Re: [PATCH, libgcc, ARM] __gnu_f2h_internal inaccuracy

2012-12-13 Thread nick clifton 

Hi John,


John Tytgatj...@bass-software.com


* config/arm/fp16.c (__gnu_f2h_internal): Fix inaccuracy.



Approved and applied.

Cheers
  Nick

Re: [PATCH, libgcc, ARM] __gnu_f2h_internal inaccuracy

2012-12-11 Thread John Tytgat 
Ping ? Paul, seen that you've contributed fp16.c together with Sandra
Loosemore, perhaps you can review this patch please ?

John.

In message ab11eef452...@hobbes.bass-software.com
  John Tytgat j...@bass-software.com wrote:

 __gnu_f2h_internal in libgcc converts single-precision floating-point value
 to half-precision value for ARM.  I noticed there is a slight inaccuracy
 for floating-point values 2^-25 + epsilon (with epsilon  0 and  2^-26).
 Those should all be converted to 2^-24 in half-precision.
 
 And this because the mask used to implement the even-odd rounding for
 aexp = -25 is wrong.  Currently we have:
 
   /* Decimal point between bits 22 and 23.  */
   mantissa |= 0x0080;
   if (aexp  -14)
 {
   mask = 0x007f;
   if (aexp  -25)
 aexp = -26;
   else if (aexp != -25)
 mask = 24 + aexp;
 }
   else
 mask = 0x1fff;
 
 But when aexp is 25 the mask should be 0xff instead of 0x7f as the
 decimal dot in half-precision will be between bit 24 and 23 of the
 above mentioned mantissa.  Cfr. the even-odd rounding done:
 
   /* Round.  */
   if (mantissa  mask)
 {
   increment = (mask + 1)  1;
   if ((mantissa  mask) == increment)
 increment = mantissa  (increment  1);
   mantissa += increment;
   if (mantissa = 0x0100)
 {
   mantissa = 1;
   aexp++;
 }
 }
 
 Attached patch solves this problem.  I've left out the clamping of
 aexp to -26 for values less than -25 as this it not necessary.  After
 the even-odd rounding all aexp values less than -24 will result in +0. or
 -0. anyway.
 
 John Tytgat  j...@bass-software.com
 
   * config/arm/fp16.c (__gnu_f2h_internal): Fix inaccuracy.
 
 I've got a copyright assignment but no write access.
 
 John Tytgat.

-- 
John Tytgat BASS
j...@bass-software.comIndex: libgcc/config/arm/fp16.c
===
--- libgcc/config/arm/fp16.c(revision 193830)
+++ libgcc/config/arm/fp16.c(working copy)
@@ -47,11 +47,9 @@
   mantissa |= 0x0080;
   if (aexp  -14)
 {
-  mask = 0x007f;
-  if (aexp  -25)
-   aexp = -26;
-  else if (aexp != -25)
-   mask = 24 + aexp;
+  mask = 0x00ff;
+  if (aexp = -25)
+mask = 25 + aexp;
 }
   else
 mask = 0x1fff;

[PATCH, libgcc, ARM] __gnu_f2h_internal inaccuracy

2012-11-26 Thread John Tytgat 
__gnu_f2h_internal in libgcc converts single-precision floating-point value
to half-precision value for ARM.  I noticed there is a slight inaccuracy
for floating-point values 2^-25 + epsilon (with epsilon  0 and  2^-26).
Those should all be converted to 2^-24 in half-precision.

And this because the mask used to implement the even-odd rounding for
aexp = -25 is wrong.  Currently we have:

  /* Decimal point between bits 22 and 23.  */
  mantissa |= 0x0080;
  if (aexp  -14)
{
  mask = 0x007f;
  if (aexp  -25)
aexp = -26;
  else if (aexp != -25)
mask = 24 + aexp;
}
  else
mask = 0x1fff;

But when aexp is 25 the mask should be 0xff instead of 0x7f as the
decimal dot in half-precision will be between bit 24 and 23 of the
above mentioned mantissa.  Cfr. the even-odd rounding done:

  /* Round.  */
  if (mantissa  mask)
{
  increment = (mask + 1)  1;
  if ((mantissa  mask) == increment)
increment = mantissa  (increment  1);
  mantissa += increment;
  if (mantissa = 0x0100)
{
  mantissa = 1;
  aexp++;
}
}

Attached patch solves this problem.  I've left out the clamping of
aexp to -26 for values less than -25 as this it not necessary.  After
the even-odd rounding all aexp values less than -24 will result in +0. or
-0. anyway.

John Tytgat  j...@bass-software.com

* config/arm/fp16.c (__gnu_f2h_internal): Fix inaccuracy.

I've got a copyright assignment but no write access.

John Tytgat.
-- 
John Tytgat BASS
j...@bass-software.com

Re: [PATCH, libgcc, ARM] __gnu_f2h_internal inaccuracy

2012-11-26 Thread John Tytgat 
In message ab11eef452...@hobbes.bass-software.com
  John Tytgat j...@bass-software.com wrote:

 [...] Attached patch solves this problem. [...]

This time for real.

John Tytgat.
-- 
John Tytgat, in his comfy chair at home
john.tyt...@aaug.net
Index: libgcc/config/arm/fp16.c
===
--- libgcc/config/arm/fp16.c(revision 193830)
+++ libgcc/config/arm/fp16.c(working copy)
@@ -47,11 +47,9 @@
   mantissa |= 0x0080;
   if (aexp  -14)
 {
-  mask = 0x007f;
-  if (aexp  -25)
-   aexp = -26;
-  else if (aexp != -25)
-   mask = 24 + aexp;
+  mask = 0x00ff;
+  if (aexp = -25)
+mask = 25 + aexp;
 }
   else
 mask = 0x1fff;