granting procedure in general |
Free format text: ADVISORY ACTION MAILED |
|
| STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |