float4 shadowpos = mul(float4(inputpos,1.0), modelMatrix);
shadowpos = mul(shadowpos, shadowVPMatrix);
return shadowpos;

And it renders a plausible shadow map.

Now in my lighting pass, I am doing:

float shadow = 1.0;
float4 lightspacePos = mul(float4(worldpos, 1.0),shadowVPMatrix);
float shadowSample = shadowmap.Sample(sampler, lightspacePos.xy).r;
if (compareforshadow(lightspacePos.z, shadowSample ))
{
    shadow = 0.0;
}

cbuffer WVP_CBUFFER : register( b0 )
{
  float4x4 WorldViewProjection;
};

VS_OUTPUT main( in VS_INPUT Input )
{
  VS_OUTPUT Output;
  Output.Position = mul( float4( Input.Position, 1.0f ), WorldViewProjection );
  Output.TexCoord = Input.TexCoord;
  return Output;
}

Then in the pixel shader of the mesh rendering :

// Transform the view position to shadow space.
float4 PositionCS = mul( float4( PositionVS, 1.0f ), SunViewProjection );
  
// Transform from NDC space to texture space.
float2 ShadowTexCoord = 0.5f * float2( PositionCS.x, -PositionCS.y ) + 0.5f;

// Compute the shadow factor.
float ShadowFactor = ShadowMap.SampleCmpLevelZero( ShadowSampler, ShadowTexCoord, PositionCS.z - 0.006f );

0.006f is an arbitrary bias but it has to be set by scene.
When the projection is orthographic which is the case for the sun shadow map, you don't have to divide by W because it's like dividing by 1.0f.