[llvm] [AIX][TLS] Optimize the small local-exec access sequence for non-zero offsets (PR #71485)

[via llvm-commits]

================
@@ -25,43 +25,33 @@ declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull) #1
 define signext i32 @StoreArrays1() {
 ; SMALL-LOCAL-EXEC-SMALLCM64-LABEL: StoreArrays1:
 ; SMALL-LOCAL-EXEC-SMALLCM64:       # %bb.0: # %entry
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    la r3, mySmallLocalExecTLSv1[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r4, 1
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r5, 4
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r4, mySmallLocalExecTLSv1[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    la r4, mySmallLocalExecTLS2[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r5, 24(r3)
+; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r3, 1
+; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r4, 4
+; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r3, mySmallLocalExecTLSv1[TL]@le(r13)
 ; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r3, 2
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r3, 320(r4)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    la r3, mySmallLocalExecTLS3[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r4, 3
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r4, 324(r3)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    la r3, mySmallLocalExecTLS4[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    li r4, 88
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r5, 328(r3)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    la r3, mySmallLocalExecTLS5[TL]@le(r13)
-; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r4, 332(r3)
+; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r4, mySmallLocalExecTLSv1[TL]@le+24(r13)
+; SMALL-LOCAL-EXEC-SMALLCM64-NEXT:    stw r3, mySmallLocalExecTLS2[TL]@le-65216(r13)
----------------
stephenpeckham wrote:

I think the alternate way would look be better, because it's easier to debug or understand the assembly code. In addition, the "-65536" is required because of an AIX assembler limitation.  If the assembler were fixed, the extra subtraction could be eliminated.  But I could accept either output format.

https://github.com/llvm/llvm-project/pull/71485