[llvm] 8fb3f84 - BPF: Workaround InstCombine trunc+icmp => mask+icmp Optimization

[Yonghong Song via llvm-commits]

Author: Yonghong Song
Date: 2021-11-18T20:25:28-08:00
New Revision: 8fb3f84484094230d563c1e4c1088755b7cde15b

URL: https://github.com/llvm/llvm-project/commit/8fb3f84484094230d563c1e4c1088755b7cde15b
DIFF: https://github.com/llvm/llvm-project/commit/8fb3f84484094230d563c1e4c1088755b7cde15b.diff

LOG: BPF: Workaround InstCombine trunc+icmp => mask+icmp Optimization

Patch [1] added further InstCombine trunc+icmp => mask+icmp
optimization and this caused a couple of bpf selftest failure.
Previous llvm BPF backend patch [2] introduced llvm.bpf.compare
builtin to handle such situations.

This patch further added support ">" and ">=" icmp opcodes.
Tested with bpf selftests and all tests are passed including two
previously failed ones.

Note Patch [1] also added optimization if the to-be-compared
constant is negative-power-of-2 (-C) or not-of-power-of-2 (~C).
This patch didn't implement these two cases as typical bpf
program compares a scalar to a positive length or boundary value,
and this scalar later is used as a index into an array buffer
or packet buffer.

  [1] https://reviews.llvm.org/D112634
  [2] https://reviews.llvm.org/D112938

Differential Revision: https://reviews.llvm.org/D114215

Added: 
    llvm/test/CodeGen/BPF/adjust-opt-icmp4.ll

Modified: 
    llvm/lib/Target/BPF/BPFAdjustOpt.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/BPF/BPFAdjustOpt.cpp b/llvm/lib/Target/BPF/BPFAdjustOpt.cpp
index 64bedf4da1ddc..69d0bca0bd772 100644
--- a/llvm/lib/Target/BPF/BPFAdjustOpt.cpp
+++ b/llvm/lib/Target/BPF/BPFAdjustOpt.cpp
@@ -131,10 +131,10 @@ bool BPFAdjustOptImpl::adjustICmpToBuiltin() {
 
         auto ConstOp1Val = ConstOp1->getValue().getZExtValue();
         auto Op = Icmp->getPredicate();
-        if (Op == ICmpInst::ICMP_ULT) {
+        if (Op == ICmpInst::ICMP_ULT || Op == ICmpInst::ICMP_UGE) {
           if ((ConstOp1Val - 1) & ConstOp1Val)
             continue;
-        } else if (Op == ICmpInst::ICMP_ULE) {
+        } else if (Op == ICmpInst::ICMP_ULE || Op == ICmpInst::ICMP_UGT) {
           if (ConstOp1Val & (ConstOp1Val + 1))
             continue;
         } else {

diff  --git a/llvm/test/CodeGen/BPF/adjust-opt-icmp4.ll b/llvm/test/CodeGen/BPF/adjust-opt-icmp4.ll
new file mode 100644
index 0000000000000..efdab8c4c70fe
--- /dev/null
+++ b/llvm/test/CodeGen/BPF/adjust-opt-icmp4.ll
@@ -0,0 +1,85 @@
+; RUN: opt -O2 -S -mtriple=bpf-pc-linux %s -o %t1
+; RUN: llc %t1 -o - | FileCheck -check-prefixes=CHECK,CHECK-V1 %s
+; RUN: opt -O2 -S -mtriple=bpf-pc-linux %s -o %t1
+; RUN: llc %t1 -mcpu=v3 -o - | FileCheck -check-prefixes=CHECK,CHECK-V3 %s
+;
+; Source:
+;   int test1(unsigned long a) {
+;     if ((unsigned)a > 3) return 2;
+;     return 3;
+;   }
+;   int test2(unsigned long a) {
+;     if ((unsigned)a >= 4) return 2;
+;     return 3;
+;   }
+; Compilation flag:
+;   clang -target bpf -O2 -S -emit-llvm -Xclang -disable-llvm-passes test.c
+
+; Function Attrs: nounwind
+define dso_local i32 @test1(i64 %a) #0 {
+entry:
+  %retval = alloca i32, align 4
+  %a.addr = alloca i64, align 8
+  store i64 %a, i64* %a.addr, align 8, !tbaa !3
+  %0 = load i64, i64* %a.addr, align 8, !tbaa !3
+  %conv = trunc i64 %0 to i32
+  %cmp = icmp ugt i32 %conv, 3
+  br i1 %cmp, label %if.then, label %if.end
+
+if.then:                                          ; preds = %entry
+  store i32 2, i32* %retval, align 4
+  br label %return
+
+if.end:                                           ; preds = %entry
+  store i32 3, i32* %retval, align 4
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %1 = load i32, i32* %retval, align 4
+  ret i32 %1
+}
+
+; CHECK-LABEL: test1
+; CHECK-V1:    if r[[#]] > 3 goto
+; CHECK-V3:    if w[[#]] > 3 goto
+
+; Function Attrs: nounwind
+define dso_local i32 @test2(i64 %a) #0 {
+entry:
+  %retval = alloca i32, align 4
+  %a.addr = alloca i64, align 8
+  store i64 %a, i64* %a.addr, align 8, !tbaa !3
+  %0 = load i64, i64* %a.addr, align 8, !tbaa !3
+  %conv = trunc i64 %0 to i32
+  %cmp = icmp uge i32 %conv, 4
+  br i1 %cmp, label %if.then, label %if.end
+
+if.then:                                          ; preds = %entry
+  store i32 2, i32* %retval, align 4
+  br label %return
+
+if.end:                                           ; preds = %entry
+  store i32 3, i32* %retval, align 4
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %1 = load i32, i32* %retval, align 4
+  ret i32 %1
+}
+
+; CHECK-LABEL: test2
+; CHECK-V1:    if r[[#]] > 3 goto
+; CHECK-V3:    if w[[#]] > 3 goto
+
+attributes #0 = { nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+
+!llvm.module.flags = !{!0, !1}
+!llvm.ident = !{!2}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 7, !"frame-pointer", i32 2}
+!2 = !{!"clang version 14.0.0 (https://github.com/llvm/llvm-project.git 930ccf0191b4a33332d924522e5676fff583f083)"}
+!3 = !{!4, !4, i64 0}
+!4 = !{!"long", !5, i64 0}
+!5 = !{!"omnipotent char", !6, i64 0}
+!6 = !{!"Simple C/C++ TBAA"}