JP5471696B2 - Image processing apparatus and image processing program - Google Patents

Description

  The present invention relates to an image processing apparatus that generates output image data based on original image data, and more particularly to an image that generates the output image data by processing the original image data in a plurality of processing units for each processing unit. The present invention relates to a processing apparatus and the like.

  As an example of an image processing apparatus, a personal computer (herein referred to as RIP-PC) that functions as a RIP (Raster Image Processor) is known. The RIP-PC performs image processing on original image data such as a drawing command and bitmap image data, and generates print data (output image data) to be supplied to a printer (image output device). The image processing includes processing for generating bitmap data based on a drawing command and various conversion processing for converting the color and resolution of the bitmap data.

  The speed at which the RIP-PC supplies print data to the printer depends on the processing speed of the processor in the RIP-PC and the data read / write speed with respect to the memory. In recent years, the printing speed of printers has been improved, such as printing in units of pages. Along with this, an improvement in the supply speed of print data is desired.

  In order to meet such a demand, a method of parallelizing image processing processors has been proposed. Patent Documents 1 and 2 describe parallel image processing processors. In RIP-PC, a configuration is known in which a plurality of GPUs (Graphic Processing Units) are provided in parallel operation in addition to a CPU (Central Processing Unit). In such RIP-PC, for example, the CPU divides original image data for one page into predetermined processing units, supplies the GPUs to each GPU, and causes each GPU to perform image processing to generate print data. Then, the print data generated by each GPU is collected, and the print data for one page is supplied to the printer.

JP-A-5-266201 JP 2002-207711 A

  By the way, in the image processing in the RIP-PC, the transfer time of the original image data from the memory of the RIP-PC to each GPU, the processing time in each GPU, and the transfer time of print data from each GPU to the memory are as follows. Occur.

  Now, let's look at the transfer time between the original image data and the print data. Since the data transfer time depends on the amount of data to be transferred, the data amount of the print data is constant depending on the number of pixels of the image that can be output by the printer. Therefore, the transfer time is constant to some extent. On the other hand, since the amount of original image data varies depending on the configuration of the original image, the transfer time of the original image data varies depending on the configuration of the original image.

  Also, when looking at the image processing time in each GPU, a difference occurs depending on the processing content. For example, a process for generating print data from original image data based on a drawing command has a longer processing time than a process for generating print data from original image data such as bitmap data, so that the processing time becomes longer.

  Then, as long as the total of the data transfer time per page and the GPU image processing time is equal to or shorter than the required printing time of the printer, the print data can be supplied to the printer with a margin. However, if the transfer time of the original image data increases or the load of image processing increases, the sum of the data transfer time and the image processing time exceeds the required printing time. Print data cannot be supplied. Then, when executing a job for printing a plurality of pages continuously, there is a problem that blank paper is output from the printer without supplying print data. Alternatively, in order to avoid this, print data may be spooled to some extent in the memory of the RIP-PC or the memory in the printer. It is also possible to stop the printer when the spooled print data amount has decreased to some extent. However, this may lead to a printing delay.

  Accordingly, an object of the present invention is to provide an image processing apparatus capable of supplying output image data in time for the operation speed of the image output apparatus regardless of the configuration of the original image data transferred to the GPU and the content of the image processing. Is to provide.

  In order to achieve the above object, according to a first aspect of the present invention, an image processing device that generates output image data for an image output device based on original image data for each page has a virtual storage area. A first storage unit; a first storage region where stored data is saved in the virtual storage region; and a second storage region where the data is not saved. A second storage unit having a high speed of reading and writing; a plurality of processing units that process the original image data for each page in a processing unit smaller than the page; and generating the output image data; and the first storage area A transfer unit that alternately transfers the stored original image data for each page to the plurality of processing units for each processing unit to generate the output image data; and the original image data for the processing unit is the processing unit Switch to A transfer time deriving unit that derives a transfer time is provided and a processing time deriving unit that derives the processing time during which the output image data is generated by the processing unit. The transfer unit is configured to transfer the original image data from the first storage area to the processing unit based on a ratio of the transfer time to the processing time. When generating the output image data based on the original image data, the original image data of the processing unit for transfer to the processing unit is transferred from the first storage area to the second storage area, and One of the second transfer modes for transferring the original image data from the second storage area to the processing unit is selected and executed.

  According to the above aspect, in the second transfer mode, since the second storage area in which data is not saved is used in the virtual storage area, the first transfer using the second storage area in which data is saved is performed. Data transfer can be performed faster than in the mode. Therefore, output image data can be supplied in time for the operation speed of the image output apparatus regardless of the configuration of the original image data transferred to the processing unit and the content of the image processing. Further, when executing the second transfer mode, when the processing unit generates the output image data based on the original image data, the original image data of the processing unit to be transferred to the processing unit is stored. Transfer from the first storage area to the second storage area. Therefore, since the original image data transmission of the next processing unit can be executed at an early timing, the time required for the image processing can be reduced correspondingly, and the delay in the supply of the output image data can be reliably prevented. Further, the transfer time deriving unit derives a transfer time for transferring the original image data of the processing unit to the processing unit, and the processing time deriving unit derives a processing time in the processing unit. Therefore, since the first and second transfer modes can be controlled according to the processing speed of the transfer unit and the processing unit, it is possible to prevent a delay in the output image data supply without depending on the operating environment.

  According to another preferred aspect of the above aspect, the transfer unit is configured to perform the first transfer mode when the ratio is a first ratio, and the first transfer mode when the ratio is a second ratio that is greater than the first ratio. 2 transfer mode is selected. Therefore, when the ratio of the data transfer time to the image processing time is large, that is, when the amount of data to be transferred is large, the data transfer time can be shortened by the second transfer mode, and the output image data supply delay is ensured. Can be prevented.

  According to another preferred aspect of the above aspect, when the transfer unit executes a single job that outputs output image data of a plurality of pages to the image output device, the ratio exceeds a threshold value. Switching from the first transfer mode to the second transfer mode. In addition, when the ratio becomes equal to or less than the threshold, the second transfer mode is switched to the first transfer mode. Therefore, the job can be completed without interrupting the output from the image output apparatus.

It is a figure explaining the structure of the image processing apparatus in this embodiment. It is a figure which shows typically the data transfer between RAM20 and GPU14_1, 14_2. It is a sequence diagram explaining a normal transfer mode and a high-speed transfer mode. It is a figure explaining the sequence of normal transfer mode. It is a figure explaining the processing sequence of high-speed transfer mode. FIG. 10 is a flowchart illustrating an operation procedure when the personal computer 2 executes a print job.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the technical scope of the present invention is not limited to these embodiments, but extends to the matters described in the claims and equivalents thereof.

  FIG. 1 is a diagram illustrating the configuration of an image processing apparatus according to this embodiment. Here, the image processing apparatus includes a personal computer 2 that functions as a RIP. Further, the image output device is constituted by a printer 4. The personal computer 2 generates original image data for each page, performs image processing on the original image data, and generates print data (output image data). Then, the personal computer 2 supplies the generated print data to the printer 4. The printer 4 is, for example, a page printer. The printer 4 prints an output image for each page on printing paper with ink or toner based on the print data.

  Each part of the personal computer 2 will be described.

  The CPU 6 operates according to a startup program stored in a ROM (Read Only Memory) 8 and a control program / processing program stored in the hard disk 10. The control program includes an operating system OS. The processing program includes an application program APP such as a drawing / document creation program for generating original image data, and a printer driver program PDR for performing image processing.

  The RAM 20 temporarily stores data for various calculations when the CPU 6 executes the control program / processing program. The hard disk 10 also has a virtual storage area for temporarily saving data stored in the RAM 20. In this embodiment, the CPU 6 that performs the operations described later in accordance with the printer driver program PDR is the “transfer unit”, the “transfer time deriving unit”, and the “processing time deriving unit”, and the printer driver program PDR is the “image processing program”. The RAM 20 corresponds to a “second storage unit”, and the hard disk 10 corresponds to a “first storage unit”.

  When executing the application program APP, the CPU 6 displays various images and characters on the display 9 connected via the interface 12. Further, the CPU 6 receives a user operation input to the operation input unit 7 via the interface 12. Such operation input includes, for example, instruction input such as drawing, document creation, or printing. Then, the CPU 6 executes processing corresponding to the input according to the application program APP and the printer driver program PDR.

  The CPU 6 stores the original image data for each page generated by the application program APP in the RAM 20. As will be described later, print data generated by the GPUs 14_1 and 14_2 and transferred to the CPU 6 is stored in the RAM 20. Then, the CPU 6 reads out the print data from the RAM 20 and transfers it to the printer 4 via the interface 12.

  At this time, the CPU 6 uses the storage area of the RAM 20 divided into a main storage area 20_1 and a page lock storage area 20_2. A method of selectively using the main storage area 20_1 and the page lock storage area 20_2 will be described in detail later. The main storage area 20_1 is an area for saving (paging or swapping) data to be stored in a virtual storage area on the hard disk 10, and corresponds to a “first storage area”. This avoids a shortage of storage capacity. On the other hand, the page lock storage area 20_2 is a page-locked area in which such data is not saved, and corresponds to the “second storage area”. Therefore, in the page lock storage area 20_2, the data reading speed is increased by the amount not accompanied by the access to the hard disk 10. The page lock storage area 20_2 can be secured fixedly or dynamically according to the situation. Here, a case where a virtual storage area is provided in the hard disk 10 will be described as an example. However, other than the hard disk 10, although the data read / write speed from the CPU 6 is inferior to that of the RAM 20, a virtual storage area can be provided as long as the storage medium has a larger capacity than the RAM 20.

  As a transfer unit, the CPU 6 transfers the original image data for each page stored in the RAM 20 to the GPUs 14_1 and 14_2 via the interface 12 for each “processing unit”. The processing unit is arbitrarily set according to the number of pixels of an image that can be printed by the printer 4, the memory capacity of the GPUs 14_1 and 14_2, and the like. For example, the memory capacity of the GPUs 14_1 and 14_2 is generally smaller than that of the RAM 20, and does not have a capacity to hold original image data for one page. Therefore, the processing unit is set to a unit smaller than one page.

  The GPUs 14_1 and 14_2 perform image processing on the transferred original image data in accordance with an instruction from the CPU 6, and generate print data. That is, the GPUs 14 </ b> _ <b> 1 and 14 </ b> _ <b> 2 perform image processing in parallel by sharing each page of original image data for each page according to the available memory capacity. By doing so, it is possible to perform image processing per page at a higher speed than when the CPU 6 processes the original image data for one page alone or with only a single GPU. Then, the GPUs 14_1 and 14_2 transfer the print data to the CPU 6. Although not shown, each of the GPUs 14_1 and 14_2 includes a memory that temporarily stores input / output data and a processor core that executes image processing according to an image processing program. In the GPUs 14_1 and 14_2, cache memory and control hardware such as branch prediction / nonsequential instruction execution are omitted. Then, the processor cores specialized for floating point arithmetic are parallelized, so that calculation resources are concentrated at high density. Hereinafter, for convenience of explanation, it is assumed that the GPUs 14_1 and 14_2 have the same specifications, and there is no difference between the GPUs 14_1 and 14_2 in terms of transfer time and processing time for the same data. However, cases where these are different are also included in the present embodiment. In FIG. 1, both the GPUs 14 </ b> _ <b> 1 and 14 </ b> _ <b> 2 are provided inside the personal computer 2, but either or both of the GPUs 14 </ b> _ <b> 1 and 14 </ b> _ <b> 2 may be provided outside the personal computer 2.

  Image processing performed by the GPUs 14_1 and 14_2 will be described. The content of the image processing varies depending on the type of original image data. When the original image data is, for example, a postscript including a drawing command in a vector format, the image processing for the original image data is analyzed and the RGB (red, green, blue) gradation value is obtained for each pixel. Processing for generating RGB bitmap data, the gradation value of each pixel of the RGB bitmap data is converted into a color system of the printer 4, for example, CMYK (cyan, magenta, yellow, black) gradation values, and CMYK bitmap data Based on the color conversion processing for generating image data, the halftone processing for binarizing CMYK bitmap data in accordance with the printing resolution of the printer 4, and the bitmap data subjected to the halftone processing. Print data having raster data indicating a recording state and a sub-scan feed amount is generated. Processing, and the like. When the original image data is bitmap data, the image processing for the original image data includes color conversion or halftone processing and print data generation processing. In either case, image processing is performed so that print data is finally generated.

  Next, the printer 4 will be described in detail. In the printer 4, the CPU 23 integrally controls the printing operation according to the control program stored in the ROM 25. In addition, the CPU 23 receives print data transferred from the personal computer 2 via the interface 22 and spools it in the RAM 24. The RAM 24 is a page buffer that can store, for example, one or more pages of print data. Then, the CPU 23 supplies the print data read from the RAM 24 to the print engine 26. The print engine 26 forms dots with a recording material such as ink or toner for each pixel constituting the output image, and prints the output image on printing paper.

  In addition to the printer 4, for example, an image output device can be configured by a projector. In this case, the projector projects and outputs the output image by irradiating light of a color corresponding to the gradation value for each pixel constituting the output image image.

  In a modification of the present embodiment, the personal computer 2 may acquire original image data from an external storage device connected by the interface 12. The external storage device is composed of a CD-ROM (Compact Disk ROM) drive unit, an MO (Magneto Optic) drive unit, an FD (Floppy (registered trademark) Disk) drive unit, and a USB (Universal Serial Bus) interface. CD, MO, FD, etc. The original image data is read from a storage medium or a USB memory.

  In yet another modification, the personal computer 2 acquires original image data from another information processing apparatus. For example, the personal computer 2 constitutes a printer server in a network environment. The configuration in that case is illustrated by a broken line. The personal computer 2 is connected to a network 40 such as a LAN (Local Area Network) by a network interface 13. Another personal computer 42 is connected to the network 40. The other personal computer 42 transfers the original image data generated by the application program and the original image data read from the storage medium to the personal computer 2 via the network 40. Here, it is of course possible for a plurality of other personal computers 42 to transfer the original image data to the personal computer 2. Alternatively, a configuration in which the personal computer 2 and another personal computer 42 are connected by peer-to-peer without using the network 40 is also possible.

  In the personal computer 2, the CPU 6 receives the original image data transferred from the other personal computer 42 via the network interface 13 and the interface 12 and stores it in the RAM 20. Then, as described above, the original image data is transferred to the GPUs 14_1 and 14_2 for image processing. In this case, the original image data may be any of a drawing command, RGB bitmap data, CMYK bitmap data, and halftone processed bitmap data. Depending on each case, the GPUs 14_1 and 14_2 perform image processing that finally generates print data.

  Here, data transfer between the RAM 20 and the GPUs 14_1 and 14_2 during image processing will be described. In this embodiment, two transfer modes are executed.

  FIG. 2 is a diagram schematically illustrating data transfer between the RAM 20 and the GPUs 14_1 and 14_2. 2A corresponds to the “first transfer mode” (hereinafter referred to as the normal transfer mode), and FIG. 2B corresponds to the “second transfer mode” (hereinafter referred to as the high-speed transfer mode).

  First, the normal transfer mode will be described. As shown in FIG. 2A, the CPU 6 first stores the original image data GD for one page generated or transferred from another computer 42 in the main storage area 20_1 of the RAM 20. Then, the CPU 6 transfers the original image data GD to the GPUs 14_1 and 14_2 for each processing unit. Here, the original image data GD_1 corresponds to a processing unit for the GPU 14_1, and the original image data GD_2 corresponds to a processing unit for the GPU 14_2. That is, the CPU 6 transfers the original image data GD_1 to the GPU 14_1 (M2), and transfers the original image data GD_2 to the GPU 14_2 (M4).

  Then, since the GPU 14_1 performs image processing to generate print data PD_1 and transfers it to the CPU 6, the CPU 6 stores it in the main storage area 20_1 (M6). On the other hand, since the GPU 14_2 performs image processing to generate print data PD_2 and transfers it to the CPU 6, the CPU 6 stores it in the main storage area 20_1 (M8). By executing such an operation for all the original image data GD for one page, print data PD for one page is generated.

  Next, the high-speed transfer mode will be described. As shown in FIG. 2B, the CPU 6 transfers (or duplicates) the original image data GD for one page stored in the main storage area 20_1 of the RAM 20 to the page lock storage area 20_2 for each processing unit. Store. Here, the page lock storage area 20_2 is divided into a first page lock storage area 20_21 for the GPU 14_1 and a second storage area 20_22 for the GPU 14_2. Therefore, the original image data GD_1 for the GPU 14_1 is stored in the first page lock storage area 20_21 (M10), and the original image data GD_2 for the GPU 14_2 is stored in the second page lock storage area 20_22 (M12).

  Then, the CPU 6 reads the original image data GD_1 from the first page lock storage area 20_21 and transfers it to the GPU 14_1 (M14), and reads the original image data GD_2 from the second page lock storage area 20_22 and transfers it to the GPU 14_2 (M16). .

  Then, the GPU 14_1 performs image processing to generate print data PD_1 and transfers it to the CPU 6. The CPU 6 stores this in the first page lock storage area 20_21 (M18). On the other hand, since the GPU 14_2 performs image processing to generate print data PD_2 and transfers it to the CPU 6, the CPU 6 stores it in the second page lock storage area 20_22 (M20).

  Then, the CPU 6 reads the print data PD_1 and PD_2 from the page lock storage area 20_2, and transfers (or duplicates) the print data PD_1 and PD_2 to the main storage area 20_1 for storage (M22, M24). By executing such an operation for all the original image data GD for one page, print data PD for one page is generated.

  FIG. 3 is a sequence diagram illustrating an operation procedure in the normal transfer mode and the high-speed transfer mode. Here, for the sake of clarity of explanation, data transfer for only the GPU 14_1 will be described as an example. However, the description here also applies to the GPU 14_2.

  In FIG. 3A corresponding to the normal transfer mode, the data transfer sequence (1) from the main storage area 20_1 of the RAM 20 to the GPU 14_1 and the main storage area 20_1 from the GPU 14_1 are shown for each processing unit with respect to the time axis on the horizontal axis. A data transfer sequence (2) to the image processing and an image processing sequence (3) in the GPU 14_1 are shown. As shown in FIG. 3A, in the normal transfer mode, for one processing unit, a data transfer sequence (1) from the main storage area 20_1 to the GPU 14_1, an image processing sequence (3) in the GPU 14_1, and a main storage area from the GPU 14_1. This is repeated in the order of the data transfer sequence (2) to 20_1.

  In FIG. 3B corresponding to the high-speed transfer mode, the data transfer sequence (1) from the main storage area 20_1 to the first page lock storage area 20_21 for each processing unit with respect to the time axis on the horizontal axis, the first Data transfer sequence (2) from the page lock storage area 20_21 to the GPU 14_1, data transfer sequence (3) from the GPU 14_1 to the first page lock storage area 20_21, and data transfer from the first page lock storage area 20_21 to the main storage area 20_1 A sequence (4) and an image processing sequence (5) in the GPU 14_1 are shown. As shown in FIG. 3B, in the high-speed transfer mode, the data transfer sequence (1) from the main storage area 20_1 to the first page lock storage area 20_21 and the first page lock storage area 20_21 for one processing unit. Data transfer sequence (2) to the GPU 14_1, image processing sequence (5) in the GPU 14_1, data transfer sequence (3) from the GPU 14_1 to the first page lock storage area 20_21, and from the first page lock storage area 20_21 to the main storage area 20_1 The data transfer sequence (4) is executed in this order, and this is repeated.

  Here, in the high-speed transfer mode, data transfer with the GPU 14_1 is performed using the page-locked first page lock storage area 20_21 which does not perform paging in the virtual storage area, and thus the main storage area which performs paging in the virtual storage area Data can be read faster than the normal transfer mode in which data is transferred from 20_1 to the GPU 14_1. Therefore, the data transfer time is shortened.

  Further, as shown in FIG. 3B, when image processing is performed on the GPU 14_1 for one processing unit, the next processing unit of the original image data is transferred from the main storage area 20_1 to the first page lock storage area 20_21. Transfer (indicated by arrow A1 in (1)). The print data once stored in the first page lock storage area 20_21 is transferred to the main storage area 20_1 when the image processing is performed on the GPU 14_1 for the next processing unit (indicated by the arrow A2 in (4)). . Therefore, when the generated print data is transferred from the GPU 14_1, transfer of the next processing unit of the original image data can be started immediately. Therefore, the overall image processing time can be shortened compared with the normal transfer mode, and even if the amount of original image data increases and the transfer time becomes longer, print data can be supplied to the printer 4 without delay. it can.

  Next, the normal transfer mode and the high-speed transfer mode when the GPUs 14_1 and 14_2 are operated in parallel will be described.

  FIG. 4 is a diagram for explaining the sequence of the normal transfer mode. 4A and 4B, the data transfer sequence (1) from the main storage area 20_1 to the GPU 14_1 of the RAM 20 and the processing from the GPU 14_1 to the main storage area 20_1 for each processing unit with respect to the time axis on the horizontal axis. The data transfer sequence (2) and the image processing sequence (3) in the GPU 14_1 are shown, the data transfer sequence (1 ′) from the main storage area 20_1 to the GPU 14_2 of the RAM 20, and the data from the GPU 14_2 to the main storage area 20_1 A transfer sequence (2 ′) and an image processing sequence (3 ′) in the GPU 14_2 are shown. Here, the data transfer from the main storage area 20_1 of the RAM 20 to the GPU 14_1 and the data transfer from the GPU 14_1 to the main storage area 20_1 of the RAM 20 are executed so as to overlap with the image processing in the GPU 14_2. At the same time, data transfer from the main storage area 20_1 of the RAM 20 to the GPU 14_2 and data transfer from the GPU 14_2 to the main storage area 20_1 of the RAM 20 are executed so as to overlap with image processing in the GPU 14_1.

  FIG. 4A particularly shows a case where the data transfer time is equal to or shorter than the image processing time. Here, the image processing time T24 in the GPU 14_2 includes a total T13 of the data transfer time T11 from the main storage area 20_1 to the GPU 14_1 and the data transfer time T12 from the GPU 14_1 to the main storage area 20_1. On the other hand, the image processing time T14 in the GPU 14_1 includes a total T23 of the data transfer time T21 from the main storage area 20_1 to the GPU 14_2 and the data transfer time T22 from the GPU 14_2 to the main storage area 20_1. By doing so, the waiting time of the GPUs 14_1 and 14_2 is minimized, and the processing time as a whole is shortened. In addition, the load on the data bus is distributed.

  FIG. 4B shows a case where the data transfer time is longer than the processing time. Here, the image processing time T24 in the GPU 14_2 is included in the total time T13 of the data transfer time T11 from the main storage area 20_1 to the GPU 14_1 and the data transfer time T12 from the GPU 14_1 to the main storage area 20_1. On the other hand, the image processing time T14 in the GPU 14_1 is included in the total time T23 of the data transfer time T21 from the main storage area 20_1 to the GPU 14_2 and the data transfer time T22 from the GPU 14_2 to the main storage area 20_1. By doing so, the waiting time of the CPU 6 is minimized. Therefore, the processing time as a whole is shortened. In addition, the load on the data bus is distributed.

  Here, focusing on the relationship between the type of the original image data and the data transfer time, the amount of data is larger when the original image data is bitmap data than when the original image data is a drawing command such as a vector format. The transfer time becomes longer. On the other hand, the image processing time is greater when the bitmap data is generated from the conversion command such as color conversion than when the bitmap data is generated based on the drawing command and the color conversion is performed. Smaller and shorter processing time. That is, the relationship between the data transfer time and the image processing time is as follows. When the data amount of the original image data is small, the data transfer time is short while the image processing time is long and the data amount of the original image data is large. The relationship that the image transfer time is short while the data transfer time is long generally holds.

  In FIG. 4B, regarding the GPUs 14_1 and 14_2, the processing speed is increased by overlapping the data transfer to one side and the image processing on the other side. However, for example, when the original image data is different for each page in one print job, if the amount of transferred data increases from the middle and the transfer time becomes longer for the image processing time, the image processing time There is a risk that the data transfer will not be completed. In such a case, the high-speed transfer mode described below is advantageous because the transfer time can be shortened.

  FIG. 5 is a diagram for explaining the processing sequence in the high-speed transfer mode. The data transfer sequence (1) from the main storage area 20_1 to the first page lock storage area 20_21 and the data transfer sequence (2 from the first page lock storage area 20_21 to the GPU 14_1) for each processing unit with respect to the horizontal time axis. ), A data transfer sequence (3) from the GPU 14_1 to the first page lock storage area 20_21, a data transfer sequence (4) from the first page lock storage area 20_21 to the main storage area 20_1, and an image processing sequence (5) in the GPU 14_1. Further, the data transfer sequence (1 ′) from the main storage area 20_1 to the second page lock storage area 20_22, the data transfer sequence (2 ′) from the second page lock storage area 20_22 to the GPU 14_2, and the GPU 14_2 to the second 2 page lock A data transfer sequence (3 ′) to the storage area 20_22, a data transfer sequence (4 ′) from the second page lock storage area 20_22 to the main storage area 20_1, and an image processing sequence (5 ′) in the GPU 14_2 are shown.

  Here, the data transfer time T15 from the main storage area 20_1 to the first page lock storage area 20_21, the data transfer time T16 from the first page lock storage area 20_21 to the GPU 14_1, and the GPU 14_1 to the first page lock storage area 20_21. The data transfer is executed such that the total data transfer time T17 is included in the image processing time T24 in the GPU 14_2. On the other hand, the data transfer time T25 from the main storage area 20_1 to the second page lock storage area 20_22, the data transfer time T26 from the second page lock storage area 20_22 to the GPU 14_2, and the GPU 14_2 to the second page lock storage area 20_22. The data transfer is executed such that the total data transfer time T27 is included in the image processing time T14 in the GPU 14_1. By doing so, in addition to shortening the data transfer time, the waiting time of the CPU 6 and the GPUs 14_1 and 14_2 is minimized, and the image processing time as a whole is shortened. Therefore, even when the amount of data to be transferred increases, data transfer can be executed within the image processing time, and a delay in supplying print data can be prevented. In addition, the load on the data bus is reduced.

  Next, switching between the normal transfer mode and the high-speed transfer mode will be described. As described above, when the data amount of the original image data is small, the relationship between the data transfer time and the image processing time is that the data transfer time is short while the image processing time is long and the data amount of the original image data is large. In this case, the relationship that the image processing time is shortened while the data transfer time is long is generally satisfied. Focusing on this, in the present embodiment, the normal transfer mode and the high-speed transfer mode are switched based on the ratio of the transfer time for transferring the original image data to the GPUs 14_1 and 14_2 to the image processing time in the GPUs 14_1 and 14_2. By doing so, it is possible to keep the supply speed of the print data at a desired speed or more and supply the print data in time for the print speed of the printer 4.

  FIG. 6 is a flowchart for explaining an operation procedure when the personal computer 2 executes a print job. The procedure of FIG. 6 shows an operation procedure in the CPU 6 and the GPUs 14_1 and 14_2.

  When an instruction to start printing is input or an instruction to start printing is sent from another personal computer (S2), the CPU 6 sets the normal transfer mode (S3) and starts image processing.

  In the normal transfer mode, the CPU 6 transfers the original image data to the GPUs 14_1 and 14_2 for each processing unit, and instructs image processing (S4).

  Upon receiving the original image data and the image processing instruction (S6), the GPUs 14_1 and 14_2 perform image processing in response thereto (S8) and generate print data. Then, the GPUs 14_1 and 14_2 transfer the print data to the CPU 6 (S10).

  During this time, the CPU 6 operates a timer when executing step S4, and measures the time required for data transfer after the transfer is completed (S11). The CPU 6 receives the print data (S12), measures the elapsed time after the transfer of the original image data, and measures the image processing time in the GPUs 14_1 and 14_2 (S13). Here, the data amount of the print data is derived in advance according to the number of pixels of an image that can be printed by the printer 4 and the processing capability of the GPUs 14_1 and 14_2. Accordingly, the transfer time of the print data is derived in advance by calculation. Here, the CPU 6 for deriving the transfer time of the original image data and the print data corresponds to the transfer time deriving unit. The image processing time is derived by subtracting the print data transfer time from the elapsed time after the original image data transfer. Here, the CPU 6 for deriving the image processing time corresponds to a processing time deriving unit.

  Next, when the transfer to processing time ratio ((original image data transfer time + print data transfer time) / image processing time) exceeds a predetermined threshold (YES in S14), the CPU 6 The transfer mode is set (S15) and the process is continued. In this case, the data transfer time is longer than the image processing time, which means that the data amount of the original image data has increased. Therefore, the supply of print data is prevented from being delayed by switching to the high-speed transfer mode. Here, the threshold value for the determination can be arbitrarily set by obtaining a threshold value that is assumed to increase the probability that the supply of the print data will be delayed by experiments and calculations. Alternatively, the transfer to processing time ratio may be (original image data transfer time / image processing time).

  On the other hand, when the transfer to processing time ratio is equal to or smaller than the predetermined threshold (NO in S14), the processing is continued in the normal transfer mode. Here, it is possible to execute control such that the page lock storage area 20_2 is secured in the RAM 20 when the high-speed transfer mode is executed, and is released when the normal transfer mode is executed. By doing so, it is possible to avoid pressing the storage area of the RAM 20 with the page-lock storage area 20_2 that is page-locked. Therefore, when the CPU 20 is executing another control program or processing program, the RAM 20 can be used efficiently, and a reduction in processing efficiency can be prevented. Even when another personal computer 2 constitutes a printer server on the network and original image data is transmitted from a plurality of personal computers 42 on the network, the storage area of the RAM 20 can be used effectively. it can.

  The CPU 6 repeats steps S4 to S15 until one page of print data is accumulated (NO in S16), and when one page of print data is accumulated (YES in S16), this is output to the printer 4. (S17). It should be noted that the number of pages serving as a criterion for determination in step S16 may be a plurality of pages that are arbitrarily set according to the procedure of the RAM 20. By doing so, the print data can also be spooled on the personal computer 2 side.

  Then, the CPU 6 repeats the above steps S4 to S17 until the process for the last page ends (NO in S18), and ends the process when the process for the last page ends (YES in S18).

  According to the above procedure, the normal transfer mode and the high-speed transfer mode can be switched even during one print job. As a result, even if the image configuration is different for each page or processing unit, and therefore the transfer time of the original image data is different, the supply of print data is delayed by increasing the data transfer time. Can be prevented. Therefore, it is possible to avoid a case where blank paper is output from the printer 4 or to temporarily stop printing of the printer 4, and it is possible to prevent the printing from being delayed.

  According to this embodiment, even if the processing performance of the CPU or GPU is different for each personal computer and the data transfer time or the image processing time is different, the normal transfer mode high-speed transfer mode is switched according to these ratios. The above effects can be obtained without depending on the environment.

  In the above description, an example is described in which image processing is performed using two GPUs. However, this embodiment can be applied even when image processing is performed using three or more GPUs.

  As described above, according to the present embodiment, the output image data is output in time for the operation speed of the image output apparatus regardless of the image configuration of the original image data transferred to the GPU and the content of the image processing. It becomes possible to supply.

2: Personal computer 4: Printer 6: CPU 14_1, 14_2: GPU
20: RAM 20_1: Main storage area 20_2: Page lock storage area

Claims (5)

In an image processing device that generates output image data for an image output device based on original image data for each page,
A first storage unit having a virtual storage area;
A second storage area in which stored data is saved in the virtual storage area and a second storage area in which the data is not saved; and a data read / write speed is higher than that in the first storage unit. A storage unit of
A plurality of processing units for processing the original image data for each page in a processing unit smaller than the page, and generating the output image data;
A transfer unit for generating the output image data by alternately transferring the original image data for each page stored in the first storage area to the plurality of processing units for each processing unit;
A transfer time deriving unit for deriving a transfer time for transferring the original image data of the processing unit to the processing unit;
A processing time deriving unit for deriving a processing time for generating the output image data in the processing unit,
The transfer unit transfers the original image data from the first storage area to the processing unit based on a ratio of the transfer time to the processing time, and the processing unit transmits the original image. When generating the output image data based on the data, the original image data of the processing unit to be transferred to the processing unit is transferred from the first storage area to the second storage area. An image processing apparatus comprising: selecting and executing one of second transfer modes in which data is transferred from the second storage area to the processing unit. In claim 1,
The transfer unit selects the first transfer mode when the ratio is a first ratio, and selects the second transfer mode when the ratio is a second ratio larger than the first ratio. An image processing apparatus. In claim 1 or 2,
The transfer unit executes a second transfer from the first transfer mode when the ratio exceeds a threshold value while executing a single job for outputting output image data of a plurality of pages to the image output device. An image processing apparatus characterized by switching to a mode. In claim 3,
The transfer unit switches from the second transfer mode to the first transfer mode when the ratio falls below the threshold during execution of the single job. . In an image processing program for causing an image processing device to generate output image data for an image output device based on original image data for each page,
The image processing apparatus is
A first storage unit having a virtual storage area;
A second storage area in which stored data is saved in the virtual storage area and a second storage area in which the data is not saved; and a data read / write speed is higher than that in the first storage unit. A storage unit of
A plurality of processing units for processing the original image data for each page in a processing unit smaller than the page and generating the output image data;
A transfer procedure for generating the output image data by alternately transferring the original image data for each page stored in the first storage area to the plurality of processing units for each processing unit;
A procedure for deriving a transfer time for transferring the original image data of the processing unit to the processing unit;
Causing the image processing apparatus to execute a procedure for deriving a processing time for generating the output image data in the processing unit;
In the transfer procedure, based on a ratio of the transfer time to the processing time, a first transfer mode in which the original image data is transferred from the first storage area to the processing unit, and the processing unit has the original image When generating the output image data based on the data, the original image data of the processing unit to be transferred to the processing unit is transferred from the first storage area to the second storage area. An image processing program that selects and executes one of the second transfer modes for transferring data from the second storage area to the processing unit.

Images