bf0c398eeec8424b98dd89f766438740786f1f2b
1 /**************************************************************************
3 * Copyright 2008 VMware, Inc.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
30 #include "pipe/p_config.h"
31 #include "util/u_math.h"
32 #include "x86/common_x86_features.h"
34 #if defined(PIPE_ARCH_SSE)
35 #include <xmmintrin.h>
36 /* This is defined in pmmintrin.h, but it can only be included when -msse3 is
37 * used, so just define it here to avoid further. */
38 #define _MM_DENORMALS_ZERO_MASK 0x0040
42 /** 2^x, for x in [-1.0, 1.0) */
43 float pow2_table
[POW2_TABLE_SIZE
];
50 for (i
= 0; i
< POW2_TABLE_SIZE
; i
++)
51 pow2_table
[i
] = exp2f((i
- POW2_TABLE_OFFSET
) / POW2_TABLE_SCALE
);
55 /** log2(x), for x in [1.0, 2.0) */
56 float log2_table
[LOG2_TABLE_SIZE
];
63 for (i
= 0; i
< LOG2_TABLE_SIZE
; i
++)
64 log2_table
[i
] = (float) log2(1.0 + i
* (1.0 / LOG2_TABLE_SCALE
));
69 * One time init for math utilities.
74 static boolean initialized
= FALSE
;
83 * Fetches the contents of the fpstate (mxcsr on x86) register.
85 * On platforms without support for it just returns 0.
88 util_fpstate_get(void)
92 #if defined(PIPE_ARCH_SSE)
101 /* TODO: this was copied from u_cpu_detection. It's another case of duplication
102 * between gallium and core mesa, and it would be nice to get rid of that
103 * duplication as well.
105 #if defined(PIPE_ARCH_X86)
106 PIPE_ALIGN_STACK
static inline bool sse2_has_daz(void)
111 uint32_t pad2
[128-8];
112 } PIPE_ALIGN_VAR(16) fxarea
;
114 fxarea
.mxcsr_mask
= 0;
115 #if defined(PIPE_CC_GCC)
116 __asm
__volatile ("fxsave %0" : "+m" (fxarea
));
117 #elif defined(PIPE_CC_MSVC) || defined(PIPE_CC_ICL)
120 fxarea
.mxcsr_mask
= 0;
122 return !!(fxarea
.mxcsr_mask
& (1 << 6));
127 * Make sure that the fp treats the denormalized floating
128 * point numbers as zero.
130 * This is the behavior required by D3D10. OpenGL doesn't care.
133 util_fpstate_set_denorms_to_zero(unsigned current_mxcsr
)
135 #if defined(PIPE_ARCH_SSE)
137 /* Enable flush to zero mode */
138 current_mxcsr
|= _MM_FLUSH_ZERO_MASK
;
139 /* x86_64 cpus always have daz, as do cpus with sse3 in fact, there's
140 * basically only a handful of very early pentium 4's that have sse2 but
143 # if !defined(PIPE_ARCH_x86_64) && !defined(PIPE_ARCH_SSSE3)
144 if (sse2_has_daz()) {
146 /* Enable denormals are zero mode */
147 current_mxcsr
|= _MM_DENORMALS_ZERO_MASK
;
148 # if !defined(PIPE_ARCH_x86_64) && !defined(PIPE_ARCH_SSSE3)
151 util_fpstate_set(current_mxcsr
);
154 return current_mxcsr
;
158 * Set the state of the fpstate (mxcsr on x86) register.
160 * On platforms without support for it's a noop.
163 util_fpstate_set(unsigned mxcsr
)
165 #if defined(PIPE_ARCH_SSE)