[PATCH 0/1] MR512: vkd3d-shader/spirv: Always emit clip/cull builtins as an array.

[Conor McCarthy (＠cmccarthy)]

Not known to occur anywhere, but it is a regression from release 1.7. We need a test for SV_ClipDistance in a pixel shader, but it's a bit late to add that before the release.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512

[PATCH 1/1] vkd3d-shader/spirv: Always emit clip/cull builtins as an array.

[Conor McCarthy]

From: Conor McCarthy <cmccarthy(a)codeweavers.com>

Clip/cull distance can appear as input in pixel shaders, and the
array size must not be forced to zero.
---
 libs/vkd3d-shader/spirv.c                | 5 ++++-
 libs/vkd3d-shader/vkd3d_shader_private.h | 5 +++++
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c
index 2197b74dd..f72875e1a 100644
--- a/libs/vkd3d-shader/spirv.c
+++ b/libs/vkd3d-shader/spirv.c
@@ -4714,8 +4714,11 @@ static uint32_t spirv_compiler_emit_input(struct spirv_compiler *compiler,
 
     array_sizes[0] = (reg->type == VKD3DSPR_PATCHCONST ? 0 : compiler->input_control_point_count);
     array_sizes[1] = signature_element->register_count;
-    if (array_sizes[1] == 1 && !vsir_sysval_semantic_is_tess_factor(signature_element->sysval_semantic))
+    if (array_sizes[1] == 1 && !vsir_sysval_semantic_is_tess_factor(signature_element->sysval_semantic)
+            && (!vsir_sysval_semantic_is_clip_cull(signature_element->sysval_semantic) || array_sizes[0]))
+    {
         array_sizes[1] = 0;
+    }
 
     write_mask = signature_element->mask;
 
diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h
index d3989672b..267b467ee 100644
--- a/libs/vkd3d-shader/vkd3d_shader_private.h
+++ b/libs/vkd3d-shader/vkd3d_shader_private.h
@@ -980,6 +980,11 @@ static inline bool vsir_sysval_semantic_is_tess_factor(enum vkd3d_shader_sysval_
         && sysval_semantic <= VKD3D_SHADER_SV_TESS_FACTOR_LINEDEN;
 }
 
+static inline bool vsir_sysval_semantic_is_clip_cull(enum vkd3d_shader_sysval_semantic sysval_semantic)
+{
+    return sysval_semantic == VKD3D_SHADER_SV_CLIP_DISTANCE || sysval_semantic == VKD3D_SHADER_SV_CULL_DISTANCE;
+}
+
 struct signature_element *vsir_signature_find_element_for_reg(const struct shader_signature *signature,
         unsigned int reg_idx, unsigned int write_mask);
 void shader_signature_cleanup(struct shader_signature *signature);
-- 
GitLab

https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512

[Conor McCarthy (＠cmccarthy)]

More accurately, not known to occur in any game supported upstream, but some Diablo IV SM 6 shaders have this issue.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512#note_55188

[Zebediah Figura (＠zfigura)]

Oops. Good catch.
> We need a test for SV_ClipDistance in a pixel shader, but it's a bit late to add that before the release.

Is this on your list then? If not I can try to write one myself.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512#note_55229

[Conor McCarthy (＠cmccarthy)]

On Thu Dec  7 01:33:54 2023 +0000, Zebediah Figura wrote:
> Oops. Good catch.
> > We need a test for SV_ClipDistance in a pixel shader, but it's a bit
> > late to add that before the release.
> > Is this on your list then? If not I can try to write one myself.
> > It's a "when I get around to it" item, so please feel free to write one.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512#note_55308

[PATCH 0/1] MR512: vkd3d-shader/spirv: Always emit clip/cull builtins as an array. - closed

[Conor McCarthy (＠cmccarthy)]

This merge request was closed by Conor McCarthy.
-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512

[Conor McCarthy (＠cmccarthy)]

Went upstream with the release MR.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/512#note_55309