webgpu: Support any component buffer #7426
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
qjia7
force-pushed
the
support_any_component_buffer
branch
from
092cc9f to
bffb703
Compare
gyagp
reviewed
Feb 28, 2023
| @@ -24,6 +24,7 @@ import {computeDispatch, flatDispatchLayout} from './webgpu_util'; | |||
| export class BinaryOpProgram implements WebGPUProgram { | |||
| dispatch: [number, number, number]; | |||
| dispatchLayout: {x: number[]}; | |||
| outputComponent = 1; | |||
| @@ -32,6 +32,7 @@ export class Conv2DDerInputProgram implements WebGPUProgram { | |||
| size = false; | |||
| isVec4 = false; | |||
| workPerThread = 1; | |||
| outputComponent = 1; | |||
| @@ -36,6 +36,7 @@ export class MatMulSplitKProgram implements WebGPUProgram { | |||
| transposeB: boolean; | |||
| atomic = true; | |||
| isVec4 = false; | |||
| outputComponent = 1; | |||
| @@ -67,6 +68,21 @@ export const compileProgram = | |||
| return pipeline; | |||
| }; | |||
|
|
|||
| export const typeSnippet = (component: number, type = 'f') => { | |||
There was a problem hiding this comment.
I'd prefer passing f32, i32, etc. as type. We may extend this to f16?
| return 'vec4<i32>'; | ||
| default: | ||
| throw new Error(`${component}-component is not supported.`); | ||
| } |
There was a problem hiding this comment.
if and else can be merged to simplify the impl.
| @@ -159,7 +159,7 @@ export class Conv2DMMProgram implements WebGPUProgram { | |||
| dispatchLayout: {x: number[], y: number[], z: number[]}; | |||
| dispatch: [number, number, number]; | |||
| variableNames = ['x', 'W']; | |||
| variableTypes: string[]; | |||
| variableComponents: number[]; | |||
There was a problem hiding this comment.
In future, can we change this to inputNames and inputComponents?
| @@ -176,6 +176,7 @@ export class Conv2DMMProgram implements WebGPUProgram { | |||
| tileInner: number; | |||
| innerElementSize: number; | |||
| isVec4?: boolean; | |||
| outputComponent = 1; | |||
| @@ -143,7 +144,8 @@ export class Conv2DDerInputMMProgram implements WebGPUProgram { | |||
| this.elementsPerThread); | |||
|
|
|||
| if (this.isVec4) { | |||
| program.isVec4, | ||
| program.variableComponents ? | ||
| program.variableComponents[i] : | ||
| program.outputComponent ? program.outputComponent : 1, |
There was a problem hiding this comment.
Can we just pass program.outputComponent here?
gyagp
reviewed
Feb 28, 2023
gyagp
approved these changes
Mar 1, 2023
| @@ -158,6 +174,8 @@ function makeShader( | |||
| const prefixSnippets: string[] = []; | |||
| const flatWorkgroupSize = program.workgroupSize[0] * | |||
| program.workgroupSize[1] * program.workgroupSize[2]; | |||
| program.outputComponent = | |||
| program.outputComponent ? program.outputComponent : 1; | |||
There was a problem hiding this comment.
It's not good to set a class member outside. We can leave this for future refactoring.
| @@ -67,6 +68,21 @@ export const compileProgram = | |||
| return pipeline; | |||
| }; | |||
|
|
|||
| export const typeSnippet = (component: number, type = 'f32') => { | |||
There was a problem hiding this comment.
The parameter orders of typeSnippet and dataTypeToGPUType are different. We may also refactor this part in future.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.
This change is