From cde8661bc17dcbaf98769d44cb71f2cf2b1297d0 Mon Sep 17 00:00:00 2001
From: Philip Rebohle <philip.rebohle@tu-dortmund.de>
Date: Sun, 3 Nov 2024 10:53:39 +0100
Subject: [PATCH] [dxvk] Add helper to compute clear values for compressed
 formats

---
 src/dxvk/dxvk_util.h | 100 ++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 99 insertions(+), 1 deletion(-)

diff --git a/src/dxvk/dxvk_util.h b/src/dxvk/dxvk_util.h
index 967458faf..edeb22398 100644
--- a/src/dxvk/dxvk_util.h
+++ b/src/dxvk/dxvk_util.h
@@ -386,5 +386,103 @@ namespace dxvk::util {
   
   bool isDualSourceBlendFactor(
           VkBlendFactor               factor);
-  
+
+  /**
+   * \brief Computes unsigned normalized value
+   *
+   * Doesn't necessarily do perfect rounding.
+   * \param [in] float Float value
+   * \param [in] bits Bit count
+   * \returns Normalized value
+   */
+  inline uint32_t computeUnorm(float f, uint32_t bits) {
+    f = std::max(f, 0.0f);
+    f = std::min(f, 1.0f);
+    return uint32_t((f * float((1u << bits) - 1u)) + 0.5f);
+  }
+
+  /**
+   * \brief Computes signed normalized value
+   *
+   * Doesn't necessarily do perfect rounding.
+   * \param [in] float Float value
+   * \param [in] bits Bit count
+   * \returns Normalized value
+   */
+  inline uint32_t computeSnorm(float f, uint32_t bits) {
+    f = std::max(f, -1.0f);
+    f = std::min(f,  1.0f);
+    return int32_t((f * float((1u << (bits - 1u)) - 1u)) + (f < 0.0f ? -0.5f : 0.5f));
+  }
+
+  /**
+   * \brief Computes clear payload for compressed image blocks
+   *
+   * Ignores whether the format is sRGB or not. If this is relevant,
+   * the input color must be converted into the correct space first.
+   * Does not support BC6H or BC7 formats.
+   * \param [in] format Image format
+   * \param [in] color Clear color, as floats
+   * \returns Block data as unsigned integers
+   */
+  inline VkClearColorValue encodeClearBlockValue(
+          VkFormat                    format,
+    const VkClearColorValue&          color) {
+    VkClearColorValue result = { };
+
+    switch (format) {
+      case VK_FORMAT_BC1_RGB_SRGB_BLOCK:
+      case VK_FORMAT_BC1_RGB_UNORM_BLOCK:
+      case VK_FORMAT_BC1_RGBA_SRGB_BLOCK:
+      case VK_FORMAT_BC1_RGBA_UNORM_BLOCK: {
+        // Encode clear color as color0, table will be all zeroes
+        result.uint32[0] = (computeUnorm(color.float32[2], 5) <<  0)
+                         | (computeUnorm(color.float32[1], 6) <<  5)
+                         | (computeUnorm(color.float32[0], 5) << 11);
+      } return result;
+
+      case VK_FORMAT_BC2_SRGB_BLOCK:
+      case VK_FORMAT_BC2_UNORM_BLOCK: {
+        // Alpha is encoded in the first four bytes as four-bit
+        // values. The color portion is identical to BC1.
+        uint32_t alpha = 0x11111111u * computeUnorm(color.float32[3], 4);
+        result.uint32[0] = alpha;
+        result.uint32[1] = alpha;
+        result.uint32[2] = (computeUnorm(color.float32[2], 5) <<  0)
+                         | (computeUnorm(color.float32[1], 6) <<  5)
+                         | (computeUnorm(color.float32[0], 5) << 11);
+      } return result;
+
+      case VK_FORMAT_BC3_UNORM_BLOCK:
+      case VK_FORMAT_BC3_SRGB_BLOCK: {
+        // Encode alpha as alpha0, color portion identical to BC1
+        result.uint32[0] = computeUnorm(color.float32[3], 8);
+        result.uint32[2] = (computeUnorm(color.float32[2], 5) <<  0)
+                         | (computeUnorm(color.float32[1], 6) <<  5)
+                         | (computeUnorm(color.float32[0], 5) << 11);
+      } return result;
+
+      case VK_FORMAT_BC4_SNORM_BLOCK: {
+        result.uint32[0] = computeSnorm(color.float32[0], 8);
+      } return result;
+
+      case VK_FORMAT_BC4_UNORM_BLOCK: {
+        result.uint32[0] = computeUnorm(color.float32[0], 8);
+      } return result;
+
+      case VK_FORMAT_BC5_SNORM_BLOCK: {
+        result.uint32[0] = computeSnorm(color.float32[0], 8);
+        result.uint32[2] = computeSnorm(color.float32[1], 8);
+      } return result;
+
+      case VK_FORMAT_BC5_UNORM_BLOCK: {
+        result.uint32[0] = computeUnorm(color.float32[0], 8);
+        result.uint32[2] = computeUnorm(color.float32[1], 8);
+      } return result;
+
+      default:
+        return color;
+    }
+  }
+
 }