Skip to content

Commit 9fcfd7c

Browse files
NavNTCMPNavNTCMP
committed
v4.1 updates
1 parent 71eb7cd commit 9fcfd7c

File tree

295 files changed

+30624
-10833
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

295 files changed

+30624
-10833
lines changed

‎applications/_libs/cmakelists.txt‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -6,6 +6,8 @@ add_subdirectory(cmp_fileio)
66
add_subdirectory(cmp_gui)
77
add_subdirectory(cmp_math)
88
add_subdirectory(cmp_mesh)
9+
if (OPTION_BUILD_DRACO)
910
add_subdirectory(cmp_meshcompressor)
11+
endif()
1012
add_subdirectory(cmp_meshoptimizer)
1113
add_subdirectory(gpu_decode)

‎applications/_libs/cmp_fileio/cmakelists.txt‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -11,9 +11,9 @@ target_sources(CMP_FileIO PRIVATE
1111
)
1212

1313
target_include_directories(CMP_FileIO PUBLIC
14-
15-
${LEGACY_LOCATION}
1614
./
15+
${LEGACY_LOCATION}
16+
${OpenEXR_INCLUDE_DIRS}
1717
)
1818

1919
set_target_properties(CMP_FileIO PROPERTIES FOLDER ${FOLDER_NAME})

‎applications/_libs/cmp_math/cmakelists.txt‎

Lines changed: 12 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,11 @@
11

22
add_library(CMP_Math)
33

4-
# This is the current location of these files. They should be moved here
5-
set(LEGACY_LOCATION ../../../cmp_framework/common)
4+
if(CMP_HOST_WINDOWS)
5+
target_compile_definitions(CMP_Math PUBLIC
6+
-DCMP_USE_XMMINTRIN
7+
)
8+
endif()
69

710
target_sources(CMP_Math PRIVATE
811

@@ -19,18 +22,20 @@ target_sources(CMP_Math PRIVATE
1922
jmlvec2.h
2023
jmlvec3.h
2124
jrtcommon.h
22-
${LEGACY_LOCATION}/mathmacros.h
25+
${PROJECT_SOURCE_DIR}/cmp_framework/common/mathmacros.h
2326
tootlepch.h
2427
)
2528

2629
target_include_directories(CMP_Math PUBLIC
27-
2830
./
31+
${PROJECT_SOURCE_DIR}/cmp_core/shaders
32+
${PROJECT_SOURCE_DIR}/cmp_core/source
2933
)
3034

3135
target_link_libraries(CMP_Math PRIVATE
32-
33-
CMP_Core
36+
# CMP_Core
3437
)
3538

36-
set_target_properties(CMP_Math PROPERTIES FOLDER ${FOLDER_NAME})
39+
set_target_properties(CMP_Math PROPERTIES
40+
FOLDER "Libs"
41+
)

‎applications/_libs/cmp_math/cmp_math_common.cpp‎

Lines changed: 37 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
//=====================================================================
2-
// Copyright 2018 (c), Advanced Micro Devices, Inc. All rights reserved.
2+
// Copyright 2020 (c), Advanced Micro Devices, Inc. All rights reserved.
33
//
44
// Permission is hereby granted, free of charge, to any person obtaining a copy
55
// of this software and associated documentation files(the "Software"), to deal
@@ -34,7 +34,8 @@ float cpu_sqrtf(float * pIn) {
3434
return sqrtf(*pIn);
3535
}
3636

37-
#ifndef _LINUX
37+
#ifdef CMP_USE_XMMINTRIN
38+
#ifndef __linux__
3839
//---------------------------------------------
3940
// SSE: Computes square root of a float value
4041
//---------------------------------------------
@@ -45,6 +46,7 @@ float sse_sqrtf( float *pIn ) {
4546
return val.m128_f32[0];
4647
}
4748
#endif
49+
#endif
4850

4951
//-------------------------------------------------
5052
// CPU: Computes 1 / (square root of a float value)
@@ -57,16 +59,29 @@ float cpu_rsqf(float *f) {
5759
return 0.0f;
5860
}
5961

60-
#ifndef _LINUX
62+
#ifdef CMP_USE_XMMINTRIN
63+
#ifndef __linux__
6164
//-------------------------------------------------
6265
// SSE: Computes 1 / (square root of a float value)
6366
//-------------------------------------------------
64-
float sse_rsqf(float *v) {
67+
#ifdef CMP_USE_RSQ_RSQR
68+
float sse_rsqf(float* v)
69+
{
6570
__m128 val = _mm_load1_ps(v);
66-
val = _mm_rsqrt_ss(val);
71+
val = _mm_rsqrt_ss(val);
6772
float frsq = val.m128_f32[0];
68-
return (0.5f * frsq) * (3.0f - (*v * frsq) * frsq);
73+
return (0.5f * frsq) * (3.0f - (*v * frsq) * frsq);
6974
};
75+
#else
76+
float sse_rsqf(float *v) {
77+
__m128 val = _mm_set_ss(*v); // Copy float and zero the upper 3 elements
78+
__m128 val1 = _mm_set_ss(1.0f);
79+
val = _mm_sqrt_ss(val);
80+
val = _mm_div_ss(val1, val);
81+
return ( val.m128_f32[0] );
82+
};
83+
#endif
84+
#endif
7085
#endif
7186

7287
//---------------------------------------------
@@ -76,13 +91,15 @@ float cpu_minf(float l1, float r1) {
7691
return (l1 < r1 ? l1 : r1);
7792
}
7893

79-
#ifndef _LINUX
94+
#ifdef CMP_USE_XMMINTRIN
95+
#ifndef __linux__
8096
float sse_minf( float a, float b ) {
8197
// Branchless SSE min.
8298
_mm_store_ss( &a, _mm_min_ss(_mm_set_ss(a),_mm_set_ss(b)) );
8399
return a;
84100
}
85101
#endif
102+
#endif
86103

87104
//---------------------------------------------
88105
// CPU: Computes max of two float values
@@ -91,13 +108,15 @@ float cpu_maxf(float l1, float r1) {
91108
return (l1 > r1 ? l1 : r1);
92109
}
93110

94-
#ifndef _LINUX
111+
#ifdef CMP_USE_XMMINTRIN
112+
#ifndef __linux__
95113
float sse_maxf( float a, float b ) {
96114
// Branchless SSE max.
97115
_mm_store_ss( &a, _mm_max_ss(_mm_set_ss(a),_mm_set_ss(b)) );
98116
return a;
99117
}
100118
#endif
119+
#endif
101120

102121
//================================================
103122
// Clamp the value in the range [minval .. maxval]
@@ -111,12 +130,14 @@ float cpu_clampf(float value, float minval, float maxval) {
111130
return value;
112131
}
113132

114-
#ifndef _LINUX
133+
#ifdef CMP_USE_XMMINTRIN
134+
#ifndef __linux__
115135
float sse_clampf( float val, float minval, float maxval ) {
116136
_mm_store_ss( &val, _mm_min_ss( _mm_max_ss(_mm_set_ss(val),_mm_set_ss(minval)), _mm_set_ss(maxval) ) );
117137
return val;
118138
}
119139
#endif
140+
#endif
120141

121142
void cpu_averageRGB(unsigned char *src_rgba_block) {
122143
float medianR = 0.0f, medianG = 0.0f, medianB = 0.0f;
@@ -184,7 +205,8 @@ float cpu_lerp2(CMP_Vec4uc C1, CMP_Vec4uc CA, CMP_Vec4uc CB, CMP_Vec4uc C2, CMP_
184205
return float(min1+min2);
185206
}
186207

187-
#ifndef _LINUX
208+
#ifdef CMP_USE_XMMINTRIN
209+
#ifndef __linux__
188210
float sse_lerp2(CMP_Vec4uc C1, CMP_Vec4uc CA, CMP_Vec4uc CB, CMP_Vec4uc C2, CMP_MATH_BYTE *encode1, CMP_MATH_BYTE *encode2) {
189211
// Initial Setup
190212
__m128 iC1, iC2, iCA, iCB; //Load auchars into _m128
@@ -301,6 +323,7 @@ void cmp_set_fma3_features() {
301323
cmp_lerp2 = fma_lerp2;
302324
}
303325
#endif
326+
#endif
304327

305328

306329
void cmp_set_cpu_features() {
@@ -313,7 +336,9 @@ void cmp_set_cpu_features() {
313336
cmp_sqrtf = cpu_sqrtf;
314337
}
315338

316-
#ifndef _LINUX
339+
340+
#ifdef CMP_USE_XMMINTRIN
341+
#ifndef __linux__
317342
void cmp_set_sse2_features() {
318343
cmp_clampf = sse_clampf;
319344
cmp_lerp2 = sse_lerp2;
@@ -323,6 +348,7 @@ void cmp_set_sse2_features() {
323348
cmp_sqrtf = sse_sqrtf;
324349
}
325350
#endif
351+
#endif
326352

327353
//---------------------------------
328354
// User Interface to the CMP_MATH

‎applications/_libs/cmp_math/cmp_math_common.h‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -43,10 +43,12 @@
4343
typedef unsigned char CMP_MATH_BYTE;
4444
typedef unsigned int CMP_MATH_DWORD;
4545

46-
#ifndef _LINUX
46+
#ifdef CMP_USE_XMMINTRIN
47+
#ifndef __linux__
4748
extern void cmp_set_fma3_features();
4849
extern void cmp_set_sse2_features();
4950
#endif
51+
#endif
5052

5153
extern void cmp_set_cpu_features();
5254

@@ -61,7 +63,7 @@ extern float cpu_rsqf(float *f);
6163
extern float cpu_sqrtf(float * pIn);
6264

6365

64-
#ifndef _LINUX
66+
#ifndef __linux__
6567
extern float sse_clampf(float value, float minval, float maxval);
6668
extern float sse_lerp2(CMP_Vec4uc C1, CMP_Vec4uc CA, CMP_Vec4uc CB, CMP_Vec4uc C2, CMP_MATH_BYTE *encode1, CMP_MATH_BYTE *encode2);
6769
extern float sse_maxf(float l1, float r1);

‎applications/_libs/cmp_math/cmp_math_cpuid.cpp‎

Lines changed: 5 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -30,7 +30,7 @@
3030
void cmp_cpuid(int cpuInfo[4], int function_id) {
3131
// subfunction_id = 0
3232
#ifdef _WIN32
33-
__cpuidex(cpuInfo, function_id, 0);
33+
__cpuidex(cpuInfo, function_id, 0); // defined in intrin.h
3434
#else
3535
// To Do
3636
//__cpuid_count(0, function_id, cpuInfo[0], cpuInfo[1], cpuInfo[2], cpuInfo[3]);
@@ -47,7 +47,7 @@ cmp_cpufeatures cmp_get_cpufeatures() {
4747
cpu.feature[i] = 0;
4848
}
4949

50-
#ifndef _LINUX
50+
#ifndef __linux__
5151
cmp_cpuid(cpuInfo,0);
5252
int nIds = cpuInfo[0];
5353

@@ -118,7 +118,8 @@ void cmp_autodected_cpufeatures(CMP_MATH_BYTE set) {
118118
if ((set & CMP_MATH_USE_CPU) > 0) return;
119119

120120

121-
#ifndef _LINUX
121+
#ifdef CMP_USE_XMMINTRIN
122+
#ifndef __linux__
122123
// Auto detect CPU features to enable
123124
for (int i = 0; i<SSP_SSE_COUNT; i++) {
124125
if (cpu.feature[i] > 0) {
@@ -139,6 +140,7 @@ void cmp_autodected_cpufeatures(CMP_MATH_BYTE set) {
139140
}
140141
}
141142
#endif
143+
#endif
142144

143145
}
144146

‎applications/_libs/cmp_math/jml.h‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -24,7 +24,7 @@
2424
#ifndef _JML_H_
2525
#define _JML_H_
2626

27-
#ifdef _LINUX
27+
#ifdef __linux__
2828
#define ALIGN16
2929
#else
3030
// helpful alias for 16-byte alignment

‎applications/_libs/cmp_math/jmlfuncs.h‎

Lines changed: 15 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
//=====================================================================
2-
// Copyright 2006-2018 (c), Advanced Micro Devices, Inc. All rights reserved.
2+
// Copyright 2006-2020 (c), Advanced Micro Devices, Inc. All rights reserved.
33
//
44
// Permission is hereby granted, free of charge, to any person obtaining a copy
55
// of this software and associated documentation files(the "Software"), to deal
@@ -124,14 +124,27 @@ inline Vec3d Cross(const Vec3d& v1, const Vec3d& v2) {
124124
/// Returns a random float between 0 and 1
125125
float RandomFloat();
126126

127+
#ifdef CMP_USE_RSQ_RSQR
128+
inline float FastRCP(float v)
129+
{
130+
__m128 a = _mm_load_ss(&v);
131+
__m128 Ra0 = _mm_rcp_ps(a);
132+
__m128 result = _mm_sub_ps(_mm_add_ps(Ra0, Ra0), _mm_mul_ps(_mm_mul_ps(Ra0, a), Ra0));
133+
float x;
134+
_mm_store_ss(&x, result);
135+
return x;
136+
}
137+
#else
127138
inline float FastRCP(float v) {
128139
__m128 a = _mm_load_ss(&v);
129-
__m128 Ra0 = _mm_rcp_ps(a);
140+
__m128 val1 = _mm_set_ps(1.0f, 1.0f, 1.0f, 1.0f);
141+
__m128 Ra0 = _mm_div_ps(val1, a); //__m128 Ra0 = _mm_rcp_ps(a);
130142
__m128 result = _mm_sub_ps(_mm_add_ps(Ra0, Ra0), _mm_mul_ps(_mm_mul_ps(Ra0, a), Ra0));
131143
float x;
132144
_mm_store_ss(&x, result);
133145
return x;
134146
};
147+
#endif
135148

136149
/// Transformation of a point (w=1) by an arbitrary matrix, multiplies from right
137150
/// It is safe for pPoint to equal pPointOut

‎applications/_libs/cmp_math/jmlscalar.h‎

Lines changed: 25 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -27,23 +27,43 @@
2727
#include <stdlib.h>
2828

2929
namespace JML {
30+
3031
inline float RandomFloat() {
3132
return (float)rand() / (float) RAND_MAX;
3233
};
3334

34-
35-
inline float FastSQRT(float v) {
35+
#ifdef CMP_USE_RSQ_RSQR
36+
inline float FastSQRT(float v)
37+
{
3638
__m128 val = _mm_load1_ps(&v);
37-
val = _mm_sqrt_ss(val);
39+
val = _mm_sqrt_ss(val);
3840
return val.m128_f32[0];
3941
};
4042

41-
inline float FastRSQ(float v) {
43+
inline float FastRSQ(float v)
44+
{
4245
__m128 val = _mm_load1_ps(&v);
43-
val = _mm_rsqrt_ss(val);
46+
val = _mm_rsqrt_ss(val);
4447
float frsq = val.m128_f32[0];
48+
return (0.5f * frsq) * (3.0f - (v * frsq) * frsq);
49+
};
50+
#else
51+
inline float FastSQRT(float v) {
52+
__m128 val = _mm_set_ss(v);
53+
val = _mm_sqrt_ss(val);
54+
return (val.m128_f32[0]);
55+
56+
};
57+
58+
inline float FastRSQ(float v) {
59+
__m128 val = _mm_set_ss(v);
60+
__m128 val1 = _mm_set_ss(1.0f);
61+
val = _mm_sqrt_ss(val);
62+
val = _mm_div_ss(val1, val);
63+
float frsq = val.m128_f32[0];
4564
return (0.5f * frsq) * (3.0f - (v * frsq) * frsq);
4665
};
66+
#endif
4767

4868
};
4969

0 commit comments

Comments
 (0)