#ifndef GMX_SIMD_IMPL_INTEL_MIC_H
#define GMX_SIMD_IMPL_INTEL_MIC_H
+#include "config.h"
+
#include <math.h>
+
#include <immintrin.h>
/* Intel Xeon Phi, or
#define gmx_simd_xor_f(a, b) _mm512_castsi512_ps(_mm512_xor_epi32(_mm512_castps_si512(a), _mm512_castps_si512(b)))
#define gmx_simd_rsqrt_f _mm512_rsqrt23_ps
#define gmx_simd_rcp_f _mm512_rcp23_ps
-#define gmx_simd_fabs_f(x) gmx_simd_andnot_f(_mm512_set1_ps(-0.0), x)
+#define gmx_simd_fabs_f(x) gmx_simd_andnot_f(_mm512_set1_ps(GMX_FLOAT_NEGZERO), x)
#define gmx_simd_fneg_f(x) _mm512_addn_ps(x, _mm512_setzero_ps())
#define gmx_simd_max_f _mm512_gmax_ps
#define gmx_simd_min_f _mm512_gmin_ps
#define gmx_simd_xor_d(a, b) _mm512_castsi512_pd(_mm512_xor_epi32(_mm512_castpd_si512(a), _mm512_castpd_si512(b)))
#define gmx_simd_rsqrt_d(x) _mm512_cvtpslo_pd(_mm512_rsqrt23_ps(_mm512_cvtpd_pslo(x)))
#define gmx_simd_rcp_d(x) _mm512_cvtpslo_pd(_mm512_rcp23_ps(_mm512_cvtpd_pslo(x)))
-#define gmx_simd_fabs_d(x) gmx_simd_andnot_d(_mm512_set1_pd(-0.0), x)
+#define gmx_simd_fabs_d(x) gmx_simd_andnot_d(_mm512_set1_pd(GMX_DOUBLE_NEGZERO), x)
#define gmx_simd_fneg_d(x) _mm512_addn_pd(x, _mm512_setzero_pd())
#define gmx_simd_max_d _mm512_gmax_pd
#define gmx_simd_min_d _mm512_gmin_pd
#define gmx_simd4_or_f(a, b) _mm512_castsi512_ps(_mm512_mask_or_epi32(_mm512_undefined_epi32(), gmx_simd4_mask, _mm512_castps_si512(a), _mm512_castps_si512(b)))
#define gmx_simd4_xor_f(a, b) _mm512_castsi512_ps(_mm512_mask_xor_epi32(_mm512_undefined_epi32(), gmx_simd4_mask, _mm512_castps_si512(a), _mm512_castps_si512(b)))
#define gmx_simd4_rsqrt_f(a) _mm512_mask_rsqrt23_ps(_mm512_undefined_ps(), gmx_simd4_mask, a)
-#define gmx_simd4_fabs_f(x) gmx_simd4_andnot_f(_mm512_set1_ps(-0.0), x)
+#define gmx_simd4_fabs_f(x) gmx_simd4_andnot_f(_mm512_set1_ps(GMX_FLOAT_NEGZERO), x)
#define gmx_simd4_fneg_f(x) _mm512_mask_addn_ps(_mm512_undefined_ps(), gmx_simd4_mask, x, _mm512_setzero_ps())
#define gmx_simd4_max_f(a, b) _mm512_mask_gmax_ps(_mm512_undefined_ps(), gmx_simd4_mask, a, b)
#define gmx_simd4_min_f(a, b) _mm512_mask_gmin_ps(_mm512_undefined_ps(), gmx_simd4_mask, a, b)
#define gmx_simd4_or_d(a, b) _mm512_castsi512_pd(_mm512_mask_or_epi32(_mm512_undefined_epi32(), mask_loh, _mm512_castpd_si512(a), _mm512_castpd_si512(b)))
#define gmx_simd4_xor_d(a, b) _mm512_castsi512_pd(_mm512_mask_xor_epi32(_mm512_undefined_epi32(), mask_loh, _mm512_castpd_si512(a), _mm512_castpd_si512(b)))
#define gmx_simd4_rsqrt_d(a) _mm512_mask_cvtpslo_pd(_mm512_undefined_pd(), gmx_simd4_mask, _mm512_mask_rsqrt23_ps(_mm512_undefined_ps(), gmx_simd4_mask, _mm512_mask_cvtpd_pslo(_mm512_undefined_ps(), gmx_simd4_mask, x)))
-#define gmx_simd4_fabs_d(x) gmx_simd4_andnot_d(_mm512_set1_pd(-0.0), x)
+#define gmx_simd4_fabs_d(x) gmx_simd4_andnot_d(_mm512_set1_pd(GMX_DOUBLE_NEGZERO), x)
#define gmx_simd4_fneg_d(x) _mm512_mask_addn_pd(_mm512_undefined_pd(), gmx_simd4_mask, x, _mm512_setzero_pd())
#define gmx_simd4_max_d(a, b) _mm512_mask_gmax_pd(_mm512_undefined_pd(), gmx_simd4_mask, a, b)
#define gmx_simd4_min_d(a, b) _mm512_mask_gmin_pd(_mm512_undefined_pd(), gmx_simd4_mask, a, b)
#define mask_hih _mm512_int2mask(0xFF00)
/* load store float */
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_loadu_f_mic(const float * m)
{
return _mm512_loadunpackhi_ps(_mm512_loadunpacklo_ps(_mm512_undefined_ps(), m), m+16);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd_storeu_f_mic(float * m, __m512 s)
{
_mm512_packstorelo_ps(m, s);
}
/* load store fint32 */
-static gmx_inline __m512i
+static gmx_inline __m512i gmx_simdcall
gmx_simd_loadu_fi_mic(const gmx_int32_t * m)
{
return _mm512_loadunpackhi_epi32(_mm512_loadunpacklo_epi32(_mm512_undefined_epi32(), m), m+16);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd_storeu_fi_mic(gmx_int32_t * m, __m512i s)
{
_mm512_packstorelo_epi32(m, s);
}
/* load store double */
-static gmx_inline __m512d
+static gmx_inline __m512d gmx_simdcall
gmx_simd_loadu_d_mic(const double * m)
{
return _mm512_loadunpackhi_pd(_mm512_loadunpacklo_pd(_mm512_undefined_pd(), m), m+8);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd_storeu_d_mic(double * m, __m512d s)
{
_mm512_packstorelo_pd(m, s);
}
/* load store dint32 */
-static gmx_inline __m512i
+static gmx_inline __m512i gmx_simdcall
gmx_simd_loadu_di_mic(const gmx_int32_t * m)
{
return _mm512_mask_loadunpackhi_epi32(_mm512_mask_loadunpacklo_epi32(_mm512_undefined_epi32(), mask_loh, m), mask_loh, m+16);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd_storeu_di_mic(gmx_int32_t * m, __m512i s)
{
_mm512_mask_packstorelo_epi32(m, mask_loh, s);
}
/* load store simd4 */
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd4_loadu_f_mic(const float * m)
{
return _mm512_mask_loadunpackhi_ps(_mm512_mask_loadunpacklo_ps(_mm512_undefined_ps(), gmx_simd4_mask, m), gmx_simd4_mask, m+16);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd4_storeu_f_mic(float * m, __m512 s)
{
_mm512_mask_packstorelo_ps(m, gmx_simd4_mask, s);
_mm512_mask_packstorehi_ps(m+16, gmx_simd4_mask, s);
}
-static gmx_inline __m512d
+static gmx_inline __m512d gmx_simdcall
gmx_simd4_loadu_d_mic(const double * m)
{
return _mm512_mask_loadunpackhi_pd(_mm512_mask_loadunpacklo_pd(_mm512_undefined_pd(), gmx_simd4_mask, m), gmx_simd4_mask, m+8);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd4_storeu_d_mic(double * m, __m512d s)
{
_mm512_mask_packstorelo_pd(m, gmx_simd4_mask, s);
}
/* extract */
-static gmx_inline gmx_int32_t
+static gmx_inline gmx_int32_t gmx_simdcall
gmx_simd_extract_fi_mic(gmx_simd_fint32_t a, int index)
{
int r;
return r;
}
-static gmx_inline gmx_int32_t
+static gmx_inline gmx_int32_t gmx_simdcall
gmx_simd_extract_di_mic(gmx_simd_dint32_t a, int index)
{
int r;
/* This is likely faster than the built in scale operation (lat 8, t-put 3)
* since we only work on the integer part and use shifts. TODO: check. given that scale also only does integer
*/
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_set_exponent_f_mic(__m512 a)
{
__m512i iexp = gmx_simd_cvt_f2i(a);
*/
}
-static gmx_inline __m512d
+static gmx_inline __m512d gmx_simdcall
gmx_simd_set_exponent_d_mic(__m512d a)
{
const __m512i expbias = _mm512_set1_epi32(1023);
return _mm512_castsi512_pd(iexp);
}
-static gmx_inline void
+static gmx_inline void gmx_simdcall
gmx_simd_cvt_f2dd_mic(__m512 f, __m512d * d0, __m512d * d1)
{
__m512i i1 = _mm512_permute4f128_epi32(_mm512_castps_si512(f), _MM_PERM_CDCD);
*d1 = _mm512_cvtpslo_pd(_mm512_castsi512_ps(i1));
}
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_cvt_dd2f_mic(__m512d d0, __m512d d1)
{
__m512 f0 = _mm512_cvtpd_pslo(d0);
return _mm512_mask_permute4f128_ps(f0, mask_hih, f1, PERM_LOW2HIGH);
}
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_exp2_f_mic(__m512 x)
{
return _mm512_exp223_ps(_mm512_cvtfxpnt_round_adjustps_epi32(x, _MM_ROUND_MODE_NEAREST, _MM_EXPADJ_24));
}
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_exp_f_mic(__m512 x)
{
- /* only 59ulp accuracy so we need to do extra an iteration
- Using: http://yacas.sourceforge.net/Algochapter5.html 5.4 Method 3 */
- __m512 r = gmx_simd_exp2_f(_mm512_mul_ps(x, _mm512_set1_ps(1.44269504088896341)));
- __mmask16 m = _mm512_cmpneq_ps_mask(r, _mm512_setzero_ps());
- __m512 t = _mm512_mask_fnmadd_ps(_mm512_mask_log2ae23_ps(_mm512_undefined_ps(), m, r), m, _mm512_set1_ps(0.693147180559945286226764), x);
- return _mm512_mask_fmadd_ps(r, m, t, r);
+ const gmx_simd_float_t argscale = gmx_simd_set1_f(1.44269504088896341f);
+ const gmx_simd_float_t invargscale = gmx_simd_set1_f(-0.69314718055994528623f);
+ __m512 xscaled = _mm512_mul_ps(x, argscale);
+ __m512 r = gmx_simd_exp2_f_mic(xscaled);
+
+ /* gmx_simd_exp2_f_mic() provides 23 bits of accuracy, but we ruin some of that
+ * with the argument scaling due to single-precision rounding, where the
+ * rounding error is amplified exponentially. To correct this, we find the
+ * difference between the scaled argument and the true one (extended precision
+ * arithmetics does not appear to be necessary to fulfill our accuracy requirements)
+ * and then multiply by the exponent of this correction since exp(a+b)=exp(a)*exp(b).
+ * Note that this only adds two instructions (and maybe some constant loads).
+ */
+ x = gmx_simd_fmadd_f(invargscale, xscaled, x);
+ /* x will now be a _very_ small number, so approximate exp(x)=1+x.
+ * We should thus apply the correction as r'=r*(1+x)=r+r*x
+ */
+ r = gmx_simd_fmadd_f(r, x, r);
+ return r;
}
-static gmx_inline __m512
+static gmx_inline __m512 gmx_simdcall
gmx_simd_log_f_mic(__m512 x)
{
return _mm512_mul_ps(_mm512_set1_ps(0.693147180559945286226764), _mm512_log2ae23_ps(x));
biod.pnpi.spb.ru / alexxy / gromacs.git / blobdiff