Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
~ ~ 3 ~
VIDEO M~MORY SYSTEM WITH INTERMEDIATE BUFFER
Inventor: Chien-Chih Yu
BACKGROUND OF THE INVENTION
F ld of the Invention
The present invention relates to a computer
display memory system, and more specifically, to such a
system including an intermediate buffer which
substantially reduces the memory accessing time of an
external processing unit by eliminating the need to
insert wait cycles.
Description of the Prior Art
Many computer display systems today include a
video memory system for storing the data which
represents the image to be displayed. In Fig. 1, a
typical video memory system 20 is illustrated,
comprising a video memory 12, a video display control
unit (VDCU) 14, and a video pxocessing means 15. The
external processing unit (CPU) 10 sends data and control
siynals to the video memory 12 to update the information
stored therein. The VDCU 14 periodically causes the
memory 12 to output data to the video processing means
15. Video processing means 15 then converts the data
into signals that the external display means 16 can
recoynize, and outputs these signals to the display
means 16 in response to the control signals on lines 17
from VDCU 14. In this manner, the information on the
display screen is periodically refreshed. The video
memory system 20 may either be incorporated directly
into the motherboard of a host computer (not shown) or
may take the form of an add-on card. Its general
. : :
:, , :: : : .
operation is substantially the same for either
configuration.
As shown in Fig. 1, both the CPU 10 and the
VDCU 14 of a typical video memory system have direct
access to the memory 12. As a result, a method is
needed to designate the time slots during which the CPU
lO and the VDCU 14 may respectively access the memory
12. Otherwise, they may try to access the memory
simultaneously, an undesirable result. Time allocation
is commonly achieved through the use of a time slot
reference signal such as that shown in Fig. 2A wherein
CPU 10 may only access the memory 12 between times t
and t3 and between times t5 and t6. The time slots
between t3, tS and between t6, t7 are allocated to VDCU
l~. This type of procedure inherently introduces delays
into the system. To illustrate this point, suppose that
the CPU 10 pulls its read/write line low (not shown in
Fig. 1) when it is ready to write a set of data to
memory 12. As long as the read/write line is pulled low
at the beginning o~ a CPU time slot (times t1 or tS f
Fig. 2A), the write operation is performed without delay
(Fig. 2B). However, for the general case where the
read/write signal is not synchronized with the beginning
of a CPU time slot, the CPU must insert wait cycles in
order to properly access the memory. For example, if
the CPU wishes to write to the memory during a VDCU time
slot as shown in Fig. 2C, it must insert wait cycles
between times t4 and t6 to hold the read/write line low
in order to ensure that data is written properly into
the memory 12. Also, if the read/write line is pulled
low during a CPU time slot but the remaining tlme is
insufficient to complete the write operation (Fig. 2D),
the CPU must insert wait cycles to hold the read/write
line low until the end of the next CPU time slot t6.
The introduction of these wait cycles causes substantial
delays. Thus, there exists a need for a video memory
- . . ,:. - : ,
'; ! : ;; , ~,,
`, :,, '; , " ~',, ' ' :
' " :' ' ' ' .~
3 r~ ~
system that allows the CPU to perform its function
without the need to insert wait cycles.
SUMMARY OF THE INVENTION
The present invention eliminates the need to
insert wait cycles for write operations by interposing
an intermediate buffer, between the CPU and the video
memory, which functions to intercept the address, data,
and control signals from the CPU. For read operations,
the buffer simply relays the intercepted address signals
directly to the video memory. For write operations,
however, the buffer stores the address and data sent by
the CPU and retains this information within its memory
until the beginning of the next CPU time slot, at which
time it accesses the video memory via the stored address
and writes the data thereto. The invention allows the
CPU to perform write operations regardless of whether
the time slot is a CPU time slot, thereby, obviating the
need to introduce wait cycles. It is true that wait
cycles may still need to be inserted for read
operations, but since eighty percent of the operations
performed by the CPU are write operations and only
twenty percent are read operations, the present
invention nonetheless saves the CPU a substantial amount
of time. The invention is likewise applicable to avoid
wait cycles during read operations as well.
A preferred embodiment of the invention
comprises a memory means, a VDCU connected to the memory
means, a video means connected to both the memory means
and the VDCU, and an intermediate buffer interposed
between an external CPU and the memory means. The
external CPU operates to generate a read/write control
signal and address and data signals. The intermediate
buffer intercepts these signals, and if a write
operation is desired, the address and data signals are
stored in the buffer. The buffer then monitors the time
:, ~
':
slot reference signal until it detects a CPU time slot,
at which time, the address and data signals are sent to
the memory means. The memory means then stores the data
signals at the addresses indicated by the address
signals. For read operations, the intermediate buffer
simply relays the address signals to the memory means.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is a block diagram of a prior art video
memory system.
Fig. 2A shows a typical time slot reference
signal employed to designate time slots during which the
CPU and the VDCU may access the video memory used in the
system of Fig. 1.
Figs. 2B-2D show the possible time variations
between the read/write signal from the CPU and the time
slot reference signal in the system of Fig. 1.
Fig. 3 is a functional block diagram of a
video memory system to illustrate the invention.
Fig. 4 shows a preferred embodiment of the
system of Fig. 3.
DETAILED DESCRIPTION OF THE INVENTION
A block diagram depicting the basic components
of the invention is shown in Fig. 3, comprising
intermediate buffer 34, random access video memory 38,
VDCU 42, and video processing means 39. Video memory
38, VDCU 42, and video processing means 39 are of
regular construction and are found in many typical prior
art computer display systems; thus, they will not be
discussed in detail herein. The external CPU 30 sends
data, representing a portion of the image to be
displayed, to the video memory 38 where the information
is stored. This continues until the entire image is
stored within the video memory 38. Thereafter, the main
function of CPU 30 is to update the stored image. To
. ' ' : " ~,, "
,:
"
~ ~ 3 .~ J ~
update a portion of the video memory 38, the CPU sends
out an address on lines 31 and a read/write control
signal on line 33 indicating the nature of the
operation. For write operations, CPU 30 also sends out
data signals on d~ta lines 32.
Address lines 31, data lines 32, and control
line 33 are all connected to intermediate buffer 34.
A time slot reference signal 52 such as that shown in
Fig. 2A, dictating the times during which the CPU 30 and
the VDCU 42 may access video memory 38, is also supplied
to intermediate buffer 34. Depending on the status of
the read/write line 33, buffer 34 performs different
functions.
For read operations, buffer 34 receives a read
address on lines 31 and a read signal on line 33. The
internal logic of buffer 34 reads the signal on control
line 33 and, realizing that a read operation is desired,
relays the read address to memory 38 via memory address
lines 35. Also, the read/write control signal is
relayed to memory 38 via memory control line 37. Data
from the addressed location in the video memory 38 is
then put onto the memory data lines 36, said data
travelling from the memory 38 through buffer 34 to
eventually arrive at CPU 30 via CPU data lines 32.
Since the intermediate buffer 34 only acts as a relay
for read operations, CPU 30 is respon~ible for ensuring
that a read operation only takes place during a time
slot which has been allocated to the CPU. Otherwise,
bus conflict with the VDCU 42 may result. Thus, wait
cycles may need to be inserted for read operations.
However, since only about twenty percent of the CPU's
operations are read operations, the delaying effect of
the wait cycles is not significant.
During write operations, the CPU 30 sends to
the intermediate buffer 34 a write address on lines 31,
write data on lines 32, and a write control signal on
, ~ , .. . .
r
line 33. The internal logic of buffer 34 receives the
write control signal and responds by initiating a series
of operations. The internal logic first scans the
memory portion of the buffer 34 to ascertain which
locations may be written into. Once an appropriate
location is found, the address and data signals on lines
31 and 32, respectively, are stored. Thereafter, the
internal logic checks the status of the time slot
reference signal 52 to ascertain whether it indicates a
CPU time slot. If so, the stored address and data
signals are put onto lines 35 and 36, respectively, and
a write control signal is sent on line 37. This results
in the data being written into the appropriate address
in video memory 38. However, if signal 52 indicates a
VDCU time slot or if there is insufficient time to
complete a write operation to the memory, the buffer 34
will wait until the beginning of the next CPU time slot
to write the data to the memory 3~. Since several sets
of addresses and data may be stored by the buffer 34
before the next CPU time slot, the intermediate buffer
34 preferably outputs stored data in a first-in-first-
out (FIFO) fashion so that no data remains within the
buffer 34 for an extended period of time. Once the data
in memory 38 is updated during a CPU time slot, the
VDCU, during the next VDCU time slot, causes the updated
data to be transferred from the memory 38 to the video
processing means 39. The video processing means in turn
converts the data into video signals and~ in response to
the control signals from VDCU 42 on lines 53, outputs
the video signals to an external display means 46,
thereby, refreshing the display screen with updated
data.
Note that due to the presence of intermediate
buffer 34, the CPU 30 need not generate any wait cycles.
It is free to perform write operations at any time
without regard to the status of the time slot reference
. .~ , :,, ,
signal 52. The waiting formerly performed by the CPU is
now done by the intermediate buffer. This significantly
reduces the memory access time of the CPU and allows it
to run more efficiently.
A preferred embodiment of intermediate buffer
34 is illustrated in more detail in Fig. 4, comprising
buffer register means 60, control means 70, and a
read/write multiplexer 80. Buffer register means 60
contains a plurality of storage registers 62 with each
register capable of storing a set of address and data
signals. Register means 60 receives as input CPU
address lines 31, CPU data lines 32, read/write control
line 33, and several control lines 65 from control means
70. In response to a write signal on line 33, register
means 60 stores within one of its registers 62 the
address and data siynals appearing on lines 31 and 32.
This occurs irrespective of the status of the time slot
reference signal 52. The specific register in which the
data is stored is dictated by the control signals on
lines 65 sent by control means 70. The data remains
stored until a WRITE DATA signal on lines 65 instructs
the register means 60 to output data from one of its
registers. Again, the control signals on lines 65
specify which register is to be accessed.
The read/write multiplexer 80 is attached to
two sets of input lines, one set from register means 60
and the other set from the CPU 30. Multiplexer 80
receives from register means 60 address lines 74 and
data lines 75. Multiplexer 80 also receives a write
control signal on line 7~ from control means 70. From
the CPU 30, multiplexer 80 receives address lines 31,
data lines 32, and read/write line 33. In response to
a MUX control signal from control means 70 on line 68,
multiplexer 30 behaves as a switch to selectively
connect one of the sets of input lines to the address
lines 35, data lines 36, and read/write line 37 of the
. : . ;
.' ',;' :,
:, , : ~
,: .
, .
~3 3 ~
video memory 38. Thus, by using line 68, control means
70 can control which device (the CPU or the register
means) is connected to the video memory 38.
The control means 70 receives as inputs the
CPU read/write control signal on line 33, buffer
register status signals on lines 81, and the time slot
reference signal on line 52. From these inputs signals,
the internal logic of the control means 70 generates the
appropriate signals to control the register means 60 and
the read/write multiplexer 80. To ensure that a
register 62 containing fresh data is not overwritten,
control means 70 must monitor the status of the
registers 62. This may be achie~ed through the use of
a status register 66 which receives, as inputs, buffer
register status signals on lines 81 from register means
50. Register 66 preferably contains as many bits 67 as
there are registers 62 in the register means 60 so that
each bit 67 corresponds to a specific register ~2.
Consequently, each bit may be usedl as a flag to indicate
whether the register contains fresh data. Whenever a
register is written into, an appropriate signal is sent
on lines 81 instructing control means 70 to set the flag
corresponding to that register to indicate that the
register is full. Conversely, if data is read out of a
register, its corresponding flag is cleared by control
means 70 in response to an appropriate signal on lines
81 from register means 60. Thus, by scanning the status
register 66, control means 70 may quickly ascertain
which register may be written into.
Control means 70 is also equipped with
sequential logic means 72 so that it may perform the
FIF0 function. Each time data is written into register
means 60, sequential logic 72 records the location of
the register as well as the sequence in which each
register 62 was loaded. This allows the data to be
outputted in the same sequence as it was loaded. In
,
.
2~ 3
other words, sequential logic 72 makes it possible to
perform the desired first-in-first-out function. To
output data, control means 70 checks the status of the
time slot reference signal 52 to determine whether the
status indicates a time slot allocated for the CPU. If
so, a WRITE DATA signal is issued on one of the control
lines 65 instructing the register means 60 to output a
set of address and data signals. Sequential logic means
72 also sends out several control signals on lines 65
controlling the sequence in which the registers 62 are
selected. Control means 70 then sends a ~JX control
signal on line 68 instructing multiplexer 80 to connect
the output lines of register means 60 with the input
lines of memory 38. The data on lines 75 is thus
written into memory 38. For read operations, control
means 70 instructs multiplexer 80 via line 68 to connect
CPU 30 with the memory 38, thereby, allowing the CPU to
read data from memory 38 through multiplexer 80.
Although the invention has been described with
reference to a specific embodiment, it should not be
construed to be so limited. It will be clear to one of
ordinary skill in the art with the benefit of this
disclosure that many modifications may be made without
departing from the spirit o~ the inventionO Thus, the
video processing logic 15 and display 16 may, for
example, be combined into a single unit if desired. The
invention should not be limited by the specific
embodiments used to illustrate them but only by the
scope of the appended claims.
,~:
. .