Re: [Mingw-w64-public] Implement fused multiply-add (FMA) funcitons for x86 families properly

2017-01-19 Thread lhmouse 
> So you have decided that __builtins can't be used then?  That's too bad.
Yes it results in a call to `fma()` on x64. Can't test it on ARM though.

> I know almost nothing about the guts of floating point, so I'm prepared
> to defer to your judgement, but here's what I think:
> 
> Let me propose an alternative for fma.c:
> ... ...
> In other words, remove all the platform specific code.  This (greatly)
> simplifies this file.  You were already using fmal for x86.  And it
> doesn't lose anything for ARM, since both fma() and fmal() use the exact
> same inline asm.  Why have the exact same (hard to maintain) code in 2
> places?
Keeping asm code in fmaf.c but not in fma.c seems style inconsistency.
However the contrary is doable: In the case of ARM, call `fma()` in `fmal()`.

> As for fmaf, what about:
> ... ...
> The case here is less compelling, but I assert that if fmal is
> supported, it can always be used to calculate fmaf.  If there is a
> shorter/more efficient method (such as there is with ARM), it can be
> added here.
Fair enough. Updated.

> As for fmal, I have a question about your code.  Not the implementation,
> but the design.  Looking at https://en.wikipedia.org/wiki/Long_double,
> it says "Microsoft Windows with Visual C++ also sets the processor in
> double-precision mode by default."  Since (it appears?) you aren't
> following _controlfp_s, won't this give use a different answer than fmal
> from msvcr120.dll?
MSVC doesn't support 80-bit `long double` (it is 64 bits there) so
the results can't equal unless it fits into 64 bits precisely.
My FMA algorithm is basically splitting both operands into two 32-bit ones,
multiplying them using elementary arithmetics then adding the four 64-bit
results altogether: (a+b)(c+d) = ac+(bc+ad)+bd. So the precision of x87
indeed affects the result.
I doubt whether it is necessary to save the x87 control word and set it to
64-bit precision before the calcuation and restore it thereafter. MinGW-w64
already sets it to 64-bit precision during CRT initialization, and if people
set it lower they ain't going to need `fma()` either.

An interesting look at https://msdn.microsoft.com/en-us/library/c9676k6h.aspx
reminds me that _PC_64 isn't supported on x64. Sounds incredible, no? Does
`_controlfp_s()` return an error if we try to set _PC_64 on 0x64? I have no
idea. Nevertheless the precision flags can be set and restored using inline
assembly - yet another dirty solution.

> More nits:
>
> s/whecher/whether
> s/#x86_Extended_Precision_Format/#x86_extended_precision_format
Fixed. The bookmark to wikipedia was copied from my broswer half a year ago
at least and it probably was modified.

--   
Best regards,
lh_mouse
2017-01-20
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot___
Mingw-w64-public mailing list
Mingw-w64-public@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public

Re: [Mingw-w64-public] Implement fused multiply-add (FMA) funcitons for x86 families properly

2017-01-19 Thread David Wohlferd 
So you have decided that __builtins can't be used then?  That's too bad.

I know almost nothing about the guts of floating point, so I'm prepared 
to defer to your judgement, but here's what I think:

Let me propose an alternative for fma.c:



/**
  * This file has no copyright assigned and is placed in the Public Domain.
  * This file is part of the mingw-w64 runtime package.
  * No warranty is given; refer to the file DISCLAIMER.PD within this 
package.
  */
double fma(double x, double y, double z);

long double fmal(long double x, long double y, long double z);

double fma(double x, double y, double z){
   return (double)fmal(x, y, z);
}



In other words, remove all the platform specific code.  This (greatly) 
simplifies this file.  You were already using fmal for x86.  And it 
doesn't lose anything for ARM, since both fma() and fmal() use the exact 
same inline asm.  Why have the exact same (hard to maintain) code in 2 
places?

As for fmaf, what about:



/**
  * This file has no copyright assigned and is placed in the Public Domain.
  * This file is part of the mingw-w64 runtime package.
  * No warranty is given; refer to the file DISCLAIMER.PD within this 
package.
  */
float fmaf(float x, float y, float z);

#if defined(_ARM_) || defined(__arm__)

float fmaf(float x, float y, float z){
   __asm__ (
 "fmacs %0, %1, %2 \n"
 : "+t"(z)
 : "t"(x), "t"(y)
   );
   return z;
}

#else

long double fmal(long double x, long double y, long double z);

float fmaf(float x, float y, float z){
   return (float)fmal(x, y, z);
}

#endif



The case here is less compelling, but I assert that if fmal is 
supported, it can always be used to calculate fmaf.  If there is a 
shorter/more efficient method (such as there is with ARM), it can be 
added here.

As for fmal, I have a question about your code.  Not the implementation, 
but the design.  Looking at https://en.wikipedia.org/wiki/Long_double, 
it says "Microsoft Windows with Visual C++ also sets the processor in 
double-precision mode by default."  Since (it appears?) you aren't 
following _controlfp_s, won't this give use a different answer than fmal 
from msvcr120.dll?

More nits:

s/whecher/whether
s/#x86_Extended_Precision_Format/#x86_extended_precision_format

dw


--
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
___
Mingw-w64-public mailing list
Mingw-w64-public@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public

Re: [Mingw-w64-public] Implement fused multiply-add (FMA) funcitons for x86 families properly

2017-01-19 Thread lhmouse 
New patch attached.
This patch fixes ARM functions and adds a check in `fpu_fma()` for potential 
NaN or INF results.


--   
Best regards,
lh_mouse
2017-01-19






From 3c55daec84dac190b9e3cb032371960e1acbc38f Mon Sep 17 00:00:00 2001
From: LH_Mouse 
Date: Wed, 18 Jan 2017 19:35:43 +0800
Subject: [PATCH] mingw-w64-crt/math/fma{,f,l}.c: Implement fused multiply-add
 (FMA) funcitons for x86 families properly. mingw-w64-crt/Makefile.am:
 Likewise. mingw-w64-crt/math/fma{,f}.S: Merge into corresponding C files with
 the same names, respectively.

---
 mingw-w64-crt/Makefile.am |   4 +-
 mingw-w64-crt/math/fma.S  |  42 -
 mingw-w64-crt/math/fma.c  |  31 ++
 mingw-w64-crt/math/fmaf.S |  43 --
 mingw-w64-crt/math/fmaf.c |  31 ++
 mingw-w64-crt/math/fmal.c | 146 --
 6 files changed, 205 insertions(+), 92 deletions(-)
 delete mode 100644 mingw-w64-crt/math/fma.S
 create mode 100644 mingw-w64-crt/math/fma.c
 delete mode 100644 mingw-w64-crt/math/fmaf.S
 create mode 100644 mingw-w64-crt/math/fmaf.c

diff --git a/mingw-w64-crt/Makefile.am b/mingw-w64-crt/Makefile.am
index 44360db..5eba234 100644
--- a/mingw-w64-crt/Makefile.am
+++ b/mingw-w64-crt/Makefile.am
@@ -227,7 +227,6 @@ src_libmingwex=\
   \
   math/_chgsignl.S  math/ceil.Smath/ceilf.S  math/ceill.S  
   math/copysignl.S \
   math/floor.S  math/floorf.S  math/floorl.S \
-  math/fma.Smath/fmaf.S\
   math/nearbyint.S  math/nearbyintf.S  math/nearbyintl.S \
   math/trunc.S  math/truncf.S  \
   math/cbrt.c   \
@@ -235,7 +234,8 @@ src_libmingwex=\
   math/coshf.c  math/coshl.c   math/erfl.c   \
   math/expf.c   \
   math/fabs.c   math/fabsf.c   math/fabsl.c  math/fdim.c   
   math/fdimf.c math/fdiml.c \
-  math/fmal.c   math/fmax.cmath/fmaxf.c  math/fmaxl.c  
   math/fmin.c  math/fminf.c \
+  math/fma.cmath/fmaf.cmath/fmal.c   \
+  math/fmax.c   math/fmaxf.c   math/fmaxl.c  math/fmin.c   
   math/fminf.c \
   math/fminl.c  math/fp_consts.c   math/fp_constsf.c \
   math/fp_constsl.c math/fpclassify.c  math/fpclassifyf.c
math/fpclassifyl.c   math/frexpf.c\
   math/hypotf.c math/hypot.c  math/hypotl.c  math/isnan.c  
math/isnanf.cmath/isnanl.c\
diff --git a/mingw-w64-crt/math/fma.S b/mingw-w64-crt/math/fma.S
deleted file mode 100644
index 74becde..000
--- a/mingw-w64-crt/math/fma.S
+++ /dev/null
@@ -1,42 +0,0 @@
-/**
- * This file has no copyright assigned and is placed in the Public Domain.
- * This file is part of the mingw-w64 runtime package.
- * No warranty is given; refer to the file DISCLAIMER.PD within this package.
- */
-#include <_mingw_mac.h>
-
-   .file   "fma.S"
-   .text
-#ifdef __x86_64__
-   .align 8
-#else
-   .align 4
-#endif
-   .p2align 4,,15
-   .globl __MINGW_USYMBOL(fma)
-   .def__MINGW_USYMBOL(fma);   .scl2;  .type   32; .endef
-__MINGW_USYMBOL(fma):
-#if defined(_AMD64_) || defined(__x86_64__)
-   subq$56, %rsp
-   movsd   %xmm0,(%rsp)
-   movsd   %xmm1,16(%rsp)
-   movsd   %xmm2,32(%rsp)
-   fldl(%rsp)
-   fmull   16(%rsp)
-   fldl32(%rsp)
-   faddp
-   fstpl   (%rsp)
-   movsd   (%rsp),%xmm0
-   addq$56, %rsp
-   ret
-#elif defined(_ARM_) || defined(__arm__)
-   fmacd d2, d0, d1
-   fcpyd d0, d2
-   bx  lr
-#elif defined(_X86_) || defined(__i386__)
-   fldl4(%esp)
-   fmull   12(%esp)
-   fldl20(%esp)
-   faddp
-   ret
-#endif
diff --git a/mingw-w64-crt/math/fma.c b/mingw-w64-crt/math/fma.c
new file mode 100644
index 000..00f100c
--- /dev/null
+++ b/mingw-w64-crt/math/fma.c
@@ -0,0 +1,31 @@
+/**
+ * This file has no copyright assigned and is placed in the Public Domain.
+ * This file is part of the mingw-w64 runtime package.
+ * No warranty is given; refer to the file DISCLAIMER.PD within this package.
+ */
+double fma(double x, double y, double z);
+
+#if defined(_AMD64_) || defined(__x86_64__) || defined(_X86_) || 
defined(__i386__)
+
+long double fmal(long double x, long double y, long double z);
+
+double fma(double x, double y, double z){
+  return (double)fmal(x, y, z);
+}
+
+#elif defined(_ARM_) || defined(__arm__)
+
+double fma(double x, double y, double z){
+  __asm__ (
+"fmacd %0, %1, %2 \n"
+: "+w"(z)
+: "w"(x), "w"(y)
+  );
+  return z;
+}
+
+#else
+
+#error This platform is not supported.
+
+#endif
diff --git a/mingw-w64-crt/math/fmaf.S b/mingw-w64-crt/math/fmaf.S
deleted file mode 100644
index 6bc7ef0..000
--- a/mingw-w64-crt/math/fmaf.S
+++ /dev/null
@@ -1,43 +0,0 @@
-/**
- * This