/* { dg-require-effective-target arm_fp16_ok } */
/* { dg-options "-O2" } */
/* { dg-add-options arm_fp16_ieee } */
+/* { dg-final { check-function-bodies "**" "" "" } } */
/* Test __fp16 arguments and return value in registers (hard-float). */
void
swap (__fp16, __fp16);
+/*
+** F:
+** ...
+** (
+
+Below block is for non-armv8.1
+** (
+** vmov\.f32 (s[3-9]|s1[0-5]), s0
+** ...
+** vmov\.f32 s0, s1
+** ...
+** vmov\.f32 s1, \1
+** |
+** vmov\.f32 (s[3-9]|s1[0-5]), s1
+** ...
+** vmov\.f32 s1, s0
+** ...
+** vmov\.f32 s0, \2
+** )
+** vstr\.32 s2, \[sp, #4\] @ int
+** bl swap
+** vldr\.32 s2, \[sp, #4\] @ int
+** vmov\.f32 s0, s2
+
+** |
+
+Below block is for armv8.1
+** (
+** vmov (s[3-9]|s1[0-5]), s0 @ __fp16
+** ...
+** vmov s0, s1 @ __fp16
+** ...
+** vmov s1, \3 @ __fp16
+** |
+** vmov (s[3-9]|s1[0-5]), s1 @ __fp16
+** ...
+** vmov s1, s0 @ __fp16
+** ...
+** vmov s0, \4 @ __fp16
+** )
+** vstr\.32 s2, \[sp, #4\] @ int
+** bl swap
+** vldr\.16 s0, \[sp, #4\]
+
+** )
+** ...
+*/
__fp16
F (__fp16 a, __fp16 b, __fp16 c)
{
swap (b, a);
return c;
}
-
-/* { dg-final { scan-assembler {vmov\.f32\ts[0-9]+, s1} } } */
-/* { dg-final { scan-assembler {vmov\.f32\ts1, s0} } } */
-/* { dg-final { scan-assembler {vmov\.f32\ts[0-9]+, s2+} } } */
-/* { dg-final { scan-assembler-times {vmov\.f32\ts0, s[0-9]+} 2 } } */
/* { dg-options "-mfloat-abi=softfp -O2 -mno-long-calls" } */
/* { dg-add-options arm_fp16_ieee } */
/* { dg-skip-if "incompatible float-abi" { arm*-*-* } { "-mfloat-abi=hard" } } */
+/* { dg-final { check-function-bodies "**" "" "" } } */
/* Test __fp16 arguments and return value in registers (softfp). */
void
swap (__fp16, __fp16);
+/*
+** F:
+** ...
+** (
+** mov r3, r0 @ __fp16
+** ...
+** mov r0, r1 @ __fp16
+** ...
+** mov r1, r3 @ __fp16
+** |
+** mov r3, r1 @ __fp16
+** ...
+** mov r1, r0 @ __fp16
+** ...
+** mov r0, r3 @ __fp16
+** )
+** ...
+*/
+/*
+** F: { target arm_little_endian }
+** ...
+** str r2, \[sp, #4\]
+** bl swap
+** ldrh r0, \[sp, #4\] @ __fp16
+** ...
+*/
__fp16
F (__fp16 a, __fp16 b, __fp16 c)
{
swap (b, a);
return c;
}
-
-/* The swap must include two moves out of r0/r1 and two moves in. */
-/* { dg-final { scan-assembler-times {mov\tr[0-9]+, r[01]} 2 } } */
-/* { dg-final { scan-assembler-times {mov\tr[01], r[0-9]+} 2 } } */
-/* c should be spilled around the call. */
-/* { dg-final { scan-assembler {str\tr2, ([^\n]*).*ldrh\tr0, \1} { target arm_little_endian } } } */
/* { dg-require-effective-target arm_fp16_alternative_ok } */
/* { dg-options "-O2" } */
/* { dg-add-options arm_fp16_alternative } */
+/* { dg-final { check-function-bodies "**" "" "" } } */
/* Test __fp16 arguments and return value in registers (hard-float). */
void
swap (__fp16, __fp16);
+/*
+** F:
+** ...
+** (
+
+Below block is for non-armv8.1
+** (
+** vmov\.f32 (s[3-9]|s1[0-5]), s0
+** ...
+** vmov\.f32 s0, s1
+** ...
+** vmov\.f32 s1, \1
+** |
+** vmov\.f32 (s[3-9]|s1[0-5]), s1
+** ...
+** vmov\.f32 s1, s0
+** ...
+** vmov\.f32 s0, \2
+** )
+** vstr\.32 s2, \[sp, #4\] @ int
+** bl swap
+** vldr\.32 s2, \[sp, #4\] @ int
+** vmov\.f32 s0, s2
+
+** |
+
+Below block is for armv8.1
+** (
+** vmov (s[3-9]|s1[0-5]), s0
+** ...
+** vmov s0, s1
+** ...
+** vmov s1, \3
+** |
+** vmov (s[3-9]|s1[0-5]), s1
+** ...
+** vmov s1, s0
+** ...
+** vmov s0, \4
+** )
+** vstr\.32 s2, \[sp, #4\] @ int
+** bl swap
+** vldr\.16 s0, \[sp, #4\]
+
+** )
+** ...
+*/
__fp16
F (__fp16 a, __fp16 b, __fp16 c)
{
swap (b, a);
return c;
}
-
-/* { dg-final { scan-assembler {vmov\.f32\ts[0-9]+, s1} } } */
-/* { dg-final { scan-assembler {vmov\.f32\ts1, s0} } } */
-/* { dg-final { scan-assembler {vmov\.f32\ts[0-9]+, s2+} } } */
-/* { dg-final { scan-assembler-times {vmov\.f32\ts0, s[0-9]+} 2 } } */
-
/* { dg-options "-mfloat-abi=softfp -O2" } */
/* { dg-add-options arm_fp16_alternative } */
/* { dg-skip-if "incompatible float-abi" { arm*-*-* } { "-mfloat-abi=hard" } } */
+/* { dg-final { check-function-bodies "**" "" "" } } */
/* Test __fp16 arguments and return value in registers (softfp). */
void
swap (__fp16, __fp16);
+/*
+** F:
+** ...
+** (
+** mov r3, r0 @ __fp16
+** ...
+** mov r0, r1 @ __fp16
+** ...
+** mov r1, r3 @ __fp16
+** |
+** mov r3, r1 @ __fp16
+** ...
+** mov r1, r0 @ __fp16
+** ...
+** mov r0, r3 @ __fp16
+** )
+** ...
+*/
+/*
+** F: { target arm_little_endian }
+** ...
+** str r2, \[sp, #4\]
+** bl swap
+** ldrh r0, \[sp, #4\] @ __fp16
+** ...
+*/
__fp16
F (__fp16 a, __fp16 b, __fp16 c)
{
swap (b, a);
return c;
}
-
-/* The swap must include two moves out of r0/r1 and two moves in. */
-/* { dg-final { scan-assembler-times {mov\tr[0-9]+, r[01]} 2 } } */
-/* { dg-final { scan-assembler-times {mov\tr[01], r[0-9]+} 2 } } */
-/* c should be spilled around the call. */
-/* { dg-final { scan-assembler {str\tr2, ([^\n]*).*ldrh\tr0, \1} { target arm_little_endian } } } */