[X86] Add test cases to show missed opportunities to use a sign extended 8 or 32...

When we shift the AND mask over we should shift in sign bits instead of zero bits. The scale in the LEA will shift these bits out so it doesn't matter whether we mask the bits off or not. Using sign bits will potentially allow a sign extended immediate to be used.

Also add some other test cases for cases that are currently optimal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@357845 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/fold-and-shift-x86_64.ll b/test/CodeGen/X86/fold-and-shift-x86_64.ll
new file mode 100644 (file)
index 0000000..8aa4b68
--- /dev/null
+++ b/test/CodeGen/X86/fold-and-shift-x86_64.ll
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
+
+define i8 @t1(i8* %X, i64 %i) {
+; CHECK-LABEL: t1:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movabsq $4611686018427387649, %rax # imm = 0x3FFFFFFFFFFFFF01
+; CHECK-NEXT:    andq %rsi, %rax
+; CHECK-NEXT:    movb (%rdi,%rax,4), %al
+; CHECK-NEXT:    retq
+
+entry:
+  %tmp2 = shl i64 %i, 2
+  %tmp4 = and i64 %tmp2, -1020
+  %tmp7 = getelementptr i8, i8* %X, i64 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}
+
+define i8 @t2(i8* %X, i64 %i) {
+; CHECK-LABEL: t2:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movabsq $4611686018427387890, %rax # imm = 0x3FFFFFFFFFFFFFF2
+; CHECK-NEXT:    andq %rsi, %rax
+; CHECK-NEXT:    movb (%rdi,%rax,4), %al
+; CHECK-NEXT:    retq
+
+entry:
+  %tmp2 = shl i64 %i, 2
+  %tmp4 = and i64 %tmp2, -56
+  %tmp7 = getelementptr i8, i8* %X, i64 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}
+
+define i8 @t3(i8* %X, i64 %i) {
+; CHECK-LABEL: t3:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movl %esi, %eax
+; CHECK-NEXT:    movb (%rdi,%rax,4), %al
+; CHECK-NEXT:    retq
+
+entry:
+  %tmp2 = shl i64 %i, 2
+  %tmp4 = and i64 %tmp2, 17179869180
+  %tmp7 = getelementptr i8, i8* %X, i64 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}
+
+define i8 @t4(i8* %X, i64 %i) {
+; CHECK-LABEL: t4:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    andl $-2, %esi
+; CHECK-NEXT:    movb (%rdi,%rsi,4), %al
+; CHECK-NEXT:    retq
+
+entry:
+  %tmp2 = shl i64 %i, 2
+  %tmp4 = and i64 %tmp2, 17179869176
+  %tmp7 = getelementptr i8, i8* %X, i64 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}
+
+define i8 @t5(i8* %X, i64 %i) {
+; CHECK-LABEL: t5:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    andl $-250002, %esi # imm = 0xFFFC2F6E
+; CHECK-NEXT:    movb (%rdi,%rsi,4), %al
+; CHECK-NEXT:    retq
+
+entry:
+  %tmp2 = shl i64 %i, 2
+  %tmp4 = and i64 %tmp2, 17178869176
+  %tmp7 = getelementptr i8, i8* %X, i64 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}

diff --git a/test/CodeGen/X86/fold-and-shift.ll b/test/CodeGen/X86/fold-and-shift.ll
index a434c1f9c1d59c44327c9eff1de07d70272f528c..93b8bbd30809603418d73d21b729ef2f775b3ce4 100644 (file)
--- a/test/CodeGen/X86/fold-and-shift.ll
+++ b/test/CodeGen/X86/fold-and-shift.ll
@@ -89,3 +89,37 @@ entry:
   %sum.2 = add i32 %sum.1, %index
   ret i32 %sum.2
 }
+
+define i8 @t5(i8* %X, i32 %i) {
+; CHECK-LABEL: t5:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; CHECK-NEXT:    andl $1073741810, %ecx # imm = 0x3FFFFFF2
+; CHECK-NEXT:    movb (%eax,%ecx,4), %al
+; CHECK-NEXT:    retl
+
+entry:
+  %tmp2 = shl i32 %i, 2
+  %tmp4 = and i32 %tmp2, -56
+  %tmp7 = getelementptr i8, i8* %X, i32 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}
+
+define i8 @t6(i8* %X, i32 %i) {
+; CHECK-LABEL: t6:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; CHECK-NEXT:    andl $1073741569, %ecx # imm = 0x3FFFFF01
+; CHECK-NEXT:    movb (%eax,%ecx,4), %al
+; CHECK-NEXT:    retl
+
+entry:
+  %tmp2 = shl i32 %i, 2
+  %tmp4 = and i32 %tmp2, -1020
+  %tmp7 = getelementptr i8, i8* %X, i32 %tmp4
+  %tmp9 = load i8, i8* %tmp7
+  ret i8 %tmp9
+}