[AArch64][SME] Disable tail calls for callees that require saving ZT0

[MacDue]

We may need to load ZT0 after the call, so we can't perform a tail call.

[llvmbot]

@llvm/pr-subscribers-backend-aarch64

Author: Benjamin Maxwell (MacDue)

Changes

We may need to load ZT0 after the call, so we can't perform a tail call.

---

Full diff: https://github.com/llvm/llvm-project/pull/165371.diff

2 Files Affected:

• (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+2-1)
• (modified) llvm/test/CodeGen/AArch64/sme-zt0-state.ll (+18)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index d16b11686e3c1..60aa61e993b26 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -9028,11 +9028,12 @@ bool AArch64TargetLowering::isEligibleForTailCallOptimization(
   CallingConv::ID CallerCC = CallerF.getCallingConv();
 
   // SME Streaming functions are not eligible for TCO as they may require
-  // the streaming mode or ZA to be restored after returning from the call.
+  // the streaming mode or ZA/ZT0 to be restored after returning from the call.
   SMECallAttrs CallAttrs =
       getSMECallAttrs(CallerF, getRuntimeLibcallsInfo(), CLI);
   if (CallAttrs.requiresSMChange() || CallAttrs.requiresLazySave() ||
       CallAttrs.requiresPreservingAllZAState() ||
+      CallAttrs.requiresPreservingZT0() ||
       CallAttrs.caller().hasStreamingBody())
     return false;
 
diff --git a/llvm/test/CodeGen/AArch64/sme-zt0-state.ll b/llvm/test/CodeGen/AArch64/sme-zt0-state.ll
index 2583a93e514a2..5b81f5dafe421 100644
--- a/llvm/test/CodeGen/AArch64/sme-zt0-state.ll
+++ b/llvm/test/CodeGen/AArch64/sme-zt0-state.ll
@@ -426,3 +426,21 @@ define void @zt0_multiple_private_za_calls(ptr %callee) "aarch64_in_zt0" nounwin
   call void %callee()
   ret void
 }
+
+define void @disable_tailcallopt(ptr %callee) "aarch64_inout_zt0" nounwind {
+; CHECK-COMMON-LABEL: disable_tailcallopt:
+; CHECK-COMMON:       // %bb.0:
+; CHECK-COMMON-NEXT:    sub sp, sp, #80
+; CHECK-COMMON-NEXT:    stp x30, x19, [sp, #64] // 16-byte Folded Spill
+; CHECK-COMMON-NEXT:    mov x19, sp
+; CHECK-COMMON-NEXT:    str zt0, [x19]
+; CHECK-COMMON-NEXT:    smstop za
+; CHECK-COMMON-NEXT:    blr x0
+; CHECK-COMMON-NEXT:    smstart za
+; CHECK-COMMON-NEXT:    ldr zt0, [x19]
+; CHECK-COMMON-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-COMMON-NEXT:    add sp, sp, #80
+; CHECK-COMMON-NEXT:    ret
+  tail call void %callee()
+  ret void
+}

[llvm-ci]

LLVM Buildbot has detected a new failure on builder llvm-clang-x86_64-darwin running on doug-worker-3 while building llvm at step 6 "test-build-unified-tree-check-all".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/23/builds/15060

Here is the relevant piece of the build log for the reference

Step 6 (test-build-unified-tree-check-all) failure: test (failure)
******************** TEST 'Clang :: Driver/x86-mtune.c' FAILED ********************
Exit Code: 127

Command Output (stdout):
--
# RUN: at line 6
/Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/clang -### -c --target=x86_64 /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c 2>&1 | /Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/FileCheck /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c -check-prefix=generic
# executed command: /Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/clang '-###' -c --target=x86_64 /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c
# note: command had no output on stdout or stderr
# executed command: /Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/FileCheck /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c -check-prefix=generic
# note: command had no output on stdout or stderr
# RUN: at line 8
/Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/clang -### -c --target=x86_64 /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c -mtune=generic 2>&1 | /Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/FileCheck /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c --check-prefixes=cpu,generic
# executed command: /Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/FileCheck /Volumes/RAMDisk/buildbot-root/x86_64-darwin/llvm-project/clang/test/Driver/x86-mtune.c --check-prefixes=cpu,generic
# .---command stderr------------
# | Could not create process (/Volumes/RAMDisk/buildbot-root/x86_64-darwin/build/bin/FileCheck) due to [Errno 24] Too many open files
# `-----------------------------
# error: command failed with exit status: 127

--

********************