tint/writer/glsl: Support for F16 type, constructor, and convertor
This patch make GLSL writer support emitting f16 types, f16 literals,
f16 constructor and convertor. Unittests are also implemented, and
end-to-end testcases are updated for GLSL extension requirement.
The GLSL writer will require `GL_AMD_gpu_shader_half_float` extension if
f16 WGSL extension is enabled, emit f16 literal as `1.23hf`, and map f16
types as follow:
WGSL type -> GLSL type
f16 -> float16_t
vec2<f16> -> f16vec2
vec3<f16> -> f16vec3
vec4<f16> -> f16vec4
mat2x2<f16> -> f16mat2
mat2x3<f16> -> f16mat2x3
mat2x4<f16> -> f16mat2x4
mat3x2<f16> -> f16mat3x2
mat3x3<f16> -> f16mat3
mat3x4<f16> -> f16mat3x4
mat4x2<f16> -> f16mat4x2
mat4x3<f16> -> f16mat4x3
mat4x4<f16> -> f16mat4
Bug: tint:1473, tint:1502
Change-Id: I7e788f82be2873911961c891a644200c1cbb74db
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/95684
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
0d3d3210d4