[PATCH, commited] Optimize setjmp/longjmp for moxie

[PATCH, commited] Optimize setjmp/longjmp for moxie

[Anthony Green]

I'm pushing the attached patch to the moxie setjmp/longjmp implementation.
We only need to save/restore callee saved registers - not registers we
expect to be trashed by function calls.

AG

diff --git a/newlib/libc/include/machine/setjmp.h
b/newlib/libc/include/machine/setjmp.h
index 78a81b5d9..55152b0d7 100644
--- a/newlib/libc/include/machine/setjmp.h
+++ b/newlib/libc/include/machine/setjmp.h
@@ -238,7 +238,7 @@ _BEGIN_STD_C
 #endif

 #ifdef __moxie__
-#define _JBLEN 19
+#define _JBLEN 10
 #endif

 #ifdef __CRX__
diff --git a/newlib/libc/machine/moxie/setjmp.S
b/newlib/libc/machine/moxie/setjmp.S
index ed261d54d..014368d64 100644
--- a/newlib/libc/machine/moxie/setjmp.S
+++ b/newlib/libc/machine/moxie/setjmp.S
@@ -14,53 +14,34 @@
 # setjmp/longjmp for moxie.  The jmpbuf looks like this:
 #
 # Register        jmpbuf offset
-#   $r0             0x00
-#   $r1             0x04
-#   $r2             0x08
-#   $r3             0x0c
-#   $r4             0x10
-#   $r5             0x14
-#   $r6             0x18
-#   $r7             0x1c
-#   $r8             0x20
-#   $r9             0x24
-#   $r10            0x28
-#   $r11            0x2c
-#   $r12            0x30
-#   $r13            0x34
-#   $fp             0x38
-#   $sp             0x3c
-#   stack frame fp  0x40
-#   stack frame ra  0x44
-#   stack frame sc  0x48
+#   $r6             0x00
+#   $r7             0x04
+#   $r8             0x08
+#   $r9             0x0c
+#   $r10            0x10
+#   $fp             0x14
+#   $sp             0x18
+#   stack frame fp  0x1c
+#   stack frame ra  0x20
+#   stack frame sc  0x25

         .text
         .global setjmp
         .type   setjmp,@function
 setjmp:
- st.l ($r0),     $r0
- sto.l 0x04($r0), $r1
- sto.l 0x08($r0), $r2
- sto.l 0x0c($r0), $r3
- sto.l 0x10($r0), $r4
- sto.l 0x14($r0), $r5
- sto.l 0x18($r0), $r6
- sto.l 0x1c($r0), $r7
- sto.l 0x20($r0), $r8
- sto.l 0x24($r0), $r9
- sto.l 0x28($r0), $r10
- sto.l 0x2c($r0), $r11
- sto.l 0x30($r0), $r12
- sto.l 0x34($r0), $r13
- sto.l 0x38($r0), $sp
- sto.l 0x3c($r0), $fp
- ldo.l   $r1,       0x0($fp)
- sto.l   0x40($r0), $r1
+ st.l ($r0),     $r6
+ sto.l 0x04($r0), $r7
+ sto.l 0x08($r0), $r8
+ sto.l 0x0c($r0), $r9
+ sto.l 0x10($r0), $r10
+ sto.l 0x14($r0), $sp
+ sto.l 0x18($r0), $fp
+ ldo.l   $r1,       0x00($fp)
+ sto.l   0x1c($r0), $r1
  ldo.l   $r1,       0x04($fp)
- sto.l   0x44($r0), $r1
+ sto.l   0x20($r0), $r1
  ldo.l   $r1,       0x08($fp)
- sto.l   0x48($r0), $r1
- ldo.l $r1,       0x04($r0)
+ sto.l   0x24($r0), $r1
  xor $r0, $r0
  ret
 .Lend1:
@@ -69,56 +50,27 @@ setjmp:
  .global longjmp
  .type longjmp,@function
 longjmp:
- ldi.l $r2, 0x00
- cmp $r1, $r2
- beq .Lreturn1
+ ldo.l $r6, 0x00($r0)
+ ldo.l $r7, 0x04($r0)
+ ldo.l $r8, 0x08($r0)
+ ldo.l $r9, 0x0c($r0)
+ ldo.l $r10, 0x10($r0)
+ ldo.l $sp, 0x14($r0)
+ ldo.l $fp, 0x18($r0)
+ ldo.l   $r2, 0x1c($r0)
+ sto.l   0x0($fp), $r2
+ ldo.l   $r2, 0x20($r0)
+ sto.l   0x4($fp), $r2
+ ldo.l   $r2, 0x24($r0)
+ sto.l   0x8($fp), $r2
  ldo.l $r2, 0x08($r0)
- ldo.l $r3, 0x0c($r0)
- ldo.l $r4, 0x10($r0)
- ldo.l $r5, 0x14($r0)
- ldo.l $r6, 0x18($r0)
- ldo.l $r7, 0x1c($r0)
- ldo.l $r8, 0x20($r0)
- ldo.l $r9, 0x24($r0)
- ldo.l $r10, 0x28($r0)
- ldo.l $r11, 0x2c($r0)
- ldo.l $r12, 0x30($r0)
- ldo.l $r13, 0x34($r0)
- ldo.l $sp, 0x38($r0)
- ldo.l $fp, 0x3c($r0)
- ldo.l   $r1, 0x40($r0)
- sto.l   0x0($fp), $r1
- ldo.l   $r1, 0x44($r0)
- sto.l   0x4($fp), $r1
- ldo.l   $r1, 0x48($r0)
- sto.l   0x8($fp), $r1
- ldo.l $r1, 0x04($r0)
  mov $r0, $r1
+ xor $r2, $r2
+ cmp $r0, $r2
+ beq .Lreturn1
  ret
 .Lreturn1:
- ldo.l $r1, 0x04($r0)
- ldo.l $r2, 0x08($r0)
- ldo.l $r3, 0x0c($r0)
- ldo.l $r4, 0x10($r0)
- ldo.l $r5, 0x14($r0)
- ldo.l $r6, 0x18($r0)
- ldo.l $r7, 0x1c($r0)
- ldo.l $r8, 0x20($r0)
- ldo.l $r9, 0x24($r0)
- ldo.l $r10, 0x28($r0)
- ldo.l $r11, 0x2c($r0)
- ldo.l $r12, 0x30($r0)
- ldo.l $r13, 0x34($r0)
- ldo.l $sp, 0x38($r0)
- ldo.l $fp, 0x3c($r0)
- ldo.l   $r1, 0x40($r0)
- sto.l   0x0($fp), $r1
- ldo.l   $r1, 0x44($r0)
- sto.l   0x4($fp), $r1
- ldo.l   $r1, 0x48($r0)
- sto.l   0x8($fp), $r1
- ldo.l $r1, 0x04($r0)
- ldi.l $r0, 0x01
+ inc     $r0, 1
  ret
 .Lend2:
  .size longjmp,.Lend2-longjmp