[OMPT][test] Fix mismatch error between the current address and the return address for LoongArch

[Ami-zhang]

Summary

This patch fixes tests masked.c and master.c on LoongArch.

[protze.joachim]

Your change only allows offsets of 8 and 12 byte. But the comment indicates that 4 byte offset might still be possible? I think, you should just print the 12 byte offset in addition.

[Ami-zhang]

On LoongArch64, the following failures happened when i perform check-openmp:

            4: 0: ompt_event_master_end: codeptr_ra=0x555556134c40
check:89'0                                                        X error: no match found
check:89'1                                                          with "MASTER_ID" equal to "0"
check:89'2                                                          with "RETURN_ADDRESS_END" equal to "0x555556134c40"
            5: 0: current_address=0x555556134c48 or 0x555556134c44

The macro print_current_address(id) prints the exact address that a previously called runtime function returns to.
As the above shown, the values of curren_address can not match the value of codeptr_ra (that is return address ) .
Disassemble it and generate the following codes:

0000000000000b98 <.omp_outlined.>:
...
 c3c:   57fc97ff        bl              -876(0xffffc94) # 8d0 <__kmpc_end_master@plt>
 c40:   50000400        b               4(0x4)  # c44 <.omp_outlined.+0xac>
 c44:   03400000        andi            $zero, $zero, 0x0
 c48:   50000400        b               4(0x4)  # c4c <.omp_outlined.+0xb4>
 c4c:   1a000104        pcalau12i       $a0, 8(0x8)
 c50:   02c2c084        addi.d          $a0, $a0, 176(0xb0)

And here the value of codeptr_ra is c40, the value of current_address is c48 or c44 caused by previously ((char *)addr) - 4 or ((char *)addr) - 8) wrong calculation.
But in D132925 which follows D59880 for RISCV64, why doesn't the current_address follow ((char *)addr) - 8 or ((char *)addr) - 12 in RISCV64? @prcups

[protze.joachim]

I did not argue against adding ((char *)addr) - 12. But following the description in the comment, I think, ((char *)addr) - 4 should stay there, as this might still be valid for other cases (e.g., compiling with other optimization constraints?).

[Ami-zhang]

In D138785#3956297, @protze.joachim wrote:

I did not argue against adding ((char *)addr) - 12. But following the description in the comment, I think, ((char *)addr) - 4 should stay there, as this might still be valid for other cases (e.g., compiling with other optimization constraints?).

Yes, when compiling with optimization, ((char *)addr) - 4 is valid.
I have added it.
Thank you.

[Ami-zhang]

Addressed @protze.joachim's comment.

[xen0n]

Looks good from LoongArch side. Thanks!

[SixWeining]

I just added for LoongArch to the title since this change is LoongArch specific.

[Ami-zhang]

In D138785#3962878, @SixWeining wrote:

I just added for LoongArch to the title since this change is LoongArch specific.

Thank you.

[SixWeining]

LGTM from the LoongArch side.

[protze.joachim]

lgtm

openmp/runtime/test/ompt/callback.h

 // appear (adding 4 more bytes) when the NOP is referenced elsewhere (ie.
 // another branch).
 #define print_possible_return_addresses(addr) \
   printf("%" PRIu64 ": current_address=%p or %p\n", \
          ompt_get_thread_data()->value, ((char *)addr) - 8, ((char *)addr) - 12)
 #endif
 #elif KMP_ARCH_LOONGARCH64
 // On LoongArch64 the NOP instruction is 4 bytes long, can be followed by
-// inserted jump instruction (another 4 bytes long).
+// inserted jump instruction (another 4 bytes long). And an additional jump
+// instruction may appear (adding 4 more bytes) when the NOP is referenced
+// elsewhere (ie. another branch).
 #define print_possible_return_addresses(addr)                                  \
-  printf("%" PRIu64 ": current_address=%p or %p\n",                            \
+  printf("%" PRIu64 ": current_address=%p or %p or %p\n",                      \
          ompt_get_thread_data()->value, ((char *)addr) - 4,                    \
-         ((char *)addr) - 8)
+         ((char *)addr) - 8, ((char *)addr) - 12)
 #else
 #error Unsupported target architecture, cannot determine address offset!
 #endif
 
 
 // This macro performs a somewhat similar job to print_current_address(), except
 // that it discards a certain number of nibbles from the address and only prints
 // the most significant bits / nibbles. This can be used for cases where the