Hi, This is Haitao who is working with Niko, Dan and Ivan on SIMD128 API (https://github.com/johnmccutchan/ecmascript_simd). Jandem pointed out the similarity between Float32x4 (4 floats in a 128-bit SIMD register) operations and Float32 optimization. After reading the Float32 optimization commit (https://bugzilla.mozilla.org/show_bug.cgi?id=888109), I have a question on the Float32 codegen.
Why did not we use 4-byte stack slot and 4-byte snapshot slot for float32 and use movss instead of movsd? If we introduced FLOAT32_REG type in the LDefinition::Type and FLOAT32_REG in the LAllocation::Kind, it should be relatively easier to add Float32x4_REG. Dan worked out a series of patch on this at https://bugzilla.mozilla.org/show_bug.cgi?id=949668, please have a look there if you are interested with this question. Thanks -Haitao _______________________________________________ dev-tech-js-engine-internals mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals
