src/amd/addrlib/src/core/addrelemlib.h

   1 /*
   2  * Copyright © 2007-2018 Advanced Micro Devices, Inc.
   3  * All Rights Reserved.
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining
   6  * a copy of this software and associated documentation files (the
   7  * "Software"), to deal in the Software without restriction, including
   8  * without limitation the rights to use, copy, modify, merge, publish,
   9  * distribute, sub license, and/or sell copies of the Software, and to
  10  * permit persons to whom the Software is furnished to do so, subject to
  11  * the following conditions:
  12  *
  13  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  14  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  15  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  16  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  17  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  18  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
  19  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  20  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  21  *
  22  * The above copyright notice and this permission notice (including the
  23  * next paragraph) shall be included in all copies or substantial portions
  24  * of the Software.
  25  */
  26
  27 /**
  28 ****************************************************************************************************
  29 * @file  addrelemlib.h
  30 * @brief Contains the class for element/pixel related functions.
  31 ****************************************************************************************************
  32 */
  33
  34 #ifndef __ELEM_LIB_H__
  35 #define __ELEM_LIB_H__
  36
  37 #include "addrinterface.h"
  38 #include "addrobject.h"
  39 #include "addrcommon.h"
  40
  41 namespace Addr
  42 {
  43
  44 class Lib;
  45
  46 // The masks for property bits within the Properties INT_32
  47 union ComponentFlags
  48 {
  49     struct
  50     {
  51         UINT_32 byteAligned    : 1;    ///< all components are byte aligned
  52         UINT_32 exportNorm     : 1;    ///< components support R6xx NORM compression
  53         UINT_32 floatComp      : 1;    ///< there is at least one floating point component
  54     };
  55
  56     UINT_32 value;
  57 };
  58
  59 // Copy from legacy lib's NumberType
  60 enum NumberType
  61 {
  62     // The following number types have the range [-1..1]
  63     ADDR_NO_NUMBER,         // This component doesn't exist and has no default value
  64     ADDR_EPSILON,           // Force component value to integer 0x00000001
  65     ADDR_ZERO,              // Force component value to integer 0x00000000
  66     ADDR_ONE,               // Force component value to floating point 1.0
  67     // Above values don't have any bits per component (keep ADDR_ONE the last of these)
  68
  69     ADDR_UNORM,             // Unsigned normalized (repeating fraction) full precision
  70     ADDR_SNORM,             // Signed normalized (repeating fraction) full precision
  71     ADDR_GAMMA,             // Gamma-corrected, full precision
  72
  73     ADDR_UNORM_R5XXRB,      // Unsigned normalized (repeating fraction) for r5xx RB
  74     ADDR_SNORM_R5XXRB,      // Signed normalized (repeating fraction) for r5xx RB
  75     ADDR_GAMMA_R5XXRB,      // Gamma-corrected for r5xx RB (note: unnormalized value)
  76     ADDR_UNORM_R5XXBC,      // Unsigned normalized (repeating fraction) for r5xx BC
  77     ADDR_SNORM_R5XXBC,      // Signed normalized (repeating fraction) for r5xx BC
  78     ADDR_GAMMA_R5XXBC,      // Gamma-corrected for r5xx BC (note: unnormalized value)
  79
  80     ADDR_UNORM_R6XX,        // Unsigned normalized (repeating fraction) for R6xx
  81     ADDR_UNORM_R6XXDB,      // Unorms for 24-bit depth: one value differs from ADDR_UNORM_R6XX
  82     ADDR_SNORM_R6XX,        // Signed normalized (repeating fraction) for R6xx
  83     ADDR_GAMMA8_R6XX,       // Gamma-corrected for r6xx
  84     ADDR_GAMMA8_R7XX_TP,    // Gamma-corrected for r7xx TP 12bit unorm 8.4.
  85
  86     ADDR_U4FLOATC,          // Unsigned float: 4-bit exponent, bias=15, no NaN, clamp [0..1]
  87     ADDR_GAMMA_4SEG,        // Gamma-corrected, four segment approximation
  88     ADDR_U0FIXED,           // Unsigned 0.N-bit fixed point
  89
  90     // The following number types have large ranges (LEAVE ADDR_USCALED first or fix Finish routine)
  91     ADDR_USCALED,           // Unsigned integer converted to/from floating point
  92     ADDR_SSCALED,           // Signed integer converted to/from floating point
  93     ADDR_USCALED_R5XXRB,    // Unsigned integer to/from floating point for r5xx RB
  94     ADDR_SSCALED_R5XXRB,    // Signed integer to/from floating point for r5xx RB
  95     ADDR_UINT_BITS,         // Keep in unsigned integer form, clamped to specified range
  96     ADDR_SINT_BITS,         // Keep in signed integer form, clamped to specified range
  97     ADDR_UINTBITS,          // @@ remove Keep in unsigned integer form, use modulus to reduce bits
  98     ADDR_SINTBITS,          // @@ remove Keep in signed integer form, use modulus to reduce bits
  99
 100     // The following number types and ADDR_U4FLOATC have exponents
 101     // (LEAVE ADDR_S8FLOAT first or fix Finish routine)
 102     ADDR_S8FLOAT,           // Signed floating point with 8-bit exponent, bias=127
 103     ADDR_S8FLOAT32,         // 32-bit IEEE float, passes through NaN values
 104     ADDR_S5FLOAT,           // Signed floating point with 5-bit exponent, bias=15
 105     ADDR_S5FLOATM,          // Signed floating point with 5-bit exponent, bias=15, no NaN/Inf
 106     ADDR_U5FLOAT,           // Signed floating point with 5-bit exponent, bias=15
 107     ADDR_U3FLOATM,          // Unsigned floating point with 3-bit exponent, bias=3
 108
 109     ADDR_S5FIXED,           // Signed 5.N-bit fixed point, with rounding
 110
 111     ADDR_END_NUMBER         // Used for range comparisons
 112 };
 113
 114 // Copy from legacy lib's AddrElement
 115 enum ElemMode
 116 {
 117     // These formats allow both packing an unpacking
 118     ADDR_ROUND_BY_HALF,      // add 1/2 and truncate when packing this element
 119     ADDR_ROUND_TRUNCATE,     // truncate toward 0 for sign/mag, else toward neg
 120     ADDR_ROUND_DITHER,       // Pack by dithering -- requires (x,y) position
 121
 122     // These formats only allow unpacking, no packing
 123     ADDR_UNCOMPRESSED,       // Elements are not compressed: one data element per pixel/texel
 124     ADDR_EXPANDED,           // Elements are split up and stored in multiple data elements
 125     ADDR_PACKED_STD,         // Elements are compressed into ExpandX by ExpandY data elements
 126     ADDR_PACKED_REV,         // Like ADDR_PACKED, but X order of pixels is reverved
 127     ADDR_PACKED_GBGR,        // Elements are compressed 4:2:2 in G1B_G0R order (high to low)
 128     ADDR_PACKED_BGRG,        // Elements are compressed 4:2:2 in BG1_RG0 order (high to low)
 129     ADDR_PACKED_BC1,         // Each data element is uncompressed to a 4x4 pixel/texel array
 130     ADDR_PACKED_BC2,         // Each data element is uncompressed to a 4x4 pixel/texel array
 131     ADDR_PACKED_BC3,         // Each data element is uncompressed to a 4x4 pixel/texel array
 132     ADDR_PACKED_BC4,         // Each data element is uncompressed to a 4x4 pixel/texel array
 133     ADDR_PACKED_BC5,         // Each data element is uncompressed to a 4x4 pixel/texel array
 134     ADDR_PACKED_ETC2_64BPP,  // ETC2 formats that use 64bpp to represent each 4x4 block
 135     ADDR_PACKED_ETC2_128BPP, // ETC2 formats that use 128bpp to represent each 4x4 block
 136     ADDR_PACKED_ASTC,        // Various ASTC formats, all are 128bpp with varying block sizes
 137
 138     // These formats provide various kinds of compression
 139     ADDR_ZPLANE_R5XX,        // Compressed Zplane using r5xx architecture format
 140     ADDR_ZPLANE_R6XX,        // Compressed Zplane using r6xx architecture format
 141     //@@ Fill in the compression modes
 142
 143     ADDR_END_ELEMENT         // Used for range comparisons
 144 };
 145
 146 enum DepthPlanarType
 147 {
 148     ADDR_DEPTH_PLANAR_NONE = 0, // No plane z/stencl
 149     ADDR_DEPTH_PLANAR_R600 = 1, // R600 z and stencil planes are store within a tile
 150     ADDR_DEPTH_PLANAR_R800 = 2, // R800 has separate z and stencil planes
 151 };
 152
 153 /**
 154 ****************************************************************************************************
 155 *   PixelFormatInfo
 156 *
 157 *   @brief
 158 *       Per component info
 159 *
 160 ****************************************************************************************************
 161 */
 162 struct PixelFormatInfo
 163 {
 164     UINT_32             compBit[4];
 165     NumberType          numType[4];
 166     UINT_32             compStart[4];
 167     ElemMode            elemMode;
 168     UINT_32             comps;          ///< Number of components
 169 };
 170
 171 /**
 172 ****************************************************************************************************
 173 * @brief This class contains asic indepentent element related attributes and operations
 174 ****************************************************************************************************
 175 */
 176 class ElemLib : public Object
 177 {
 178 protected:
 179     ElemLib(Lib* pAddrLib);
 180
 181 public:
 182
 183     /// Makes this class virtual
 184     virtual ~ElemLib();
 185
 186     static ElemLib* Create(
 187         const Lib* pAddrLib);
 188
 189     /// The implementation is only for R6xx/R7xx, so make it virtual in case we need for R8xx
 190     BOOL_32 PixGetExportNorm(
 191         AddrColorFormat colorFmt,
 192         AddrSurfaceNumber numberFmt, AddrSurfaceSwap swap) const;
 193
 194     /// Below method are asic independent, so make them just static.
 195     /// Remove static if we need different operation in hwl.
 196
 197     VOID    Flt32ToDepthPixel(
 198         AddrDepthFormat format, const ADDR_FLT_32 comps[2], UINT_8 *pPixel) const;
 199
 200     VOID    Flt32ToColorPixel(
 201         AddrColorFormat format, AddrSurfaceNumber surfNum, AddrSurfaceSwap surfSwap,
 202         const ADDR_FLT_32 comps[4], UINT_8 *pPixel) const;
 203
 204     static VOID    Flt32sToInt32s(
 205         ADDR_FLT_32 value, UINT_32 bits, NumberType numberType, UINT_32* pResult);
 206
 207     static VOID    Int32sToPixel(
 208         UINT_32 numComps, UINT_32* pComps, UINT_32* pCompBits, UINT_32* pCompStart,
 209         ComponentFlags properties, UINT_32 resultBits, UINT_8* pPixel);
 210
 211     VOID    PixGetColorCompInfo(
 212         AddrColorFormat format, AddrSurfaceNumber number, AddrSurfaceSwap swap,
 213         PixelFormatInfo* pInfo) const;
 214
 215     VOID    PixGetDepthCompInfo(
 216         AddrDepthFormat format, PixelFormatInfo* pInfo) const;
 217
 218     UINT_32 GetBitsPerPixel(
 219         AddrFormat format, ElemMode* pElemMode = NULL,
 220         UINT_32* pExpandX = NULL, UINT_32* pExpandY = NULL, UINT_32* pBitsUnused = NULL);
 221
 222     static VOID    SetClearComps(
 223         ADDR_FLT_32 comps[4], BOOL_32 clearColor, BOOL_32 float32);
 224
 225     VOID    AdjustSurfaceInfo(
 226         ElemMode elemMode, UINT_32 expandX, UINT_32 expandY,
 227         UINT_32* pBpp, UINT_32* pBasePitch, UINT_32* pWidth, UINT_32* pHeight);
 228
 229     VOID    RestoreSurfaceInfo(
 230         ElemMode elemMode, UINT_32 expandX, UINT_32 expandY,
 231         UINT_32* pBpp, UINT_32* pWidth, UINT_32* pHeight);
 232
 233     /// Checks if depth and stencil are planar inside a tile
 234     BOOL_32 IsDepthStencilTilePlanar()
 235     {
 236         return (m_depthPlanarType == ADDR_DEPTH_PLANAR_R600) ? TRUE : FALSE;
 237     }
 238
 239     /// Sets m_configFlags, copied from AddrLib
 240     VOID    SetConfigFlags(ConfigFlags flags)
 241     {
 242         m_configFlags = flags;
 243     }
 244
 245     static BOOL_32 IsCompressed(AddrFormat format);
 246     static BOOL_32 IsBlockCompressed(AddrFormat format);
 247     static BOOL_32 IsExpand3x(AddrFormat format);
 248     static BOOL_32 IsMacroPixelPacked(AddrFormat format);
 249
 250 protected:
 251
 252     static VOID    GetCompBits(
 253         UINT_32 c0, UINT_32 c1, UINT_32 c2, UINT_32 c3,
 254         PixelFormatInfo* pInfo,
 255         ElemMode elemMode = ADDR_ROUND_BY_HALF);
 256
 257     static VOID    GetCompType(
 258         AddrColorFormat format, AddrSurfaceNumber numType,
 259         PixelFormatInfo* pInfo);
 260
 261     static VOID    GetCompSwap(
 262         AddrSurfaceSwap swap, PixelFormatInfo* pInfo);
 263
 264     static VOID    SwapComps(
 265         UINT_32 c0, UINT_32 c1, PixelFormatInfo* pInfo);
 266
 267 private:
 268
 269     UINT_32             m_fp16ExportNorm;   ///< If allow FP16 to be reported as EXPORT_NORM
 270     DepthPlanarType     m_depthPlanarType;
 271
 272     ConfigFlags         m_configFlags;      ///< Copy of AddrLib's configFlags
 273     Addr::Lib* const    m_pAddrLib;         ///< Pointer to parent addrlib instance
 274 };
 275
 276 } //Addr
 277
 278 #endif
 279