From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeffrey A Law <law@cygnus.com>
To: gcc-patches@gcc.gnu.org
Subject: Minor testsuite improvement
Date: Tue, 31 Aug 1999 17:01:00 -0000
Message-id: <20455.936143593@upchuck.cygnus.com>
X-SW-Source: 1999-08/msg01021.html

This speeds up native testing of the c-torture execute tests by a few
percent for native tests.

Basically there's an optimization in the test path which greatly helps
embedded or simulated targets, but loses for native targets because the
costs of the various steps is radically different.

        * lib/c-torture.exp: Avoid the "compare executables" optimization
        when testing native.

Index: c-torture.exp
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/testsuite/lib/c-torture.exp,v
retrieving revision 1.11
diff -c -3 -p -r1.11 c-torture.exp
*** c-torture.exp	1999/06/30 01:36:43	1.11
--- c-torture.exp	1999/08/31 23:58:48
*************** proc c-torture-execute { src args } {
*** 189,196 ****
              set ignore_me [eval $torture_eval_before_execute]
  	}
  	
  	set skip 0;
! 	if [info exists oldexec] {
  	    if { [remote_file build cmp $oldexec $execname] == 0 } {
  		set skip 1;
  	    }
--- 189,210 ----
              set ignore_me [eval $torture_eval_before_execute]
  	}
  	
+ 
+ 	# Sometimes we end up creating identical executables for two
+ 	# consecutive sets of different of compiler options.
+ 	#
+ 	# In such cases we know the result of this test will be identical
+ 	# to the result of the last test.
+ 	#
+ 	# So in cases where the time to load and run/simulate the test
+ 	# is relatively high, compare the two binaries and avoid rerunning
+ 	# tests if the executables are identical.
+ 	#
+ 	# Do not do this for native testing since the cost to load/execute
+ 	# the test is fairly small and the comparison step actually slows
+ 	# the entire process down because it usually does not "hit".
  	set skip 0;
! 	if { ![isnative] && [info exists oldexec] } {
  	    if { [remote_file build cmp $oldexec $execname] == 0 } {
  		set skip 1;
  	    }



From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeffrey A Law <law@cygnus.com>
To: gcc-patches@gcc.gnu.org
Subject: Minor testsuite improvement
Date: Tue, 31 Aug 1999 22:41:00 -0000
Message-ID: <20455.936143593@upchuck.cygnus.com>
X-SW-Source: 1999-08n/msg01021.html
Message-ID: <19990831224100.UF0tnJAczTub2vKqqeMIAlMeCUPe_GoIZde0Zdx3P2Q@z>

This speeds up native testing of the c-torture execute tests by a few
percent for native tests.

Basically there's an optimization in the test path which greatly helps
embedded or simulated targets, but loses for native targets because the
costs of the various steps is radically different.

        * lib/c-torture.exp: Avoid the "compare executables" optimization
        when testing native.

Index: c-torture.exp
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/testsuite/lib/c-torture.exp,v
retrieving revision 1.11
diff -c -3 -p -r1.11 c-torture.exp
*** c-torture.exp	1999/06/30 01:36:43	1.11
--- c-torture.exp	1999/08/31 23:58:48
*************** proc c-torture-execute { src args } {
*** 189,196 ****
              set ignore_me [eval $torture_eval_before_execute]
  	}
  	
  	set skip 0;
! 	if [info exists oldexec] {
  	    if { [remote_file build cmp $oldexec $execname] == 0 } {
  		set skip 1;
  	    }
--- 189,210 ----
              set ignore_me [eval $torture_eval_before_execute]
  	}
  	
+ 
+ 	# Sometimes we end up creating identical executables for two
+ 	# consecutive sets of different of compiler options.
+ 	#
+ 	# In such cases we know the result of this test will be identical
+ 	# to the result of the last test.
+ 	#
+ 	# So in cases where the time to load and run/simulate the test
+ 	# is relatively high, compare the two binaries and avoid rerunning
+ 	# tests if the executables are identical.
+ 	#
+ 	# Do not do this for native testing since the cost to load/execute
+ 	# the test is fairly small and the comparison step actually slows
+ 	# the entire process down because it usually does not "hit".
  	set skip 0;
! 	if { ![isnative] && [info exists oldexec] } {
  	    if { [remote_file build cmp $oldexec $execname] == 0 } {
  		set skip 1;
  	    }