[webkit-reviews] review granted: [Bug 195771] [WHLSL] Implement zero-filling of types : [Attachment 371047] patch
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Wed Jun 5 10:01:44 PDT 2019
Myles C. Maxfield <mmaxfield at apple.com> has granted review:
Bug 195771: [WHLSL] Implement zero-filling of types
https://bugs.webkit.org/show_bug.cgi?id=195771
Attachment 371047: patch
https://bugs.webkit.org/attachment.cgi?id=371047&action=review
--- Comment #5 from Myles C. Maxfield <mmaxfield at apple.com> ---
Comment on attachment 371047
--> https://bugs.webkit.org/attachment.cgi?id=371047
patch
View in context: https://bugs.webkit.org/attachment.cgi?id=371047&action=review
Does the Metal compiler still produce warnings now? GPURenderPipelineMetal.mm
has a FIXME with a link to this bug.
> Source/WebCore/Modules/webgpu/WHLSL/WHLSLAutoInitializeVariables.cpp:54
> + // Skip argument declarations.
Cool.
> Source/WebCore/Modules/webgpu/WHLSL/WHLSLAutoInitializeVariables.cpp:66
> + printStream.print(TypeDumper(*type));
🤔 This seems unwise, though I do understand how it isn't observable. Are you
confident this is a good idea?
> Source/WebCore/Modules/webgpu/WHLSL/WHLSLChecker.cpp:-360
> - auto* overload = resolveFunctionOverloadImpl(*getterFuncs,
argumentTypeReferences, nullptr);
Excellent.
> Source/WebCore/Modules/webgpu/WHLSL/AST/WHLSLVariableDeclaration.h:73
> + void setInitializer(UniqueRef<Expression> expression)
Doesn't this need to take an rvalue reference? UniqueRef is move-only (I
think???)
> Source/WebCore/Modules/webgpu/WHLSL/Metal/WHLSLFunctionWriter.cpp:453
>
m_stringBuilder.append(makeString(m_typeNamer.mangledNameForType(*variableDecla
ration.type()), ' ', variableName, ";\n"));
Is this dead code now?
> Source/WebCore/Modules/webgpu/WHLSL/Metal/WHLSLNativeFunctionWriter.cpp:110
> + stringBuilder.append(makeString(" for (size_t i = 0; i <
sizeof(", metalReturnName, "); ++i) {\n"));
> + stringBuilder.append(" ptr[i] = 0;\n");
> + stringBuilder.append(" }\n");
Can we remove the { } and pull it up into one line?
More information about the webkit-reviews
mailing list