# US 2025/0147803 A1 · Synchronizing scheduling tasks with atomic alu

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0147803 A1
- Document ID: US20250147803A1
- Status: Published application
- Kind: A1
- Application number: 19/017,668
- Filed: 2025-01-12
- Published: 2025-05-08
- Earliest claimed: 2017-06-16
- Assignees: Imagination Technologies Limited
- Applicants: Imagination Technologies Limited
- Inventors: Ollie Mower; Yoong-Chert Foo
- Web page: https://patentgrep.com/patent/US20250147803A1

## CPC

- G06F 9/4881: {Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues}
- G06F 7/575: Basic arithmetic logic units, i.e. devices selectable to perform either addition, subtraction or one of several logical operations, using, at least partially, the same circuitry
- G06F 9/3001: {Arithmetic instructions}
- G06F 9/3004: {to perform operations on memory}
- G06F 9/30087: {Synchronisation or serialisation instructions}
- G06F 9/3009: {Thread control instructions}
- G06F 9/3016: {Decoding the operand specifier, e.g. specifier format}
- G06F 9/485: {Task life-cycle, e.g. stopping, restarting, resuming execution (G06F9/4881 takes precedence)}
- G06F 9/522: {Barrier synchronisation}
- G06F 15/17325: {Synchronisation; Hardware support therefor (intertask synchronisation G06F9/52)}
- G06F 15/76: Architectures of general purpose stored program computers (with program plugboard G06F15/08; multicomputers G06F15/16)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method of synchronizing a group of scheduled tasks within a parallel processing unit into a known state is described. The method uses a synchronization instruction in a scheduled task which triggers, in response to decoding of the instruction, an instruction decoder to place the scheduled task into a non-active state and forward the decoded synchronization instruction to an atomic ALU for execution. When the atomic ALU executes the decoded synchronization instruction, the atomic ALU performs an operation and check on data assigned to the group ID of the scheduled task and if the check is passed, all scheduled tasks having the particular group ID are removed from the non-active state.

## Claims

### Claim 1

A method of synchronizing a group of scheduled tasks into a known state within a parallel processing unit, the method comprising:
in response to decoding, in an instruction decoder, a synchronization instruction in a scheduled task from a group of tasks, placing the scheduled task into a non-active state wherein no instructions in the scheduled task are executed except for the synchronization instruction, wherein the decoded synchronization instruction causes an ALU that executes the synchronization instruction to perform a check on data assigned to the group of tasks.

### Claim 2 (depends on 1)

The method according to claim 1, further comprising:
in response to receiving, at a scheduler, a message from an ALU indicating that the data assigned to the group of tasks has passed a check, removing all scheduled tasks in the group from the non-active state.

### Claim 3 (depends on 1)

The method according to claim 1, further comprising:
in response to receiving, at the scheduler, a message from an ALU indicating that the data assigned to the group of tasks has failed a check, leaving the scheduled task in the non-active state.

### Claim 4 (depends on 1)

The method according to claim 1, further comprising, at the scheduler and dependent upon an outcome of the check, either removing all scheduled tasks in the group from the non-active state or leaving the scheduled task in the non-active state.

### Claim 5 (depends on 1)

The method according to claim 1, wherein the ALU is an atomic ALU.

### Claim 6 (depends on 1)

The method according to claim 1, further comprising:
performing, in the ALU, an operation on data assigned to the group of tasks;
performing, in the ALU, a check on the updated data assigned to the group of tasks; and
in response to the updated data passing the check, sending the message from the ALU to the scheduler indicating that the data assigned to the group of tasks passed the check.

### Claim 7 (depends on 6)

The method according to claim 6, further comprising:
in response to the updated data passing the check, resetting the data assigned to the group of tasks.

### Claim 8

A scheduling module comprising:
one or more queues arranged to store scheduled tasks; and
hardware logic arranged, in response to receiving an instruction from an instruction decoder decoding a synchronization instruction in a scheduled task from a group of tasks, to place the scheduled task into a non-active state wherein no instructions in the scheduled task are executed except for the synchronization instruction;
wherein the decoded synchronization instruction causes an ALU that executes the synchronization instruction to perform a check on data assigned to the group of tasks.

### Claim 9 (depends on 8)

The scheduling module according to claim 8, wherein the scheduling module further comprises hardware logic arranged, in response to receiving a message from an ALU indicating that the data assigned to the group of tasks passed a check, to remove all scheduled tasks in the group from the non-active state.

### Claim 10 (depends on 8)

The scheduling module according to claim 8, wherein the scheduling module further comprises hardware logic arranged, in response to receiving a message from an ALU indicating that the data assigned to the group of tasks has failed a check, to leave the scheduled task in the non-active state.

### Claim 11 (depends on 8)

The scheduling module according to claim 8, wherein the scheduling module further comprises hardware logic arranged to, dependent upon an outcome of the check, either remove all scheduled tasks in the group from the non-active state or leave the scheduled task in the non-active state.

### Claim 12 (depends on 8)

The scheduling module according to claim 8, wherein the ALU is an atomic ALU.

### Claim 13

A processor comprising:
a scheduling module including:
one or more queues arranged to store scheduled tasks; and
hardware logic arranged, in response to receiving an instruction from an instruction decoder decoding a synchronization instruction in a scheduled task from a group of tasks, to place the scheduled task into a non-active state wherein no instructions in the scheduled task are executed except for the synchronization instruction;
wherein the decoded synchronization instruction causes an ALU to perform a check on data assigned to the group of tasks.

### Claim 14 (depends on 13)

The processor according to claim 13, wherein the scheduling module further comprises hardware logic arranged, in response to receiving a message from an ALU indicating that the data assigned to the group of tasks passed a check, to remove all scheduled tasks in the group from the non-active state.

### Claim 15 (depends on 13)

The processor according to claim 13, wherein the scheduling module further comprises hardware logic arranged, in response to receiving a message from an ALU indicating that the data assigned to the group of tasks has failed a check, to leave the scheduled task in the non-active state.

### Claim 16 (depends on 13)

The processor according to claim 13, wherein the scheduling module further comprises hardware logic arranged to, dependent upon an outcome of the check, either remove all scheduled tasks in the group from the non-active state or leave the scheduled task in the non-active state.

### Claim 17 (depends on 13)

The processor according to claim 13, wherein the ALU is an atomic ALU.

### Claim 18 (depends on 13)

The processor according to claim 13, further comprising:
a processing block comprising an ALU arranged to perform an operation on data assigned to the group of tasks and to perform a check on the updated data assigned to the group of tasks, and in response to the updated data passing the check, to send a message from the ALU to the scheduling module indicating that the data assigned to the group of tasks passed the check.

### Claim 19 (depends on 18)

The processor according to claim 18, wherein the ALU is further arranged, in response to the updated data passing the check, to reset the data assigned to the group of tasks.
