https://gcc.gnu.org/g:f2794c206f112547907ed010b541146dc005d37e
commit r16-3346-gf2794c206f112547907ed010b541146dc005d37e
Author: Pan Li <pan2...@intel.com>
Date: Wed Aug 13 13:55:27 2025 +0800
Match: Add form 3 for unsigned SAT_MUL
This patch would like to try to match the the unsigned
SAT_MUL form 3, aka below:
#define DEF_SAT_U_MUL_FMT_3(NT, WT) \
NT __attribute__((noinline)) \
sat_u_mul_##NT##_from_##WT##_fmt_3 (NT a, NT b) \
{ \
WT x = (WT)a * (WT)b; \
if ((x >> sizeof(a) * 8) == 0) \
return (NT)x; \
else \
return (NT)-1; \
}
While WT is T is uint16_t, uint32_t, uint64_t and uint128_t,
and NT is is uint8_t, uint16_t, uint32_t and uint64_t.
gcc/ChangeLog:
* match.pd: Add form 3 for unsigned SAT_MUL.
Signed-off-by: Pan Li <pan2...@intel.com>
Diff:
---
gcc/match.pd | 27 ++++++++++++++++++++++++++-
1 file changed, 26 insertions(+), 1 deletion(-)
diff --git a/gcc/match.pd b/gcc/match.pd
index 66e8a7874493..b1d7a3a1b73f 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -3629,12 +3629,37 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
if (__builtin_mul_overflow (X, Y, &result))
return -1;
else
- return -(T)overflow_p | result;
+ return result;
}
while T can be uint8_t, uint16_t, uint32_t and uint64_t. */
(cond^ (ne (imagpart (IFN_MUL_OVERFLOW:c@2 @0 @1)) integer_zerop)
integer_minus_onep (realpart @2))
(if (types_match (type, @0, @1))))
+ (match (unsigned_integer_sat_mul @0 @1)
+ /* SAT_U_MUL (X, Y) = {
+ WT x = (WT)a * (WT)b;
+ if ((x >> sizeof(a) * 8) == 0)
+ return (T)x;
+ else
+ return (T)-1;
+ }
+ while WT is uint128_t, T is uint8_t, uint16_t, uint32_t or uint64_t. */
+ (convert (cond^ (eq (rshift (mult:c@3 (convert @0) (convert @1))
+ INTEGER_CST@2)
+ integer_zerop)
+ @3 INTEGER_CST@4))
+ (if (types_match (type, @0, @1))
+ (with
+ {
+ unsigned prec = TYPE_PRECISION (type);
+ unsigned widen_prec = TYPE_PRECISION (TREE_TYPE (@3));
+ wide_int c4 = wi::to_wide (@4);
+ wide_int max = wi::mask (prec, false, widen_prec);
+ bool c4_is_max_p = wi::eq_p (c4, max);
+ unsigned c2 = tree_to_uhwi (@2);
+ bool c2_is_type_precision_p = c2 == prec;
+ }
+ (if (widen_prec > prec && c2_is_type_precision_p && c4_is_max_p)))))
)
/* The boundary condition for case 10: IMM = 1:
