This view uses a bar chart to visualize CPU utilization for each of the processor nodes. Each node is represented as a bar on the chart. As you play back the trace file or monitor nodes online, the length of the bars change to show their relative value of CPU utilization. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous CPU utilization. The hatched fill pattern is the average CPU utilization. In addition, VT draws a vertical line to the right of each bar to show the highest level of CPU utilization so far reached by each node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous value of CPU utilization
- -average value of CPU utilization
- -maximum value of CPU utilization
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents processor utilization as a percentage of 100. You can adjust this so that the view represents processor utilization as a percentage of some other value. For example, say you ran your program on four processor nodes - none of which ever exceeded 10 percent utilization. You might want to adjust the view so that it represents processor utilization as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value
This view uses a strip graph, or a group of strip graphs, to visualize CPU utilization for the processor nodes. This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. A line on this graph shows the average value of CPU utilization across all processor nodes. Individual display mode shows a separate strip graph for each processor node's CPU utilization. When in this mode, each graph contains two lines - one showing the individual processor node's CPU utilization and, for comparison, one showing the aggregate CPU utilization. To distinguish between the two, the area between the individual CPU utilization line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on a line representing either aggregate or individual CPU utilization.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate CPU utilization at that point in time
- -the individual processor node's CPU utilization at that point in time if the cursor is on a line representing an individual node's CPU utilization.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
- Displaying a Text Representation of the Graph Information
- You can open a Statistics Window containing a text representation of the information - the mean, the standard deviation, and so on - shown in the view. To open the Statistics Window:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Show Statistics
- The Statistics Window opens.
This view is similar to the processor utilization bar chart. It too visualizes CPU utilization for each of the processor nodes during the program's run. In this view, however, the processor nodes are laid out as bars in a two-dimensional grid and their CPU utilization values raise the bars up along the z-axis. As the bars rise up along the z-axis, they appear in different colors according to your display spectrum. This is an instantaneous view, and is for trace visualization only.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the bars representing a processor node.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous CPU utilization
- Changing the Display Angle
- You can change the angle at which the view displays the 3-D Bar Chart. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the angle field
- TYPE IN 0, 1, or 2. Specifying 0 gives you a side view of the chart, while specifying 2 angles the chart as if you were looking down on it. 1 is in between these two extremes.
This view uses three overlapping polygons to show CPU utilization for each of the processor nodes, and the overall processor load balance. This is an instantaneous view.
The largest of the polygons represents 100 percent utilization for all of the processor nodes. Within this polygon, VT draws each processor node as a spoke starting at the polygon's center and extending out to the rim.
VT draws the second polygon inside the first. This polygon represents the instantaneous CPU utilization for each of the processor nodes. On each node's spoke, VT draws a point which represents the current CPU utilization for that node. VT then connects the points to form a polygon with a solid fill pattern. The more regular the polygon, the better your processor load balance.
The third polygon is similar to the second. It shows the average CPU utilization, however, and has a hatched fill pattern.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the spokes representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -the instantaneous value of CPU utilization
- -the average value of CPU utilization.
- The window remains open only while you hold the mouse button down.
This view visualizes message passing events between processor nodes. The view uses a small circle to represent each processor node on which your program was run. Messages sent between nodes appear as an arc joining two of the circles. This is an instantaneous view. Note: this view is for trace visualization only.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the small circles in the view.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -number of messages sent from that node
- -number of messages received by that node.
- The window remains open only while you hold the mouse button down.
- Toggling Between Instantaneous and Cumulative Presentation
- This view lets you toggle between an instantaneous and a cumulative presentation of the message passing information. An instantaneous presentation is the default - the arc representing the message passing event is visible only for the duration of the event. With a cumulative presentation, the arcs remain visible past the completion of the event. A cumulative presentation lets you see the overall communication pattern among the processor nodes.
- If you want a cumulative presentation of the message passing information:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Cumulative
This view uses a bar chart to visualizes the type and duration of communication events. Each bar represents a processor node on which your program was run, and the chart's horizontal axis represents a range of time. Each bar in the chart will be made up of a number of colored blocks. Each block represents a communication event involving the processor. The size of the block represents the event's duration, and its color indicates the type of event. For example, blocking sends will all be shown in one color, non-blocking sends in another, broadcasts in another, and so on. In addition, messages sent between processor nodes appear as a line between the bars on the chart. This is a steaming view. Note: this view is for trace visualization only.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the blocks representing a communication event.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -the type of communication event
- -the time the event occurred.
- The window remains open only while you hold the mouse button down.
- Using the Search Feature
- This view has a search feature that helps you to easily examine the communication events and processor nodes you are interested in. For example, say you are interested in any blocking sends between the nodes 2 and 6. You would:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Search
- The Search window opens.
- The Search window
- The window contains a separate area for each type of communication event. Each area contains a text entry field and a selection button. The color shown in each area corresponds to the color used for that type of communication event in the view.
- PRESS the blocking send button.
- FOCUS on the text entry field for blocking sends.
- TYPE IN 2,6
- To see the first blocking send involving the two nodes:
- PRESS First
- To see the next blocking send involving the two nodes:
- PRESS Next
- To see the previous blocking send involving the two nodes:
- PRESS Previous
- To close the Search window:
- PRESS Done
This view uses a grid to visualize message sends between processor nodes. Each processor node has both a row and a column on the table. The rows represent when processor nodes send a message, and the columns represent when processor nodes receive a message. The rectangle intersections on the grid represent the message path between the sending node (the row) and the the receiving node (the column). As you play back your trace file, the message path rectangles light up to show you the message sends. This is an instantaneous view. Note: this view is for trace visualization only.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the rectangles representing message paths between processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number of the sending node.
- -node number of the receiving node.
- -number of messages sent.
- The window remains open only while you hold the mouse button down.
- Toggling Between Instantaneous and Cumulative Presentation
- This view lets you toggle between an instantaneous and a cumulative presentation of the message passing information. An instantaneous presentation is the default - the message path rectangles stay lit only for the duration of the message passing event. With a cumulative presentation, the message path rectangles remain lit past the completion of the event. A cumulative presentation lets you see the overall communication pattern among the processor nodes.
If you want a cumulative presentation of the message passing information:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Cumulative
This view shows you the C or FORTRAN source code of the program associated with the most recent event. A series of colored bars across the top of the display represent the processor nodes that ran the program. As you play back the trace file, the bars move through the code to show you each processor node's position. To be more specific, for each time a processor node entered into a communication function such as a blocking send, an environment initialization, or an application marker call, its bar moves to that line in the source code. In order to use this view, you must have compiled your program with the -g flag. This is an instantaneous view. This view is for trace visualization only. NOTE: If the Source Code view indicates an error with finding the .file, you may need
to place your source code file in the directory where VT was started and name it ".file".
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any one of the bars representing a processor node.
- PRESS the left mouse button.
- A window opens displaying the:
- -list of nodes at that line
- -name of the source code file
- -line number you are at in the source code file.
- Using the Automatic Scrolling Capability
- This view has an automatic scrolling capability that automatically scrolls the view to include the most recent trace event. If this capability is off, you must manually scroll the view. To toggle the automatic scrolling capability on and off:
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Parameters
- The view's Parameters Window opens.
- PRESS the autoscroll toggle button on the Parameter Window.
- PRESS Apply
- PRESS OK
This view uses a bar chart to visualize disk reads - the number of times processes acquire information from the system's hard disk. This refers to reads of cached information rather than the actual physical device. Each node is represented by a bar on the chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of disk reads. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of disk reads. The hatched fill pattern is the average number of disk reads. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of disk reads so far reached for each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of disk reads
- -average number of disk reads
- -maximum number of disk reads so far recorded
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents disk reads as a percentage of 100. You can adjust this so that the view represents disk reads as a percentage of some other value. For example, say the amount of disk reads never exceeds 10 percent. You might want to adjust the view so that it represents disk reads as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize disk reads - the number of times processes acquire information from the system's hard disk. This refers to reads of cached information rather than the actual physical device. If you are using this view for trace visualization, the processor nodes depicted are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of disk reads across all processor nodes. Individual display mode shows a separate graph for each processor node's disk reads. When in this mode, each graph contains two lines - one showing the individual processor node's disk reads, and, for comparison, one showing the aggregate number of disk reads. To distinguish between the two, the area between the individual disk read line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual disk reads.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of disk reads at that point in time
- -the individual processor node's number of disk reads at that point in time If the cursor is on a line representing an individual node's disk reads:
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize disk transfers - the number of times the system transfers blocks of read/write data to and from the hard disk. Each node is represented by a bar on the chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of disk transfers. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of disk transfers. The hatched fill pattern is the average number of disk transfers. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of disk transfers so far reached for each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of disk transfers
- -average number of disk transfers
- -maximum number of disk transfers so far recorded
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents disk transfers as a percentage of 100. You can adjust this so that the view represents disk transfers as a percentage of some other value. For example say the amount of disk transfers never exceeds 10 percent. You might want to adjust the view so that it represents disk transfers as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize disk transfers - the number of times the system transfers blocks of read/write data to and from the hard disk. If you are using this view for trace visualization, the processor nodes depicted are the ones that ran your program. If you are using this view for performance monitoring, they are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of disk transfers across all processor nodes. Individual display mode shows a separate graph for each processor node's disk transfers. When in this mode, each graph contains two lines - one showing the individual processor node's disk transfers, and, for comparison, one showing the aggregate number of disk transfers. To distinguish between the two, the area between the individual disk transfers line and the graph's horizontal axis has a solid fill. This is a streaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual disk transfers.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of disk transfers at that point in time.
- -the individual processor node's number of disk transfers at that point in time
- If the cursor is on a line representing an individual node's disk transfers
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize disk writes - the number of times processes write information to disk. Each node is represented by a bar on the chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of disk writes. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of disk writes. The hatched fill pattern is the average number of disk writes. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of disk writes so far reached for each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of disk writes
- -average number of disk writes
- -maximum number of disk writes so far recorded
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents the number of disk writes as a percentage of 100. You can adjust this so that the view represents disk writes as a percentage of some other value. For example, say the number of disk writes never exceeds 10 percent. You might want to adjust the view so that it represents disk writes as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize disk writes - the number of times processes write information to disk. If you are using this view for trace visualization, the processor nodes it depicts are the ones that ran your program. If you are using this view for performance monitoring, they are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of disk writes across all processor nodes. Individual display mode shows a separate graph for each processor node's disk writes. When in this mode, each graph contains two lines - one showing the individual processor node's disk writes, and, for comparison, one showing the aggregate number of disk writes. To distinguish between the two, the area between the individual disk writes line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual disk writes.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of disk writes at that point in time
- -the individual processor node's number of disk writes at that point in time If the cursor is on a line representing an individual node's disk writes
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize the number of TCP/IP packets received by processor nodes. Each node is represented by a bar on the chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of packets received. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of packets received. The hatched fill pattern is the average number of packets received. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of packets received so far reached by each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of packets received
- -average number of packets received
- -maximum instantaneous number of packets so far received
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents packets received as a percentage of 100. You can adjust this so that the view represents packets received as a percentage of some other value. For example, say the number of packets received never exceeds 10 percent. You might want to adjust the view so that it represents packets received as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize the number of TCP/IP packets received by processor nodes. If you are using this view for trace visualization, the processor nodes it depicts are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of packets received across all processor nodes. Individual display mode shows a separate graph for each processor node's packets received. When in this mode, each graph contains two lines - one showing the individual processor node's packets received, and, for comparison, one showing the aggregate number of packets received. To distinguish between the two, the area between the individual packets received line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual packets received.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of packets received at that point in time
- -the individual processor node's number of packets received at that point in time if the cursor is on a line representing an individual node's packets received.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize the number of TCP/IP packets sent by processor nodes. Each node is represented as a bar on this chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of packets sent. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of packets sent. The hatched fill pattern is the average number of packets sent. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of packets sent so far by each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of packets sent
- -average number of packets sent
- -maximum instantaneous number of packets so far sent
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
By default, this view represents packets sent as a percentage of 100. You can adjust this so that the view represents packets sent as a percentage of some other value. For example, say the number of packets sent never exceeds 10 percent. You might want to adjust the view so that it represents packets sent as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize the number of TCP/IP packets sent by processor nodes. If you are using this view for trace visualization, the processor nodes it depicts are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of packets sent across all processor nodes. Individual display mode shows a separate graph for each processor node's packets sent. When in this mode, each graph contains two lines - one showing the individual processor node's packets sent, and, for comparison, one showing the aggregate number of packets sent. To distinguish between the two, the area between the individual packets sent line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual packets sent.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of packets sent at that point in time
- -the individual processor node's number of packets sent at that point in time if the cursor is on a line representing an individual node's packets sent.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize context switches - the number of times the processes are swapped in and out of active execution. Each node is represented as a bar on the chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of context switches. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of context switches. The hatched fill pattern is the average number of context switches. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of context switches so far reached for each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of context switches
- -average number of context switches
- -maximum number of context switches so far recorded
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents the number of context switches as a percentage of 100. You can adjust this so that the view represents context switches as a percentage of some other value. For example, say the number of context switches never exceeds 10 percent. You might want to adjust the view so that it represents context switches as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize context switches - the number of times processes are swapped in and out of active execution. If you are using this view for trace visualization, the processor nodes depicted in this view are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of context switches across all processor nodes. Individual display mode shows a separate graph for each processor node's context switches. When in this mode, each graph contains two lines - one showing the individual processor node's context switches, and, for comparison, one showing the aggregate number of context switches. To distinguish between the two, the area between the line for the individual node and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual context switches.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of context switches at that point in time
- -the individual processor node's number of context switches at that point in time if the cursor is on a line representing an individual node's context switches.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize page faults. Each node is represented as a bar on this chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of page faults. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of page faults. The hatched fill pattern is the average number of page faults. In addition, VT draws a vertical line to the right of each bar to show the highest instantaneous number of page faults so far by each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of page faults
- -average number of page faults
- -maximum number of page faults so far recorded
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents page faults as a percentage of 100. You can adjust this so that the view represents page faults as a percentage of some other value. For example, say the number of page faults never exceeds 10 percent. You might want to adjust the view so that it represents page faults as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize page faults. If you are using this view for trace visualization, the processor nodes it depicts are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of page faults across all processor nodes. Individual display mode shows a separate graph for each processor node's page faults. When in this mode, each graph contains two lines - one showing the individual processor node's page faults, and, for comparison, one showing the aggregate number of page faults. To distinguish between the two, the area between the individual page faults line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual page faults.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of page faults at that point in time
- -the individual processor node's number of page faults at that point in time if the cursor is on a line representing an individual node's page faults.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a bar chart to visualize the number of times processor nodes invoke kernel subroutines. Each node is represented as a bar on this chart. If you are using this view for trace visualization, these are the processor nodes that ran your program. If you are using this view for performance monitoring, these are the selected nodes. The length of the bars change to show the number of system calls. Each bar has both a solid and a hatched fill pattern. The solid fill pattern represents the instantaneous number of system calls. The hatched fill pattern is the average number of system calls. In addition, VT draws a vertical line to the right of each bar to show the highest number of system calls so far reached by each processor node. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any of the bars representing processor nodes.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -instantaneous number of system calls
- -average number of system calls
- -maximum instantaneous number of system calls so far reached
- The window remains open only while you hold the mouse button down.
- Adjusting the Default Maximum Value
- By default, this view represents the number of system calls as a percentage of 100. You can adjust this so that the view represents system calls as a percentage of some other value. For example, say the number of system calls never exceeds 10 percent. You might want to adjust the view so that it represents system calls as a percentage of 10 instead of 100. To do this:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- SELECT Parameters
- The view's Parameters window opens.
- FOCUS on the barMaxValue text entry field.
- TYPE IN 10
- PRESS Apply
- The view calibrates to the new maximum value.
This view uses a strip graph, or a group of strip graphs, to visualize the number of times processor nodes invoke kernel subroutines. If you are using this view for trace visualization, the processor nodes it depicts are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes.
This view has two display modes - aggregate and individual. By default, the view is in aggregate display mode - it uses a single strip graph to represent the average number of system calls across all processor nodes. Individual display mode shows a separate graph for each processor node's system calls. When in this mode, each graph contains two lines - one showing the individual processor node's system calls, and, for comparison, one showing the aggregate number of system calls. To distinguish between the two, the area between the individual system calls line and the graph's horizontal axis has a solid fill. This is a steaming view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over a point on the line representing either aggregate or individual system calls.
- PRESS the left mouse button.
- A window opens displaying:
- -the aggregate number of system calls at that point in time.
- -the individual processor node's number of system calls at that point in time if the cursor is on a line representing an individual node's system calls.
- The window remains open only while you hold the mouse button down.
- Toggling Between Display Modes
- This view has two display modes - aggregate and individual. By default the view is in aggregate display mode - it uses a single strip graph. To display a separate strip graph for each node:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Individual
This view uses a series of pie charts to summarize system activity. Each pie chart represents a processor node, and is segmented to show the percentage of time the node spent:
-
- -idle
- -executing a program
- -executing a kernel function
- -waiting for some resource to become available. For example, waiting for a disk to become available.
If you are using this view for trace visualization, the processor nodes shown are the ones that ran your program. If you are using this view for performance monitoring, these are the selected nodes. This is an instantaneous view.
- Displaying Additional Data in the View
- You may display additional data in this view. To do this:
- PLACE the mouse cursor over any pie chart.
- PRESS the left mouse button.
- A window opens displaying the:
- -node number
- -system state represented. This includes:
- -user time
- -kernel time
- -wait time
- -idle time
- -percent of time spent in that state
- Toggling Between Instantaneous and Cumulative Presentation
- The view lets you toggle between instantaneous and cumulative presentation of system activity. By default, the pie charts summarize instantaneous system activity. A cumulative presentation shows overall system activity for each processor node.
- If you want a cumulative system summary:
- PLACE the mouse cursor over the view.
- PRESS the right mouse button.
- A selection menu appears.
- SELECT Cumulative
©
Copyright1999, Maui High Performance Computing Center.
All rights reserved.