TIS-100 Reference Manual Text Version
Documentation should be searchable.
Overview
The Tessellated Intelligence System is a massively parallel computer architecture comprised of non/uniformly interconnected heterogeneous nodes. The Tessellated Intelligence System is ideal for applications requiring complex data stream processing, such as automated financial trading, bulk data collection, and civilian behavioral analysis.
Note: Notes like this one will appear in this manual to indicate scenarios requiring special attention and to refer to other documents that contain more information on a topic.
System Architecture and Organization
The Tessellated Intelligence System consists of a large number of independent nodes connected on a local basis. (Refer to the model-specific manual to find the precise node population counts present on a particular device.) Node types can be broadly classified as processing or storage, with several variants within each category.
Generally, nodes are connected to up to four neighbors via ports. Ports enable lightweight message/passing communication between nodes. Communication over ports is coordinated by allowing either node to issue a read or write to a port and blocking until the request is filled by the corresponding node.
Note: If two nodes issue the same communication command (read or write) on the connection between them, the nodes will deadlock and a hardware fault will occur. Refer to the separate document “Tessellated Intelligence System Best Practices - Patterns of Node Communication” for details on how to use ports effectively and safely.
Note: If a node issues a communication command and it is never fulfilled by the corresponding node, the node will deadlock and a hardware fault will occur. (Exceptions to this rule exist; refer to the documentation of specific node types for details.) Refer to the separate document “Tessellated Intelligence System Best Practices - Patterns of Node Communication” for details on how to use ports effectively and safely.
Note: This document does not describe timing or throughput for node communication operations or instructions, as these values vary by model and hardware revision. Refer to the model/specific manual for a detailed description of performance characteristics of a particular device.
Node Type T20 - Reserved
Note: This node type identifier is restricted to specific models of the Tessellated Intelligence System and will not be described in this document. Documentation for node type T20 is distributed only with systems containing this node type. Unauthorized requests for copies of documentation describing this node are reported to the state security bureau, as required by law.???
Node Type T21 - Basic Execution Node
1. Architecture
The Basic Execution Node is responsible for coordinating the behavior of the Tessellated Intelligence System. Processing can occur within the Basic Execution Node, or can be delegated to specialized processing and storage nodes.
The Basic Execution Node executes a program specified in the Basic Execution Node Instruction Set. A Basic Execution Node program specifies computational and communication operations to perform. Operations are performed sequentially, beginning with the first instruction in the program. After executing the last instruction of the program, execution automatically continues to the first instruction. This behavior supports the common usage of Basic Execution Nodes, in which programs are written to operate in a continuous loop.
In addition to the communication ports common to all Tessellated Intelligence System nodes, the Basic Execution Node contains a number of registers that are used in the execution of its program. No additional memory is available on the Basic Execution Node; if additional storage is required, the node should coordinate with another Basic Execution Node or a storage node.
All registers store integer values between -999 and 999 (inclusive). The representation of register values is implementationrdefined, and knowledge of the representation is not required to program the Basic Execution Node.
1-1. ACC
Type: Internal
Description: ACC is the primary storage register for a Basic Execution Node. ACC is used as the implicit source or destination operand of many instructions, including arithmetic and conditional instructions.
1-2. BAK
Type: Internal (non/addressable)
Description: BAK is temporary storage for values in ACC. It is only accessible through the SAV and SWP instructions, and cannot be read or written directly .
1-3. NIL
Type: Internal (special)
Description: Reading NIL produces the value zero. Writing to NIL has no effect. NIL can be used as a destination operand to execute an instruction for its side effects only, discarding the result.
1-4. LEFT, RIGHT, UP, DOWN
Type: Port
Description: The four communication registers UP, DOWN, LEFT, and RIGHT correspond to the four ports that all Basic Execution Nodes use to communicate with topologically adjacent nodes. Some ports will be disconnected on certain nodes within the hardware, and will block indefinitely if a READ or WRITE command is issued. Refer to the interconnection diagram for the node to determine which ports are available for use.
1-5. ANY
Type: Port (pseudoport)
Description: When ANY is used as the source of an instruction, the instruction will read the first value that becomes available on any port. When ANY is used as the destination of an instruction, the result of the instruction will be sent to the first node that reads from any port on this node.
1.6. LAST
Type: Port (pseudoport)
Description: LAST refers to the port last read or written using the ANY pseudoport. It otherwise behaves identically to explicitly specifying a port. Reading from or writing to LAST before it has been set by a successful read or write using the ANY pseudoport will result in implementation-defined behavior. Refer to the separate document “Tessellated Intelligence System Best Practices - Patterns of Node Communication” for sample code demonstrating the use of the LAST pseudoport.
2. Instruction Set
<SRC>
and <DST>
instruction parameters may specify a port or internal register. Any use of a port will block until the corresponding node connected to that port completes the communication by reading or writing a value. Additionally, a <SRC>
parameter may be a literal integer value between -999 and 999 (inclusive). BAK cannot be specified as a <SRC>
or <DST>
operand. The value of BAK is only accessible through special instructions SAV and SWP. <LABEL>
parameters are arbitrary textual names used to specify jump targets within the program.
2-1. Comments
Syntax: # COMMENT TEXT
Description: All text including and after the comment symbol (#) is ignored.
Note: Text placed after two comment symbols (##) will be used as the title of the program in which it occurs, and is displayed in the debugger to make browsing programs easier.
2-2. Labels
Syntax: <LABEL>
:
Description: Labels are used to identify targets for jump instructions. When used as a jump target, the instruction following the label will be executed next.
Examples:
2-3. NOP
Syntax: NOP
Equivalent syntax: ADD NIL
Description: NOP is a pseudo-instruction that has no effect on the node’s internal state or communication ports. NOP is automatically converted to the instruction ADD NIL.
2-4. MOV
Syntax: MOV <SRC>, <DST>
Description: <SRC>
is read and the resulting value is written to <DST>
.
Examples:
2-5. SWP
Syntax: SWP
Description: The values of ACC and BAK are exchanged.
2-6. SAV
Syntax: SAV
Description: The value of ACC is written to BAK.
2-7. ADD
Syntax: ADD <SRC>
Description: The value of <SRC>
is added to the value of ACC and the result is stored to ACC.
Examples:
2.8. SUB
Syntax: SUB <SRC>
Description: The value of <SRC>
is subtracted from the value of ACC and the result is stored to ACC.
Examples:
2-9. NEG
Syntax: NEG
Description: The value of ACC is arithmetically negated. A value of zero remains the same.
2-10. JMP
Syntax: JMP <LABEL>
Description: Transfer execution unconditionally. The instruction after the label <LABEL>
will be executed next.
2-11. JEZ
Syntax: JEZ <LABEL>
Description: Transfer execution conditionally. The instruction after the label <LABEL>
will be executed next if the value of ACC is zero.
2-12. JNZ
Syntax: JNZ <LABEL>
Description: Transfer execution conditionally. The instruction after the label <LABEL>
will be executed next if the value of ACC is not zero.
2-13. JGZ
Syntax: JGZ <LABEL> 1
Description: Transfer execution conditionally. The instruction after the label <LABEL>
will be executed next if the value of ACC is positive (greater than Zero).
2-14. JLZ
Syntax: JLZ <LABEL>
Description: Transfer execution conditionally. The instruction after the label <LABEL>
will be executed next if the value of ACC is negative (less than zero).
2-15. JRO
Syntax: JRO <SRC>
Description: Transfer execution unconditionally. The instruction at the offset specified by <SRC>
relative to the current instruction will be executed next.
Example Programs
The following sample program reads a sequence of values from the LEFT port, doubling each value read and writing that to the RIGHT port. Because of the automatic looping behavior of the Basic Execution Node, it continues to the first instruction after executing the last instruction.
The following sample program reads a sequence of values from the UP port, writing positive values to the RIGHT port and negative values to the LEFT port. Zero values are discarded.
Node Type T30 - Stack Memory Node
1. Architecture
The Stack Memory Node enables read/write access to a large number of values according to a simple stack-based communication protocol. (Refer to the model-specific manual to find the capacity of the Stack Memory Nodes on a particular device.)
2. Communication Protocol
All interaction with the Stack Memory Node is performed through ports. Writing to the Stack Memory Node adds the value to the top of the stack. If the stack is full, the write will block until space becomes available. Reading from the Stack Memory Node removes the top value from the stack and produces that value. If the stack is empty, the read will block until a value is available.
Stack Memory Nodes are typically connected to multiple other nodes, and can be used by any connected node. Simultaneous reads and writes to a Stack Memory Node resolve in an undefined order, but each individual communication will behave according to the described communication protocol. For more information on using storage nodes from multiple nodes effectively and predictably, refer to the separate document “Tessellated Intelligence System Best Practices & Patterns of Node Communication”.
Node Type T31 - Random Access Memory Node
Note: The Random Access Memory Node is not yet available in standard Tessellated Intelligence System devices. Emulators and prototype hardware are available to interested users. The specification and behavior is not yet finalized and therefore is omitted from this document.
TODO
- Figure out who sold TIS-100 to swap meet dealer
Rebuild signal multiplier- Look for book of mico-optimisation tips
- Renew license plate tabs
Embedded Interactive Debugger
1. Keyboard Shortcuts
The interactive debugger contains the following keyboard shortcuts:
Control-Z: Undo last change
Control-Y: Redo last change
Control-X: Cut selected text to clipboard
Control-C: Copy selected text to clipboard
Control-V: Paste clipboard text
Control-Arrow: Navigate to the adjacent execution node
F1: View instruction set quick reference
F2: View anti-tamper certification status ??
F5: Begin running the current program
F6: Step or pause the current program
2. Breakpoints
To set a breakpoint, place an exclamation mark (!) at the beginning of a line. When a breakpoint is set, the program will be paused before that line is executed, allowing you to easily debug code that would be too tedious to step through one instruction at a time.
Visualization Module
1. Visualization Module Usage
The TIS-100 contains a Visualization module that allows programs to programmatically create and display images. The module contents can be modified by sending command sequences, which consist of the starting X coordinate, the starting Y coordinate, one or more color values, and a terminating negative value (often -I). The coordinate system starts at (0, 0), which is located in the topleft of the display area.
The visualization module supports the following colors:
0: Black
1: Dark grey
2: Bright grey
3: White
4: Red
2. Visualization Module Resolution
The standard TIS-100 visualization module is 30 characters wide and 18 characters tall.
The “image console sandbox” contains a larger visualization module that is 36 characters wide and 22 characters tall.