msvcrt: Add a non-optimized implementation for the SSE2 math functions.

diff --git a/dlls/msvcr100/msvcr100.spec b/dlls/msvcr100/msvcr100.spec
index 9b4cecda13c10ab79bfe06023721a899d29ac104..a601f2382f7248b0e24c2d4321681e68661267c6 100644 (file)
--- a/dlls/msvcr100/msvcr100.spec
+++ b/dlls/msvcr100/msvcr100.spec
@@ -413,27 +413,27 @@
 @ stub __iswcsymf
 @ stub __lconv
 @ cdecl __lconv_init() msvcrt.__lconv_init
-@ stub __libm_sse2_acos
-@ stub __libm_sse2_acosf
-@ stub __libm_sse2_asin
-@ stub __libm_sse2_asinf
-@ stub __libm_sse2_atan
-@ stub __libm_sse2_atan2
-@ stub __libm_sse2_atanf
-@ stub __libm_sse2_cos
-@ stub __libm_sse2_cosf
-@ stub __libm_sse2_exp
-@ stub __libm_sse2_expf
-@ stub __libm_sse2_log
-@ stub __libm_sse2_log10
-@ stub __libm_sse2_log10f
-@ stub __libm_sse2_logf
-@ stub __libm_sse2_pow
-@ stub __libm_sse2_powf
-@ stub __libm_sse2_sin
-@ stub __libm_sse2_sinf
-@ stub __libm_sse2_tan
-@ stub __libm_sse2_tanf
+@ cdecl -arch=i386 __libm_sse2_acos() msvcrt.__libm_sse2_acos
+@ cdecl -arch=i386 __libm_sse2_acosf() msvcrt.__libm_sse2_acosf
+@ cdecl -arch=i386 __libm_sse2_asin() msvcrt.__libm_sse2_asin
+@ cdecl -arch=i386 __libm_sse2_asinf() msvcrt.__libm_sse2_asinf
+@ cdecl -arch=i386 __libm_sse2_atan() msvcrt.__libm_sse2_atan
+@ cdecl -arch=i386 __libm_sse2_atan2() msvcrt.__libm_sse2_atan2
+@ cdecl -arch=i386 __libm_sse2_atanf() msvcrt.__libm_sse2_atanf
+@ cdecl -arch=i386 __libm_sse2_cos() msvcrt.__libm_sse2_cos
+@ cdecl -arch=i386 __libm_sse2_cosf() msvcrt.__libm_sse2_cosf
+@ cdecl -arch=i386 __libm_sse2_exp() msvcrt.__libm_sse2_exp
+@ cdecl -arch=i386 __libm_sse2_expf() msvcrt.__libm_sse2_expf
+@ cdecl -arch=i386 __libm_sse2_log() msvcrt.__libm_sse2_log
+@ cdecl -arch=i386 __libm_sse2_log10() msvcrt.__libm_sse2_log10
+@ cdecl -arch=i386 __libm_sse2_log10f() msvcrt.__libm_sse2_log10f
+@ cdecl -arch=i386 __libm_sse2_logf() msvcrt.__libm_sse2_logf
+@ cdecl -arch=i386 __libm_sse2_pow() msvcrt.__libm_sse2_pow
+@ cdecl -arch=i386 __libm_sse2_powf() msvcrt.__libm_sse2_powf
+@ cdecl -arch=i386 __libm_sse2_sin() msvcrt.__libm_sse2_sin
+@ cdecl -arch=i386 __libm_sse2_sinf() msvcrt.__libm_sse2_sinf
+@ cdecl -arch=i386 __libm_sse2_tan() msvcrt.__libm_sse2_tan
+@ cdecl -arch=i386 __libm_sse2_tanf() msvcrt.__libm_sse2_tanf
 @ extern __mb_cur_max msvcrt.__mb_cur_max
 @ cdecl -arch=i386 __p___argc() msvcrt.__p___argc
 @ cdecl -arch=i386 __p___argv() msvcrt.__p___argv

diff --git a/dlls/msvcr80/msvcr80.spec b/dlls/msvcr80/msvcr80.spec
index 4eaed927f746bc2b4df2606690edba5b5a7a96b5..b4787ad5e72471b74c05aec232a3a85c830cbf6f 100644 (file)
--- a/dlls/msvcr80/msvcr80.spec
+++ b/dlls/msvcr80/msvcr80.spec
@@ -231,27 +231,27 @@
 @ extern __lc_handle msvcrt.__lc_handle
 @ stub __lconv
 @ cdecl __lconv_init() msvcrt.__lconv_init
-@ stub __libm_sse2_acos
-@ stub __libm_sse2_acosf
-@ stub __libm_sse2_asin
-@ stub __libm_sse2_asinf
-@ stub __libm_sse2_atan
-@ stub __libm_sse2_atan2
-@ stub __libm_sse2_atanf
-@ stub __libm_sse2_cos
-@ stub __libm_sse2_cosf
-@ stub __libm_sse2_exp
-@ stub __libm_sse2_expf
-@ stub __libm_sse2_log
-@ stub __libm_sse2_log10
-@ stub __libm_sse2_log10f
-@ stub __libm_sse2_logf
-@ stub __libm_sse2_pow
-@ stub __libm_sse2_powf
-@ stub __libm_sse2_sin
-@ stub __libm_sse2_sinf
-@ stub __libm_sse2_tan
-@ stub __libm_sse2_tanf
+@ cdecl -arch=i386 __libm_sse2_acos() msvcrt.__libm_sse2_acos
+@ cdecl -arch=i386 __libm_sse2_acosf() msvcrt.__libm_sse2_acosf
+@ cdecl -arch=i386 __libm_sse2_asin() msvcrt.__libm_sse2_asin
+@ cdecl -arch=i386 __libm_sse2_asinf() msvcrt.__libm_sse2_asinf
+@ cdecl -arch=i386 __libm_sse2_atan() msvcrt.__libm_sse2_atan
+@ cdecl -arch=i386 __libm_sse2_atan2() msvcrt.__libm_sse2_atan2
+@ cdecl -arch=i386 __libm_sse2_atanf() msvcrt.__libm_sse2_atanf
+@ cdecl -arch=i386 __libm_sse2_cos() msvcrt.__libm_sse2_cos
+@ cdecl -arch=i386 __libm_sse2_cosf() msvcrt.__libm_sse2_cosf
+@ cdecl -arch=i386 __libm_sse2_exp() msvcrt.__libm_sse2_exp
+@ cdecl -arch=i386 __libm_sse2_expf() msvcrt.__libm_sse2_expf
+@ cdecl -arch=i386 __libm_sse2_log() msvcrt.__libm_sse2_log
+@ cdecl -arch=i386 __libm_sse2_log10() msvcrt.__libm_sse2_log10
+@ cdecl -arch=i386 __libm_sse2_log10f() msvcrt.__libm_sse2_log10f
+@ cdecl -arch=i386 __libm_sse2_logf() msvcrt.__libm_sse2_logf
+@ cdecl -arch=i386 __libm_sse2_pow() msvcrt.__libm_sse2_pow
+@ cdecl -arch=i386 __libm_sse2_powf() msvcrt.__libm_sse2_powf
+@ cdecl -arch=i386 __libm_sse2_sin() msvcrt.__libm_sse2_sin
+@ cdecl -arch=i386 __libm_sse2_sinf() msvcrt.__libm_sse2_sinf
+@ cdecl -arch=i386 __libm_sse2_tan() msvcrt.__libm_sse2_tan
+@ cdecl -arch=i386 __libm_sse2_tanf() msvcrt.__libm_sse2_tanf
 @ extern __mb_cur_max msvcrt.__mb_cur_max
 @ cdecl -arch=i386 __p___argc() msvcrt.__p___argc
 @ cdecl -arch=i386 __p___argv() msvcrt.__p___argv

diff --git a/dlls/msvcr90/msvcr90.spec b/dlls/msvcr90/msvcr90.spec
index a2b8359e839394feddd3169bdda62ca50ee6bd26..809265b0554a899472a5196e303518cf18b372cf 100644 (file)
--- a/dlls/msvcr90/msvcr90.spec
+++ b/dlls/msvcr90/msvcr90.spec
@@ -228,27 +228,27 @@
 @ extern __lc_handle msvcrt.__lc_handle
 @ stub __lconv
 @ cdecl __lconv_init() msvcrt.__lconv_init
-@ stub __libm_sse2_acos
-@ stub __libm_sse2_acosf
-@ stub __libm_sse2_asin
-@ stub __libm_sse2_asinf
-@ stub __libm_sse2_atan
-@ stub __libm_sse2_atan2
-@ stub __libm_sse2_atanf
-@ stub __libm_sse2_cos
-@ stub __libm_sse2_cosf
-@ stub __libm_sse2_exp
-@ stub __libm_sse2_expf
-@ stub __libm_sse2_log
-@ stub __libm_sse2_log10
-@ stub __libm_sse2_log10f
-@ stub __libm_sse2_logf
-@ stub __libm_sse2_pow
-@ stub __libm_sse2_powf
-@ stub __libm_sse2_sin
-@ stub __libm_sse2_sinf
-@ stub __libm_sse2_tan
-@ stub __libm_sse2_tanf
+@ cdecl -arch=i386 __libm_sse2_acos() msvcrt.__libm_sse2_acos
+@ cdecl -arch=i386 __libm_sse2_acosf() msvcrt.__libm_sse2_acosf
+@ cdecl -arch=i386 __libm_sse2_asin() msvcrt.__libm_sse2_asin
+@ cdecl -arch=i386 __libm_sse2_asinf() msvcrt.__libm_sse2_asinf
+@ cdecl -arch=i386 __libm_sse2_atan() msvcrt.__libm_sse2_atan
+@ cdecl -arch=i386 __libm_sse2_atan2() msvcrt.__libm_sse2_atan2
+@ cdecl -arch=i386 __libm_sse2_atanf() msvcrt.__libm_sse2_atanf
+@ cdecl -arch=i386 __libm_sse2_cos() msvcrt.__libm_sse2_cos
+@ cdecl -arch=i386 __libm_sse2_cosf() msvcrt.__libm_sse2_cosf
+@ cdecl -arch=i386 __libm_sse2_exp() msvcrt.__libm_sse2_exp
+@ cdecl -arch=i386 __libm_sse2_expf() msvcrt.__libm_sse2_expf
+@ cdecl -arch=i386 __libm_sse2_log() msvcrt.__libm_sse2_log
+@ cdecl -arch=i386 __libm_sse2_log10() msvcrt.__libm_sse2_log10
+@ cdecl -arch=i386 __libm_sse2_log10f() msvcrt.__libm_sse2_log10f
+@ cdecl -arch=i386 __libm_sse2_logf() msvcrt.__libm_sse2_logf
+@ cdecl -arch=i386 __libm_sse2_pow() msvcrt.__libm_sse2_pow
+@ cdecl -arch=i386 __libm_sse2_powf() msvcrt.__libm_sse2_powf
+@ cdecl -arch=i386 __libm_sse2_sin() msvcrt.__libm_sse2_sin
+@ cdecl -arch=i386 __libm_sse2_sinf() msvcrt.__libm_sse2_sinf
+@ cdecl -arch=i386 __libm_sse2_tan() msvcrt.__libm_sse2_tan
+@ cdecl -arch=i386 __libm_sse2_tanf() msvcrt.__libm_sse2_tanf
 @ extern __mb_cur_max msvcrt.__mb_cur_max
 @ cdecl -arch=i386 __p___argc() msvcrt.__p___argc
 @ cdecl -arch=i386 __p___argv() msvcrt.__p___argv

diff --git a/dlls/msvcrt/math.c b/dlls/msvcrt/math.c
index eac3f087685587ce9a31b0073967d242328779a4..c0fc11cbbbce7b524c7bb90d1177552b0da3fa96 100644 (file)
--- a/dlls/msvcrt/math.c
+++ b/dlls/msvcrt/math.c
@@ -1764,4 +1764,235 @@ void _safe_fprem1(void)
   TRACE("(): stub\n");
 }
 
+/***********************************************************************
+ *             __libm_sse2_acos   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_acos(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = acos( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_acosf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_acosf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = acosf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_asin   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_asin(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = asin( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_asinf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_asinf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = asinf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_atan   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_atan(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = atan( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_atan2   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_atan2(void)
+{
+    double d1, d2;
+    __asm__ __volatile__( "movq %%xmm0,%0; movq %%xmm1,%1 " : "=m" (d1), "=m" (d2) );
+    d1 = atan2( d1, d2 );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d1) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_atanf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_atanf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = atanf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_cos   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_cos(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = cos( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_cosf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_cosf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = cosf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_exp   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_exp(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = exp( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_expf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_expf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = expf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_log   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_log(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = log( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_log10   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_log10(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = log10( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_log10f   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_log10f(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = log10f( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_logf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_logf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = logf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_pow   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_pow(void)
+{
+    double d1, d2;
+    __asm__ __volatile__( "movq %%xmm0,%0; movq %%xmm1,%1 " : "=m" (d1), "=m" (d2) );
+    d1 = pow( d1, d2 );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d1) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_powf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_powf(void)
+{
+    float f1, f2;
+    __asm__ __volatile__( "movd %%xmm0,%0; movd %%xmm1,%1" : "=g" (f1), "=g" (f2) );
+    f1 = powf( f1, f2 );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f1) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_sin   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_sin(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = sin( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_sinf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_sinf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = sinf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_tan   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_tan(void)
+{
+    double d;
+    __asm__ __volatile__( "movq %%xmm0,%0" : "=m" (d) );
+    d = tan( d );
+    __asm__ __volatile__( "movq %0,%%xmm0" : : "m" (d) );
+}
+
+/***********************************************************************
+ *             __libm_sse2_tanf   (MSVCRT.@)
+ */
+void __cdecl __libm_sse2_tanf(void)
+{
+    float f;
+    __asm__ __volatile__( "movd %%xmm0,%0" : "=g" (f) );
+    f = tanf( f );
+    __asm__ __volatile__( "movd %0,%%xmm0" : : "g" (f) );
+}
+
 #endif  /* __i386__ */

diff --git a/dlls/msvcrt/msvcrt.spec b/dlls/msvcrt/msvcrt.spec
index 7fd3dfab8db984d675ccd275d11ae087c05c1d7c..5582e6344cfb11843b0df0f287d9dd90d6c32932 100644 (file)
--- a/dlls/msvcrt/msvcrt.spec
+++ b/dlls/msvcrt/msvcrt.spec
@@ -196,27 +196,27 @@
 @ extern __lc_collate_cp MSVCRT___lc_collate_cp
 @ extern __lc_handle MSVCRT___lc_handle
 @ cdecl __lconv_init()
-# stub __libm_sse2_acos
-# stub __libm_sse2_acosf
-# stub __libm_sse2_asin
-# stub __libm_sse2_asinf
-# stub __libm_sse2_atan
-# stub __libm_sse2_atan2
-# stub __libm_sse2_atanf
-# stub __libm_sse2_cos
-# stub __libm_sse2_cosf
-# stub __libm_sse2_exp
-# stub __libm_sse2_expf
-# stub __libm_sse2_log
-# stub __libm_sse2_log10
-# stub __libm_sse2_log10f
-# stub __libm_sse2_logf
-# stub __libm_sse2_pow
-# stub __libm_sse2_powf
-# stub __libm_sse2_sin
-# stub __libm_sse2_sinf
-# stub __libm_sse2_tan
-# stub __libm_sse2_tanf
+@ cdecl -arch=i386 __libm_sse2_acos()
+@ cdecl -arch=i386 __libm_sse2_acosf()
+@ cdecl -arch=i386 __libm_sse2_asin()
+@ cdecl -arch=i386 __libm_sse2_asinf()
+@ cdecl -arch=i386 __libm_sse2_atan()
+@ cdecl -arch=i386 __libm_sse2_atan2()
+@ cdecl -arch=i386 __libm_sse2_atanf()
+@ cdecl -arch=i386 __libm_sse2_cos()
+@ cdecl -arch=i386 __libm_sse2_cosf()
+@ cdecl -arch=i386 __libm_sse2_exp()
+@ cdecl -arch=i386 __libm_sse2_expf()
+@ cdecl -arch=i386 __libm_sse2_log()
+@ cdecl -arch=i386 __libm_sse2_log10()
+@ cdecl -arch=i386 __libm_sse2_log10f()
+@ cdecl -arch=i386 __libm_sse2_logf()
+@ cdecl -arch=i386 __libm_sse2_pow()
+@ cdecl -arch=i386 __libm_sse2_powf()
+@ cdecl -arch=i386 __libm_sse2_sin()
+@ cdecl -arch=i386 __libm_sse2_sinf()
+@ cdecl -arch=i386 __libm_sse2_tan()
+@ cdecl -arch=i386 __libm_sse2_tanf()
 @ extern __mb_cur_max MSVCRT___mb_cur_max
 @ cdecl -arch=i386 __p___argc()
 @ cdecl -arch=i386 __p___argv()