Errors in CRC-32C accleration on Z

[Spencer-Comin]

While benchmarking (see benchmark below) the changes at #16379, I found a ~2.5% error rate in my 15B testcase. Out of 200 tests, I got 3 NullPointerExceptions and 2 segfaults. Looking into the coredumps produced @r30shah and I were able to isolate the cause of the segfaults (both segfaults failed identically). I've annotated the disassembly to illustrate the problem:

l         %r10, 8(%r1)                  ; Incoming checksum value is loaded into %r10
cgite     %r8, 0
sllg      %r2, %r8, 3 
lgr       %r3, %r2
xr        %r4, %r4
l         %r0, 4(%r2) 
llgfr     %r4, %r4
llgfr     %r0, %r0
la        %r8, 0x10(%r4, %r3)
sgrk      %r9, %r0, %r4
clgijl    %r9, 0x10, 0x3ff64b91c00      ; (1) Jump to OOL call to original Java impl if less than 16B of data
larl      %r14, 0x3ff64b91d80
vlm       %v9, %v14, 0x80(%r14)
vzero     %v0
vlvgf     %v0, %r10, 3
stg       %r10, 0xb0(%r5)               ; %r10 is spilled here
lg        %r10, 0xa8(%r5)               ; If jump at (1) is taken, this spilled is missed, and an incorrect value is left in %r10 

...                                     ; rest of CRC-32C acceleration impl

clgijlh   %r9, 0, 0x3ff64b91c00         ; Jump to OOL call to original Java impl if remaining data  
st        %r8, 8(%r1)  <<< ^+3678       ; OOL returns to here
l         %r8, 0x20c(%r10)              ; Segfault occurs dereferencing %r10 here

It looks like it should be a matter of fixing some register dependencies and internal control flow to prevent the problematic spill from happening where it does. Hopefully resolving the segfaults will also resolve the NullPointerExceptions as well.

Here is the benchmark I used:

import org.openjdk.jmh.annotations.*;

import java.util.zip.CRC32C;
import java.util.Random;

@Fork(value = 1)
@Warmup(iterations = 3)
@Measurement(iterations = 10)
public class CRC32CBenchmark {
    @State(Scope.Thread)
    public static class TestValues {
        public CRC32C crc;
        public byte[] array;

        @Param({"15", "48", "4096", "16384"})
        public int arraySize;

        @Setup(Level.Trial)
        public void setup() {
            crc = new CRC32C();
            Random rng = new Random(0xdeadbeef);

            array = new byte[arraySize];
            rng.nextBytes(array);
        }
    }

    @Benchmark
    public void updateCRC32C(TestValues tv) {
        tv.crc.update(tv.array);
    }
}

[Spencer-Comin]

Here is the full jitdump from the segfault corresponding to the disassembly above: jitdump.20230113.134230.57555.0004.txt