diff options
| -rw-r--r-- | src/glsl/Makefile.am | 1 | ||||
| -rw-r--r-- | src/glsl/ir_constant_expression.cpp | 1 | ||||
| -rw-r--r-- | src/glsl/nir/nir_constant_expressions.py | 1 | ||||
| -rw-r--r-- | src/mesa/drivers/dri/i965/brw_sampler_state.c | 1 | ||||
| -rw-r--r-- | src/mesa/main/format_utils.h | 1 | ||||
| -rw-r--r-- | src/mesa/main/imports.c | 148 | ||||
| -rw-r--r-- | src/mesa/main/imports.h | 7 | ||||
| -rw-r--r-- | src/mesa/main/mipmap.c | 1 | ||||
| -rw-r--r-- | src/mesa/main/texcompress_bptc.c | 1 | ||||
| -rw-r--r-- | src/mesa/tnl/t_draw.c | 1 | ||||
| -rw-r--r-- | src/util/Makefile.sources | 2 | ||||
| -rw-r--r-- | src/util/half_float.c | 177 | ||||
| -rw-r--r-- | src/util/half_float.h | 41 |
13 files changed, 228 insertions, 155 deletions
diff --git a/src/glsl/Makefile.am b/src/glsl/Makefile.am index 3265391..347919b 100644 --- a/src/glsl/Makefile.am +++ b/src/glsl/Makefile.am @@ -160,6 +160,7 @@ glsl_compiler_SOURCES = \ glsl_compiler_LDADD = \ libglsl.la \ $(top_builddir)/src/libglsl_util.la \ + $(top_builddir)/src/util/libmesautil.la \ $(PTHREAD_LIBS) glsl_test_SOURCES = \ diff --git a/src/glsl/ir_constant_expression.cpp b/src/glsl/ir_constant_expression.cpp index 309b6b7..67ed360 100644 --- a/src/glsl/ir_constant_expression.cpp +++ b/src/glsl/ir_constant_expression.cpp @@ -36,6 +36,7 @@ #include <math.h> #include "main/core.h" /* for MAX2, MIN2, CLAMP */ #include "util/rounding.h" /* for _mesa_roundeven */ +#include "util/half_float.h" #include "ir.h" #include "glsl_types.h" #include "program/hash_table.h" diff --git a/src/glsl/nir/nir_constant_expressions.py b/src/glsl/nir/nir_constant_expressions.py index 8fd9b10..2ba8554 100644 --- a/src/glsl/nir/nir_constant_expressions.py +++ b/src/glsl/nir/nir_constant_expressions.py @@ -29,6 +29,7 @@ template = """\ #include <math.h> #include "main/core.h" #include "util/rounding.h" /* for _mesa_roundeven */ +#include "util/half_float.h" #include "nir_constant_expressions.h" #if defined(__SUNPRO_CC) diff --git a/src/mesa/drivers/dri/i965/brw_sampler_state.c b/src/mesa/drivers/dri/i965/brw_sampler_state.c index c2db5f6..6d73444 100644 --- a/src/mesa/drivers/dri/i965/brw_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_sampler_state.c @@ -44,6 +44,7 @@ #include "main/macros.h" #include "main/samplerobj.h" +#include "util/half_float.h" /** * Emit a 3DSTATE_SAMPLER_STATE_POINTERS_{VS,HS,GS,DS,PS} packet. diff --git a/src/mesa/main/format_utils.h b/src/mesa/main/format_utils.h index 618f43d..378997b 100644 --- a/src/mesa/main/format_utils.h +++ b/src/mesa/main/format_utils.h @@ -34,6 +34,7 @@ #include "imports.h" #include "macros.h" #include "util/rounding.h" +#include "util/half_float.h" extern const mesa_array_format RGBA32_FLOAT; extern const mesa_array_format RGBA8_UBYTE; diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index 350e675..230ebbc 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -307,154 +307,6 @@ _mesa_bitcount_64(uint64_t n) } #endif - -/** - * Convert a 4-byte float to a 2-byte half float. - * - * Not all float32 values can be represented exactly as a float16 value. We - * round such intermediate float32 values to the nearest float16. When the - * float32 lies exactly between to float16 values, we round to the one with - * an even mantissa. - * - * This rounding behavior has several benefits: - * - It has no sign bias. - * - * - It reproduces the behavior of real hardware: opcode F32TO16 in Intel's - * GPU ISA. - * - * - By reproducing the behavior of the GPU (at least on Intel hardware), - * compile-time evaluation of constant packHalf2x16 GLSL expressions will - * result in the same value as if the expression were executed on the GPU. - */ -GLhalfARB -_mesa_float_to_half(float val) -{ - const fi_type fi = {val}; - const int flt_m = fi.i & 0x7fffff; - const int flt_e = (fi.i >> 23) & 0xff; - const int flt_s = (fi.i >> 31) & 0x1; - int s, e, m = 0; - GLhalfARB result; - - /* sign bit */ - s = flt_s; - - /* handle special cases */ - if ((flt_e == 0) && (flt_m == 0)) { - /* zero */ - /* m = 0; - already set */ - e = 0; - } - else if ((flt_e == 0) && (flt_m != 0)) { - /* denorm -- denorm float maps to 0 half */ - /* m = 0; - already set */ - e = 0; - } - else if ((flt_e == 0xff) && (flt_m == 0)) { - /* infinity */ - /* m = 0; - already set */ - e = 31; - } - else if ((flt_e == 0xff) && (flt_m != 0)) { - /* NaN */ - m = 1; - e = 31; - } - else { - /* regular number */ - const int new_exp = flt_e - 127; - if (new_exp < -14) { - /* The float32 lies in the range (0.0, min_normal16) and is rounded - * to a nearby float16 value. The result will be either zero, subnormal, - * or normal. - */ - e = 0; - m = _mesa_lroundevenf((1 << 24) * fabsf(fi.f)); - } - else if (new_exp > 15) { - /* map this value to infinity */ - /* m = 0; - already set */ - e = 31; - } - else { - /* The float32 lies in the range - * [min_normal16, max_normal16 + max_step16) - * and is rounded to a nearby float16 value. The result will be - * either normal or infinite. - */ - e = new_exp + 15; - m = _mesa_lroundevenf(flt_m / (float) (1 << 13)); - } - } - - assert(0 <= m && m <= 1024); - if (m == 1024) { - /* The float32 was rounded upwards into the range of the next exponent, - * so bump the exponent. This correctly handles the case where f32 - * should be rounded up to float16 infinity. - */ - ++e; - m = 0; - } - - result = (s << 15) | (e << 10) | m; - return result; -} - - -/** - * Convert a 2-byte half float to a 4-byte float. - * Based on code from: - * http://www.opengl.org/discussion_boards/ubb/Forum3/HTML/008786.html - */ -float -_mesa_half_to_float(GLhalfARB val) -{ - /* XXX could also use a 64K-entry lookup table */ - const int m = val & 0x3ff; - const int e = (val >> 10) & 0x1f; - const int s = (val >> 15) & 0x1; - int flt_m, flt_e, flt_s; - fi_type fi; - float result; - - /* sign bit */ - flt_s = s; - - /* handle special cases */ - if ((e == 0) && (m == 0)) { - /* zero */ - flt_m = 0; - flt_e = 0; - } - else if ((e == 0) && (m != 0)) { - /* denorm -- denorm half will fit in non-denorm single */ - const float half_denorm = 1.0f / 16384.0f; /* 2^-14 */ - float mantissa = ((float) (m)) / 1024.0f; - float sign = s ? -1.0f : 1.0f; - return sign * mantissa * half_denorm; - } - else if ((e == 31) && (m == 0)) { - /* infinity */ - flt_e = 0xff; - flt_m = 0; - } - else if ((e == 31) && (m != 0)) { - /* NaN */ - flt_e = 0xff; - flt_m = 1; - } - else { - /* regular */ - flt_e = e + 112; - flt_m = m << 13; - } - - fi.i = (flt_s << 31) | (flt_e << 23) | flt_m; - result = fi.f; - return result; -} - /*@}*/ diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h index 9024758..042147f 100644 --- a/src/mesa/main/imports.h +++ b/src/mesa/main/imports.h @@ -396,13 +396,6 @@ _mesa_flsll(uint64_t n) #endif } - -extern GLhalfARB -_mesa_float_to_half(float f); - -extern float -_mesa_half_to_float(GLhalfARB h); - static inline bool _mesa_half_is_negative(GLhalfARB h) { diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index ab16c28..5046995 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -37,6 +37,7 @@ #include "texstore.h" #include "image.h" #include "macros.h" +#include "util/half_float.h" #include "../../gallium/auxiliary/util/u_format_rgb9e5.h" #include "../../gallium/auxiliary/util/u_format_r11g11b10f.h" diff --git a/src/mesa/main/texcompress_bptc.c b/src/mesa/main/texcompress_bptc.c index f0f6553..26e5915 100644 --- a/src/mesa/main/texcompress_bptc.c +++ b/src/mesa/main/texcompress_bptc.c @@ -30,6 +30,7 @@ #include "texcompress.h" #include "texcompress_bptc.h" #include "util/format_srgb.h" +#include "util/half_float.h" #include "texstore.h" #include "macros.h" #include "image.h" diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c index c130ab3..6f29abb 100644 --- a/src/mesa/tnl/t_draw.c +++ b/src/mesa/tnl/t_draw.c @@ -35,6 +35,7 @@ #include "main/mtypes.h" #include "main/macros.h" #include "main/enums.h" +#include "util/half_float.h" #include "t_context.h" #include "tnl.h" diff --git a/src/util/Makefile.sources b/src/util/Makefile.sources index e45431d..a871146 100644 --- a/src/util/Makefile.sources +++ b/src/util/Makefile.sources @@ -3,6 +3,8 @@ MESA_UTIL_FILES := \ debug.c \ debug.h \ format_srgb.h \ + half_float.c \ + half_float.h \ hash_table.c \ hash_table.h \ list.h \ diff --git a/src/util/half_float.c b/src/util/half_float.c new file mode 100644 index 0000000..4df64c2 --- /dev/null +++ b/src/util/half_float.c @@ -0,0 +1,177 @@ +/* + * Mesa 3-D graphics library + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <math.h> +#include <assert.h> +#include "half_float.h" +#include "rounding.h" + +typedef union { float f; int32_t i; uint32_t u; } fi_type; + +/** + * Convert a 4-byte float to a 2-byte half float. + * + * Not all float32 values can be represented exactly as a float16 value. We + * round such intermediate float32 values to the nearest float16. When the + * float32 lies exactly between to float16 values, we round to the one with + * an even mantissa. + * + * This rounding behavior has several benefits: + * - It has no sign bias. + * + * - It reproduces the behavior of real hardware: opcode F32TO16 in Intel's + * GPU ISA. + * + * - By reproducing the behavior of the GPU (at least on Intel hardware), + * compile-time evaluation of constant packHalf2x16 GLSL expressions will + * result in the same value as if the expression were executed on the GPU. + */ +uint16_t +_mesa_float_to_half(float val) +{ + const fi_type fi = {val}; + const int flt_m = fi.i & 0x7fffff; + const int flt_e = (fi.i >> 23) & 0xff; + const int flt_s = (fi.i >> 31) & 0x1; + int s, e, m = 0; + uint16_t result; + + /* sign bit */ + s = flt_s; + + /* handle special cases */ + if ((flt_e == 0) && (flt_m == 0)) { + /* zero */ + /* m = 0; - already set */ + e = 0; + } + else if ((flt_e == 0) && (flt_m != 0)) { + /* denorm -- denorm float maps to 0 half */ + /* m = 0; - already set */ + e = 0; + } + else if ((flt_e == 0xff) && (flt_m == 0)) { + /* infinity */ + /* m = 0; - already set */ + e = 31; + } + else if ((flt_e == 0xff) && (flt_m != 0)) { + /* NaN */ + m = 1; + e = 31; + } + else { + /* regular number */ + const int new_exp = flt_e - 127; + if (new_exp < -14) { + /* The float32 lies in the range (0.0, min_normal16) and is rounded + * to a nearby float16 value. The result will be either zero, subnormal, + * or normal. + */ + e = 0; + m = _mesa_lroundevenf((1 << 24) * fabsf(fi.f)); + } + else if (new_exp > 15) { + /* map this value to infinity */ + /* m = 0; - already set */ + e = 31; + } + else { + /* The float32 lies in the range + * [min_normal16, max_normal16 + max_step16) + * and is rounded to a nearby float16 value. The result will be + * either normal or infinite. + */ + e = new_exp + 15; + m = _mesa_lroundevenf(flt_m / (float) (1 << 13)); + } + } + + assert(0 <= m && m <= 1024); + if (m == 1024) { + /* The float32 was rounded upwards into the range of the next exponent, + * so bump the exponent. This correctly handles the case where f32 + * should be rounded up to float16 infinity. + */ + ++e; + m = 0; + } + + result = (s << 15) | (e << 10) | m; + return result; +} + + +/** + * Convert a 2-byte half float to a 4-byte float. + * Based on code from: + * http://www.opengl.org/discussion_boards/ubb/Forum3/HTML/008786.html + */ +float +_mesa_half_to_float(uint16_t val) +{ + /* XXX could also use a 64K-entry lookup table */ + const int m = val & 0x3ff; + const int e = (val >> 10) & 0x1f; + const int s = (val >> 15) & 0x1; + int flt_m, flt_e, flt_s; + fi_type fi; + float result; + + /* sign bit */ + flt_s = s; + + /* handle special cases */ + if ((e == 0) && (m == 0)) { + /* zero */ + flt_m = 0; + flt_e = 0; + } + else if ((e == 0) && (m != 0)) { + /* denorm -- denorm half will fit in non-denorm single */ + const float half_denorm = 1.0f / 16384.0f; /* 2^-14 */ + float mantissa = ((float) (m)) / 1024.0f; + float sign = s ? -1.0f : 1.0f; + return sign * mantissa * half_denorm; + } + else if ((e == 31) && (m == 0)) { + /* infinity */ + flt_e = 0xff; + flt_m = 0; + } + else if ((e == 31) && (m != 0)) { + /* NaN */ + flt_e = 0xff; + flt_m = 1; + } + else { + /* regular */ + flt_e = e + 112; + flt_m = m << 13; + } + + fi.i = (flt_s << 31) | (flt_e << 23) | flt_m; + result = fi.f; + return result; +} diff --git a/src/util/half_float.h b/src/util/half_float.h new file mode 100644 index 0000000..64f2042 --- /dev/null +++ b/src/util/half_float.h @@ -0,0 +1,41 @@ +/* + * Mesa 3-D graphics library + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef _HALF_FLOAT_H_ +#define _HALF_FLOAT_H_ + +#include <stdint.h> + +#ifdef __cplusplus +extern "C" { +#endif + +uint16_t _mesa_float_to_half(float val); +float _mesa_half_to_float(uint16_t val); + +#ifdef __cplusplus +} /* extern C */ +#endif + +#endif /* _HALF_FLOAT_H_ */ |