Skip to content

[AArch64] Disable FastISel/GlobalISel for ZT0 state #82768

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Feb 28, 2024

Conversation

sdesmalen-arm
Copy link
Collaborator

For __arm_new("zt0") we need to have special setup code in the prologue. For calls that don't preserve zt0, we need to emit code preserve ZT0 around the call.
This is only emitted by SelectionDAG ISel at the moment.

@llvmbot
Copy link
Member

llvmbot commented Feb 23, 2024

@llvm/pr-subscribers-backend-aarch64

Author: Sander de Smalen (sdesmalen-arm)

Changes

For __arm_new("zt0") we need to have special setup code in the prologue. For calls that don't preserve zt0, we need to emit code preserve ZT0 around the call.
This is only emitted by SelectionDAG ISel at the moment.


Full diff: https://github.com/llvm/llvm-project/pull/82768.diff

4 Files Affected:

  • (modified) llvm/lib/Target/AArch64/AArch64FastISel.cpp (+2-1)
  • (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+2-1)
  • (modified) llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp (+2-1)
  • (modified) llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll (+63-2)
diff --git a/llvm/lib/Target/AArch64/AArch64FastISel.cpp b/llvm/lib/Target/AArch64/AArch64FastISel.cpp
index 635beeed0df833..49bcab588e5204 100644
--- a/llvm/lib/Target/AArch64/AArch64FastISel.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FastISel.cpp
@@ -5179,7 +5179,8 @@ FastISel *AArch64::createFastISel(FunctionLoweringInfo &FuncInfo,
                                         const TargetLibraryInfo *LibInfo) {
 
   SMEAttrs CallerAttrs(*FuncInfo.Fn);
-  if (CallerAttrs.hasZAState() || CallerAttrs.hasStreamingInterfaceOrBody() ||
+  if (CallerAttrs.hasZAState() || CallerAttrs.hasZT0State() ||
+      CallerAttrs.hasStreamingInterfaceOrBody() ||
       CallerAttrs.hasStreamingCompatibleInterface())
     return nullptr;
   return new AArch64FastISel(FuncInfo, LibInfo);
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 3b92e95d7c2876..b86661c8a4e0d4 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -25892,7 +25892,8 @@ bool AArch64TargetLowering::fallBackToDAGISel(const Instruction &Inst) const {
     auto CallerAttrs = SMEAttrs(*Inst.getFunction());
     auto CalleeAttrs = SMEAttrs(*Base);
     if (CallerAttrs.requiresSMChange(CalleeAttrs) ||
-        CallerAttrs.requiresLazySave(CalleeAttrs))
+        CallerAttrs.requiresLazySave(CalleeAttrs) ||
+        CallerAttrs.requiresPreservingZT0(CalleeAttrs))
       return true;
   }
   return false;
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp b/llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp
index 3dc3d31a34e84e..26dbad713594aa 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp
@@ -535,7 +535,8 @@ bool AArch64CallLowering::fallBackToDAGISel(const MachineFunction &MF) const {
   }
 
   SMEAttrs Attrs(F);
-  if (Attrs.hasZAState() || Attrs.hasStreamingInterfaceOrBody() ||
+  if (Attrs.hasZAState() || Attrs.hasZT0State() ||
+      Attrs.hasStreamingInterfaceOrBody() ||
       Attrs.hasStreamingCompatibleInterface())
     return true;
 
diff --git a/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll b/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
index 2a78012045ff42..5c4c5a10447311 100644
--- a/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
+++ b/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -fast-isel=true -global-isel=false -fast-isel-abort=0 -mtriple=aarch64-linux-gnu -mattr=+sme < %s \
+; RUN: llc -fast-isel=true -global-isel=false -fast-isel-abort=0 -mtriple=aarch64-linux-gnu -mattr=+sme2 < %s \
 ; RUN:     | FileCheck %s --check-prefixes=CHECK-COMMON,CHECK-FISEL
-; RUN: llc -fast-isel=false -global-isel=true -global-isel-abort=0 -mtriple=aarch64-linux-gnu -mattr=+sme < %s \
+; RUN: llc -fast-isel=false -global-isel=true -global-isel-abort=0 -mtriple=aarch64-linux-gnu -mattr=+sme2 < %s \
 ; RUN:     | FileCheck %s --check-prefixes=CHECK-COMMON,CHECK-GISEL
 
 
@@ -447,3 +447,64 @@ define float @frem_call_sm_compat(float %a, float %b) "aarch64_pstate_sm_compati
   %res = frem float %a, %b
   ret float %res
 }
+
+;
+; Check ZT0 State
+;
+
+declare double @zt0_shared_callee(double) "aarch64_inout_zt0"
+
+define double  @zt0_new_caller_to_zt0_shared_callee(double %x) nounwind noinline optnone "aarch64_new_zt0" {
+; CHECK-COMMON-LABEL: zt0_new_caller_to_zt0_shared_callee:
+; CHECK-COMMON:       // %bb.0: // %prelude
+; CHECK-COMMON-NEXT:    sub sp, sp, #80
+; CHECK-COMMON-NEXT:    str x30, [sp, #64] // 8-byte Folded Spill
+; CHECK-COMMON-NEXT:    mrs x8, TPIDR2_EL0
+; CHECK-COMMON-NEXT:    cbz x8, .LBB13_2
+; CHECK-COMMON-NEXT:    b .LBB13_1
+; CHECK-COMMON-NEXT:  .LBB13_1: // %save.za
+; CHECK-COMMON-NEXT:    mov x8, sp
+; CHECK-COMMON-NEXT:    str zt0, [x8]
+; CHECK-COMMON-NEXT:    bl __arm_tpidr2_save
+; CHECK-COMMON-NEXT:    ldr zt0, [x8]
+; CHECK-COMMON-NEXT:    msr TPIDR2_EL0, xzr
+; CHECK-COMMON-NEXT:    b .LBB13_2
+; CHECK-COMMON-NEXT:  .LBB13_2: // %entry
+; CHECK-COMMON-NEXT:    smstart za
+; CHECK-COMMON-NEXT:    zero { zt0 }
+; CHECK-COMMON-NEXT:    bl zt0_shared_callee
+; CHECK-COMMON-NEXT:    mov x8, #4631107791820423168 // =0x4045000000000000
+; CHECK-COMMON-NEXT:    fmov d1, x8
+; CHECK-COMMON-NEXT:    fadd d0, d0, d1
+; CHECK-COMMON-NEXT:    smstop za
+; CHECK-COMMON-NEXT:    ldr x30, [sp, #64] // 8-byte Folded Reload
+; CHECK-COMMON-NEXT:    add sp, sp, #80
+; CHECK-COMMON-NEXT:    ret
+entry:
+  %call = call double @zt0_shared_callee(double %x)
+  %add = fadd double %call, 4.200000e+01
+  ret double %add;
+}
+
+define double  @zt0_new_caller_to_normal_callee(double %x) nounwind noinline optnone "aarch64_inout_zt0" {
+; CHECK-COMMON-LABEL: zt0_new_caller_to_normal_callee:
+; CHECK-COMMON:       // %bb.0: // %entry
+; 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:    bl normal_callee
+; CHECK-COMMON-NEXT:    smstart za
+; CHECK-COMMON-NEXT:    ldr zt0, [x19]
+; CHECK-COMMON-NEXT:    mov x8, #4631107791820423168 // =0x4045000000000000
+; CHECK-COMMON-NEXT:    fmov d1, x8
+; CHECK-COMMON-NEXT:    fadd d0, d0, d1
+; CHECK-COMMON-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-COMMON-NEXT:    add sp, sp, #80
+; CHECK-COMMON-NEXT:    ret
+entry:
+  %call = call double @normal_callee(double %x)
+  %add = fadd double %call, 4.200000e+01
+  ret double %add;
+}

@davemgreen
Copy link
Collaborator

Is the acle_sme_zero.c failing test unrelated? This looks sensible otherwise.

For __arm_new("zt0") we need to have special setup code in the prologue.
For calls that don't preserve zt0, we need to emit code preserve ZT0 around
the call.
This is only emitted by SelectionDAG ISel at the moment.
@sdesmalen-arm
Copy link
Collaborator Author

Is the acle_sme_zero.c failing test unrelated? This looks sensible otherwise.

Thanks. Yes that failing test was unrelated. I've rebased the patch so that pre-commit will turn green.

ret double %add;
}

define double @zt0_new_caller_to_normal_callee(double %x) nounwind noinline optnone "aarch64_inout_zt0" {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should this be called zt0_shared_caller_to_normal_callee since the caller has the inout attribute?

@sdesmalen-arm sdesmalen-arm merged commit 41427b0 into llvm:main Feb 28, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants