util: implement R11G11B10_FLOAT pack/unpack functions

Reviewed-by: Brian Paul <[email protected]>

4 files changed, 336 insertions, 1 deletions

diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv
index 990bcff7234..a8baad111f1 100644
--- a/src/gallium/auxiliary/util/u_format.csv
+++ b/src/gallium/auxiliary/util/u_format.csv
@@ -147,7 +147,7 @@ PIPE_FORMAT_G8R8_G8B8_UNORM      , subsampled, 2, 1, x32 ,     ,     ,     , xyz
 
 # some special formats not fitting anywhere else
 PIPE_FORMAT_R10G10B10A2_USCALED   , plain,      1,  1, u10 , u10 , u10 , u2  , xyzw, rgb
-PIPE_FORMAT_R11G11B10_FLOAT       , plain,      1,  1, f11 , f11 , f10 ,     , xyz1, rgb
+PIPE_FORMAT_R11G11B10_FLOAT       , other,      1,  1, x32 ,     ,     ,     , xyz1, rgb
 PIPE_FORMAT_R9G9B9E5_FLOAT        , other,      1,  1, x32 ,     ,     ,     , xyz1, rgb
 PIPE_FORMAT_R1_UNORM              , other,      8,  1, x8  ,     ,     ,     , x001, rgb
 # A.k.a. D3DFMT_CxV8U8
diff --git a/src/gallium/auxiliary/util/u_format_other.c b/src/gallium/auxiliary/util/u_format_other.c
index 1beb61868eb..c23f4ee4ac5 100644
--- a/src/gallium/auxiliary/util/u_format_other.c
+++ b/src/gallium/auxiliary/util/u_format_other.c
@@ -29,6 +29,7 @@
 #include "u_math.h"
 #include "u_format_other.h"
 #include "u_format_rgb9e5.h"
+#include "u_format_r11g11b10f.h"
 
 
 void
@@ -150,6 +151,124 @@ util_format_r9g9b9e5_float_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_strid
 
 
 void
+util_format_r11g11b10_float_unpack_rgba_float(float *dst_row, unsigned dst_stride,
+                                        const uint8_t *src_row, unsigned src_stride,
+                                        unsigned width, unsigned height)
+{
+   unsigned x, y;
+   for(y = 0; y < height; y += 1) {
+      float *dst = dst_row;
+      const uint8_t *src = src_row;
+      for(x = 0; x < width; x += 1) {
+         uint32_t value = *(const uint32_t *)src;
+#ifdef PIPE_ARCH_BIG_ENDIAN
+         value = util_bswap32(value);
+#endif
+         r11g11b10f_to_float3(value, dst);
+         dst[3] = 1; /* a */
+         src += 4;
+         dst += 4;
+      }
+      src_row += src_stride;
+      dst_row += dst_stride/sizeof(*dst_row);
+   }
+}
+
+void
+util_format_r11g11b10_float_pack_rgba_float(uint8_t *dst_row, unsigned dst_stride,
+                                      const float *src_row, unsigned src_stride,
+                                      unsigned width, unsigned height)
+{
+   unsigned x, y;
+   for(y = 0; y < height; y += 1) {
+      const float *src = src_row;
+      uint8_t *dst = dst_row;
+      for(x = 0; x < width; x += 1) {
+         uint32_t value = float3_to_r11g11b10f(src);
+#ifdef PIPE_ARCH_BIG_ENDIAN
+         value = util_bswap32(value);
+#endif
+         *(uint32_t *)dst = value;
+         src += 4;
+         dst += 4;
+      }
+      dst_row += dst_stride;
+      src_row += src_stride/sizeof(*src_row);
+   }
+}
+
+void
+util_format_r11g11b10_float_fetch_rgba_float(float *dst, const uint8_t *src,
+                                       unsigned i, unsigned j)
+{
+   uint32_t value = *(const uint32_t *)src;
+#ifdef PIPE_ARCH_BIG_ENDIAN
+   value = util_bswap32(value);
+#endif
+   r11g11b10f_to_float3(value, dst);
+   dst[3] = 1; /* a */
+}
+
+
+void
+util_format_r11g11b10_float_unpack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
+                                         const uint8_t *src_row, unsigned src_stride,
+                                         unsigned width, unsigned height)
+{
+   unsigned x, y;
+   float p[3];
+   for(y = 0; y < height; y += 1) {
+      uint8_t *dst = dst_row;
+      const uint8_t *src = src_row;
+      for(x = 0; x < width; x += 1) {
+         uint32_t value = *(const uint32_t *)src;
+#ifdef PIPE_ARCH_BIG_ENDIAN
+         value = util_bswap32(value);
+#endif
+         r11g11b10f_to_float3(value, p);
+         dst[0] = float_to_ubyte(p[0]); /* r */
+         dst[1] = float_to_ubyte(p[1]); /* g */
+         dst[2] = float_to_ubyte(p[2]); /* b */
+         dst[3] = 255; /* a */
+         src += 4;
+         dst += 4;
+      }
+      src_row += src_stride;
+      dst_row += dst_stride/sizeof(*dst_row);
+   }
+}
+
+
+void
+util_format_r11g11b10_float_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
+                                       const uint8_t *src_row, unsigned src_stride,
+                                       unsigned width, unsigned height)
+{
+   unsigned x, y;
+   float p[3];
+   for(y = 0; y < height; y += 1) {
+      const uint8_t *src = src_row;
+      uint8_t *dst = dst_row;
+      for(x = 0; x < width; x += 1) {
+         uint32_t value;
+         p[0] = ubyte_to_float(src[0]);
+         p[1] = ubyte_to_float(src[1]);
+         p[2] = ubyte_to_float(src[2]);
+         value = float3_to_r11g11b10f(p);
+#ifdef PIPE_ARCH_BIG_ENDIAN
+         value = util_bswap32(value);
+#endif
+         *(uint32_t *)dst = value;
+         src += 4;
+         dst += 4;
+      }
+      dst_row += dst_stride;
+      src_row += src_stride/sizeof(*src_row);
+   }
+}
+
+
+void
 util_format_r1_unorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
                                   const uint8_t *src_row, unsigned src_stride,
                                   unsigned width, unsigned height)
diff --git a/src/gallium/auxiliary/util/u_format_other.h b/src/gallium/auxiliary/util/u_format_other.h
index 98c97be33fa..2f6a908bbe8 100644
--- a/src/gallium/auxiliary/util/u_format_other.h
+++ b/src/gallium/auxiliary/util/u_format_other.h
@@ -57,6 +57,32 @@ util_format_r9g9b9e5_float_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_strid
                                        const uint8_t *src_row, unsigned src_stride,
                                        unsigned width, unsigned height);
 
+
+void
+util_format_r11g11b10_float_unpack_rgba_float(float *dst_row, unsigned dst_stride,
+                                        const uint8_t *src_row, unsigned src_stride,
+                                        unsigned width, unsigned height);
+
+void
+util_format_r11g11b10_float_pack_rgba_float(uint8_t *dst_row, unsigned dst_stride,
+                                      const float *src_row, unsigned src_stride,
+                                      unsigned width, unsigned height);
+
+void
+util_format_r11g11b10_float_fetch_rgba_float(float *dst, const uint8_t *src,
+                                       unsigned i, unsigned j);
+
+void
+util_format_r11g11b10_float_unpack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
+                                         const uint8_t *src_row, unsigned src_stride,
+                                         unsigned width, unsigned height);
+
+void
+util_format_r11g11b10_float_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
+                                       const uint8_t *src_row, unsigned src_stride,
+                                       unsigned width, unsigned height);
+
+
 void
 util_format_r1_unorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
                                   const uint8_t *src_row, unsigned src_stride,
diff --git a/src/gallium/auxiliary/util/u_format_r11g11b10f.h b/src/gallium/auxiliary/util/u_format_r11g11b10f.h
new file mode 100644
index 00000000000..c4181d0e34e
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_r11g11b10f.h
@@ -0,0 +1,190 @@
+/*
+ * Copyright (C) 2011 Marek Olšák <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/* Based on code from The OpenGL Programming Guide / 7th Edition, Appendix J.
+ * Available here: http://www.opengl-redbook.com/appendices/
+ * The algorithm in the book contains a bug though, which is fixed in the code
+ * below.
+ */
+
+#define UF11_EXPONENT_BIAS   15
+#define UF11_EXPONENT_BITS   0x1F
+#define UF11_EXPONENT_SHIFT  6
+#define UF11_MANTISSA_BITS   0x3F
+#define UF11_MANTISSA_SHIFT  (23 - UF11_EXPONENT_SHIFT)
+#define UF11_MAX_EXPONENT    (UF11_EXPONENT_BITS << UF11_EXPONENT_SHIFT)
+
+#define UF10_EXPONENT_BIAS   15
+#define UF10_EXPONENT_BITS   0x1F
+#define UF10_EXPONENT_SHIFT  5
+#define UF10_MANTISSA_BITS   0x3F
+#define UF10_MANTISSA_SHIFT  (23 - UF10_EXPONENT_SHIFT)
+#define UF10_MAX_EXPONENT    (UF10_EXPONENT_BITS << UF10_EXPONENT_SHIFT)
+
+#define F32_INFINITY         0x7f800000
+
+static INLINE unsigned f32_to_uf11(float val)
+{
+   uint32_t f32 = (*(uint32_t *) &val);
+   uint16_t uf11 = 0;
+
+   /* Decode little-endian 32-bit floating-point value */
+   int sign = (f32 >> 16) & 0x8000;
+   /* Map exponent to the range [-127,128] */
+   int exponent = ((f32 >> 23) & 0xff) - 127;
+   int mantissa = f32 & 0x007fffff;
+
+   if (sign) return 0;
+
+   if (exponent == 128) { /* Infinity or NaN */
+      uf11 = UF11_MAX_EXPONENT;
+      if (mantissa) uf11 |= (mantissa & UF11_MANTISSA_BITS);
+   }
+   else if (exponent > 15) { /* Overflow - flush to Infinity */
+      uf11 = UF11_MAX_EXPONENT;
+   }
+   else if (exponent > -15) { /* Representable value */
+      exponent += UF11_EXPONENT_BIAS;
+      mantissa >>= UF11_MANTISSA_SHIFT;
+      uf11 = exponent << UF11_EXPONENT_SHIFT | mantissa;
+   }
+
+   return uf11;
+}
+
+static INLINE float uf11_to_f32(uint16_t val)
+{
+   union {
+      float f;
+      uint32_t ui;
+   } f32;
+
+   int exponent = (val & 0x07c0) >> UF11_EXPONENT_SHIFT;
+   int mantissa = (val & 0x003f);
+
+   f32.f = 0.0;
+
+   if (exponent == 0) {
+      if (mantissa != 0) {
+         const float scale = 1.0 / (1 << 20);
+         f32.f = scale * mantissa;
+      }
+   }
+   else if (exponent == 31) {
+      f32.ui = F32_INFINITY | mantissa;
+   }
+   else {
+      float scale, decimal;
+      exponent -= 15;
+      if (exponent < 0) {
+         scale = 1.0 / (1 << -exponent);
+      }
+      else {
+         scale = 1 << exponent;
+      }
+      decimal = 1.0 + (float) mantissa / 64;
+      f32.f = scale * decimal;
+   }
+
+   return f32.f;
+}
+
+static INLINE unsigned f32_to_uf10(float val)
+{
+   uint32_t f32 = (*(uint32_t *) &val);
+   uint16_t uf10 = 0;
+
+   /* Decode little-endian 32-bit floating-point value */
+   int sign = (f32 >> 16) & 0x8000;
+   /* Map exponent to the range [-127,128] */
+   int exponent = ((f32 >> 23) & 0xff) - 127;
+   int mantissa = f32 & 0x007fffff;
+
+   if (sign) return 0;
+
+   if (exponent == 128) { /* Infinity or NaN */
+      uf10 = UF10_MAX_EXPONENT;
+      if (mantissa) uf10 |= (mantissa & UF10_MANTISSA_BITS);
+   }
+   else if (exponent > 15) { /* Overflow - flush to Infinity */
+      uf10 = UF10_MAX_EXPONENT;
+   }
+   else if (exponent > -15) { /* Representable value */
+      exponent += UF10_EXPONENT_BIAS;
+      mantissa >>= UF10_MANTISSA_SHIFT;
+      uf10 = exponent << UF10_EXPONENT_SHIFT | mantissa;
+   }
+
+   return uf10;
+}
+
+static INLINE float uf10_to_f32(uint16_t val)
+{
+   union {
+      float f;
+      uint32_t ui;
+   } f32;
+
+   int exponent = (val & 0x07c0) >> UF10_EXPONENT_SHIFT;
+   int mantissa = (val & 0x003f);
+
+   f32.f = 0.0;
+
+   if (exponent == 0) {
+      if (mantissa != 0) {
+         const float scale = 1.0 / (1 << 20);
+         f32.f = scale * mantissa;
+      }
+   }
+   else if (exponent == 31) {
+      f32.ui = F32_INFINITY | mantissa;
+   }
+   else {
+      float scale, decimal;
+      exponent -= 15;
+      if (exponent < 0) {
+         scale = 1.0 / (1 << -exponent);
+      }
+      else {
+         scale = 1 << exponent;
+      }
+      decimal = 1.0 + (float) mantissa / 32;
+      f32.f = scale * decimal;
+   }
+
+   return f32.f;
+}
+
+static INLINE unsigned float3_to_r11g11b10f(const float rgb[3])
+{
+   return ( f32_to_uf11(rgb[0]) & 0x7ff) |
+          ((f32_to_uf11(rgb[1]) & 0x7ff) << 11) |
+          ((f32_to_uf10(rgb[2]) & 0x3ff) << 22);
+}
+
+static INLINE void r11g11b10f_to_float3(unsigned rgb, float retval[3])
+{
+   retval[0] = uf11_to_f32( rgb        & 0x7ff);
+   retval[1] = uf11_to_f32((rgb >> 11) & 0x7ff);
+   retval[2] = uf10_to_f32((rgb >> 22) & 0x3ff);
+}