[AArch64][GlobalISel] Enable memcpy family combines on minsize functions

[paquette]

Summary

The combines in tryCombineMemCpyFamily have heuristics (e.g. TLI.getMaxStoresPerMemset) which consider size. So, theoretically, enabling these combines on minsize functions shouldn't be harmful.

With this enabled we save 0.9% geomean on CTMark at -Oz, and 5.1% on Bullet. There are no code size regressions.

[aemerson]

Wow, not running this at all for -Oz? LGTM.

llvm/lib/Target/AArch64/GISel/AArch64PreLegalizerCombiner.cpp

case TargetOpcode::G_SHUFFLE_VECTOR:
     return Helper.tryCombineShuffleVector(MI);
   case TargetOpcode::G_MEMCPY:
   case TargetOpcode::G_MEMMOVE:
   case TargetOpcode::G_MEMSET: {
     // If we're at -O0 set a maxlen of 32 to inline, otherwise let the other
     // heuristics decide.
     unsigned MaxLen = EnableOpt ? 0 : 32;
     // Try to inline memcpy type calls if optimizations are enabled.
-    if (!EnableMinSize && Helper.tryCombineMemCpyFamily(MI, MaxLen))
+    if (Helper.tryCombineMemCpyFamily(MI, MaxLen))
       return true;
     if (Opc == TargetOpcode::G_MEMSET)
       return llvm::AArch64GISelUtils::tryEmitBZero(MI, B, EnableMinSize);
     return false;
   }
   }
 
   return false;

llvm/test/CodeGen/AArch64/GlobalISel/inline-memset.mir

--- |
   }
 
   define void @test_ms4_const_both_unaligned(i8* nocapture %dst) local_unnamed_addr #0 {
   entry:
     tail call void @llvm.memset.p0i8.i64(i8* align 1 %dst, i8 64, i64 18, i1 false)
     ret void
   }
 
+  define void @minsize(i8* nocapture %dst) minsize { unreachable }
+
   declare void @llvm.stackprotector(i8*, i8**) #2
 
   attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="cyclone" "target-features"="+aes,+crypto,+fp-armv8,+neon,+sha2,+zcm,+zcz" "unsafe-fp-math"="false" "use-soft-float"="false" }
   attributes #1 = { argmemonly nounwind }
 
 ...
 ---
 name:            test_ms1
bb.1.entry:
     ; CHECK: RET_ReallyLR
     %0:_(p0) = COPY $x0
     %1:_(s8) = G_CONSTANT i8 64
     %2:_(s64) = G_CONSTANT i64 18
     G_MEMSET %0(p0), %1(s8), %2(s64), 1 :: (store 1 into %ir.dst)
     RET_ReallyLR
 
 ...
+---
+name:            minsize
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $w1, $x0
+    ; CHECK-LABEL: name: minsize
+    ; CHECK: liveins: $w1, $x0
+    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $w1
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[COPY1]](s32)
+    ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[TRUNC]](s8)
+    ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 72340172838076673
+    ; CHECK: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[ZEXT]], [[C]]
+    ; CHECK: G_STORE [[MUL]](s64), [[COPY]](p0) :: (store 8 into %ir.dst, align 1)
+    ; CHECK: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+    ; CHECK: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
+    ; CHECK: G_STORE [[MUL]](s64), [[PTR_ADD]](p0) :: (store 8 into %ir.dst + 8, align 1)
+    ; CHECK: RET_ReallyLR
+    %0:_(p0) = COPY $x0
+    %1:_(s32) = COPY $w1
+    %3:_(s64) = G_CONSTANT i64 16
+    %2:_(s8) = G_TRUNC %1(s32)
+    G_MEMSET %0(p0), %2(s8), %3(s64), 1 :: (store 1 into %ir.dst)
+    RET_ReallyLR