diff options
| -rw-r--r-- | target-arm/neon_helper.c | 38 |
1 files changed, 21 insertions, 17 deletions
diff --git a/target-arm/neon_helper.c b/target-arm/neon_helper.c index c3ac96a099..7df925ad31 100644 --- a/target-arm/neon_helper.c +++ b/target-arm/neon_helper.c @@ -1514,9 +1514,13 @@ uint64_t HELPER(neon_addl_saturate_s64)(uint64_t a, uint64_t b) return result; } -#define DO_ABD(dest, x, y, type) do { \ - type tmp_x = x; \ - type tmp_y = y; \ +/* We have to do the arithmetic in a larger type than + * the input type, because for example with a signed 32 bit + * op the absolute difference can overflow a signed 32 bit value. + */ +#define DO_ABD(dest, x, y, intype, arithtype) do { \ + arithtype tmp_x = (intype)(x); \ + arithtype tmp_y = (intype)(y); \ dest = ((tmp_x > tmp_y) ? tmp_x - tmp_y : tmp_y - tmp_x); \ } while(0) @@ -1524,12 +1528,12 @@ uint64_t HELPER(neon_abdl_u16)(uint32_t a, uint32_t b) { uint64_t tmp; uint64_t result; - DO_ABD(result, a, b, uint8_t); - DO_ABD(tmp, a >> 8, b >> 8, uint8_t); + DO_ABD(result, a, b, uint8_t, uint32_t); + DO_ABD(tmp, a >> 8, b >> 8, uint8_t, uint32_t); result |= tmp << 16; - DO_ABD(tmp, a >> 16, b >> 16, uint8_t); + DO_ABD(tmp, a >> 16, b >> 16, uint8_t, uint32_t); result |= tmp << 32; - DO_ABD(tmp, a >> 24, b >> 24, uint8_t); + DO_ABD(tmp, a >> 24, b >> 24, uint8_t, uint32_t); result |= tmp << 48; return result; } @@ -1538,12 +1542,12 @@ uint64_t HELPER(neon_abdl_s16)(uint32_t a, uint32_t b) { uint64_t tmp; uint64_t result; - DO_ABD(result, a, b, int8_t); - DO_ABD(tmp, a >> 8, b >> 8, int8_t); + DO_ABD(result, a, b, int8_t, int32_t); + DO_ABD(tmp, a >> 8, b >> 8, int8_t, int32_t); result |= tmp << 16; - DO_ABD(tmp, a >> 16, b >> 16, int8_t); + DO_ABD(tmp, a >> 16, b >> 16, int8_t, int32_t); result |= tmp << 32; - DO_ABD(tmp, a >> 24, b >> 24, int8_t); + DO_ABD(tmp, a >> 24, b >> 24, int8_t, int32_t); result |= tmp << 48; return result; } @@ -1552,8 +1556,8 @@ uint64_t HELPER(neon_abdl_u32)(uint32_t a, uint32_t b) { uint64_t tmp; uint64_t result; - DO_ABD(result, a, b, uint16_t); - DO_ABD(tmp, a >> 16, b >> 16, uint16_t); + DO_ABD(result, a, b, uint16_t, uint32_t); + DO_ABD(tmp, a >> 16, b >> 16, uint16_t, uint32_t); return result | (tmp << 32); } @@ -1561,22 +1565,22 @@ uint64_t HELPER(neon_abdl_s32)(uint32_t a, uint32_t b) { uint64_t tmp; uint64_t result; - DO_ABD(result, a, b, int16_t); - DO_ABD(tmp, a >> 16, b >> 16, int16_t); + DO_ABD(result, a, b, int16_t, int32_t); + DO_ABD(tmp, a >> 16, b >> 16, int16_t, int32_t); return result | (tmp << 32); } uint64_t HELPER(neon_abdl_u64)(uint32_t a, uint32_t b) { uint64_t result; - DO_ABD(result, a, b, uint32_t); + DO_ABD(result, a, b, uint32_t, uint64_t); return result; } uint64_t HELPER(neon_abdl_s64)(uint32_t a, uint32_t b) { uint64_t result; - DO_ABD(result, a, b, int32_t); + DO_ABD(result, a, b, int32_t, int64_t); return result; } #undef DO_ABD |