diff mbox

MIPS: Add the lo register to the clobber list in the madd-8.c and msub-8.c testcases

Message ID 0DA23CC379F5F945ACB41CF394B98277210F9AA5@LEMAIL01.le.imgtec.org
State New
Headers show

Commit Message

Andrew Bennett July 21, 2015, 2:56 p.m. UTC 
Hi,

The lo register is not listed in the clobber list in the inline asm statement
for the madd-8.c and msub-8.c testcases.  This means that when building for the 
n64 ABI GCC is free to use the lo register instead of the stack when saving/restoring 
the clobbered registers.  Then then means that it decides to use the msub/madd
instruction to perform the "x - y * z" operation rather than using mul; addu/subu
which the test is looking for.

The following patch therefore adds the lo register to the clobber list
for the madd-8.c and msub-8.c testcases.  The patch has been tested on the 
mti/img elf/linux-gnu toolchains, and there have been no new regressions.

The patch and ChangeLog are below.

Ok to commit?


Many thanks,



Andrew



testsuite/
	gcc.target/mips/madd-8.c: Add lo register to clobber list. 
 	gcc.target/mips/msub-8.c: Ditto

Comments

Moore, Catherine Aug. 24, 2015, 9:42 p.m. UTC | #1 
HI Andrew,

> -----Original Message-----
> From: Andrew Bennett [mailto:Andrew.Bennett@imgtec.com]
> Sent: Tuesday, July 21, 2015 10:56 AM
> To: gcc-patches@gcc.gnu.org
> Cc: Matthew Fortune; Moore, Catherine
> Subject: [PATCH] MIPS: Add the lo register to the clobber list in the madd-8.c
> and msub-8.c testcases
> 
> Hi,
> 
> The lo register is not listed in the clobber list in the inline asm statement for
> the madd-8.c and msub-8.c testcases.  This means that when building for the
> n64 ABI GCC is free to use the lo register instead of the stack when
> saving/restoring the clobbered registers.  Then then means that it decides to
> use the msub/madd instruction to perform the "x - y * z" operation rather
> than using mul; addu/subu which the test is looking for.
> 
> The following patch therefore adds the lo register to the clobber list for the
> madd-8.c and msub-8.c testcases.  The patch has been tested on the mti/img
> elf/linux-gnu toolchains, and there have been no new regressions.
> 
> The patch and ChangeLog are below.
> 
> Ok to commit?
> 
> 
Yes, this looks OK.
Andrew Bennett Aug. 28, 2015, 1:35 p.m. UTC | #2 
> Yes, this looks OK.

Committed as SVN 227299.

Regards,


Andrew
diff mbox

Patch

diff --git a/gcc/testsuite/gcc.target/mips/madd-8.c b/gcc/testsuite/gcc.target/mips/madd-8.c
index 794a6ff..56c1947 100644
--- a/gcc/testsuite/gcc.target/mips/madd-8.c
+++ b/gcc/testsuite/gcc.target/mips/madd-8.c
@@ -11,6 +11,6 @@  f2 (int x, int y, int z)
   asm volatile ("" ::: "$1", "$2", "$3", "$4", "$5", "$6", "$7", "$8", "$9",
 		"$10", "$11", "$12", "$13", "$14", "$15", "$16", "$17",
 		"$18", "$19", "$20", "$21", "$22", "$23", "$24", "$25",
-		"$31");
+		"$31", "lo");
   return x * y + z;
 }
diff --git a/gcc/testsuite/gcc.target/mips/msub-8.c b/gcc/testsuite/gcc.target/mips/msub-8.c
index a66307f..b0f1523 100644
--- a/gcc/testsuite/gcc.target/mips/msub-8.c
+++ b/gcc/testsuite/gcc.target/mips/msub-8.c
@@ -11,6 +11,6 @@  f2 (int x, int y, int z)
   asm volatile ("" ::: "$1", "$2", "$3", "$4", "$5", "$6", "$7", "$8", "$9",
 		"$10", "$11", "$12", "$13", "$14", "$15", "$16", "$17",
 		"$18", "$19", "$20", "$21", "$22", "$23", "$24", "$25",
-		"$31");
+		"$31", "lo");
   return x - y * z;
 }