Move result of itof double conv back to VFP reg

EABI functions to convert an int to a double register take the integer
value in core registers and also give the result in core registers.
It is thus necessary to move the result back to VFP register after the
function call. This only affected integer to double conversion because
integer to float conversion used a VFP instruction to do the conversion
and this obviously left the result in VFP register. Note that the
behavior is left untouched for !EABI as the correct behavior in this
case is unknown to the author of this patch.

2 files changed, 10 insertions, 0 deletions

diff --git a/Changelog b/Changelog
index 9a497cf..20814c1 100644
--- a/Changelog
+++ b/Changelog
@@ -65,6 +65,7 @@ Bug fixes:
 - fix NaN comparison (Thomas Preud'homme)
 - use libtcc for static linking with runtime library (Thomas Preud'homme)
 - fix negation of 0.0 and -0.0 values (Thomas Preud'homme)
+- fix integer to double conversion on ARM (Thomas Preud'homme)
 
 version 0.9.26:
 
diff --git a/arm-gen.c b/arm-gen.c
index 9611dca..c746e91 100644
--- a/arm-gen.c
+++ b/arm-gen.c
@@ -1979,8 +1979,17 @@ ST_FUNC void gen_cvt_itof1(int t)
       vpush_global_sym(func_type, func);
       vswap();
       gfunc_call(1);
+#if defined(TCC_ARM_VFP) && defined(TCC_ARM_EABI)
+      r=get_reg(RC_FLOAT);
+      r2=vfpr(r);
+      o(0xEE000B10|(r2<<16)); /* vmov.32 dr2[0], r0 */
+      o(0xEE201B10|(r2<<16)); /* vmov.32 dr2[1], r1 */
+      vpushi(0);
+      vtop->r=r;
+#else
       vpushi(0);
       vtop->r=TREG_F0;
+#endif
       return;
     }
   }