From: Pan Li <[email protected]>
There are 3 kinds of widen_mul during the unsigned SAT_MUL pattern, aka
* widen_mul directly, like _3 w* _4
* convert and the widen_mul, like (uint64_t)_3 *w (uint64_t)_4
* convert and then mul, like (uint64_t)_3 * (uint64_t)_4
All of them will be referenced during different forms of unsigned
SAT_MUL pattern match, but actually we can wrap them into a helper
which present the "widening_mul" sematics. With this helper, some
unnecessary pattern and duplicated code could be eliminated. Like
min based pattern, this patch focus on bit_ior based pattern.
The below test suites are passed for this patch:
1. The rv64gcv fully regression tests.
2. The x86 bootstrap tests.
3. The x86 fully regression tests.
gcc/ChangeLog:
* match.pd: Leverage usmul_widen_mult by bit_ior based
unsigned SAT_MUL pattern.
Signed-off-by: Pan Li <[email protected]>
---
gcc/match.pd | 44 +++++++++-----------------------------------
1 file changed, 9 insertions(+), 35 deletions(-)
diff --git a/gcc/match.pd b/gcc/match.pd
index 0ea86d97416..3cd9ab1e9b0 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -3726,53 +3726,27 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
bool c2_is_type_precision_p = c2 == prec;
}
(if (widen_prec > prec && c2_is_type_precision_p && c4_is_max_p)))))
- (for mult_op (mult widen_mult)
- (match (unsigned_integer_sat_mul @0 @1)
- /* SAT_U_MUL (X, Y) = {
- WT x = (WT)a * (WT)b;
- NT hi = x >> (sizeof(NT) * 8);
- NT lo = (NT)x;
- return lo | -!!hi;
- } while WT is uint128_t, uint64_t, uint32_t, uint16_t,
- and T is uint64_t, uint32_t, uint16_t, uint8_t. */
- (convert1?
- (bit_ior
- (convert?
- (negate
- (convert (ne (convert2? (rshift @3 INTEGER_CST@2)) integer_zerop))))
- (convert (mult_op:c@3 (convert@4 @0) (convert@5 @1)))))
- (if (types_match (type, @0, @1))
- (with
- {
- unsigned prec = TYPE_PRECISION (type);
- unsigned widen_prec = TYPE_PRECISION (TREE_TYPE (@3));
- unsigned cvt4_prec = TYPE_PRECISION (TREE_TYPE (@4));
- unsigned cvt5_prec = TYPE_PRECISION (TREE_TYPE (@5));
-
- bool widen_mult_p = mult_op == WIDEN_MULT_EXPR && cvt4_prec == cvt5_prec
- && widen_prec == cvt5_prec * 2;
- bool mult_p = mult_op == MULT_EXPR && cvt4_prec == cvt5_prec
- && cvt4_prec == widen_prec && widen_prec > prec;
- bool c2_is_type_precision_p = tree_to_uhwi (@2) == prec;
- }
- (if (c2_is_type_precision_p && (mult_p || widen_mult_p)))))))
(match (unsigned_integer_sat_mul @0 @1)
+ /* SAT_U_MUL (X, Y) = {
+ WT x = (WT)a * (WT)b;
+ NT hi = x >> (sizeof(NT) * 8);
+ NT lo = (NT)x;
+ return lo | -!!hi;
+ } while WT is uint128_t, uint64_t, uint32_t, uint16_t,
+ and T is uint64_t, uint32_t, uint16_t, uint8_t. */
(convert1?
(bit_ior
(convert?
(negate
(convert (ne (convert2? (rshift @3 INTEGER_CST@2)) integer_zerop))))
- (convert (widen_mult:c@3 @0 @1))))
+ (convert (usmul_widen_mult@3 @0 @1))))
(if (types_match (type, @0, @1))
(with
{
unsigned prec = TYPE_PRECISION (type);
- unsigned widen_prec = TYPE_PRECISION (TREE_TYPE (@3));
-
bool c2_is_type_precision_p = tree_to_uhwi (@2) == prec;
- bool widen_mult_p = prec * 2 == widen_prec;
}
- (if (c2_is_type_precision_p && widen_mult_p)))))
+ (if (c2_is_type_precision_p)))))
)
/* The boundary condition for case 10: IMM = 1:
--
2.43.0
