H.J. Lu wrote:
On Thu, Mar 12, 2009 at 8:58 AM, H.J. Lu <hjl.to...@gmail.com> wrote:
Index: i386.md
===================================================================
--- i386.md (revision 144796)
+++ i386.md (working copy)
@@ -20813,7 +20813,7 @@
[(match_dup 0)
(match_operand:SI 2 "memory_operand" "")]))
(clobber (reg:CC FLAGS_REG))])]
- "operands[0] != operands[1]
+ "REGNO (operands[0]) != REGNO (operands[1])
&& GENERAL_REGNO_P (REGNO (operands[0]))
&& GENERAL_REGNO_P (REGNO (operands[1]))"
[(set (match_dup 0) (match_dup 4))
@@ -20829,7 +20829,7 @@
(match_operator 3 "commutative_operator"
[(match_dup 0)
(match_operand 2 "memory_operand" "")]))]
- "operands[0] != operands[1]
+ "REGNO (operands[0]) != REGNO (operands[1])
&& ((MMX_REG_P (operands[0]) && MMX_REG_P (operands[1]))
|| (SSE_REG_P (operands[0]) && SSE_REG_P (operands[1])))"
[(set (match_dup 0) (match_dup 2))
now.
It is revision 144817.
Thanks. I just rebuilt GCC using revision 144821 and rebuilt HIRLAM
with that compiler.
It doesn't exhibit the Infinity's anymore. A typical time step output is:
NSTEP 3 BEGINS HERE:
1800s into the forecast; valid on 20090312 at 09:30:00
-------------------------------
ABS(DPS)/3H = 1.351563 MB
MAXWIND = 76.612305 M/S IN (X,Y,LEV)= 27 34 10
STPS = 10072.463867187500
STQ = 4.852974414825
STS = 0.014612672850
STPE = 2496469248.000000000000
STKE = 1680526.000000000000
STTE = 2498149888.000000000000
TOTAL RAIN RATE(mm/d)= 0.171960353851
STRATIFORM PRECIP RATE= 0.169621527195
CONVECTIVE PRECIP RATE= 0.002338820370
CWPATH = 0.016851592809
COV2D = 0.151605650783
HIGH CLOUDS = 0.079861477017
MEDIUM CLOUDS = 0.041859772056
LOW CLOUDS = 0.073108613491
LOWEST LEVEL CLOUDS/FOG = 0.007746003568
AVERAGED CLOUD BASE= 625.465087890625
AVERAGED CLOUD TOP= 1203.065795898438
SENF = -35.196342468262
LATF = -99.297454833984
MOMF = 0.244225323200
Thanks go out to Paolo Bonzini and H.J. Lu for quickly identifying the
problem (i.e., beyond the "OMG it's full of NaNs" that I displayed).
My concern was that a regression would go out in the wild because not
enough people are testing floating point intensive programs (SPEC not
withstanding).
Fortunately, Paolo and H.J. took my anger in stride and produced a fix
within 24 hours.
Wherefore thanks and kudo's.
Cheers,
--
Toon Moene - e-mail: t...@moene.org (*NEW*) - phone: +31 346 214290
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
At home: http://moene.org/~toon/
Progress of GNU Fortran: http://gcc.gnu.org/gcc-4.4/changes.html
