RFR: 8227222 : vmTestbase/jit/FloatingPoint/gen_math/Loops04/Loops04.java hits assert
Bhateja, Jatin
jatin.bhateja at intel.com
Tue Jul 9 03:54:34 UTC 2019
Hi,
In max/min intrinsification for floating point types [1] for reduction pattern we use vand/vor instructions operating over 32/64 bits operands .
Crash dump for the test case shows that the target is AVX512 but does not support AVX512VL feature.
RA constraints the allocation set of an operand by using the register masks associated with operand's register classes. Thus using legacy register classes in the instruction patterns for reduction case will guide allocator to assign register from lower register bank (0-15).
Please find below a patch for the same.
http://cr.openjdk.java.net/~jbhateja/8227222/webrev.00/
Thanks,
Jatin
[1] https://bugs.openjdk.java.net/browse/JDK-8217561
More information about the hotspot-compiler-dev
mailing list