<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/99113>99113</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            Implement the `f32tof16` HLSL Function
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            metabug,
            backend:DirectX,
            HLSL,
            backend:SPIR-V,
            bot:HLSL
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          farzonl
      </td>
    </tr>
</table>

<pre>
    - [ ] Implement `f32tof16` clang builtin,
- [ ] Link `f32tof16` clang builtin with `hlsl_intrinsics.h`
- [ ] Add sema checks for `f32tof16` to `CheckHLSLBuiltinFunctionCall` in `SemaChecking.cpp`
- [ ] Add codegen for `f32tof16` to `EmitHLSLBuiltinExpr` in `CGBuiltin.cpp`
- [ ] Add codegen tests to `clang/test/CodeGenHLSL/builtins/f32tof16.hlsl`
- [ ] Add sema tests to `clang/test/SemaHLSL/BuiltIns/f32tof16-errors.hlsl`
- [ ] Create the `int_dx_f32tof16` intrinsic in `IntrinsicsDirectX.td`
- [ ] Create the `DXILOpMapping` of `int_dx_f32tof16` to  `130` in `DXIL.td`
- [ ] Create the  `f32tof16.ll` and `f32tof16_errors.ll` tests in `llvm/test/CodeGen/DirectX/`
- [ ] Create the `int_spv_f32tof16` intrinsic in `IntrinsicsSPIRV.td`
- [ ] In SPIRVInstructionSelector.cpp create the `f32tof16` lowering and map  it to `int_spv_f32tof16` in `SPIRVInstructionSelector::selectIntrinsic`.
- [ ] Create SPIR-V backend test case in `llvm/test/CodeGen/SPIRV/hlsl-intrinsics/f32tof16.ll`

## DirectX

| DXIL Opcode | DXIL OpName | Shader Model | Shader Stages |
| ----------- | ----------- | ------------ | ------------- |
| 130 | LegacyF32ToF16 | 6.0 | () |

## SPIR-V

# PackHalf2x16:

## Description:
**PackHalf2x16**  
  
Result is the unsigned integer obtained by converting the components of
a two-component floating-point vector to the 16-bit **OpTypeFloat**, and
then packing these two 16-bit integers into a 32-bit unsigned integer.
The first vector component specifies the 16 least-significant bits of
the result; the second component specifies the 16 most-significant
bits.  
  
The *v* operand must be a vector of 2 components whose type is a 32-bit
floating-point.  
  
*Result Type* must be a 32-bit integer type.

<table>
<colgroup>
<col style="width: 20%" />
<col style="width: 20%" />
<col style="width: 20%" />
<col style="width: 20%" />
<col style="width: 20%" />
</colgroup>
<thead>
<tr>
<th>Number</th>
<th>Operand 1</th>
<th>Operand 2</th>
<th>Operand 3</th>
<th>Operand 4</th>
</tr>
</thead>
<tbody>
<tr>
<td class="tableblock halign-left valign-top"><p>58</p></td>
<td
class="tableblock halign-left valign-top"><p><em>&lt;id&gt;</em><br />
<em>v</em></p></td>
<td></td>
<td></td>
<td></td>
</tr>
</tbody>
</table>



## Test Case(s)

 
 ### Example 1
```hlsl
//dxc f32tof16_test.hlsl -T lib_6_8 -enable-16bit-types -O0

export uint4 fn(float4 p1) {
    return f32tof16(p1);
}
```
## HLSL:

Converts an input into a float16 type.

## Syntax

``` syntax
uint f32tof16(
  in float value
);
```

## Parameters

<dl> <dt>

*value* \[in\]
</dt> <dd>

Type: **float**

The input value.

</dd> </dl>

## Return value

Type: **uint**

The converted value, stored in the low-half of the uint.

## Remarks

The following overloaded versions are also available:

``` syntax
uint2 f32tof16(float2 value);
uint3 f32tof16(float3 value);
uint4 f32tof16(float4 value);
```

### Minimum Shader Model

This function is supported in the following shader models.



| Shader Model                                                        | Supported |
|---------------------------------------------------------------------|-----------|
| [Shader Model 4](https://github.com/MicrosoftDocs/win32/blob/docs/desktop-src//direct3dhlsl/dx-graphics-hlsl-sm4.md) and higher shader models | yes       |



 

Although this function is available in [Shader Model 4](https://github.com/MicrosoftDocs/win32/blob/docs/desktop-src//direct3dhlsl/dx-graphics-hlsl-sm4.md) and higher shader models, because it is emulated in 4.0 and 4.1, it is less perfomant on these shader models than it is on [Shader Model 5](https://github.com/MicrosoftDocs/win32/blob/docs/desktop-src//direct3dhlsl/d3d11-graphics-reference-sm5.md).

This function is supported in the following types of shaders:



| Vertex | Hull | Domain | Geometry | Pixel | Compute |
|--------|------|--------|----------|-------|---------|
| x      | x    | x      | x        | x     | x       |



 

## See also

<dl> <dt>

[Intrinsic Functions](https://github.com/MicrosoftDocs/win32/blob/docs/desktop-src//direct3dhlsl/dx-graphics-hlsl-intrinsic-functions.md)
</dt> <dt>

[Shader Model 5](https://github.com/MicrosoftDocs/win32/blob/docs/desktop-src//direct3dhlsl/d3d11-graphics-reference-sm5.md)
</dt> </dl>
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJzUWF9v2zgS_zTMCyFDomwnfvCDY8fbAO22aIJi3wJKGlm8UKRAUo5zn_4w1P_Yde_2Dne3gdGKw-HM8Df_SHJrxUEBrMninix2N7x2hTbrnJu_ayVvEp29rwNKFveULHb0sawklKAcJcswj5nTebQky5CmkqsDTWohnVCEbUm4GVZ9Fur16gL6JlyBHIW08kUoZ4SyIrWzgizDiahNllELJadpAemrpbk2HyU7jZQtzn_6_PT5vlGxr1XqhFZbLiVyCYVcT1ByzynUYZZW1SV1qc7gAOrnqh5K4UaaHk6VGTRsf2vJvxLvwDrbSvTgELZHGmH7rc7gN1Cog7B9i5klbN_ZMkPcfgrVFcm4_1asN_NxIjYAY7SxF6VvDXAH1BWAYoVyL9npZYxN78UWiMfeqzthIHV_zFx2Xejuj8fPX6svvKqEOqBInf9El9MUZ6I4HIDH1ddVjL05a6KCq2xMfWkBaCYbHBvpUh7LMwcRtm_3Rtj-14DZ6vhPIvb07fH7jwubeVTUTz0q60zt4_sJJKROGww3mk50jnVJ_QZGqIPfcckrSoVrQ-SyaT5bfqKLxBsSb6wf9UaTZTi7BAAKCX7QhKevoDIPKk25hV8A63UTtsdYDIYKMU4C2QYp_lhMWEw7bzS02y3FoKBfK8w5Ohr_zstm_FTwDAz9ojOQY8KT4wewSGkFBcMf_cX4nBCMBEVx6Oc_w4Gn7_uYPet9tPSk5ayZIuyOsFW3pt9cA2RPot94-vqJy5ydoiV6ZAIE2NSICp3WTjH8TZZ4CqUkbP75DraWjgrrw6dWvk1kGKRwAEN14rhAQvJOU62OYByGE_Kmuqy0AuUs1TkJN5y6Nx30VJpLzZE5qLRQjh59DGHw4eJoGSTC0caar9XzewV75G8IhG0xYkm4cQUoWnFfuXGhBdTSLW-txGx1mnIaM0_-uAsM0OcCaC6M7Q0ZDLUVpCIXYFvLqARuXYAyRC5SrhxNRLdLZDEeMxLf-wUWUq2ya_JKPRVHwg0KnA1eQOsI2xzRNboC49O1to4mQHlnsc4pG6P-VmiE470CdF-3exJupsiP1BDW-RsBR2WDkha7zvEodtYGV7x1PJFA4odmlGp5MLquxgRq3Tuy7AhjbyJzBYk3lIWELQhjFAvlX5mZsP3Zpl0BPBtGZjRB4off6zIB0yz1hNHc19bD0fVpdn06vj49P5vGkZmMpjvAQ-DF_WR4jLO2AcqHQiJ1-koLLsVBBRJyR4_Nt9MVYQyXxluEanHXqKoaEiodacQM_9OiSbyFEv9nS5-MIiNsecCvRlEzGW8TM3GlJx-nLFcM_Dep55hPUEbCKLXCaTl_xqa55RYIu7OErZpJn80NBzI9nDie1mmEs8uw-fmjHMrZE7bPTintjzrYcf1JjwbPVIrkZflyRwNQaEUQLRPhAkx9S4OvbZOFU6WNo7VQbk5zRdidry9zWkVNv7r35YVSA642qldF2J3nQIdgE9yNDRw26c-lXR_bNi3GUq6oUFXtusruVUbLSVlq--O7cvzUkjrx1HZUNHtskrdVKLJlZBN6qRhfNXiBna1jK3tF37jhJTgwti-LmSTxA8UPN3iQbRqBbEPJYksW90L5_3e9yz23X5b1y3xBjjdtR8xHvbCZLqAFxAsfKjOKy1px-C1Hhnizvzde6Tf5URUC9FFT2-kha5exLbVOG99RfVOT-i0ouMyxJ_lzA_aZD2pLbl7tIDPXUuo3bOP6CEZqnqF4MFZoZSk3QLm0mvIjF7JJic1Vp7KxVz1erLO2dSNyxWdc8QWu-RnX_APXxZDAfX4RSpR1OTlSdpsWlubtfRQ7tK0rzKQBxQES26wucbWdTUrBx-Pqn_zzcnoD-oNp8J_4m8oZDr1kcT8xfY5JwO4K5yqL7vXV6SBcUSezVON94ItIjbY6dzvtD_1vQsUML8NSJxjdDTUD--p0FViTthXOXwDizJc9LHjBwfCqEKkN_EXClvNZmWG1ws5YiEMBZoq5x-cd7IDWxAm-WjTfG-kKXR8K6j76tw9df8f56-wd0zuBlNd4O_NXAShrydtAnc-a-_J8FiFjwyDBWlqByXWJx2Ot2rP5FFNXYBn3C_Q5Iov_FiJxFkUDKAZyMKBSCGy5aJCZ_esJ2_RInbc7tkO16rP2B1bQkw-sT7Vs7pk7XXIMjtst_Q10Cc68-8E3cWpvoltdVrWD8wztPy_RpoMxecjG01AITpOPyed4MCZfyYiuEUNTwq93x8V9_3RAu8c6-7_Kjf6BIejcbpuIuNCqJ3v4Pw3kc7O7I8FNto6zVbziN7COblkULVar2_imWLMwW94leZpwWIUJLBZhmN2leTqP40UIt8sbsWYhm4e30TJaLVZxPItXKzbnwBd5Pr_N4yWZh1ByIWdSHsuZNocbYW0N69UqiuIbyROQ1j87M1aC40l9wGM82xLG2tchEm_6B7V2pnmtPGNrH0P6Ce1IvOl4F7sbs0YjgqQ-WDIPpbDODmY54SSsh7ftC-9lKKoPy5vayPUVr7ZvWF5jZfTfIHWE7f3m0Z3N_o9r9o8AAAD__0-j7XA">