initial commit

1 files changed, 92 insertions, 0 deletions

diff --git a/gcc/testsuite/gcc.wendy/gnu13.c b/gcc/testsuite/gcc.wendy/gnu13.c
new file mode 100755
index 0000000..d7db170
--- /dev/null
+++ b/gcc/testsuite/gcc.wendy/gnu13.c
@@ -0,0 +1,92 @@
+/*
+Date: Tue, 21 Apr 87 09:58:10 PDT
+From: aks@acc-sb-unix.arpa (Alan Stebbens)
+To: bug-gcc@PREP.AI.MIT.EDU
+Subject: Optimizer doesn't.
+
+Dear GCC guys:
+
+The following simple C program illustrates a problem with the
+current version of GCC.  What happens is that D3 is used doubly:
+once for loop limit testing, and once for long constant stores.
+Also, a loop limit test is being performed for the first iteration
+even when both the initial value and the ending values are known
+constants.  Most optimizers can avoid the initial test if these
+values are known.
+
+Thanks for your time.
+
+Alan Stebbens (aks@acc-sb-unix.ARPA)
+
+==================== tst1.c ====================
+*/
+int array[100];
+int darray[10][10];
+
+main()
+{
+   register int i, j;
+   for (i = 0; i < 100; i++) array[i] = 3;
+   for (i = 0; i < 10; i++)
+      for (j = 0; j < 10; j++)
+	 darray[i][j] = 3;
+   return 0;
+}
+
+/*
+==================== tst1.s ====================
+.comm _array,400
+.comm _darray,400
+.globl _main
+.text
+	.even
+_main:
+	link a6,#0
+	moveml #0x3030,sp@-
+	clrl d2
+	lea _array,a0
+	moveq #100,d3	<------ D3 now has limit of the test
+	cmpl d2,d3
+	jle L19
+L2:
+	moveq #3,d3	<----- D3 is being reused!!
+	movl d3,a0@(d2:l:4)
+	addql #1,d2
+	moveq #100,d3	<----- D3 is being reloaded!
+	cmpl d2,d3
+	jgt L2
+L19:
+	clrl d2
+	lea _darray,a2
+	moveq #10,d3
+	cmpl d2,d3
+	jle L18
+L7:
+	clrl d1
+	movl d2,d0
+	asll #5,d0
+	movl d0,a3
+	lea a3@(d2:l:8),a1
+	moveq #10,d3
+	cmpl d1,d3
+	jle L17
+L10:
+	lea a1@(d1:l:4),a0
+	moveq #3,d3
+	movl d3,a2@(a0:l)
+	addql #1,d1
+	moveq #10,d3
+	cmpl d1,d3
+	jgt L10
+L17:
+	addql #1,d2
+	moveq #10,d3
+	cmpl d2,d3
+	jgt L7
+L18:
+	moveml a6@(-16),#0xc0c
+	unlk a6
+	rts
+
+Alan Stebbens (aks@acc-sb-unix.ARPA)
+*/