[webkit-changes] [WebKit/WebKit] dd88d9: [WGSL] quantizeToF16 generates invalid code for ve...

Tadeu Zagallo noreply at github.com
Fri Jan 5 05:45:56 PST 2024


  Branch: refs/heads/main
  Home:   https://github.com/WebKit/WebKit
  Commit: dd88d938e7386c6e2501dca6241afdece3abf458
      https://github.com/WebKit/WebKit/commit/dd88d938e7386c6e2501dca6241afdece3abf458
  Author: Tadeu Zagallo <tzagallo at apple.com>
  Date:   2024-01-05 (Fri, 05 Jan 2024)

  Changed paths:
    M Source/WebGPU/WGSL/Metal/MetalFunctionWriter.cpp

  Log Message:
  -----------
  [WGSL] quantizeToF16 generates invalid code for vectors
https://bugs.webkit.org/show_bug.cgi?id=267108
rdar://120500374

Reviewed by Mike Wyrzykowski.

The code generator was always emitting `float(half(x))` for quantizeToF16, but
it should to be `floatN(halfN(x))` if the argument is a vector of size N.

* Source/WebGPU/WGSL/Metal/MetalFunctionWriter.cpp:
(WGSL::Metal::emitQuantizeToF16):

Canonical link: https://commits.webkit.org/272685@main




More information about the webkit-changes mailing list