JPH0721240A - Device for timing considering arrangement - Google Patents

Abstract

PURPOSE:To shorten a critical path and to set the number of nets crossing a cut line to a minimum. CONSTITUTION:A critical path extraction device 101 extracting the critical path, a cut line generation device 103 generating the cut line, a merging opponent selection device 102 deciding the merging opponents of respective blocks from the connection degree of the respective blocks and critical path information, a merging device 104 merging the blocks from the merging opponents of the respective blocks, which are obtained by the merging opponent selection device, and a pairwise device 105 exchanging a pair so that the number of the nets crossing the cut line becomes the minimum are provided.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION This invention relates to layout design, and more particularly to timing sensitive placement.

[0002]

2. Description of the Related Art In the case where a net circuit operating speed limit is given which indicates a region, a block belonging to the region, a coupling relation between the blocks, and a coupling relation between blocks other than the region and blocks belonging to the region, A cut line that divides the area into two is created, and the problem of determining the position of the block that shortens the critical path length while minimizing the number of nets that cross the cut line is given.

A conventional placement device of this type is shown in FIG. This arrangement device is a cut line creation device 202 that creates a cut line, a merge partner selection device that determines a merge partner of each block from each block coupling degree, and a merge process that performs merge processing of blocks obtained by 201 and the merge partner selection device 201. A pairwise device 204 for performing pair exchange so that the number of nets crossing the device 203 and the cut line is minimized.
Was composed of. (Masato Edahiro
and Takeshi Yoshimura, "N
ew Placemmet and Global R
outingAlgorithms for Stan
dard cell Layouts ", Proc. 2
7th DA Conf pp. 642-645, (1
990)).

[0004]

Therefore, in the past, with respect to the above problem, the number of nets crossing the cut line is minimized by selecting a merge partner only by the degree of connection of each block and performing merge processing. As such, the pair exchange was performed, and the result that satisfied the critical path could not be obtained.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems and to provide a timing consideration placement device that shortens the critical path and minimizes the number of nets that cross the cut line.

[0006]

SUMMARY OF THE INVENTION A timing consideration placement device according to the present invention comprises a netlist showing a set of blocks belonging to a certain region of a circuit and a connection relationship between the blocks, an operating speed limit of the circuit, and the circuit. The internal delay value is input, the timing analysis is performed, the required time and the arrival time of each net are obtained, the margin of each net is obtained, and a critical path extraction device that extracts a critical path based on the margin, the netlist, A cut line generation device that inputs a layout area of the circuit, generates a cut line in the layout area, and determines a block position of the block, and a block position, the netlist, and the margin merged block By inputting the area limit, the degree of coupling between the blocks is obtained, and the block that satisfies the area limitation of the block is calculated by the degree of coupling and the margin. Calculating a merge cost between blocks and deciding a block to be merged with the delay consideration merge partner selection device and the netlist as inputs, and merging the blocks decided by the delay consideration merge partner selection device to update the netlist From a merging device, a netlist updated by the merging device, the cutline, and a pairwise device that inputs the block position and performs block exchange processing so that the number of nets crossing the cutline is minimized. It is characterized by being configured.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an embodiment of the timing consideration placement device of the present invention. The overall flow of the present invention is shown in FIG.

A netlist (representing a set of blocks and connections between blocks (net)) as an input, and a circuit operation speed limit (how many ns must operate),
When the internal delay value of the circuit is given, the critical path extraction device 101 performs timing analysis and obtains the required time and arrival time of each net. Timing analysis is the latest arrival time for each net from the input to the output, that is, the arrival time, and the delay from the output to the input until the delay from the input to the net. The time to be expressed, that is, the required time is obtained. Arrival time (request time) is the internal delay (and wiring delay) of the block on the path from the input (output) to the net.
It can be obtained by adding (subtracting). Figure 7
The result of delay analysis when the operation speed limit of the circuit is 22 is shown in (a). (Required time and arrival time is Robert
B. Hitchcock, Gordon L. et al. Smit
h, David D. Cheng, “Timing A
analysis of Computer Hardw
are ", IBM J. Res. Development. vo
11 January 1982, pp. 100-10
5, (1982). Can be obtained by After that, the slack (request time-arrival time) of each net is obtained, and the margin of each net is obtained by performing the slack distribution processing (12-1). What is slack distribution processing?
Take out a net whose slack is not 0 and divide the slack value into nets on the path that passes through that net,
This is a process of allocating as a margin. When the margin is allocated, the margin is added to the wiring delay, and the delay analysis is performed again. After that, a slack distribution process is performed. This process is repeated until the slack value of all nets becomes almost zero. (The margin is, for example, Peter S. Hauge,
Ravi Nair, Ellen J, Yoffa,
"Circuit Placement for Pr
editable Performance ", Pr
oc. ICCAD-87. , Pp. 88-91, (19
87). Can be obtained by ) FIG. 7B shows the result of distribution of the margin. A net with a negative margin becomes a net on the critical path. The smaller the margin, the tighter the limit. Here, the critical path is a path (a series of nets) having a negative request time-arrival time.

Next, the cut line creation device 103 inputs a netlist and a placement area, creates a cut line in the placement area, and determines whether each block belongs to the right or left of the cut line. (12-2).

Next, in the delay-considered merge partner selection device 102, the block position (whether the block belongs to the right or left of the cutline) and the netlist (updated by the merge process) determined by the cutline creation device 103. May be a netlist) and the critical path extraction device 10
Using the margin of each net obtained in 1 and the area limit of the merged block as an input, the degree of coupling between each block (how many nets exist between block i and block j) is first obtained. In the example of FIG. 8, the coupling degree between the block A and the block B is 1, and the coupling degree between the block B and the block C is 2. Next, the merge cost is calculated between the blocks satisfying the block area restriction based on the coupling degree and the margin, and the block to be merged is determined. (FIGS. 11 and 12-3).

A method for obtaining the merge cost cost (i, j) when the blocks i, j are merged will be described below. The degree of connection of the net between the block i and the block j is defined as nonect (i, j). A set of nets connecting block i and block j

[0012]

[Equation 1]

[0013] Afford the margin of net i
(I). In the delay consideration merge partner selection device, a set of all blocks (blocks i, j) belonging to the same area
On the other hand, if the sum of the areas of the blocks i and j is less than the limit value, the margin cost:

[0014]

[Equation 2]

The merge cost between the blocks is determined by calculating Where α and β are arbitrary variables.

In the blocks to be merged, the merge costs are sorted in descending order, and the value becomes the largest.

The merging device 104 inputs the netlist and the merge target block determined by the delay consideration merging partner selection device 102 and the target number of blocks, and performs merging processing (12-4). The merging process is a process of updating the netlist with two input blocks as one block. FIG. 9 shows a case where blocks a and b are merged to form block c. If the number of blocks has reached the target number of blocks or the merge target block determined by the delay consideration merge partner selection apparatus is empty, control is performed to the pairwise apparatus 105, and otherwise, control is performed to the delay consideration merge partner selection apparatus 102. Moving.

In the pairwise device 105, the net list updated by the merging device and the cut line and block position obtained by the cut line creation device are used as inputs, and one block is taken out from each side of the cut line, and a net that crosses the cut line is taken. Block exchange processing (pairwise) so that the number of blocks is minimized (the number of nets that cross the cut line is minimum and the number of critical paths that cross the cut line is minimum) (FIGS. 10 and 12-6). ).
Pairwise includes (Masato Edahiro a
nd takeshi Yoshimura, “New
Placemmet and Global Rou
toning Algorithms for Stand
ard cell Layouts ”. Proc. 27
th DA Conf pp. 642-645, (19
90)) can be used.

FIG. 14 shows the pass passage density. In this example, a path passing through nets 1, 3, 5 and nets 2, 4,
There is a path through 5. Block 1 and block 5
When the constraint between the two is severe, shortening the net 5 can shorten two paths, but shortening the net 1 can shorten only one path. In this way, by shortening the net, many paths can be shortened, that is, by shortening the net through which many paths pass, a certain net causes multiple critical paths to remain without satisfying the delay. It is possible to prevent it.

In the example of FIG. 3, the number of paths passing through the net 1 is one, the number of paths passing through the net 2 is one, and the number of paths passing through the net 3 is one. , Net 4
The number of paths passing through the network is one, and the number of paths passing through the net 5 is two. Therefore, the net 5 should be shortened.

FIG. 13 is a block diagram of a delay consideration placement device according to a second embodiment of the present invention.

The overall flow is shown in FIG.

In the path passage density extraction device, a net list (a set of blocks and a net representing a connection between blocks) is input and the number of paths passing through each net is obtained. First, the number of paths from the starting point to each net is obtained. (Level each net by the number of nets from the starting point (Fig. 15). Next, add the number of paths from the starting point of the nets that have entered the blocks in which the nets are in order from the lowest level, It is assumed that the number of paths from the start point of the net is 1. In the net 5 of FIG.
Is the number of passes from the starting point. If the block from which the net is output is the block at the start point, the number of paths from the start point of the net is set to 1. ) Next, count the number of passes from the end point to each block. (The same processing is performed from the end point side as on the start point side. (FIG. 16).) Next, in each block, the number of paths from the start point to each block is multiplied by the number of paths from the end point to each block to obtain each Determine the number of passes through the block (FIG. 17).

In the critical path extraction device, when the netlist, the operating speed limit of the circuit (how many ns it has to operate), and the internal delay value of the circuit are given, the timing Analyze and calculate the required time and arrival time of each net. Timing analysis is the time to arrive at each net latest from the input to the output, that is, the arrival time, and how long it takes to reach the net from the input to the net from the output to the input. The time to be expressed, that is, the required time is obtained. The arrival time (request time) can be obtained by adding (subtracting) the internal delay (and placement delay) of blocks on the path from the input (output) to the net. FIG. 7A shows the result of delay analysis when the operation speed limit of the circuit is 22. (Required time and arrival time is Robert
B. Hitchcock, Gordon L. et al. Smit
h, David D. Cheng, “Timing A
analysis of Computer Hard-
were ", IBM J, Res. Development. v.
ol1 January 1982, pp. 100-1
05, (1982). Can be obtained by After that, the slack (request time-arrival time) of each net is calculated, and the margin of each net is calculated by performing the slack distribution process. With slack distribution processing, slack is 0
This is a process of taking out a net other than the above and dividing the slack value into nets on a path passing through the net as a margin. When the margin is allocated, the margin is added to the wiring delay, and the delay analysis is performed again. After that, a slack distribution process is performed. This process is repeated until the slack values of all the nets become almost zero. (The margin is, for example, Peter S. Hauge, RaviN
air, Ellen J, Yoffa, "Circui"
t Placement for Predicttab
le Performance ”, Proc. ICCA
D-87. , Pp. 88-91, (1987). Can be obtained by ) FIG. 7B shows the result of distribution of the margin. A net with a negative margin is a net on the critical path. The smaller the margin, the tighter the limit. Here, the critical path is the required time −
A path (a series of nets) with a negative arrival time.

Next, in the cut line creation device, the net list and the placement area are input, and a cut line is created in the placement area, and it is determined whether each block belongs to the right side or the left side of the cut line.

In the pairwise device, the number of paths passing through each net obtained by the pass passage density extraction device, the margin of each net obtained by the critical path extraction device, the cut list obtained by the net list and the cut line creation device And the block position as input, extract blocks one by one from the cut line area so that the number of nets that cross the cut line is minimized (the number of nets that cross the cut line is the minimum, and the critical path that crosses the cut line is Block exchange processing (pairwise) is performed so that the number of nets with the minimum number and high pass passage density across the cut line is minimized. For pairwise, (Masato E
dahiro and takeshiYoshimu
ra, "New Placemmet and Glo
balRouting Algorithms for
Standard cell Layouts ".P
roc. 27th DA Conf pp. 642-6
45, (1990)) can be used. That is, the cost exch for exchanging block i and block j
When the set of nets connected to the block i and the block j is E2, the age (i, j) is defined as follows.

[0027]

Here, "aford (l)" represents the margin of the net, and "form (l, i)" is the number of nets that cross the cut line that changes by moving the pin i of the net l to the opposite region. And path (l) is
The degree of influence of the net 1 on the path is shown. Also, P (l) is
It represents a set of pins connected to the net l. α is an arbitrary constant.

Exchange (i, j) is calculated for all sets of blocks, and the one with the highest cost is the target of exchange and the exchange is performed. This exchange processing is performed until the number of nets that cross the cut line, the number of critical paths that cross the cut line, and the number of nets that have high pass passage density that cross the cut line do not decrease due to the exchange processing.

[0030]

[Effect] FIG. 3 is a result of extracting a critical path by creating a cut line for one input example with respect to the embodiment of FIG. In FIG. 3, circles numbered 1, 2, ... Represent blocks (3-1). The line connecting the blocks represents a net (3-2). The thick line represents the critical path (3-
3). 3-4 represents a cut line.

In the conventional method, when the merge partner is selected, it is determined only by the degree of connection, so that block 1 and block 2 are combined, block 3 and block 4 are combined,
Blocks 5 and 6 are combined, and blocks 7 and 8 are combined (FIG. 4). If pairwise is performed in this state, the number of nets that cross the cut line cannot be reduced any more. At this time, the critical path crosses the cut line three times and is long.

However, according to the present invention, when a merge partner is selected, not only the degree of coupling but also the critical path information extracted by the critical path extraction device is used to select the merge partner for delay consideration by the selection device. Block 1 and block 4 are combined, block 2 and block 3 are combined, block 6 and block 7 are combined, and block 5 and block 8 are combined (FIG. 5). On the other hand, when pair exchange is performed, the result of FIG. 6 is obtained. In this case, the critical path crosses the cut line only once, and a short critical path is obtained.

As described above, according to the present invention, the number of nets crossing the cut line can be minimized, and at the same time,
The critical path can be shortened.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of the present invention.

FIG. 2 is a diagram showing a conventional placement device.

FIG. 3 is a diagram showing a result of extracting a critical path.

FIG. 4 is a diagram showing a state in which blocks are combined according to a conventional technique.

FIG. 5 is a diagram showing a combined state of blocks according to the present invention.

6 is a diagram showing a result of performing a pair exchange on the block of FIG.

7A is a diagram showing a result of delay analysis in the case where the operation speed limit of the circuit is 22. FIG. 7B is a diagram showing a result of distributing the margin.

FIG. 8 is a diagram showing an example of combining blocks.

FIG. 9 is a diagram showing a merge result.

FIG. 10 is a diagram showing a block exchange process.

FIG. 11 is a diagram showing a flow of determining blocks to be merged.

FIG. 12 is a diagram showing the overall flow of timing-considered placement according to the present invention.

FIG. 13 is a diagram showing a second embodiment of the present invention.

FIG. 14 is a diagram showing pass passage density.

FIG. 15 is a diagram showing the number of paths from the starting point to each net.

FIG. 16 is a diagram showing the number of paths from the end point to each net.

FIG. 17 is a diagram showing the number of paths passing through each net.

FIG. 18 is a diagram showing a flow of a second embodiment.

[Explanation of symbols]

 101 Critical Path Extractor 102 Delay Considering Merge Partner Selection Device 103 Cut Line Creation Device 104 Merge Device 105 Pairwise Device 201 Merge Partner Selection Device 202 Cut Line Creation Device 203 Merge Device 204 Pairwise Device

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] June 22, 1994

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 2

[Correction method] Change

[Correction content]

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0006

[Correction method] Change

[Correction content]

[0006]

A timing-considered placement apparatus according to the first aspect of the present invention comprises a netlist representing a set of blocks belonging to a certain area of a circuit and a connection relation between the blocks.
The operation speed limit of the circuit and the internal delay value of the circuit are input, timing analysis is performed, the required time and arrival time of each net are obtained, the margin of each net is obtained, and the critical path is extracted by the margin. Input the critical path extraction device, the placement area of the netlist and the circuit,
A cut line generation device that generates a cut line in the arrangement area and determines a block position of the block, and inputs the block position, the netlist, the area limit of the merged block, and the block, The connection cost between the blocks that determines the block to be merged is calculated by calculating the coupling degree between the blocks, calculating the merge cost between the blocks that satisfy the area limitation of the block based on the coupling degree and the margin, and inputting the netlist. And a merge device for merging blocks determined by the delay-considered merge partner selection device to update a netlist, a netlist updated by the merge device, the cutline, and the block position, Pairwise equipment that performs block exchange processing so that the number of nets that cross the cut line is minimized. To characterized in that it is composed of a. The timing consideration placement apparatus of the second invention inputs a set of blocks belonging to a certain area of the circuit and a netlist showing a connection relation between the blocks, an operation speed limit of the circuit, and an internal delay value of the circuit. , Perform timing analysis,
The required time and arrival time of each net are obtained, the margin of the net is obtained, and the critical path extraction device that extracts the critical path based on the margin, and the number of paths that pass through each net by inputting the net list (path passage Density), a pass passage density extraction device for determining the density, a cut line generation device for inputting the arrangement area of the netlist and the circuit, generating a cut line in the arrangement area, and determining a block position of the block, and the net. A list, the cut line, and the block position are input, and the number of nets that cross the cut line is minimum, the number of critical paths that cross the cut line is minimum, and the net with a high pass passage density that crosses the cut line is minimum. It is characterized by comprising a pairwise device for exchanging pairs as described above.

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0007

[Correction method] Change

[Correction content]

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an embodiment of the timing consideration placement device of the first invention. In addition, the entire flow of the present invention is shown in FIG.
Shown in.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0021

[Correction method] Change

[Correction content]

FIG. 13 shows an embodiment of the timing consideration placement device of the second invention.

[Procedure Amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0033

[Correction method] Change

[Correction content]

As described above, according to the present invention, the number of nets that cross the cut line can be minimized, and at the same time, the critical path can be shortened. Also,
In the second aspect of the present invention, since the exchange processing is performed with priority given to the net having the highest density, it is possible to prevent a plurality of critical paths from remaining without satisfying the delay due to a certain net. It also prevents the concentration of blocks.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location H01L 27/04 8832-4M H01L 27/04 A

Claims (2)

[Claims] 1. A netlist showing a set of blocks belonging to a certain area of a circuit and a connection relationship between the blocks, an operation speed limit of the circuit, and an internal delay value of the circuit are input to perform timing analysis, A critical path extraction device that obtains the required time and arrival time of each net, obtains the margin of each net, and extracts the critical path based on the margin, and inputs the arrangement area of the netlist and the circuit, and the arrangement area A cut line generation device that generates a cut line and determines the block position of the block, and inputs the block position, the netlist, the area limit of the merged block, and the coupling between the blocks. Block, the merge cost between blocks satisfying the area restriction of the block is calculated by the coupling degree and the margin, and the block for merging is calculated. A merge device for selecting a merge partner for determining delay, a merge device for inputting the netlist, and merging the blocks determined by the merge device for selecting delay consideration to update the netlist, and a net updated by the merge device. A timing-considered placement device comprising a list, the cut line, and a pairwise device that inputs the block position and performs block exchange processing so that the number of nets that cross the cut line is minimized. . 2. A path passage density extraction device for obtaining the number of paths passing through each net, a critical path extraction device for extracting critical paths, a cut line creation device for creating cut lines, and a minimum number of nets crossing the cut lines. A timing-considered placement device including a pairwise device that performs pair exchange so that the number of critical paths that cross the cut line is the minimum and the number of nets that have a high pass passage density that cross the cut line is the minimum.