blob: 8c25bb886892f838b046c0d5073f1eae7814153e [file] [log] [blame]
/*-------------------------------------------------------------------------
* drawElements Quality Program Tester Core
* ----------------------------------------
*
* Copyright 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*//*!
* \file
* \brief Reference Texture Implementation.
*//*--------------------------------------------------------------------*/
#include "tcuTexture.hpp"
#include "deInt32.h"
#include "deFloat16.h"
#include "deMath.h"
#include "deMemory.h"
#include "tcuTestLog.hpp"
#include "tcuSurface.hpp"
#include "tcuFloat.hpp"
#include "tcuTextureUtil.hpp"
#include "deStringUtil.hpp"
#include "deArrayUtil.hpp"
#include "tcuMatrix.hpp"
#include <limits>
namespace tcu
{
// \note No sign. Denorms are supported.
typedef Float<deUint32, 5, 6, 15, FLOAT_SUPPORT_DENORM> Float11;
typedef Float<deUint32, 5, 5, 15, FLOAT_SUPPORT_DENORM> Float10;
namespace
{
// Optimized getters for common formats.
// \todo [2012-11-14 pyry] Use intrinsics if available.
inline Vec4 readRGBA8888Float (const deUint8* ptr) { return Vec4(ptr[0]/255.0f, ptr[1]/255.0f, ptr[2]/255.0f, ptr[3]/255.0f); }
inline Vec4 readRGB888Float (const deUint8* ptr) { return Vec4(ptr[0]/255.0f, ptr[1]/255.0f, ptr[2]/255.0f, 1.0f); }
inline IVec4 readRGBA8888Int (const deUint8* ptr) { return IVec4(ptr[0], ptr[1], ptr[2], ptr[3]); }
inline IVec4 readRGB888Int (const deUint8* ptr) { return IVec4(ptr[0], ptr[1], ptr[2], 1); }
// Optimized setters.
inline void writeRGBA8888Int (deUint8* ptr, const IVec4& val)
{
ptr[0] = (deUint8)de::clamp(val[0], 0, 255);
ptr[1] = (deUint8)de::clamp(val[1], 0, 255);
ptr[2] = (deUint8)de::clamp(val[2], 0, 255);
ptr[3] = (deUint8)de::clamp(val[3], 0, 255);
}
inline void writeRGB888Int (deUint8* ptr, const IVec4& val)
{
ptr[0] = (deUint8)de::clamp(val[0], 0, 255);
ptr[1] = (deUint8)de::clamp(val[1], 0, 255);
ptr[2] = (deUint8)de::clamp(val[2], 0, 255);
}
inline void writeRGBA8888Float (deUint8* ptr, const Vec4& val)
{
ptr[0] = floatToU8(val[0]);
ptr[1] = floatToU8(val[1]);
ptr[2] = floatToU8(val[2]);
ptr[3] = floatToU8(val[3]);
}
inline void writeRGB888Float (deUint8* ptr, const Vec4& val)
{
ptr[0] = floatToU8(val[0]);
ptr[1] = floatToU8(val[1]);
ptr[2] = floatToU8(val[2]);
}
inline void writeUint24 (deUint8* dst, deUint32 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
dst[0] = (deUint8)((val & 0x0000FFu) >> 0u);
dst[1] = (deUint8)((val & 0x00FF00u) >> 8u);
dst[2] = (deUint8)((val & 0xFF0000u) >> 16u);
#else
dst[0] = (deUint8)((val & 0xFF0000u) >> 16u);
dst[1] = (deUint8)((val & 0x00FF00u) >> 8u);
dst[2] = (deUint8)((val & 0x0000FFu) >> 0u);
#endif
}
inline deUint32 readUint24 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
return (((deUint32)src[0]) << 0u) |
(((deUint32)src[1]) << 8u) |
(((deUint32)src[2]) << 16u);
#else
return (((deUint32)src[0]) << 16u) |
(((deUint32)src[1]) << 8u) |
(((deUint32)src[2]) << 0u);
#endif
}
inline deUint8 readUint32Low8 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffsetBits0To8 = 0; //!< least significant byte in the lowest address
#else
const deUint32 uint32ByteOffsetBits0To8 = 3; //!< least significant byte in the highest address
#endif
return src[uint32ByteOffsetBits0To8];
}
inline deUint8 readUint32High8 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffsetBits24To32 = 3;
#else
const deUint32 uint32ByteOffsetBits24To32 = 0;
#endif
return src[uint32ByteOffsetBits24To32];
}
inline void writeUint32Low8 (deUint8* dst, deUint8 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffsetBits0To8 = 0; //!< least significant byte in the lowest address
#else
const deUint32 uint32ByteOffsetBits0To8 = 3; //!< least significant byte in the highest address
#endif
dst[uint32ByteOffsetBits0To8] = val;
}
inline void writeUint32High8 (deUint8* dst, deUint8 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffsetBits24To32 = 3;
#else
const deUint32 uint32ByteOffsetBits24To32 = 0;
#endif
dst[uint32ByteOffsetBits24To32] = val;
}
inline deUint32 readUint32High16 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset16To32 = 2;
#else
const deUint32 uint32ByteOffset16To32 = 0;
#endif
return *(const deUint16*)(src + uint32ByteOffset16To32);
}
inline void writeUint32High16 (deUint8* dst, deUint16 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset16To32 = 2;
#else
const deUint32 uint32ByteOffset16To32 = 0;
#endif
*(deUint16*)(dst + uint32ByteOffset16To32) = val;
}
inline deUint32 readUint32Low24 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset0To24 = 0;
#else
const deUint32 uint32ByteOffset0To24 = 1;
#endif
return readUint24(src + uint32ByteOffset0To24);
}
inline deUint32 readUint32High24 (const deUint8* src)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset8To32 = 1;
#else
const deUint32 uint32ByteOffset8To32 = 0;
#endif
return readUint24(src + uint32ByteOffset8To32);
}
inline void writeUint32Low24 (deUint8* dst, deUint32 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset0To24 = 0;
#else
const deUint32 uint32ByteOffset0To24 = 1;
#endif
writeUint24(dst + uint32ByteOffset0To24, val);
}
inline void writeUint32High24 (deUint8* dst, deUint32 val)
{
#if (DE_ENDIANNESS == DE_LITTLE_ENDIAN)
const deUint32 uint32ByteOffset8To32 = 1;
#else
const deUint32 uint32ByteOffset8To32 = 0;
#endif
writeUint24(dst + uint32ByteOffset8To32, val);
}
// \todo [2011-09-21 pyry] Move to tcutil?
template <typename T>
inline T convertSatRte (float f)
{
// \note Doesn't work for 64-bit types
DE_STATIC_ASSERT(sizeof(T) < sizeof(deUint64));
DE_STATIC_ASSERT((-3 % 2 != 0) && (-4 % 2 == 0));
deInt64 minVal = std::numeric_limits<T>::min();
deInt64 maxVal = std::numeric_limits<T>::max();
float q = deFloatFrac(f);
deInt64 intVal = (deInt64)(f-q);
// Rounding.
if (q == 0.5f)
{
if (intVal % 2 != 0)
intVal++;
}
else if (q > 0.5f)
intVal++;
// else Don't add anything
// Saturate.
intVal = de::max(minVal, de::min(maxVal, intVal));
return (T)intVal;
}
inline deUint32 convertSatRteUint24 (float f)
{
const deUint32 rounded = convertSatRte<deUint32>(f);
const deUint32 maxUint24 = 0xFFFFFFu;
return de::min(rounded, maxUint24);
}
inline deUint16 convertSatRteUint10 (float f)
{
const deUint16 rounded = convertSatRte<deUint16>(f);
const deUint16 maxUint10 = 0x3FFu;
return de::min(rounded, maxUint10);
}
inline deUint16 convertSatRteUint12 (float f)
{
const deUint16 rounded = convertSatRte<deUint16>(f);
const deUint16 maxUint12 = 0xFFFu;
return de::min(rounded, maxUint12);
}
inline float channelToFloat (const deUint8* value, TextureFormat::ChannelType type)
{
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::SNORM_INT8: return de::max(-1.0f, (float)*((const deInt8*)value) / 127.0f);
case TextureFormat::SNORM_INT16: return de::max(-1.0f, (float)*((const deInt16*)value) / 32767.0f);
case TextureFormat::SNORM_INT32: return de::max(-1.0f, (float)*((const deInt32*)value) / 2147483647.0f);
case TextureFormat::UNORM_INT8: return (float)*((const deUint8*)value) / 255.0f;
case TextureFormat::UNORM_INT16: return (float)*((const deUint16*)value) / 65535.0f;
case TextureFormat::UNORM_INT24: return (float)readUint24(value) / 16777215.0f;
case TextureFormat::UNORM_INT32: return (float)*((const deUint32*)value) / 4294967295.0f;
case TextureFormat::SIGNED_INT8: return (float)*((const deInt8*)value);
case TextureFormat::SIGNED_INT16: return (float)*((const deInt16*)value);
case TextureFormat::SIGNED_INT32: return (float)*((const deInt32*)value);
case TextureFormat::SIGNED_INT64: return (float)*((const deInt64*)value);
case TextureFormat::UNSIGNED_INT8: return (float)*((const deUint8*)value);
case TextureFormat::UNSIGNED_INT16: return (float)*((const deUint16*)value);
case TextureFormat::UNSIGNED_INT24: return (float)readUint24(value);
case TextureFormat::UNSIGNED_INT32: return (float)*((const deUint32*)value);
case TextureFormat::UNSIGNED_INT64: return (float)*((const deUint64*)value);
case TextureFormat::HALF_FLOAT: return deFloat16To32(*(const deFloat16*)value);
case TextureFormat::FLOAT: return *((const float*)value);
case TextureFormat::FLOAT64: return (float)*((const double*)value);
case TextureFormat::UNORM_SHORT_10: return (float)((*((const deUint16*)value)) >> 6u) / 1023.0f;
case TextureFormat::UNORM_SHORT_12: return (float)((*((const deUint16*)value)) >> 4u) / 4095.0f;
case TextureFormat::USCALED_INT8: return (float)*((const deUint8*)value);
case TextureFormat::USCALED_INT16: return (float)*((const deUint16*)value);
case TextureFormat::SSCALED_INT8: return (float)*((const deInt8*)value);
case TextureFormat::SSCALED_INT16: return (float)*((const deInt16*)value);
default:
DE_ASSERT(DE_FALSE);
return 0.0f;
}
}
template <class T>
inline T channelToIntType (const deUint8* value, TextureFormat::ChannelType type)
{
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::SNORM_INT8: return (T)*((const deInt8*)value);
case TextureFormat::SNORM_INT16: return (T)*((const deInt16*)value);
case TextureFormat::SNORM_INT32: return (T)*((const deInt32*)value);
case TextureFormat::UNORM_INT8: return (T)*((const deUint8*)value);
case TextureFormat::UNORM_INT16: return (T)*((const deUint16*)value);
case TextureFormat::UNORM_INT24: return (T)readUint24(value);
case TextureFormat::UNORM_INT32: return (T)*((const deUint32*)value);
case TextureFormat::SIGNED_INT8: return (T)*((const deInt8*)value);
case TextureFormat::SIGNED_INT16: return (T)*((const deInt16*)value);
case TextureFormat::SIGNED_INT32: return (T)*((const deInt32*)value);
case TextureFormat::SIGNED_INT64: return (T)*((const deInt64*)value);
case TextureFormat::UNSIGNED_INT8: return (T)*((const deUint8*)value);
case TextureFormat::UNSIGNED_INT16: return (T)*((const deUint16*)value);
case TextureFormat::UNSIGNED_INT24: return (T)readUint24(value);
case TextureFormat::UNSIGNED_INT32: return (T)*((const deUint32*)value);
case TextureFormat::UNSIGNED_INT64: return (T)*((const deUint64*)value);
case TextureFormat::HALF_FLOAT: return (T)deFloat16To32(*(const deFloat16*)value);
case TextureFormat::FLOAT: return (T)*((const float*)value);
case TextureFormat::FLOAT64: return (T)*((const double*)value);
case TextureFormat::UNORM_SHORT_10: return (T)((*(((const deUint16*)value))) >> 6u);
case TextureFormat::UNORM_SHORT_12: return (T)((*(((const deUint16*)value))) >> 4u);
case TextureFormat::USCALED_INT8: return (T)*((const deUint8*)value);
case TextureFormat::USCALED_INT16: return (T)*((const deUint16*)value);
case TextureFormat::SSCALED_INT8: return (T)*((const deInt8*)value);
case TextureFormat::SSCALED_INT16: return (T)*((const deInt16*)value);
default:
DE_ASSERT(DE_FALSE);
return 0;
}
}
inline int channelToInt (const deUint8* value, TextureFormat::ChannelType type)
{
return channelToIntType<int>(value, type);
}
void floatToChannel (deUint8* dst, float src, TextureFormat::ChannelType type)
{
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::SNORM_INT8: *((deInt8*)dst) = convertSatRte<deInt8> (src * 127.0f); break;
case TextureFormat::SNORM_INT16: *((deInt16*)dst) = convertSatRte<deInt16> (src * 32767.0f); break;
case TextureFormat::SNORM_INT32: *((deInt32*)dst) = convertSatRte<deInt32> (src * 2147483647.0f); break;
case TextureFormat::UNORM_INT8: *((deUint8*)dst) = convertSatRte<deUint8> (src * 255.0f); break;
case TextureFormat::UNORM_INT16: *((deUint16*)dst) = convertSatRte<deUint16> (src * 65535.0f); break;
case TextureFormat::UNORM_INT24: writeUint24(dst, convertSatRteUint24 (src * 16777215.0f)); break;
case TextureFormat::UNORM_INT32: *((deUint32*)dst) = convertSatRte<deUint32> (src * 4294967295.0f); break;
case TextureFormat::SIGNED_INT8: *((deInt8*)dst) = convertSatRte<deInt8> (src); break;
case TextureFormat::SIGNED_INT16: *((deInt16*)dst) = convertSatRte<deInt16> (src); break;
case TextureFormat::SIGNED_INT32: *((deInt32*)dst) = convertSatRte<deInt32> (src); break;
case TextureFormat::UNSIGNED_INT8: *((deUint8*)dst) = convertSatRte<deUint8> (src); break;
case TextureFormat::UNSIGNED_INT16: *((deUint16*)dst) = convertSatRte<deUint16> (src); break;
case TextureFormat::UNSIGNED_INT24: writeUint24(dst, convertSatRteUint24 (src)); break;
case TextureFormat::UNSIGNED_INT32: *((deUint32*)dst) = convertSatRte<deUint32> (src); break;
case TextureFormat::HALF_FLOAT: *((deFloat16*)dst) = deFloat32To16 (src); break;
case TextureFormat::FLOAT: *((float*)dst) = src; break;
case TextureFormat::FLOAT64: *((double*)dst) = (double)src; break;
case TextureFormat::UNORM_SHORT_10: *((deUint16*)dst) = (deUint16)(convertSatRteUint10(src * 1023.0f) << 6u); break;
case TextureFormat::UNORM_SHORT_12: *((deUint16*)dst) = (deUint16)(convertSatRteUint12(src * 4095.0f) << 4u); break;
case TextureFormat::USCALED_INT8: *((deUint8*)dst) = convertSatRte<deUint8> (src); break;
case TextureFormat::USCALED_INT16: *((deUint16*)dst) = convertSatRte<deUint16> (src); break;
case TextureFormat::SSCALED_INT8: *((deInt8*)dst) = convertSatRte<deInt8> (src); break;
case TextureFormat::SSCALED_INT16: *((deInt16*)dst) = convertSatRte<deInt16> (src); break;
default:
DE_ASSERT(DE_FALSE);
}
}
template <typename T, typename S>
static inline T convertSat (S src)
{
S min = (S)std::numeric_limits<T>::min();
S max = (S)std::numeric_limits<T>::max();
if (src < min)
return (T)min;
else if (src > max)
return (T)max;
else
return (T)src;
}
template <typename S>
static inline deUint32 convertSatUint24 (S src)
{
S min = (S)0u;
S max = (S)0xFFFFFFu;
if (src < min)
return (deUint32)min;
else if (src > max)
return (deUint32)max;
else
return (deUint32)src;
}
template <typename S>
static inline deUint16 convertSatUint10 (S src)
{
S min = (S)0u;
S max = (S)0x3FFu;
if (src < min)
return (deUint16)min;
else if (src > max)
return (deUint16)max;
else
return (deUint16)src;
}
template <typename S>
static inline deUint16 convertSatUint12 (S src)
{
S min = (S)0u;
S max = (S)0xFFFu;
if (src < min)
return (deUint16)min;
else if (src > max)
return (deUint16)max;
else
return (deUint16)src;
}
void intToChannel (deUint8* dst, int src, TextureFormat::ChannelType type)
{
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::SNORM_INT8: *((deInt8*)dst) = convertSat<deInt8> (src); break;
case TextureFormat::SNORM_INT16: *((deInt16*)dst) = convertSat<deInt16> (src); break;
case TextureFormat::UNORM_INT8: *((deUint8*)dst) = convertSat<deUint8> (src); break;
case TextureFormat::UNORM_INT16: *((deUint16*)dst) = convertSat<deUint16> (src); break;
case TextureFormat::UNORM_INT24: writeUint24(dst, convertSatUint24 (src)); break;
case TextureFormat::SIGNED_INT8: *((deInt8*)dst) = convertSat<deInt8> (src); break;
case TextureFormat::SIGNED_INT16: *((deInt16*)dst) = convertSat<deInt16> (src); break;
case TextureFormat::SIGNED_INT32: *((deInt32*)dst) = convertSat<deInt32> (src); break;
case TextureFormat::SIGNED_INT64: *((deInt64*)dst) = convertSat<deInt64> ((deInt64)src); break;
case TextureFormat::UNSIGNED_INT8: *((deUint8*)dst) = convertSat<deUint8> ((deUint32)src); break;
case TextureFormat::UNSIGNED_INT16: *((deUint16*)dst) = convertSat<deUint16> ((deUint32)src); break;
case TextureFormat::UNSIGNED_INT24: writeUint24(dst, convertSatUint24 ((deUint32)src)); break;
case TextureFormat::UNSIGNED_INT32: *((deUint32*)dst) = convertSat<deUint32> ((deUint32)src); break;
case TextureFormat::UNSIGNED_INT64: *((deUint64*)dst) = convertSat<deUint64> ((deUint64)src); break;
case TextureFormat::HALF_FLOAT: *((deFloat16*)dst) = deFloat32To16((float)src); break;
case TextureFormat::FLOAT: *((float*)dst) = (float)src; break;
case TextureFormat::FLOAT64: *((double*)dst) = (double)src; break;
case TextureFormat::UNORM_SHORT_10: *((deUint16*)dst) = (deUint16)(convertSatUint10(src) << 6u); break;
case TextureFormat::UNORM_SHORT_12: *((deUint16*)dst) = (deUint16)(convertSatUint12(src) << 4u); break;
case TextureFormat::USCALED_INT8: *((deUint8*)dst) = convertSat<deUint8> ((deUint32)src); break;
case TextureFormat::USCALED_INT16: *((deUint16*)dst) = convertSat<deUint16> ((deUint32)src); break;
case TextureFormat::SSCALED_INT8: *((deInt8*)dst) = convertSat<deInt8> (src); break;
case TextureFormat::SSCALED_INT16: *((deInt16*)dst) = convertSat<deInt16> (src); break;
default:
DE_ASSERT(DE_FALSE);
}
}
inline float channelToUnormFloat (deUint32 src, int bits)
{
const deUint32 maxVal = (1u << bits) - 1;
// \note Will lose precision if bits > 23
return (float)src / (float)maxVal;
}
//! Extend < 32b signed integer to 32b
inline deInt32 signExtend (deUint32 src, int bits)
{
const deUint32 signBit = 1u << (bits-1);
src |= ~((src & signBit) - 1);
return (deInt32)src;
}
inline float channelToSnormFloat (deUint32 src, int bits)
{
const deUint32 range = (1u << (bits-1)) - 1;
// \note Will lose precision if bits > 24
return de::max(-1.0f, (float)signExtend(src, bits) / (float)range);
}
inline deUint32 unormFloatToChannel (float src, int bits)
{
const deUint32 maxVal = (1u << bits) - 1;
const deUint32 intVal = convertSatRte<deUint32>(src * (float)maxVal);
return de::min(intVal, maxVal);
}
inline deUint32 snormFloatToChannel (float src, int bits)
{
const deInt32 range = (deInt32)((1u << (bits-1)) - 1u);
const deUint32 mask = (1u << bits) - 1;
const deInt32 intVal = convertSatRte<deInt32>(src * (float)range);
return (deUint32)de::clamp(intVal, -range, range) & mask;
}
inline deUint32 uintToChannel (deUint32 src, int bits)
{
const deUint32 maxVal = (1u << bits) - 1;
return de::min(src, maxVal);
}
inline deUint32 intToChannel (deInt32 src, int bits)
{
const deInt32 minVal = -(deInt32)(1u << (bits-1));
const deInt32 maxVal = (deInt32)((1u << (bits-1)) - 1u);
const deUint32 mask = (1u << bits) - 1;
return (deUint32)de::clamp(src, minVal, maxVal) & mask;
}
tcu::Vec4 unpackRGB999E5 (deUint32 color)
{
const int mBits = 9;
const int eBias = 15;
deUint32 exp = color >> 27;
deUint32 bs = (color >> 18) & ((1<<9)-1);
deUint32 gs = (color >> 9) & ((1<<9)-1);
deUint32 rs = color & ((1<<9)-1);
float e = deFloatPow(2.0f, (float)((int)exp - eBias - mBits));
float r = (float)rs * e;
float g = (float)gs * e;
float b = (float)bs * e;
return tcu::Vec4(r, g, b, 1.0f);
}
bool isColorOrder (TextureFormat::ChannelOrder order)
{
DE_STATIC_ASSERT(TextureFormat::CHANNELORDER_LAST == 22);
switch (order)
{
case TextureFormat::R:
case TextureFormat::A:
case TextureFormat::I:
case TextureFormat::L:
case TextureFormat::LA:
case TextureFormat::RG:
case TextureFormat::RA:
case TextureFormat::RGB:
case TextureFormat::RGBA:
case TextureFormat::ARGB:
case TextureFormat::ABGR:
case TextureFormat::BGR:
case TextureFormat::BGRA:
case TextureFormat::sR:
case TextureFormat::sRG:
case TextureFormat::sRGB:
case TextureFormat::sRGBA:
case TextureFormat::sBGR:
case TextureFormat::sBGRA:
return true;
default:
return false;
}
}
} // anonymous
bool isValid (TextureFormat format)
{
const bool isColor = isColorOrder(format.order);
switch (format.type)
{
case TextureFormat::SNORM_INT8:
case TextureFormat::SNORM_INT16:
case TextureFormat::SNORM_INT32:
return isColor;
case TextureFormat::UNORM_INT8:
case TextureFormat::UNORM_INT16:
case TextureFormat::UNORM_INT24:
case TextureFormat::UNORM_INT32:
return isColor || format.order == TextureFormat::D;
case TextureFormat::UNORM_BYTE_44:
case TextureFormat::UNSIGNED_BYTE_44:
return format.order == TextureFormat::RG;
case TextureFormat::UNORM_SHORT_565:
case TextureFormat::UNORM_SHORT_555:
case TextureFormat::UNSIGNED_SHORT_565:
return format.order == TextureFormat::RGB || format.order == TextureFormat::BGR;
case TextureFormat::UNORM_SHORT_4444:
case TextureFormat::UNORM_SHORT_5551:
case TextureFormat::UNSIGNED_SHORT_4444:
case TextureFormat::UNSIGNED_SHORT_5551:
return format.order == TextureFormat::RGBA || format.order == TextureFormat::BGRA
|| format.order == TextureFormat::ARGB || format.order == TextureFormat::ABGR;
case TextureFormat::UNORM_SHORT_1555:
return format.order == TextureFormat::ARGB;
case TextureFormat::UNORM_INT_101010:
return format.order == TextureFormat::RGB;
case TextureFormat::SNORM_INT_1010102_REV:
case TextureFormat::UNORM_INT_1010102_REV:
case TextureFormat::SIGNED_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_1010102_REV:
case TextureFormat::USCALED_INT_1010102_REV:
case TextureFormat::SSCALED_INT_1010102_REV:
return format.order == TextureFormat::RGBA || format.order == TextureFormat::BGRA;
case TextureFormat::UNSIGNED_INT_11F_11F_10F_REV:
case TextureFormat::UNSIGNED_INT_999_E5_REV:
return format.order == TextureFormat::RGB;
case TextureFormat::UNSIGNED_INT_16_8_8:
return format.order == TextureFormat::DS;
case TextureFormat::UNSIGNED_INT_24_8:
case TextureFormat::UNSIGNED_INT_24_8_REV:
return format.order == TextureFormat::D || format.order == TextureFormat::DS;
case TextureFormat::SIGNED_INT8:
case TextureFormat::SIGNED_INT16:
case TextureFormat::SIGNED_INT32:
case TextureFormat::SSCALED_INT8:
case TextureFormat::SSCALED_INT16:
case TextureFormat::SIGNED_INT64:
return isColor;
case TextureFormat::UNSIGNED_INT8:
case TextureFormat::UNSIGNED_INT16:
case TextureFormat::UNSIGNED_INT24:
case TextureFormat::UNSIGNED_INT32:
case TextureFormat::USCALED_INT8:
case TextureFormat::USCALED_INT16:
case TextureFormat::UNSIGNED_INT64:
return isColor || format.order == TextureFormat::S;
case TextureFormat::HALF_FLOAT:
case TextureFormat::FLOAT:
case TextureFormat::FLOAT64:
return isColor || format.order == TextureFormat::D;
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
return format.order == TextureFormat::DS;
case TextureFormat::UNORM_SHORT_10:
case TextureFormat::UNORM_SHORT_12:
return isColor;
default:
DE_FATAL("Unknown format");
return 0u;
}
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
}
int getNumUsedChannels (TextureFormat::ChannelOrder order)
{
// make sure this table is updated if type table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELORDER_LAST == 22);
switch (order)
{
case TextureFormat::R: return 1;
case TextureFormat::A: return 1;
case TextureFormat::I: return 1;
case TextureFormat::L: return 1;
case TextureFormat::LA: return 2;
case TextureFormat::RG: return 2;
case TextureFormat::RA: return 2;
case TextureFormat::RGB: return 3;
case TextureFormat::RGBA: return 4;
case TextureFormat::ARGB: return 4;
case TextureFormat::ABGR: return 4;
case TextureFormat::BGR: return 3;
case TextureFormat::BGRA: return 4;
case TextureFormat::sR: return 1;
case TextureFormat::sRG: return 2;
case TextureFormat::sRGB: return 3;
case TextureFormat::sRGBA: return 4;
case TextureFormat::sBGR: return 3;
case TextureFormat::sBGRA: return 4;
case TextureFormat::D: return 1;
case TextureFormat::S: return 1;
case TextureFormat::DS: return 2;
default:
DE_ASSERT(DE_FALSE);
return 0;
}
}
int getChannelSize (TextureFormat::ChannelType type)
{
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::SNORM_INT8: return 1;
case TextureFormat::SNORM_INT16: return 2;
case TextureFormat::SNORM_INT32: return 4;
case TextureFormat::UNORM_INT8: return 1;
case TextureFormat::UNORM_INT16: return 2;
case TextureFormat::UNORM_INT24: return 3;
case TextureFormat::UNORM_INT32: return 4;
case TextureFormat::SIGNED_INT8: return 1;
case TextureFormat::SIGNED_INT16: return 2;
case TextureFormat::SIGNED_INT32: return 4;
case TextureFormat::SIGNED_INT64: return 8;
case TextureFormat::UNSIGNED_INT8: return 1;
case TextureFormat::UNSIGNED_INT16: return 2;
case TextureFormat::UNSIGNED_INT24: return 3;
case TextureFormat::UNSIGNED_INT32: return 4;
case TextureFormat::UNSIGNED_INT64: return 8;
case TextureFormat::HALF_FLOAT: return 2;
case TextureFormat::FLOAT: return 4;
case TextureFormat::FLOAT64: return 8;
case TextureFormat::UNORM_SHORT_10: return 2;
case TextureFormat::UNORM_SHORT_12: return 2;
case TextureFormat::USCALED_INT8: return 1;
case TextureFormat::USCALED_INT16: return 2;
case TextureFormat::SSCALED_INT8: return 1;
case TextureFormat::SSCALED_INT16: return 2;
default:
DE_ASSERT(DE_FALSE);
return 0;
}
}
/** Get pixel size in bytes. */
int getPixelSize (TextureFormat format)
{
const TextureFormat::ChannelOrder order = format.order;
const TextureFormat::ChannelType type = format.type;
DE_ASSERT(isValid(format));
// make sure this table is updated if format table is updated
DE_STATIC_ASSERT(TextureFormat::CHANNELTYPE_LAST == 48);
switch (type)
{
case TextureFormat::UNORM_BYTE_44:
case TextureFormat::UNSIGNED_BYTE_44:
return 1;
case TextureFormat::UNORM_SHORT_565:
case TextureFormat::UNORM_SHORT_555:
case TextureFormat::UNORM_SHORT_4444:
case TextureFormat::UNORM_SHORT_5551:
case TextureFormat::UNORM_SHORT_1555:
case TextureFormat::UNSIGNED_SHORT_565:
case TextureFormat::UNSIGNED_SHORT_4444:
case TextureFormat::UNSIGNED_SHORT_5551:
return 2;
case TextureFormat::UNORM_INT_101010:
case TextureFormat::UNSIGNED_INT_999_E5_REV:
case TextureFormat::UNSIGNED_INT_11F_11F_10F_REV:
case TextureFormat::SNORM_INT_1010102_REV:
case TextureFormat::UNORM_INT_1010102_REV:
case TextureFormat::SIGNED_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_24_8:
case TextureFormat::UNSIGNED_INT_24_8_REV:
case TextureFormat::UNSIGNED_INT_16_8_8:
case TextureFormat::USCALED_INT_1010102_REV:
case TextureFormat::SSCALED_INT_1010102_REV:
return 4;
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
return 8;
default:
return getNumUsedChannels(order) * getChannelSize(type);
}
}
int TextureFormat::getPixelSize (void) const
{
return ::tcu::getPixelSize(*this);
}
const TextureSwizzle& getChannelReadSwizzle (TextureFormat::ChannelOrder order)
{
// make sure to update these tables when channel orders are updated
DE_STATIC_ASSERT(TextureFormat::CHANNELORDER_LAST == 22);
static const TextureSwizzle INV = {{ TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle R = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle A = {{ TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_0 }};
static const TextureSwizzle I = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0 }};
static const TextureSwizzle L = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle LA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1 }};
static const TextureSwizzle RG = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle RA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_1 }};
static const TextureSwizzle RGB = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle RGBA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_3 }};
static const TextureSwizzle BGR = {{ TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle BGRA = {{ TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3 }};
static const TextureSwizzle ARGB = {{ TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_0 }};
static const TextureSwizzle ABGR = {{ TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0 }};
static const TextureSwizzle D = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ONE }};
static const TextureSwizzle S = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ZERO, TextureSwizzle::CHANNEL_ONE }};
switch (order)
{
case TextureFormat::R: return R;
case TextureFormat::A: return A;
case TextureFormat::I: return I;
case TextureFormat::L: return L;
case TextureFormat::LA: return LA;
case TextureFormat::RG: return RG;
case TextureFormat::RA: return RA;
case TextureFormat::RGB: return RGB;
case TextureFormat::RGBA: return RGBA;
case TextureFormat::ARGB: return ARGB;
case TextureFormat::ABGR: return ABGR;
case TextureFormat::BGR: return BGR;
case TextureFormat::BGRA: return BGRA;
case TextureFormat::sR: return R;
case TextureFormat::sRG: return RG;
case TextureFormat::sRGB: return RGB;
case TextureFormat::sRGBA: return RGBA;
case TextureFormat::sBGR: return BGR;
case TextureFormat::sBGRA: return BGRA;
case TextureFormat::D: return D;
case TextureFormat::S: return S;
case TextureFormat::DS:
DE_ASSERT(false); // combined formats cannot be read from
return INV;
default:
DE_ASSERT(DE_FALSE);
return INV;
}
}
const TextureSwizzle& getChannelWriteSwizzle (TextureFormat::ChannelOrder order)
{
// make sure to update these tables when channel orders are updated
DE_STATIC_ASSERT(TextureFormat::CHANNELORDER_LAST == 22);
static const TextureSwizzle INV = {{ TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle R = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle A = {{ TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle I = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle L = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle LA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle RG = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle RA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle RGB = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle RGBA = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_3 }};
static const TextureSwizzle BGR = {{ TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle BGRA = {{ TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3 }};
static const TextureSwizzle ARGB = {{ TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_2 }};
static const TextureSwizzle ABGR = {{ TextureSwizzle::CHANNEL_3, TextureSwizzle::CHANNEL_2, TextureSwizzle::CHANNEL_1, TextureSwizzle::CHANNEL_0 }};
static const TextureSwizzle D = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
static const TextureSwizzle S = {{ TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST, TextureSwizzle::CHANNEL_LAST }};
switch (order)
{
case TextureFormat::R: return R;
case TextureFormat::A: return A;
case TextureFormat::I: return I;
case TextureFormat::L: return L;
case TextureFormat::LA: return LA;
case TextureFormat::RG: return RG;
case TextureFormat::RA: return RA;
case TextureFormat::RGB: return RGB;
case TextureFormat::RGBA: return RGBA;
case TextureFormat::ARGB: return ARGB;
case TextureFormat::ABGR: return ABGR;
case TextureFormat::BGR: return BGR;
case TextureFormat::BGRA: return BGRA;
case TextureFormat::sR: return R;
case TextureFormat::sRG: return RG;
case TextureFormat::sRGB: return RGB;
case TextureFormat::sRGBA: return RGBA;
case TextureFormat::sBGR: return BGR;
case TextureFormat::sBGRA: return BGRA;
case TextureFormat::D: return D;
case TextureFormat::S: return S;
case TextureFormat::DS:
DE_ASSERT(false); // combined formats cannot be written to
return INV;
default:
DE_ASSERT(DE_FALSE);
return INV;
}
}
IVec3 calculatePackedPitch (const TextureFormat& format, const IVec3& size)
{
const int pixelSize = format.getPixelSize();
const int rowPitch = pixelSize * size.x();
const int slicePitch = rowPitch * size.y();
return IVec3(pixelSize, rowPitch, slicePitch);
}
ConstPixelBufferAccess::ConstPixelBufferAccess (void)
: m_size (0)
, m_pitch (0)
, m_divider (1,1,1)
, m_data (DE_NULL)
{
}
ConstPixelBufferAccess::ConstPixelBufferAccess (const TextureFormat& format, int width, int height, int depth, const void* data)
: m_format (format)
, m_size (width, height, depth)
, m_pitch (calculatePackedPitch(m_format, m_size))
, m_divider (1,1,1)
, m_data ((void*)data)
{
DE_ASSERT(isValid(format));
}
ConstPixelBufferAccess::ConstPixelBufferAccess (const TextureFormat& format, const IVec3& size, const void* data)
: m_format (format)
, m_size (size)
, m_pitch (calculatePackedPitch(m_format, m_size))
, m_divider (1,1,1)
, m_data ((void*)data)
{
DE_ASSERT(isValid(format));
}
ConstPixelBufferAccess::ConstPixelBufferAccess (const TextureFormat& format, int width, int height, int depth, int rowPitch, int slicePitch, const void* data)
: m_format (format)
, m_size (width, height, depth)
, m_pitch (format.getPixelSize(), rowPitch, slicePitch)
, m_divider (1,1,1)
, m_data ((void*)data)
{
DE_ASSERT(isValid(format));
}
ConstPixelBufferAccess::ConstPixelBufferAccess (const TextureFormat& format, const IVec3& size, const IVec3& pitch, const void* data)
: m_format (format)
, m_size (size)
, m_pitch (pitch)
, m_divider (1,1,1)
, m_data ((void*)data)
{
DE_ASSERT(isValid(format));
DE_ASSERT(m_format.getPixelSize() <= m_pitch.x());
}
ConstPixelBufferAccess::ConstPixelBufferAccess(const TextureFormat& format, const IVec3& size, const IVec3& pitch, const IVec3& block, const void* data)
: m_format (format)
, m_size (size)
, m_pitch (pitch)
, m_divider (block)
, m_data ((void*)data)
{
DE_ASSERT(isValid(format));
DE_ASSERT(m_format.getPixelSize() <= m_pitch.x());
}
ConstPixelBufferAccess::ConstPixelBufferAccess (const TextureLevel& level)
: m_format (level.getFormat())
, m_size (level.getSize())
, m_pitch (calculatePackedPitch(m_format, m_size))
, m_divider (1,1,1)
, m_data ((void*)level.getPtr())
{
}
PixelBufferAccess::PixelBufferAccess (const TextureFormat& format, int width, int height, int depth, void* data)
: ConstPixelBufferAccess(format, width, height, depth, data)
{
}
PixelBufferAccess::PixelBufferAccess (const TextureFormat& format, const IVec3& size, void* data)
: ConstPixelBufferAccess(format, size, data)
{
}
PixelBufferAccess::PixelBufferAccess (const TextureFormat& format, int width, int height, int depth, int rowPitch, int slicePitch, void* data)
: ConstPixelBufferAccess(format, width, height, depth, rowPitch, slicePitch, data)
{
}
PixelBufferAccess::PixelBufferAccess (const TextureFormat& format, const IVec3& size, const IVec3& pitch, void* data)
: ConstPixelBufferAccess(format, size, pitch, data)
{
}
PixelBufferAccess::PixelBufferAccess(const TextureFormat& format, const IVec3& size, const IVec3& pitch, const IVec3& block, void* data)
: ConstPixelBufferAccess(format, size, pitch, block, data)
{
}
PixelBufferAccess::PixelBufferAccess (TextureLevel& level)
: ConstPixelBufferAccess(level)
{
}
//! Swizzle generally based on channel order.
template<typename T>
Vector<T, 4> swizzleGe (const Vector<T, 4>& v, TextureFormat::ChannelOrder src, TextureFormat::ChannelOrder dst)
{
if (src == dst)
return v;
else
{
if ((src == TextureFormat::RGBA && dst == TextureFormat::ARGB) ||
(src == TextureFormat::BGRA && dst == TextureFormat::ABGR))
return v.swizzle(3, 0, 1, 2);
if ((src == TextureFormat::ARGB && dst == TextureFormat::RGBA) ||
(src == TextureFormat::ABGR && dst == TextureFormat::BGRA))
return v.swizzle(1, 2, 3, 0);
if ((src == TextureFormat::BGRA && dst == TextureFormat::ARGB) ||
(src == TextureFormat::ABGR && dst == TextureFormat::RGBA) ||
(src == TextureFormat::RGBA && dst == TextureFormat::ABGR) ||
(src == TextureFormat::ARGB && dst == TextureFormat::BGRA))
return v.swizzle(3, 2, 1, 0);
if ((src == TextureFormat::RGB && dst == TextureFormat::BGR) ||
(src == TextureFormat::BGR && dst == TextureFormat::RGB) ||
(src == TextureFormat::RGBA && dst == TextureFormat::BGRA) ||
(src == TextureFormat::BGRA && dst == TextureFormat::RGBA))
return v.swizzle(2,1,0,3);
DE_ASSERT(false);
return v;
}
}
Vec4 ConstPixelBufferAccess::getPixel (int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, m_size.x()));
DE_ASSERT(de::inBounds(y, 0, m_size.y()));
DE_ASSERT(de::inBounds(z, 0, m_size.z()));
DE_ASSERT(!isCombinedDepthStencilType(m_format.type)); // combined types cannot be accessed directly
DE_ASSERT(m_format.order != TextureFormat::DS); // combined formats cannot be accessed directly
const deUint8* pixelPtr = (const deUint8*)getPixelPtr(x, y, z);
// Optimized fomats.
if (m_format.type == TextureFormat::UNORM_INT8)
{
if (m_format.order == TextureFormat::RGBA || m_format.order == TextureFormat::sRGBA)
return readRGBA8888Float(pixelPtr);
else if (m_format.order == TextureFormat::RGB || m_format.order == TextureFormat::sRGB)
return readRGB888Float(pixelPtr);
}
#define UI8(OFFS, COUNT) ((*((const deUint8*)pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define UI16(OFFS, COUNT) ((*((const deUint16*)pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define UI32(OFFS, COUNT) ((*((const deUint32*)pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define SI32(OFFS, COUNT) signExtend(UI32(OFFS, COUNT), (COUNT))
#define UN8(OFFS, COUNT) channelToUnormFloat(UI8 (OFFS, COUNT), (COUNT))
#define UN16(OFFS, COUNT) channelToUnormFloat(UI16(OFFS, COUNT), (COUNT))
#define UN32(OFFS, COUNT) channelToUnormFloat(UI32(OFFS, COUNT), (COUNT))
#define SN32(OFFS, COUNT) channelToSnormFloat(UI32(OFFS, COUNT), (COUNT))
// Packed formats.
switch (m_format.type)
{
case TextureFormat::UNORM_BYTE_44: return Vec4(UN8 (4, 4), UN8 ( 0, 4), 0.0f, 1.0f);
case TextureFormat::UNSIGNED_BYTE_44: return UVec4(UI8 (4, 4), UI8 ( 0, 4), 0u, 1u).cast<float>();
case TextureFormat::UNORM_SHORT_565: return swizzleGe( Vec4(UN16(11, 5), UN16( 5, 6), UN16( 0, 5), 1.0f), m_format.order, TextureFormat::RGB);
case TextureFormat::UNSIGNED_SHORT_565: return swizzleGe(UVec4(UI16(11, 5), UI16( 5, 6), UI16( 0, 5), 1u), m_format.order, TextureFormat::RGB).cast<float>();
case TextureFormat::UNORM_SHORT_555: return swizzleGe( Vec4(UN16(10, 5), UN16( 5, 5), UN16( 0, 5), 1.0f), m_format.order, TextureFormat::RGB);
case TextureFormat::UNORM_SHORT_4444: return swizzleGe( Vec4(UN16(12, 4), UN16( 8, 4), UN16( 4, 4), UN16( 0, 4)), m_format.order, TextureFormat::RGBA);
case TextureFormat::UNSIGNED_SHORT_4444: return swizzleGe(UVec4(UI16(12, 4), UI16( 8, 4), UI16( 4, 4), UI16( 0, 4)), m_format.order, TextureFormat::RGBA).cast<float>();
case TextureFormat::UNORM_SHORT_5551: return swizzleGe( Vec4(UN16(11, 5), UN16( 6, 5), UN16( 1, 5), UN16( 0, 1)), m_format.order, TextureFormat::RGBA);
case TextureFormat::UNSIGNED_SHORT_5551: return swizzleGe(UVec4(UI16(11, 5), UI16( 6, 5), UI16( 1, 5), UI16( 0, 1)), m_format.order, TextureFormat::RGBA).cast<float>();
case TextureFormat::UNORM_INT_101010: return Vec4(UN32(22, 10), UN32(12, 10), UN32( 2, 10), 1.0f);
case TextureFormat::UNORM_INT_1010102_REV: return swizzleGe( Vec4(UN32( 0, 10), UN32(10, 10), UN32(20, 10), UN32(30, 2)), m_format.order, TextureFormat::RGBA);
case TextureFormat::SNORM_INT_1010102_REV: return swizzleGe( Vec4(SN32( 0, 10), SN32(10, 10), SN32(20, 10), SN32(30, 2)), m_format.order, TextureFormat::RGBA);
case TextureFormat::USCALED_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_1010102_REV: return swizzleGe( UVec4(UI32(0, 10), UI32(10, 10), UI32(20, 10), UI32(30, 2)), m_format.order, TextureFormat::RGBA).cast<float>();
case TextureFormat::SSCALED_INT_1010102_REV:
case TextureFormat::SIGNED_INT_1010102_REV: return swizzleGe( UVec4(SI32(0, 10), SI32(10, 10), SI32(20, 10), SI32(30, 2)), m_format.order, TextureFormat::RGBA).cast<float>();
case TextureFormat::UNSIGNED_INT_999_E5_REV: return unpackRGB999E5(*((const deUint32*)pixelPtr));
case TextureFormat::UNORM_SHORT_1555:
DE_ASSERT(m_format.order == TextureFormat::ARGB);
return Vec4(UN16(15, 1), UN16(10, 5), UN16(5, 5), UN16(0, 5)).swizzle(1,2,3,0); // ARGB -> RGBA
case TextureFormat::UNSIGNED_INT_11F_11F_10F_REV:
return Vec4(Float11(UI32(0, 11)).asFloat(), Float11(UI32(11, 11)).asFloat(), Float10(UI32(22, 10)).asFloat(), 1.0f);
default:
break;
}
#undef UN8
#undef UN16
#undef UN32
#undef SN32
#undef SI32
#undef UI8
#undef UI16
#undef UI32
// Generic path.
Vec4 result;
const TextureSwizzle::Channel* channelMap = getChannelReadSwizzle(m_format.order).components;
int channelSize = getChannelSize(m_format.type);
for (int c = 0; c < 4; c++)
{
switch (channelMap[c])
{
case TextureSwizzle::CHANNEL_0:
case TextureSwizzle::CHANNEL_1:
case TextureSwizzle::CHANNEL_2:
case TextureSwizzle::CHANNEL_3:
result[c] = channelToFloat(pixelPtr + channelSize*((int)channelMap[c]), m_format.type);
break;
case TextureSwizzle::CHANNEL_ZERO:
result[c] = 0.0f;
break;
case TextureSwizzle::CHANNEL_ONE:
result[c] = 1.0f;
break;
default:
DE_ASSERT(false);
}
}
return result;
}
template <typename T>
static tcu::Vector<T, 4> getPixelIntGeneric (const deUint8* pixelPtr, const tcu::TextureFormat& format)
{
tcu::Vector<T, 4> result;
// Generic path.
const TextureSwizzle::Channel* channelMap = getChannelReadSwizzle(format.order).components;
int channelSize = getChannelSize(format.type);
for (int c = 0; c < 4; c++)
{
switch (channelMap[c])
{
case TextureSwizzle::CHANNEL_0:
case TextureSwizzle::CHANNEL_1:
case TextureSwizzle::CHANNEL_2:
case TextureSwizzle::CHANNEL_3:
result[c] = channelToIntType<T>(pixelPtr + channelSize*((int)channelMap[c]), format.type);
break;
case TextureSwizzle::CHANNEL_ZERO:
result[c] = 0;
break;
case TextureSwizzle::CHANNEL_ONE:
result[c] = 1;
break;
default:
DE_ASSERT(false);
}
}
return result;
}
IVec4 ConstPixelBufferAccess::getPixelInt (int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, m_size.x()));
DE_ASSERT(de::inBounds(y, 0, m_size.y()));
DE_ASSERT(de::inBounds(z, 0, m_size.z()));
DE_ASSERT(!isCombinedDepthStencilType(m_format.type)); // combined types cannot be accessed directly
DE_ASSERT(m_format.order != TextureFormat::DS); // combined formats cannot be accessed directly
const deUint8* const pixelPtr = (const deUint8*)getPixelPtr(x, y, z);
// Optimized fomats.
if (m_format.type == TextureFormat::UNORM_INT8)
{
if (m_format.order == TextureFormat::RGBA || m_format.order == TextureFormat::sRGBA)
return readRGBA8888Int(pixelPtr);
else if (m_format.order == TextureFormat::RGB || m_format.order == TextureFormat::sRGB)
return readRGB888Int(pixelPtr);
}
#define U8(OFFS, COUNT) ((*((const deUint8* )pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define U16(OFFS, COUNT) ((*((const deUint16*)pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define U32(OFFS, COUNT) ((*((const deUint32*)pixelPtr) >> (OFFS)) & ((1<<(COUNT))-1))
#define S32(OFFS, COUNT) signExtend(U32(OFFS, COUNT), (COUNT))
switch (m_format.type)
{
case TextureFormat::UNSIGNED_BYTE_44: // Fall-through
case TextureFormat::UNORM_BYTE_44: return UVec4(U8 ( 4, 4), U8 ( 0, 4), 0u, 1u).cast<int>();
case TextureFormat::UNSIGNED_SHORT_565: // Fall-through
case TextureFormat::UNORM_SHORT_565: return swizzleGe(UVec4(U16(11, 5), U16( 5, 6), U16( 0, 5), 1).cast<int>(), m_format.order, TextureFormat::RGB);
case TextureFormat::UNORM_SHORT_555: return swizzleGe(UVec4(U16(10, 5), U16( 5, 5), U16( 0, 5), 1).cast<int>(), m_format.order, TextureFormat::RGB);
case TextureFormat::UNSIGNED_SHORT_4444: // Fall-through
case TextureFormat::UNORM_SHORT_4444: return swizzleGe(UVec4(U16(12, 4), U16( 8, 4), U16( 4, 4), U16( 0, 4)).cast<int>(), m_format.order, TextureFormat::RGBA);
case TextureFormat::UNSIGNED_SHORT_5551: // Fall-through
case TextureFormat::UNORM_SHORT_5551: return swizzleGe(UVec4(U16(11, 5), U16( 6, 5), U16( 1, 5), U16( 0, 1)).cast<int>(), m_format.order, TextureFormat::RGBA);
case TextureFormat::UNORM_INT_101010: return UVec4(U32(22, 10), U32(12, 10), U32( 2, 10), 1).cast<int>();
case TextureFormat::UNORM_INT_1010102_REV: // Fall-through
case TextureFormat::USCALED_INT_1010102_REV: // Fall-through
case TextureFormat::UNSIGNED_INT_1010102_REV: return swizzleGe(UVec4(U32( 0, 10), U32(10, 10), U32(20, 10), U32(30, 2)), m_format.order, TextureFormat::RGBA).cast<int>();
case TextureFormat::SNORM_INT_1010102_REV: // Fall-through
case TextureFormat::SSCALED_INT_1010102_REV: // Fall-through
case TextureFormat::SIGNED_INT_1010102_REV: return swizzleGe(IVec4(S32( 0, 10), S32(10, 10), S32(20, 10), S32(30, 2)), m_format.order, TextureFormat::RGBA);
case TextureFormat::UNORM_SHORT_1555:
DE_ASSERT(m_format.order == TextureFormat::ARGB);
return UVec4(U16(15, 1), U16(10, 5), U16(5, 5), U16(0, 5)).cast<int>().swizzle(1,2,3,0); // ARGB -> RGBA
default:
break; // To generic path.
}
#undef U8
#undef U16
#undef U32
#undef S32
// Generic path.
return getPixelIntGeneric<int>(pixelPtr, m_format);
}
I64Vec4 ConstPixelBufferAccess::getPixelInt64 (int x, int y, int z) const
{
// Rely on getPixelInt() for some formats.
if (m_format.type == TextureFormat::UNORM_INT8 &&
(m_format.order == TextureFormat::RGBA || m_format.order == TextureFormat::sRGBA ||
m_format.order == TextureFormat::RGB || m_format.order == TextureFormat::sRGB))
{
return getPixelInt(x, y, z).cast<deInt64>();
}
switch (m_format.type)
{
case TextureFormat::UNSIGNED_BYTE_44:
case TextureFormat::UNORM_BYTE_44:
case TextureFormat::UNSIGNED_SHORT_565:
case TextureFormat::UNORM_SHORT_565:
case TextureFormat::UNORM_SHORT_555:
case TextureFormat::UNSIGNED_SHORT_4444:
case TextureFormat::UNORM_SHORT_4444:
case TextureFormat::UNSIGNED_SHORT_5551:
case TextureFormat::UNORM_SHORT_5551:
case TextureFormat::UNORM_INT_101010:
case TextureFormat::UNORM_INT_1010102_REV:
case TextureFormat::USCALED_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_1010102_REV:
case TextureFormat::SNORM_INT_1010102_REV:
case TextureFormat::SSCALED_INT_1010102_REV:
case TextureFormat::SIGNED_INT_1010102_REV:
case TextureFormat::UNORM_SHORT_1555:
return getPixelInt(x, y, z).cast<deInt64>();
default:
break; // To generic path.
}
// Generic path.
auto pixelPtr = reinterpret_cast<const deUint8*>(getPixelPtr(x, y, z));
return getPixelIntGeneric<deInt64>(pixelPtr, m_format);
}
template<>
Vec4 ConstPixelBufferAccess::getPixelT (int x, int y, int z) const
{
return getPixel(x, y, z);
}
template<>
IVec4 ConstPixelBufferAccess::getPixelT (int x, int y, int z) const
{
return getPixelInt(x, y, z);
}
template<>
UVec4 ConstPixelBufferAccess::getPixelT (int x, int y, int z) const
{
return getPixelUint(x, y, z);
}
template<>
I64Vec4 ConstPixelBufferAccess::getPixelT (int x, int y, int z) const
{
return getPixelInt64(x, y, z);
}
template<>
U64Vec4 ConstPixelBufferAccess::getPixelT (int x, int y, int z) const
{
return getPixelUint64(x, y, z);
}
float ConstPixelBufferAccess::getPixDepth (int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
const deUint8* const pixelPtr = (const deUint8*)getPixelPtr(x, y, z);
switch (m_format.type)
{
case TextureFormat::UNSIGNED_INT_16_8_8:
DE_ASSERT(m_format.order == TextureFormat::DS);
return (float)readUint32High16(pixelPtr) / 65535.0f;
case TextureFormat::UNSIGNED_INT_24_8:
DE_ASSERT(m_format.order == TextureFormat::D || m_format.order == TextureFormat::DS);
return (float)readUint32High24(pixelPtr) / 16777215.0f;
case TextureFormat::UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::D || m_format.order == TextureFormat::DS);
return (float)readUint32Low24(pixelPtr) / 16777215.0f;
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
return *((const float*)pixelPtr);
default:
DE_ASSERT(m_format.order == TextureFormat::D); // no other combined depth stencil types
return channelToFloat(pixelPtr, m_format.type);
}
}
int ConstPixelBufferAccess::getPixStencil (int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
const deUint8* const pixelPtr = (const deUint8*)getPixelPtr(x, y, z);
switch (m_format.type)
{
case TextureFormat::UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
return (int)readUint32High8(pixelPtr);
case TextureFormat::UNSIGNED_INT_16_8_8:
case TextureFormat::UNSIGNED_INT_24_8:
DE_ASSERT(m_format.order == TextureFormat::DS);
return (int)readUint32Low8(pixelPtr);
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
return (int)readUint32Low8(pixelPtr + 4);
default:
{
DE_ASSERT(m_format.order == TextureFormat::S); // no other combined depth stencil types
return channelToInt(pixelPtr, m_format.type);
}
}
}
void PixelBufferAccess::setPixel (const Vec4& color, int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
DE_ASSERT(!isCombinedDepthStencilType(m_format.type)); // combined types cannot be accessed directly
DE_ASSERT(m_format.order != TextureFormat::DS); // combined formats cannot be accessed directly
deUint8* const pixelPtr = (deUint8*)getPixelPtr(x, y, z);
// Optimized fomats.
if (m_format.type == TextureFormat::UNORM_INT8)
{
if (m_format.order == TextureFormat::RGBA || m_format.order == TextureFormat::sRGBA)
{
writeRGBA8888Float(pixelPtr, color);
return;
}
else if (m_format.order == TextureFormat::RGB || m_format.order == TextureFormat::sRGB)
{
writeRGB888Float(pixelPtr, color);
return;
}
}
#define PN(VAL, OFFS, BITS) (unormFloatToChannel((VAL), (BITS)) << (OFFS))
#define PS(VAL, OFFS, BITS) (snormFloatToChannel((VAL), (BITS)) << (OFFS))
#define PU(VAL, OFFS, BITS) (uintToChannel((VAL), (BITS)) << (OFFS))
#define PI(VAL, OFFS, BITS) (intToChannel((VAL), (BITS)) << (OFFS))
switch (m_format.type)
{
case TextureFormat::UNORM_BYTE_44: *((deUint8 *)pixelPtr) = (deUint8)(PN(color[0], 4, 4) | PN(color[1], 0, 4)); break;
case TextureFormat::UNSIGNED_BYTE_44: *((deUint8 *)pixelPtr) = (deUint8)(PU((deUint32)color[0], 4, 4) | PU((deUint32)color[1], 0, 4)); break;
case TextureFormat::UNORM_INT_101010: *((deUint32*)pixelPtr) = PN(color[0], 22, 10) | PN(color[1], 12, 10) | PN(color[2], 2, 10); break;
case TextureFormat::UNORM_SHORT_565:
{
const Vec4 swizzled = swizzleGe(color, TextureFormat::RGB, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PN(swizzled[0], 11, 5) | PN(swizzled[1], 5, 6) | PN(swizzled[2], 0, 5));
break;
}
case TextureFormat::UNSIGNED_SHORT_565:
{
const UVec4 swizzled = swizzleGe(color.cast<deUint32>(), TextureFormat::RGB, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 11, 5) | PU(swizzled[1], 5, 6) | PU(swizzled[2], 0, 5));
break;
}
case TextureFormat::UNORM_SHORT_555:
{
const Vec4 swizzled = swizzleGe(color, TextureFormat::RGB, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PN(swizzled[0], 10, 5) | PN(swizzled[1], 5, 5) | PN(swizzled[2], 0, 5));
break;
}
case TextureFormat::UNORM_SHORT_4444:
{
const Vec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PN(swizzled[0], 12, 4) | PN(swizzled[1], 8, 4) | PN(swizzled[2], 4, 4) | PN(swizzled[3], 0, 4));
break;
}
case TextureFormat::UNSIGNED_SHORT_4444:
{
const UVec4 swizzled = swizzleGe(color.cast<deUint32>(), TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 12, 4) | PU(swizzled[1], 8, 4) | PU(swizzled[2], 4, 4) | PU(swizzled[3], 0, 4));
break;
}
case TextureFormat::UNORM_SHORT_5551:
{
const Vec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PN(swizzled[0], 11, 5) | PN(swizzled[1], 6, 5) | PN(swizzled[2], 1, 5) | PN(swizzled[3], 0, 1));
break;
}
case TextureFormat::UNORM_SHORT_1555:
{
const Vec4 swizzled = color.swizzle(3,0,1,2); // RGBA -> ARGB
*((deUint16*)pixelPtr) = (deUint16)(PN(swizzled[0], 15, 1) | PN(swizzled[1], 10, 5) | PN(swizzled[2], 5, 5) | PN(swizzled[3], 0, 5));
break;
}
case TextureFormat::UNSIGNED_SHORT_5551:
{
const UVec4 swizzled = swizzleGe(color.cast<deUint32>(), TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 11, 5) | PU(swizzled[1], 6, 5) | PU(swizzled[2], 1, 5) | PU(swizzled[3], 0, 1));
break;
}
case TextureFormat::UNORM_INT_1010102_REV:
{
const Vec4 u = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PN(u[0], 0, 10) | PN(u[1], 10, 10) | PN(u[2], 20, 10) | PN(u[3], 30, 2);
break;
}
case TextureFormat::SNORM_INT_1010102_REV:
{
const Vec4 u = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PS(u[0], 0, 10) | PS(u[1], 10, 10) | PS(u[2], 20, 10) | PS(u[3], 30, 2);
break;
}
case TextureFormat::UNSIGNED_INT_1010102_REV:
case TextureFormat::USCALED_INT_1010102_REV:
{
const UVec4 u = swizzleGe(color.cast<deUint32>(), TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PU(u[0], 0, 10) | PU(u[1], 10, 10) | PU(u[2], 20, 10) | PU(u[3], 30, 2);
break;
}
case TextureFormat::SIGNED_INT_1010102_REV:
case TextureFormat::SSCALED_INT_1010102_REV:
{
const IVec4 u = swizzleGe(color.cast<deInt32>(), TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PI(u[0], 0, 10) | PI(u[1], 10, 10) | PI(u[2], 20, 10) | PI(u[3], 30, 2);
break;
}
case TextureFormat::UNSIGNED_INT_11F_11F_10F_REV:
*((deUint32*)pixelPtr) = Float11(color[0]).bits() | (Float11(color[1]).bits() << 11) | (Float10(color[2]).bits() << 22);
break;
case TextureFormat::UNSIGNED_INT_999_E5_REV:
*((deUint32*)pixelPtr) = packRGB999E5(color);
break;
default:
{
// Generic path.
int numChannels = getNumUsedChannels(m_format.order);
const TextureSwizzle::Channel* map = getChannelWriteSwizzle(m_format.order).components;
int channelSize = getChannelSize(m_format.type);
for (int c = 0; c < numChannels; c++)
{
DE_ASSERT(deInRange32(map[c], TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3));
floatToChannel(pixelPtr + channelSize*c, color[map[c]], m_format.type);
}
break;
}
}
#undef PN
#undef PS
#undef PU
#undef PI
}
void PixelBufferAccess::setPixel (const IVec4& color, int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
DE_ASSERT(!isCombinedDepthStencilType(m_format.type)); // combined types cannot be accessed directly
DE_ASSERT(m_format.order != TextureFormat::DS); // combined formats cannot be accessed directly
deUint8* const pixelPtr = (deUint8*)getPixelPtr(x, y, z);
// Optimized fomats.
if (m_format.type == TextureFormat::UNORM_INT8)
{
if (m_format.order == TextureFormat::RGBA || m_format.order == TextureFormat::sRGBA)
{
writeRGBA8888Int(pixelPtr, color);
return;
}
else if (m_format.order == TextureFormat::RGB || m_format.order == TextureFormat::sRGB)
{
writeRGB888Int(pixelPtr, color);
return;
}
}
#define PU(VAL, OFFS, BITS) (uintToChannel((deUint32)(VAL), (BITS)) << (OFFS))
#define PI(VAL, OFFS, BITS) (intToChannel((deUint32)(VAL), (BITS)) << (OFFS))
switch (m_format.type)
{
case TextureFormat::UNSIGNED_BYTE_44: // Fall-through
case TextureFormat::UNORM_BYTE_44: *((deUint8 *)pixelPtr) = (deUint8 )(PU(color[0], 4, 4) | PU(color[1], 0, 4)); break;
case TextureFormat::UNORM_INT_101010: *((deUint32*)pixelPtr) = PU(color[0], 22, 10) | PU(color[1], 12, 10) | PU(color[2], 2, 10); break;
case TextureFormat::UNORM_SHORT_565:
case TextureFormat::UNSIGNED_SHORT_565:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGB, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 11, 5) | PU(swizzled[1], 5, 6) | PU(swizzled[2], 0, 5));
break;
}
case TextureFormat::UNORM_SHORT_555:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGB, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 10, 5) | PU(swizzled[1], 5, 5) | PU(swizzled[2], 0, 5));
break;
}
case TextureFormat::UNORM_SHORT_4444:
case TextureFormat::UNSIGNED_SHORT_4444:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 12, 4) | PU(swizzled[1], 8, 4) | PU(swizzled[2], 4, 4) | PU(swizzled[3], 0, 4));
break;
}
case TextureFormat::UNORM_SHORT_5551:
case TextureFormat::UNSIGNED_SHORT_5551:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 11, 5) | PU(swizzled[1], 6, 5) | PU(swizzled[2], 1, 5) | PU(swizzled[3], 0, 1));
break;
}
case TextureFormat::UNORM_SHORT_1555:
{
const IVec4 swizzled = color.swizzle(3,0,1,2); // RGBA -> ARGB
*((deUint16*)pixelPtr) = (deUint16)(PU(swizzled[0], 15, 1) | PU(swizzled[1], 10, 5) | PU(swizzled[2], 5, 5) | PU(swizzled[3], 0, 5));
break;
}
case TextureFormat::UNORM_INT_1010102_REV:
case TextureFormat::UNSIGNED_INT_1010102_REV:
case TextureFormat::USCALED_INT_1010102_REV:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PU(swizzled[0], 0, 10) | PU(swizzled[1], 10, 10) | PU(swizzled[2], 20, 10) | PU(swizzled[3], 30, 2);
break;
}
case TextureFormat::SNORM_INT_1010102_REV:
case TextureFormat::SIGNED_INT_1010102_REV:
case TextureFormat::SSCALED_INT_1010102_REV:
{
const IVec4 swizzled = swizzleGe(color, TextureFormat::RGBA, m_format.order);
*((deUint32*)pixelPtr) = PI(swizzled[0], 0, 10) | PI(swizzled[1], 10, 10) | PI(swizzled[2], 20, 10) | PI(swizzled[3], 30, 2);
break;
}
default:
{
// Generic path.
int numChannels = getNumUsedChannels(m_format.order);
const TextureSwizzle::Channel* map = getChannelWriteSwizzle(m_format.order).components;
int channelSize = getChannelSize(m_format.type);
for (int c = 0; c < numChannels; c++)
{
DE_ASSERT(deInRange32(map[c], TextureSwizzle::CHANNEL_0, TextureSwizzle::CHANNEL_3));
intToChannel(pixelPtr + channelSize*c, color[map[c]], m_format.type);
}
break;
}
}
#undef PU
#undef PI
}
void PixelBufferAccess::setPixDepth (float depth, int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
deUint8* const pixelPtr = (deUint8*)getPixelPtr(x, y, z);
switch (m_format.type)
{
case TextureFormat::UNSIGNED_INT_16_8_8:
DE_ASSERT(m_format.order == TextureFormat::DS);
writeUint32High16(pixelPtr, convertSatRte<deUint16>(depth * 65535.0f));
break;
case TextureFormat::UNSIGNED_INT_24_8:
DE_ASSERT(m_format.order == TextureFormat::D || m_format.order == TextureFormat::DS);
writeUint32High24(pixelPtr, convertSatRteUint24(depth * 16777215.0f));
break;
case TextureFormat::UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::D || m_format.order == TextureFormat::DS);
writeUint32Low24(pixelPtr, convertSatRteUint24(depth * 16777215.0f));
break;
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
*((float*)pixelPtr) = depth;
break;
default:
DE_ASSERT(m_format.order == TextureFormat::D); // no other combined depth stencil types
floatToChannel(pixelPtr, depth, m_format.type);
break;
}
}
void PixelBufferAccess::setPixStencil (int stencil, int x, int y, int z) const
{
DE_ASSERT(de::inBounds(x, 0, getWidth()));
DE_ASSERT(de::inBounds(y, 0, getHeight()));
DE_ASSERT(de::inBounds(z, 0, getDepth()));
deUint8* const pixelPtr = (deUint8*)getPixelPtr(x, y, z);
switch (m_format.type)
{
case TextureFormat::UNSIGNED_INT_16_8_8:
case TextureFormat::UNSIGNED_INT_24_8:
DE_ASSERT(m_format.order == TextureFormat::DS);
writeUint32Low8(pixelPtr, convertSat<deUint8>((deUint32)stencil));
break;
case TextureFormat::UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
writeUint32High8(pixelPtr, convertSat<deUint8>((deUint32)stencil));
break;
case TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
DE_ASSERT(m_format.order == TextureFormat::DS);
writeUint32Low8(pixelPtr + 4, convertSat<deUint8>((deUint32)stencil));
break;
default:
DE_ASSERT(m_format.order == TextureFormat::S); // no other combined depth stencil types
intToChannel(pixelPtr, stencil, m_format.type);
break;
}
}
static inline int imod (int a, int b)
{
int m = a % b;
return m < 0 ? m + b : m;
}
static inline int mirror (int a)
{
return a >= 0 ? a : -(1 + a);
}
// Nearest-even rounding in case of tie (fractional part 0.5), otherwise ordinary rounding.
static inline float rint (float a)
{
DE_STATIC_ASSERT((-3 % 2 != 0) && (-4 % 2 == 0));
float fracVal = deFloatFrac(a);
if (fracVal != 0.5f)
return deFloatRound(a); // Ordinary case.
float floorVal = a - fracVal;
bool roundUp = (deInt64)floorVal % 2 != 0;
return floorVal + (roundUp ? 1.0f : 0.0f);
}
static inline int wrap (Sampler::WrapMode mode, int c, int size)
{
switch (mode)
{
case tcu::Sampler::CLAMP_TO_BORDER:
return deClamp32(c, -1, size);
case tcu::Sampler::CLAMP_TO_EDGE:
return deClamp32(c, 0, size-1);
case tcu::Sampler::REPEAT_GL:
return imod(c, size);
case tcu::Sampler::REPEAT_CL:
return imod(c, size);
case tcu::Sampler::MIRRORED_ONCE:
c = deClamp32(c, -size, size);
// Fall-through
case tcu::Sampler::MIRRORED_REPEAT_GL:
return (size - 1) - mirror(imod(c, 2*size) - size);
case tcu::Sampler::MIRRORED_REPEAT_CL:
return deClamp32(c, 0, size-1); // \note Actual mirroring done already in unnormalization function.
default:
DE_ASSERT(DE_FALSE);
return 0;
}
}
// Special unnormalization for REPEAT_CL and MIRRORED_REPEAT_CL wrap modes; otherwise ordinary unnormalization.
static inline float unnormalize (Sampler::WrapMode mode, float c, int size)
{
switch (mode)
{
case tcu::Sampler::CLAMP_TO_EDGE:
case tcu::Sampler::CLAMP_TO_BORDER:
case tcu::Sampler::REPEAT_GL:
case tcu::Sampler::MIRRORED_REPEAT_GL:
case tcu::Sampler::MIRRORED_ONCE: // Fall-through (ordinary case).
return (float)size*c;
case tcu::Sampler::REPEAT_CL:
return (float)size * (c - deFloatFloor(c));
case tcu::Sampler::MIRRORED_REPEAT_CL:
return (float)size * deFloatAbs(c - 2.0f * rint(0.5f * c));
default:
DE_ASSERT(DE_FALSE);
return 0.0f;
}
}
static bool isFixedPointDepthTextureFormat (const tcu::TextureFormat& format)
{
DE_ASSERT(format.order == TextureFormat::D || format.order == TextureFormat::R);
const tcu::TextureChannelClass channelClass = tcu::getTextureChannelClass(format.type);
if (channelClass == tcu::TEXTURECHANNELCLASS_FLOATING_POINT)
return false;
else if (channelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT)
return true;
else
{
DE_ASSERT(false);
return false;
}
}
// Texel lookup with color conversion.
static inline Vec4 lookup (const ConstPixelBufferAccess& access, int i, int j, int k)
{
const TextureFormat& format = access.getFormat();
if (isSRGB(format))
{
if (format.type == TextureFormat::UNORM_INT8 && format.order == TextureFormat::sRGB)
return sRGB8ToLinear(access.getPixelUint(i, j, k));
else if (format.type == TextureFormat::UNORM_INT8 && format.order == TextureFormat::sRGBA)
return sRGBA8ToLinear(access.getPixelUint(i, j, k));
else
return sRGBToLinear(access.getPixel(i, j, k));
}
else
{
return access.getPixel(i, j, k);
}
}
// Border texel lookup with color conversion.
static inline Vec4 lookupBorder (const tcu::TextureFormat& format, const tcu::Sampler& sampler)
{
// "lookup" for a combined format does not make sense, disallow
DE_ASSERT(!isCombinedDepthStencilType(format.type));
const tcu::TextureChannelClass channelClass = tcu::getTextureChannelClass(format.type);
const bool isFloat = channelClass == tcu::TEXTURECHANNELCLASS_FLOATING_POINT;
const bool isFixed = channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT ||
channelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT;
const bool isPureInteger = channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER;
const bool isPureUnsignedInteger = channelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER;
if (isFloat || isFixed)
return sampleTextureBorder<float>(format, sampler);
else if (isPureInteger)
return sampleTextureBorder<deInt32>(format, sampler).cast<float>();
else if (isPureUnsignedInteger)
return sampleTextureBorder<deUint32>(format, sampler).cast<float>();
else
{
DE_ASSERT(false);
return Vec4(-1.0);
}
}
static inline float execCompare (const tcu::Vec4& color, Sampler::CompareMode compare, int chanNdx, float ref_, bool isFixedPoint)
{
const bool clampValues = isFixedPoint; // if comparing against a floating point texture, ref (and value) is not clamped
const float cmp = (clampValues) ? (de::clamp(color[chanNdx], 0.0f, 1.0f)) : (color[chanNdx]);
const float ref = (clampValues) ? (de::clamp(ref_, 0.0f, 1.0f)) : (ref_);
bool res = false;
switch (compare)
{
case Sampler::COMPAREMODE_LESS: res = ref < cmp; break;
case Sampler::COMPAREMODE_LESS_OR_EQUAL: res = ref <= cmp; break;
case Sampler::COMPAREMODE_GREATER: res = ref > cmp; break;
case Sampler::COMPAREMODE_GREATER_OR_EQUAL: res = ref >= cmp; break;
case Sampler::COMPAREMODE_EQUAL: res = ref == cmp; break;
case Sampler::COMPAREMODE_NOT_EQUAL: res = ref != cmp; break;
case Sampler::COMPAREMODE_ALWAYS: res = true; break;
case Sampler::COMPAREMODE_NEVER: res = false; break;
default:
DE_ASSERT(false);
}
return res ? 1.0f : 0.0f;
}
static Vec4 sampleNearest1D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, const IVec2& offset)
{
int width = access.getWidth();
int x = deFloorFloatToInt32(u)+offset.x();
// Check for CLAMP_TO_BORDER.
if (sampler.wrapS == Sampler::CLAMP_TO_BORDER && !deInBounds32(x, 0, width))
return lookupBorder(access.getFormat(), sampler);
int i = wrap(sampler.wrapS, x, width);
return lookup(access, i, offset.y(), 0);
}
static Vec4 sampleNearest2D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, const IVec3& offset)
{
int width = access.getWidth();
int height = access.getHeight();
int x = deFloorFloatToInt32(u)+offset.x();
int y = deFloorFloatToInt32(v)+offset.y();
// Check for CLAMP_TO_BORDER.
if ((sampler.wrapS == Sampler::CLAMP_TO_BORDER && !deInBounds32(x, 0, width)) ||
(sampler.wrapT == Sampler::CLAMP_TO_BORDER && !deInBounds32(y, 0, height)))
return lookupBorder(access.getFormat(), sampler);
int i = wrap(sampler.wrapS, x, width);
int j = wrap(sampler.wrapT, y, height);
return lookup(access, i, j, offset.z());
}
static Vec4 sampleNearest3D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, float w, const IVec3& offset)
{
int width = access.getWidth();
int height = access.getHeight();
int depth = access.getDepth();
int x = deFloorFloatToInt32(u)+offset.x();
int y = deFloorFloatToInt32(v)+offset.y();
int z = deFloorFloatToInt32(w)+offset.z();
// Check for CLAMP_TO_BORDER.
if ((sampler.wrapS == Sampler::CLAMP_TO_BORDER && !deInBounds32(x, 0, width)) ||
(sampler.wrapT == Sampler::CLAMP_TO_BORDER && !deInBounds32(y, 0, height)) ||
(sampler.wrapR == Sampler::CLAMP_TO_BORDER && !deInBounds32(z, 0, depth)))
return lookupBorder(access.getFormat(), sampler);
int i = wrap(sampler.wrapS, x, width);
int j = wrap(sampler.wrapT, y, height);
int k = wrap(sampler.wrapR, z, depth);
return lookup(access, i, j, k);
}
static Vec4 sampleLinear1D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, const IVec2& offset)
{
int w = access.getWidth();
int x0 = deFloorFloatToInt32(u-0.5f)+offset.x();
int x1 = x0+1;
int i0 = wrap(sampler.wrapS, x0, w);
int i1 = wrap(sampler.wrapS, x1, w);
float a = deFloatFrac(u-0.5f);
bool i0UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i0, 0, w);
bool i1UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i1, 0, w);
// Border color for out-of-range coordinates if using CLAMP_TO_BORDER, otherwise execute lookups.
Vec4 p0 = i0UseBorder ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, offset.y(), 0);
Vec4 p1 = i1UseBorder ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, offset.y(), 0);
// Interpolate.
return p0 * (1.0f - a) + p1 * a;
}
static Vec4 sampleCubic1D(const ConstPixelBufferAccess& access, const Sampler& sampler, float u, const IVec2& offset)
{
int width = access.getWidth();
tcu::IVec4 x, i;
x[0] = deFloorFloatToInt32(u - 1.5f) + offset.x();
x[1] = x[0] + 1;
x[2] = x[1] + 1;
x[3] = x[2] + 1;
for (deUint32 m = 0; m < 4; ++m)
i[m] = wrap(sampler.wrapS, x[m], width);
bool iUseBorder[4];
for (deUint32 m = 0; m < 4; ++m)
iUseBorder[m] = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i[m], 0, width);
// Catmull-Rom basis matrix
static const float crValues[16] = { 0.0f, 1.0f, 0.0f, 0.0f,
-0.5f, 0.0f, 0.5f, 0.0f,
1.0f, -2.5f, 2.0f, -0.5f,
-0.5f, 1.5f, -1.5f, 0.5f };
static const tcu::Mat4 crBasis(crValues);
float a = deFloatFrac(u - 0.5f);
tcu::Vec4 alpha(1, a, a*a, a*a*a);
tcu::Vec4 wi = alpha * crBasis;
tcu::Vec4 result(0.0f, 0.0f, 0.0f, 0.0f);
for (deUint32 m = 0; m < 4; ++m)
{
tcu::Vec4 p = (iUseBorder[m]) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i[m], offset.y(), 0);
result += wi[m] * p;
}
return result;
}
static Vec4 sampleLinear2D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, const IVec3& offset)
{
int w = access.getWidth();
int h = access.getHeight();
int x0 = deFloorFloatToInt32(u-0.5f)+offset.x();
int x1 = x0+1;
int y0 = deFloorFloatToInt32(v-0.5f)+offset.y();
int y1 = y0+1;
int i0 = wrap(sampler.wrapS, x0, w);
int i1 = wrap(sampler.wrapS, x1, w);
int j0 = wrap(sampler.wrapT, y0, h);
int j1 = wrap(sampler.wrapT, y1, h);
float a = deFloatFrac(u-0.5f);
float b = deFloatFrac(v-0.5f);
bool i0UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i0, 0, w);
bool i1UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i1, 0, w);
bool j0UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j0, 0, h);
bool j1UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j1, 0, h);
// Border color for out-of-range coordinates if using CLAMP_TO_BORDER, otherwise execute lookups.
Vec4 p00 = (i0UseBorder || j0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j0, offset.z());
Vec4 p10 = (i1UseBorder || j0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j0, offset.z());
Vec4 p01 = (i0UseBorder || j1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j1, offset.z());
Vec4 p11 = (i1UseBorder || j1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j1, offset.z());
// Interpolate.
return (p00*(1.0f-a)*(1.0f-b)) +
(p10*( a)*(1.0f-b)) +
(p01*(1.0f-a)*( b)) +
(p11*( a)*( b));
}
static Vec4 sampleCubic2D(const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, const IVec3& offset)
{
int width = access.getWidth();
int height = access.getHeight();
tcu::IVec4 x, y, i, j;
x[0] = deFloorFloatToInt32(u - 1.5f) + offset.x();
x[1] = x[0] + 1;
x[2] = x[1] + 1;
x[3] = x[2] + 1;
y[0] = deFloorFloatToInt32(v - 1.5f) + offset.y();
y[1] = y[0] + 1;
y[2] = y[1] + 1;
y[3] = y[2] + 1;
for (deUint32 m = 0; m < 4; ++m)
i[m] = wrap(sampler.wrapS, x[m], width);
for (deUint32 n = 0; n < 4; ++n)
j[n] = wrap(sampler.wrapT, y[n], height);
bool iUseBorder[4], jUseBorder[4];
for (deUint32 m = 0; m < 4; ++m)
iUseBorder[m] = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i[m], 0, width);
for (deUint32 n = 0; n < 4; ++n)
jUseBorder[n] = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j[n], 0, height);
// Catmull-Rom basis matrix
static const float crValues[16] = { 0.0f, 1.0f, 0.0f, 0.0f,
-0.5f, 0.0f, 0.5f, 0.0f,
1.0f, -2.5f, 2.0f, -0.5f,
-0.5f, 1.5f, -1.5f, 0.5f };
static const tcu::Mat4 crBasis(crValues);
float a = deFloatFrac(u - 0.5f);
float b = deFloatFrac(v - 0.5f);
tcu::Vec4 alpha (1, a, a*a, a*a*a);
tcu::Vec4 beta (1, b, b*b, b*b*b);
tcu::Vec4 wi = alpha * crBasis;
tcu::Vec4 wj = beta * crBasis;
tcu::Vec4 result(0.0f, 0.0f, 0.0f, 0.0f);
for (deUint32 n = 0; n < 4; ++n)
for (deUint32 m = 0; m < 4; ++m)
{
tcu::Vec4 p = (iUseBorder[m] || jUseBorder[n]) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i[m], j[n], offset.z());
result += wi[m] * wj[n] * p;
}
return result;
}
static float sampleLinear1DCompare (const ConstPixelBufferAccess& access, const Sampler& sampler, float ref, float u, const IVec2& offset, bool isFixedPointDepthFormat)
{
int w = access.getWidth();
int x0 = deFloorFloatToInt32(u-0.5f)+offset.x();
int x1 = x0+1;
int i0 = wrap(sampler.wrapS, x0, w);
int i1 = wrap(sampler.wrapS, x1, w);
float a = deFloatFrac(u-0.5f);
bool i0UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i0, 0, w);
bool i1UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i1, 0, w);
// Border color for out-of-range coordinates if using CLAMP_TO_BORDER, otherwise execute lookups.
Vec4 p0Clr = i0UseBorder ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, offset.y(), 0);
Vec4 p1Clr = i1UseBorder ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, offset.y(), 0);
// Execute comparisons.
float p0 = execCompare(p0Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
float p1 = execCompare(p1Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
// Interpolate.
return (p0 * (1.0f - a)) + (p1 * a);
}
static float sampleLinear2DCompare (const ConstPixelBufferAccess& access, const Sampler& sampler, float ref, float u, float v, const IVec3& offset, bool isFixedPointDepthFormat)
{
int w = access.getWidth();
int h = access.getHeight();
int x0 = deFloorFloatToInt32(u-0.5f)+offset.x();
int x1 = x0+1;
int y0 = deFloorFloatToInt32(v-0.5f)+offset.y();
int y1 = y0+1;
int i0 = wrap(sampler.wrapS, x0, w);
int i1 = wrap(sampler.wrapS, x1, w);
int j0 = wrap(sampler.wrapT, y0, h);
int j1 = wrap(sampler.wrapT, y1, h);
float a = deFloatFrac(u-0.5f);
float b = deFloatFrac(v-0.5f);
bool i0UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i0, 0, w);
bool i1UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i1, 0, w);
bool j0UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j0, 0, h);
bool j1UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j1, 0, h);
// Border color for out-of-range coordinates if using CLAMP_TO_BORDER, otherwise execute lookups.
Vec4 p00Clr = (i0UseBorder || j0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j0, offset.z());
Vec4 p10Clr = (i1UseBorder || j0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j0, offset.z());
Vec4 p01Clr = (i0UseBorder || j1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j1, offset.z());
Vec4 p11Clr = (i1UseBorder || j1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j1, offset.z());
// Execute comparisons.
float p00 = execCompare(p00Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
float p10 = execCompare(p10Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
float p01 = execCompare(p01Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
float p11 = execCompare(p11Clr, sampler.compare, sampler.compareChannel, ref, isFixedPointDepthFormat);
// Interpolate.
return (p00*(1.0f-a)*(1.0f-b)) +
(p10*( a)*(1.0f-b)) +
(p01*(1.0f-a)*( b)) +
(p11*( a)*( b));
}
static Vec4 sampleLinear3D (const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, float w, const IVec3& offset)
{
int width = access.getWidth();
int height = access.getHeight();
int depth = access.getDepth();
int x0 = deFloorFloatToInt32(u-0.5f)+offset.x();
int x1 = x0+1;
int y0 = deFloorFloatToInt32(v-0.5f)+offset.y();
int y1 = y0+1;
int z0 = deFloorFloatToInt32(w-0.5f)+offset.z();
int z1 = z0+1;
int i0 = wrap(sampler.wrapS, x0, width);
int i1 = wrap(sampler.wrapS, x1, width);
int j0 = wrap(sampler.wrapT, y0, height);
int j1 = wrap(sampler.wrapT, y1, height);
int k0 = wrap(sampler.wrapR, z0, depth);
int k1 = wrap(sampler.wrapR, z1, depth);
float a = deFloatFrac(u-0.5f);
float b = deFloatFrac(v-0.5f);
float c = deFloatFrac(w-0.5f);
bool i0UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i0, 0, width);
bool i1UseBorder = sampler.wrapS == Sampler::CLAMP_TO_BORDER && !de::inBounds(i1, 0, width);
bool j0UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j0, 0, height);
bool j1UseBorder = sampler.wrapT == Sampler::CLAMP_TO_BORDER && !de::inBounds(j1, 0, height);
bool k0UseBorder = sampler.wrapR == Sampler::CLAMP_TO_BORDER && !de::inBounds(k0, 0, depth);
bool k1UseBorder = sampler.wrapR == Sampler::CLAMP_TO_BORDER && !de::inBounds(k1, 0, depth);
// Border color for out-of-range coordinates if using CLAMP_TO_BORDER, otherwise execute lookups.
Vec4 p000 = (i0UseBorder || j0UseBorder || k0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j0, k0);
Vec4 p100 = (i1UseBorder || j0UseBorder || k0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j0, k0);
Vec4 p010 = (i0UseBorder || j1UseBorder || k0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j1, k0);
Vec4 p110 = (i1UseBorder || j1UseBorder || k0UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j1, k0);
Vec4 p001 = (i0UseBorder || j0UseBorder || k1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j0, k1);
Vec4 p101 = (i1UseBorder || j0UseBorder || k1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j0, k1);
Vec4 p011 = (i0UseBorder || j1UseBorder || k1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i0, j1, k1);
Vec4 p111 = (i1UseBorder || j1UseBorder || k1UseBorder) ? lookupBorder(access.getFormat(), sampler) : lookup(access, i1, j1, k1);
// Interpolate.
return (p000*(1.0f-a)*(1.0f-b)*(1.0f-c)) +
(p100*( a)*(1.0f-b)*(1.0f-c)) +
(p010*(1.0f-a)*( b)*(1.0f-c)) +
(p110*( a)*( b)*(1.0f-c)) +
(p001*(1.0f-a)*(1.0f-b)*( c)) +
(p101*( a)*(1.0f-b)*( c)) +
(p011*(1.0f-a)*( b)*( c)) +
(p111*( a)*( b)*( c));
}
static Vec4 sampleCubic3D(const ConstPixelBufferAccess& access, const Sampler& sampler, float u, float v, float w, const IVec3& offset)
{
int width = access.getWidth();
int height = access.getHeight();
int depth = access.getDepth();
tcu::IVec4 x, y, z, i, j, k;
x[0] = deFloorFloatToInt32(u - 1.5f) + offset.x();
x[1] = x[0] + 1;
x[2] = x[1] + 1;
x[3] = x[2] + 1;
y[0] = deFloorFloatToInt32(v - 1.5f) + offset.y();
y[1] = y[0] + 1;
y[2] = y[1] + 1;
y[3] = y[2] + 1;
z[0] = deFloorFloatToInt32(w - 1.5f) + offset.z();
z[1] = z[0] + 1;
z[2] = z[1] + 1;
z[3] = z[2] + 1;
for (deUint32 m = 0; m < 4; ++m)
i[m] = wrap(sampler.wrapS, x[m], width);
for (deUint32 n = 0; n < 4; ++n)
j[n] = wrap(sampler.wrapT, y[n], height);
for (deUint32 o = 0; o < 4; ++o)