[PATCH] D81408: [builtins] Improve compatibility with 16 bit targets

Anatoly Trosinenko via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon Jun 8 09:22:41 PDT 2020


atrosinenko created this revision.
atrosinenko added reviewers: aykevl, MaskRay, efriedma, uabelho, howard.hinnant.
Herald added a project: Sanitizers.
Herald added a subscriber: Sanitizers.

Some parts of existing codebase assume the default `int` type to be (at least) 32 bit wide. On 16 bit targets such as MSP430 this may cause Undefined Behavior or results being defined but incorrect.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D81408

Files:
  compiler-rt/lib/builtins/floatdidf.c
  compiler-rt/lib/builtins/floatundidf.c
  compiler-rt/lib/builtins/fp_lib.h


Index: compiler-rt/lib/builtins/fp_lib.h
===================================================================
--- compiler-rt/lib/builtins/fp_lib.h
+++ compiler-rt/lib/builtins/fp_lib.h
@@ -46,7 +46,7 @@
 #define REP_C UINT32_C
 #define significandBits 23
 
-static __inline int rep_clz(rep_t a) { return __builtin_clz(a); }
+static __inline int rep_clz(rep_t a) { return clzsi(a); }
 
 // 32x32 --> 64 bit multiply
 static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) {
Index: compiler-rt/lib/builtins/floatundidf.c
===================================================================
--- compiler-rt/lib/builtins/floatundidf.c
+++ compiler-rt/lib/builtins/floatundidf.c
@@ -90,7 +90,7 @@
     // a is now rounded to DBL_MANT_DIG bits
   }
   double_bits fb;
-  fb.u.s.high = ((e + 1023) << 20) |              // exponent
+  fb.u.s.high = ((su_int)(e + 1023) << 20) |      // exponent
                 ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high
   fb.u.s.low = (su_int)a;                         // mantissa-low
   return fb.f;
Index: compiler-rt/lib/builtins/floatdidf.c
===================================================================
--- compiler-rt/lib/builtins/floatdidf.c
+++ compiler-rt/lib/builtins/floatdidf.c
@@ -87,7 +87,7 @@
   }
   double_bits fb;
   fb.u.s.high = ((su_int)s & 0x80000000) |        // sign
-                ((e + 1023) << 20) |              // exponent
+                ((su_int)(e + 1023) << 20) |      // exponent
                 ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high
   fb.u.s.low = (su_int)a;                         // mantissa-low
   return fb.f;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D81408.269263.patch
Type: text/x-patch
Size: 1629 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20200608/3957de18/attachment-0001.bin>


More information about the cfe-commits mailing list