git subrepo pull (merge) lib/astc-encoder

subrepo:
  subdir:   "lib/astc-encoder"
  merged:   "dd816397"
upstream:
  origin:   "https://github.com/ARM-software/astc-encoder.git"
  branch:   "main"
  commit:   "97a5ff37"
git-subrepo:
  version:  "0.4.3"
  origin:   "https://github.com/MarkCallow/git-subrepo.git"
  commit:   "c1f1132"

lib/astc-encoder/.gitrepo

@@ -6,7 +6,7 @@
 [subrepo]
 	remote = https://github.com/ARM-software/astc-encoder.git
 	branch = main
-	commit = 2003150070c13536908714bd6eb031e678d7a2ca
+	commit = 97a5ff37fd7b885530c9c7b16a3397d2808eea3e
 	parent = 15369663a43d72972dfe0a7e3597d3ef6d90b6b9
 	method = merge
 	cmdver = 0.4.3

lib/astc-encoder/CMakeLists.txt

@@ -24,7 +24,7 @@ if(MSVC)
     add_compile_options("/wd4324") # Disable structure was padded due to alignment specifier
 endif()
 
-project(astcencoder VERSION 3.0.0)
+project(astcencoder VERSION 3.1.0)
 
 set(CMAKE_CXX_STANDARD 14)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
@@ -39,6 +39,7 @@ option(ISA_SSE41 "Enable builds for SSE4.1 SIMD")
 option(ISA_SSE2 "Enable builds for SSE2 SIMD")
 option(ISA_NEON "Enable builds for NEON SIMD")
 option(ISA_NONE "Enable builds for no SIMD")
+option(ISA_NATIVE "Enable builds for native SIMD")
 option(DECOMPRESSOR "Enable builds for decompression only")
 option(DIAGNOSTICS "Enable builds for diagnostic trace")
 option(UNITTEST "Enable builds for unit tests")
@@ -175,8 +176,9 @@ else()
     endif()
 endif()
 
-if(("${ARM64_ISA_COUNT}" EQUAL 0) AND ("${X64_ISA_COUNT}" EQUAL 0))
-    set(ISA_NONE ON)
+# If nothing more specific is set then fall back on the compiler's defaults
+if(("${ARM64_ISA_COUNT}" EQUAL 0) AND ("${X64_ISA_COUNT}" EQUAL 0) AND (NOT "${ISA_NONE}"))
+    set(ISA_NATIVE ON)
 endif()
 
 function(printopt optName optVal)
@@ -187,17 +189,18 @@ function(printopt optName optVal)
     endif()
 endfunction()
 
-printopt("AVX2 backend  " ${ISA_AVX2})
-printopt("SSE4.1 backend" ${ISA_SSE41})
-printopt("SSE2 backend  " ${ISA_SSE2})
-printopt("NEON backend  " ${ISA_NEON})
-printopt("NONE backend  " ${ISA_NONE})
+printopt("AVX2 backend   " ${ISA_AVX2})
+printopt("SSE4.1 backend " ${ISA_SSE41})
+printopt("SSE2 backend   " ${ISA_SSE2})
+printopt("NEON backend   " ${ISA_NEON})
+printopt("NONE backend   " ${ISA_NONE})
+printopt("NATIVE backend " ${ISA_NATIVE})
 if("${MACOS_BUILD}")
-    printopt("Universal bin " ${UNIVERSAL_BUILD})
+    printopt("Universal bin  " ${UNIVERSAL_BUILD})
 endif()
-printopt("Decompressor  " ${DECOMPRESSOR})
-printopt("Diagnostics   " ${DIAGNOSTICS})
-printopt("Unit tests    " ${UNITTEST})
+printopt("Decompressor   " ${DECOMPRESSOR})
+printopt("Diagnostics    " ${DIAGNOSTICS})
+printopt("Unit tests     " ${UNITTEST})
 
 # Subcomponents
 add_subdirectory(Source)

lib/astc-encoder/Docs/Building.md

@@ -175,3 +175,11 @@ make package -j16
 
 Windows packages will use the `.zip` format, other packages will use the
 `.tar.gz` format.
+
+## Integrating as a library into another project
+
+The core codec of astcenc is built as a library, and so can be easily
+integrated into other projects using CMake. An example of the CMake integration
+and the codec API usage can be found in the `./Utils/Example` directory in the
+repository. See the [Example Readme](../Utils/Example/README.md) for more
+details.

lib/astc-encoder/Docs/ChangeLog-3x.md

@@ -6,6 +6,65 @@ release of the 3.x series.
 All performance data on this page is measured on an Intel Core i5-9600K
 clocked at 4.2 GHz, running astcenc using AVX2 and 6 threads.
 
+<!-- ---------------------------------------------------------------------- -->
+## 3.1
+
+**Status:** July 2021
+
+The 3.1 release is the second release in the 3.x series. This release gives
+another performance boost, typically between 5 and 20% faster than the 3.0
+release, as well as further incremental improvements to image quality. A number
+of build system improvements make astcenc easier and faster to integrate into
+other projects as a library, including support for building unversal binaries
+on macOS. Full change list is shown below.
+
+Reminder for users of the library interface - the API is not designed to be
+binary compatible across versions, and this release is not compatible with
+earlier releases. Please update and rebuild your client-side code using the
+updated `astcenc.h` header.
+
+* **General:**
+  * **Feature:** RGB color data now supports `-perceptual` operation. The
+    current implementation is simple, weighting color channel errors by their
+    contribution to perceived luminance. This mimics the behavior of the human
+    visual system, which is most sensitive to green, then red, then blue.
+  * **Feature:** Codec supports a new low weight search mode, which is a
+    simpler weight assignment for encodings with a low number of weights in the
+    weight grid. The weight threshold can be overriden using the new
+    `-lowweightmodelimit` commadn line option.
+  * **Feature:** All platform builds now support building a native binary.
+    Native binaries automatically select the SIMD level based on the default
+    configuration of the compiler in use. Native binaries built on one machine
+    may use different SIMD options than native binaries build on another.
+  * **Feature:** macOS platform builds now support building universal binaries
+    containing both `x86_64` and `arm64` target support.
+  * **Feature:** Building the command line can be disabled when using as a
+    library in another project. Set `-DCLI=OFF` during the CMake configure
+    step.
+  * **Feature:** A standalone minimal example of the core codec API usage has
+    been added in the `./Utils/Example/` directory.
+* **Core API:**
+  * **Feature:** Config flag `ASTCENC_FLG_USE_PERCEPTUAL` works for color data.
+  * **Feature:** Config option `tune_low_weight_count_limit` added.
+  * **Feature:** New heuristic added which prunes dual weight plane searches if
+    they are unlikely to help. This heuristic is not user controllable.
+  * **Feature:** Image quality has been improved. In general we see significant
+    improvements (up to 0.2dB) for high bitrate encodings (4x4, 5x4), and a
+    smaller improvement (up to 0.1dB) for lower bitrate encodings.
+  * **Bug fix:** Arm "none" SIMD builds could be invariant with other builds.
+    This fix has also been back-ported to the 2.x LTS branch.
+
+### Performance:
+
+Key for charts:
+
+* Color = block size (see legend).
+* Letter = image format (N = normal map, G = greyscale, L = LDR, H = HDR).
+
+**Relative performance vs 3.0 release:**
+
+![Relative scores 3.1 vs 3.0](./ChangeLogImg/relative-3.0-to-3.1.png)
+
 <!-- ---------------------------------------------------------------------- -->
 ## 3.0
 

lib/astc-encoder/Source/CMakeLists.txt

@@ -37,8 +37,8 @@ if(${UNIVERSAL_BUILD})
     endif()
     include(cmake_core.cmake)
 else()
-    set(ARTEFACTS none neon avx2 sse4.1 sse2)
-    set(CONFIGS ${ISA_NONE} ${ISA_NEON} ${ISA_AVX2} ${ISA_SSE41} ${ISA_SSE2})
+    set(ARTEFACTS native none neon avx2 sse4.1 sse2)
+    set(CONFIGS ${ISA_NATIVE} ${ISA_NONE} ${ISA_NEON} ${ISA_AVX2} ${ISA_SSE41} ${ISA_SSE2})
     list(LENGTH ARTEFACTS ARTEFACTS_LEN)
     math(EXPR ARTEFACTS_LEN "${ARTEFACTS_LEN} - 1")
 

lib/astc-encoder/Source/Fuzzers/build.sh

@@ -20,6 +20,11 @@
 
 # This script is invoked by oss-fuzz from <root>/Source/
 
+# Generate a dummy verison header (normally built by CMake variable expansion)
+echo "#pragma once" > astcenccli_version.h
+echo "#define VERSION_STRING \"0.0.0\"" >> astcenccli_version.h
+echo "#define YEAR_STRING \"2021\"" >> astcenccli_version.h
+
 # Build the core project for fuzz tests to link against
 for source in ./*.cpp; do
   BASE="${source##*/}"

lib/astc-encoder/Source/UnitTest/CMakeLists.txt

@@ -25,8 +25,8 @@ if(${UNIVERSAL_BUILD})
     endif()
     include(cmake_core.cmake)
 else()
-    set(ARTEFACTS none neon avx2 sse4.1 sse2)
-    set(CONFIGS ${ISA_NONE} ${ISA_NEON} ${ISA_AVX2} ${ISA_SSE41} ${ISA_SSE2})
+    set(ARTEFACTS native none neon avx2 sse4.1 sse2)
+    set(CONFIGS ${ISA_NATIVE} ${ISA_NONE} ${ISA_NEON} ${ISA_AVX2} ${ISA_SSE41} ${ISA_SSE2})
     list(LENGTH ARTEFACTS ARTEFACTS_LEN)
     math(EXPR ARTEFACTS_LEN "${ARTEFACTS_LEN} - 1")
 

lib/astc-encoder/Source/UnitTest/test_simd.cpp

@@ -1082,11 +1082,11 @@ TEST(vfloat4, normalize_safe)
 /** @brief Test vfloat4 float_to_int. */
 TEST(vfloat4, float_to_int)
 {
-	vfloat4 a(1.1f, 1.5f, 1.6f, 4.0f);
+	vfloat4 a(1.1f, 1.5f, -1.6f, 4.0f);
 	vint4 r = float_to_int(a);
 	EXPECT_EQ(r.lane<0>(), 1);
 	EXPECT_EQ(r.lane<1>(), 1);
-	EXPECT_EQ(r.lane<2>(), 1);
+	EXPECT_EQ(r.lane<2>(), -1);
 	EXPECT_EQ(r.lane<3>(), 4);
 }
 

lib/astc-encoder/Source/astcenc_entry.cpp

@@ -672,8 +672,7 @@ astcenc_error astcenc_config_init(
 			config.v_a_stdev = 50.0f;
 		}
 	}
-
-	if (flags & ASTCENC_FLG_MAP_MASK)
+	else if (flags & ASTCENC_FLG_MAP_MASK)
 	{
 		config.v_rgba_radius = 3;
 		config.v_rgba_mean_stdev_mix = 0.03f;
@@ -682,13 +681,30 @@ astcenc_error astcenc_config_init(
 		config.v_a_mean = 0.0f;
 		config.v_a_stdev = 25.0f;
 	}
-
-	if (flags & ASTCENC_FLG_MAP_RGBM)
+	else if (flags & ASTCENC_FLG_MAP_RGBM)
 	{
 		config.rgbm_m_scale = 5.0f;
 		config.cw_a_weight = 2.0f * config.rgbm_m_scale;
 	}
-
+	else // (This is color data)
+	{
+		// This is a very basic perceptual metric for RGB color data, which weights error
+		// significance by the perceptual luminance contribution of each color channel. For
+		// luminance the usual weights to compute luminance from a linear RGB value are as
+		// follows:
+		//
+		//     l = r * 0.3 + g * 0.59 + b * 0.11
+		//
+		// ... but we scale these up to keep a better balance between color and alpha. Note
+		// that if the content is using alpha we'd recommend using the -a option to weight
+		// the color conribution by the alpha transparency.
+		if (flags & ASTCENC_FLG_USE_PERCEPTUAL)
+		{
+			config.cw_r_weight = 0.30f * 2.25f;
+			config.cw_g_weight = 0.59f * 2.25f;
+			config.cw_b_weight = 0.11f * 2.25f;
+		}
+	}
 	config.flags = flags;
 
 	return ASTCENC_SUCCESS;

lib/astc-encoder/Source/astcenc_vecmathlib_none_4.h

@@ -936,12 +936,10 @@ ASTCENC_SIMD_INLINE void storea(vfloat4 a, float* ptr)
  */
 ASTCENC_SIMD_INLINE vint4 float_to_int(vfloat4 a)
 {
-	// Casting to unsigned buys us an extra bit of precision in cases where
-	// we can use the integer as nasty bit hacks.
-	return vint4((unsigned int)a.m[0],
-	             (unsigned int)a.m[1],
-	             (unsigned int)a.m[2],
-	             (unsigned int)a.m[3]);
+	return vint4((int)a.m[0],
+	             (int)a.m[1],
+	             (int)a.m[2],
+	             (int)a.m[3]);
 }
 
 /**f

lib/astc-encoder/Source/astcenccli_toplevel_help.cpp

@@ -24,7 +24,7 @@
 
 /** @brief The version header. */
 static const char *astcenc_copyright_string =
-R"(astcenc v%s, %u-bit %s%s
+R"(astcenc v%s, %u-bit %s%s%s
 Copyright 2011-%s Arm Limited, all rights reserved
 )";
 
@@ -192,7 +192,7 @@ COMPRESSION
            The codec should optimize perceptual error, instead of direct
            RMS error. This aims to improves perceived image quality, but
            typically lowers the measured PSNR score. Perceptual methods are
-           currently only available for normal maps.
+           currently only available for normal maps and RGB color data.
 
        -array <size>
            Loads an array of <size> 2D image slices to use as a 3D image.
@@ -595,9 +595,15 @@ void astcenc_print_header()
 	const char* pcnttype = "";
 #endif
 
+#if (ASTCENC_F16C == 1)
+	const char* f16ctype = "+f16c";
+#else
+	const char* f16ctype = "";
+#endif
+
 	unsigned int bits = (int)(sizeof(void*) * 8);
 	printf(astcenc_copyright_string,
-	       VERSION_STRING, bits, simdtype, pcnttype, YEAR_STRING);
+	       VERSION_STRING, bits, simdtype, pcnttype, f16ctype, YEAR_STRING);
 }
 
 /* See header for documentation. */

lib/astc-encoder/Source/cmake_core.cmake

@@ -54,7 +54,8 @@ add_library(${ASTC_TARGET}-static
 
 target_include_directories(${ASTC_TARGET}-static
     PUBLIC
-        ${CMAKE_CURRENT_SOURCE_DIR})
+        $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
+        $<INSTALL_INTERFACE:.>)
 
 if(${CLI})
     add_executable(${ASTC_TARGET}