Float precision difference in arm64 vs x86_64

[sanikolaev]

Float precision difference in arm64 vs x86_64

[sanikolaev]

MRE

Not sure it's the only issue though:

x86_64 (dev2):

mysql> DROP TABLE IF EXISTS test; CREATE TABLE test (flt float); INSERT INTO test (id, flt) VALUES ( 1, 1.1); select flt*flt+flt from test;
--------------
DROP TABLE IF EXISTS test
--------------

Query OK, 0 rows affected (0.00 sec)

--------------
CREATE TABLE test (flt float)
--------------

Query OK, 0 rows affected (0.00 sec)

--------------
INSERT INTO test (id, flt) VALUES ( 1, 1.1)
--------------

Query OK, 1 row affected (0.00 sec)

--------------
select flt*flt+flt from test
--------------

+-------------+
| flt*flt+flt |
+-------------+
|    2.310000 |
+-------------+
1 row in set (0.00 sec)
--- 1 out of 1 results in 0ms ---

arm64 (Sergey's mac):

mysql> DROP TABLE IF EXISTS test; CREATE TABLE test (flt float); INSERT INTO test (id, flt) VALUES ( 1, 1.1); select flt*flt+flt from test;
--------------
DROP TABLE IF EXISTS test
--------------

Query OK, 0 rows affected (0.00 sec)

--------------
CREATE TABLE test (flt float)
--------------

Query OK, 0 rows affected (0.00 sec)

--------------
INSERT INTO test (id, flt) VALUES ( 1, 1.1)
--------------

Query OK, 1 row affected (0.00 sec)

--------------
select flt*flt+flt from test
--------------

+-------------+
| flt*flt+flt |
+-------------+
|  2.31000018 |
+-------------+
1 row in set (0.00 sec)
--- 1 out of 1 results in 0ms ---

Expected: same float values returned in both cases.

[sanikolaev]

Note, the following query returns the same value on arm64 and x86_64:

mysql> select 1.1*1.1+1.1;
--------------
select 1.1*1.1+1.1
--------------

+-------------+
| 1.1*1.1+1.1 |
+-------------+
|    2.310000 |
+-------------+
1 row in set (0.04 sec)