redundant 'float16_t' conversion triggers optimization internal error

When compiling the following shader with optimization, glslc throws internal error. Although the 'float16_t' conversion is redundant, it should be harmful. Compiling without optimization, or use glslang directly glslang --target-env vulkan1.3 -Os issue2.comp is OK.

The issue keeps if we replace 'float16_t' with other 16-bit types, like 'uint16_t' or 'int16_t', but the issue disappears with other types, like 'float', 'int' or 'uint'

$ glslc --target-env=vulkan1.3 -O issue2.comp
shaderc: internal error: compilation succeeded but failed to optimize: Expected input to have different bit width from Result Type: FConvert
  %33 = OpFConvert %half %31

#version 450
#extension GL_EXT_shader_16bit_storage : require
layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;

#define TYPE float16_t
layout(binding = 0) readonly buffer A { TYPE data_a[]; };
layout(binding = 1) writeonly buffer D { TYPE data_d[]; };

void main() {
    const uint i = gl_GlobalInvocationID.x;
    data_d[i] = TYPE(data_a[i]);
}

The case is minified from https://github.com/ggerganov/llama.cpp/blob/adc9ff384121f4d550d28638a646b336d051bf42/ggml_vk_generate_shaders.py#L2057C1-L2079C4, when I was curious why there exists optimization workaround.

Version

$ glslc --version
2024.0
2024.1
14.1.0
Target: SPIR-V 1.0

$ glslang --version
Glslang Version: 11:14.1.0
ESSL Version: OpenGL ES GLSL 3.20 glslang Khronos. 14.1.0
GLSL Version: 4.60 glslang Khronos. 14.1.0
SPIR-V Version 0x00010600, Revision 1
GLSL.std.450 Version 100, Revision 1
Khronos Tool ID 8
SPIR-V Generator Version 11
GL_KHR_vulkan_glsl version 100
ARB_GL_gl_spirv version 100

OS: Archlinux

google / shaderc

redundant 'float16_t' conversion triggers optimization internal error #1418

Version