What coordinate space dose Fragment shader's input in?

I use " i.position = UnityObjectToClipPos(v.position);" to convert point from model space to clip space. The "clip space" I understand is x,y,z all in [-w,w],these points which x,y,z not in [w,-w] would be clipped. Then My question is in fragment shader what space dose " i.position" in?
Shader "Unlit/CustomShader"
{
SubShader {

Pass {
CGPROGRAM
#pragma vertex MyVertexProgram
#pragma fragment MyFragmentProgram
#include "UnityCG.cginc"

struct VertexData {
float4 position : POSITION;
};
struct Interpolators {
float4 position : SV_POSITION;
};
Interpolators MyVertexProgram (VertexData v) {
Interpolators i;
i.position = UnityObjectToClipPos(v.position);
return i;
}

float4 MyFragmentProgram (Interpolators i) : SV_TARGET {
return i.position;
}
ENDCG
}
}
}
the result is object became "yellow" when the camera move too close to it(as figure 1). That means the x and y always be larger than 1.
Then I replace the return value with "return i.position/i.position.w". I thought this would convert coordinate to UDC space which [-1,1],But the result is still "yellow"(as figure 2).That means the x and y always be larger than 1.
So I confused. I thought the left half of the sphere will be a little blue as x < 0.

 

https://answers.unity.com/questions/1443941/shaders-what-is-clip-space.html

 

I wonder the input of fragment shader. Is it in clip space or NDC?

 

The result picture shows it is not clip space or NDC. Or I make mistake somewhere?

 

The XY values of SV_Position in the fragment shader is the pixel coordinates (0 to the resolution), Z is depth (0.0 to 1.0), W is 1.0, the same as VPOS (or gl_FragCoord in GLSL).

 

As I understand, the the x coordinate is from 0-1 as left to right, y coordinate is from 0-1 as bottom to top, on openGL.
If so why the color is yellow ,because the yellow is (1,1,0) which x and y is 1.following figure is using a orthographic camera, which frustum is a cube as the object. if frag shader's input i.position is in pixel coordinates then the cube's left half should be a color whose r is small such as 0.But it is yellow.@bgolus
I confused.

 

The bottom left pixel is 0,0, the top right pixel is screen resolution XY - 1. So if you're rendering at 640x480 the bottom left is 0,0 and the top right is 639,479. It's yellow because the X and Y are 1 or greater for almost every pixel on screen.

Between the output of the vertex shader and the input of the fragment shader that SV_Position value is used for rasterization. It starts as homogeneous clip space when output from the vertex shader, and is transformed into NDC and ultimately viewport space / pixel positions before getting to the fragment shader. If you want the NDC in the fragment you have to pass the clip space position as an additional semantic (like TEXCOORD#) and do the perspective divide on your own.

https://forum.unity.com/threads/wha...shader-input-sv_position.520499/#post-3413867

 

@bgolus Thanks a lot. I don't confused now. thank you