[x86] try harder to match bitwise 'or' into an LEA

The motivation for this patch starts with the epic fail example in PR18007:
https://llvm.org/bugs/show_bug.cgi?id=18007

...unfortunately, this patch makes no difference for that case, but it solves some
simpler cases. We'll get there some day. :)

The current 'or' matching code was using computeKnownBits() via 
isBaseWithConstantOffset() -> MaskedValueIsZero(), but that's an unnecessarily limited use. 
We can do more by copying the logic in ValueTracking's haveNoCommonBitsSet(), so we can 
treat the 'or' as if it was an 'add'.

There's a TODO comment here because we should lift the bit-checking logic into a helper
function, so it's not duplicated in DAGCombiner.

An example of the better LEA matching:

leal (%rdi,%rdi), %eax
andl $1, %esi
orl %esi, %eax

Becomes:

andl $1, %esi
leal (%rsi,%rdi,2), %eax

Differential Revision: http://reviews.llvm.org/D13956

llvm-svn: 252515

3 files changed, 15 insertions, 25 deletions

diff --git a/llvm/test/CodeGen/X86/or-lea.ll b/llvm/test/CodeGen/X86/or-lea.ll
index bd117207e6c..f28cc8569cf 100644
--- a/llvm/test/CodeGen/X86/or-lea.ll
+++ b/llvm/test/CodeGen/X86/or-lea.ll
@@ -8,9 +8,8 @@
 define i32 @or_shift1_and1(i32 %x, i32 %y) {
 ; CHECK-LABEL: or_shift1_and1:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    addl %edi, %edi
 ; CHECK-NEXT:    andl $1, %esi
-; CHECK-NEXT:    leal (%rsi,%rdi), %eax
+; CHECK-NEXT:    leal (%rsi,%rdi,2), %eax
 ; CHECK-NEXT:    retq
 
   %shl = shl i32 %x, 1
@@ -22,9 +21,8 @@ define i32 @or_shift1_and1(i32 %x, i32 %y) {
 define i32 @or_shift1_and1_swapped(i32 %x, i32 %y) {
 ; CHECK-LABEL: or_shift1_and1_swapped:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leal (%rdi,%rdi), %eax
 ; CHECK-NEXT:    andl $1, %esi
-; CHECK-NEXT:    orl %esi, %eax
+; CHECK-NEXT:    leal (%rsi,%rdi,2), %eax
 ; CHECK-NEXT:    retq
 
   %shl = shl i32 %x, 1
@@ -36,9 +34,8 @@ define i32 @or_shift1_and1_swapped(i32 %x, i32 %y) {
 define i32 @or_shift2_and1(i32 %x, i32 %y) {
 ; CHECK-LABEL: or_shift2_and1:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leal (,%rdi,4), %eax
 ; CHECK-NEXT:    andl $1, %esi
-; CHECK-NEXT:    orl %esi, %eax
+; CHECK-NEXT:    leal (%rsi,%rdi,4), %eax
 ; CHECK-NEXT:    retq
 
   %shl = shl i32 %x, 2
@@ -50,9 +47,8 @@ define i32 @or_shift2_and1(i32 %x, i32 %y) {
 define i32 @or_shift3_and1(i32 %x, i32 %y) {
 ; CHECK-LABEL: or_shift3_and1:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leal (,%rdi,8), %eax
 ; CHECK-NEXT:    andl $1, %esi
-; CHECK-NEXT:    orl %esi, %eax
+; CHECK-NEXT:    leal (%rsi,%rdi,8), %eax
 ; CHECK-NEXT:    retq
 
   %shl = shl i32 %x, 3
@@ -64,9 +60,8 @@ define i32 @or_shift3_and1(i32 %x, i32 %y) {
 define i32 @or_shift3_and7(i32 %x, i32 %y) {
 ; CHECK-LABEL: or_shift3_and7:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    leal (,%rdi,8), %eax
 ; CHECK-NEXT:    andl $7, %esi
-; CHECK-NEXT:    orl %esi, %eax
+; CHECK-NEXT:    leal (%rsi,%rdi,8), %eax
 ; CHECK-NEXT:    retq
 
   %shl = shl i32 %x, 3
@@ -112,9 +107,8 @@ define i32 @or_shift3_and8(i32 %x, i32 %y) {
 define i64 @or_shift1_and1_64(i64 %x, i64 %y) {
 ; CHECK-LABEL: or_shift1_and1_64:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    addq %rdi, %rdi
 ; CHECK-NEXT:    andl $1, %esi
-; CHECK-NEXT:    leaq (%rsi,%rdi), %rax
+; CHECK-NEXT:    leaq (%rsi,%rdi,2), %rax
 ; CHECK-NEXT:    retq
 
   %shl = shl i64 %x, 1
diff --git a/llvm/test/CodeGen/X86/x86-64-double-precision-shift-left.ll b/llvm/test/CodeGen/X86/x86-64-double-precision-shift-left.ll
index f2380f23b8e..75e9052c129 100644
--- a/llvm/test/CodeGen/X86/x86-64-double-precision-shift-left.ll
+++ b/llvm/test/CodeGen/X86/x86-64-double-precision-shift-left.ll
@@ -8,11 +8,9 @@
 ;    return (a << 1) | (b >> 63);
 ;}
 
-; CHECK:             lshift1:
-; CHECK:             addq    {{.*}},{{.*}}
-; CHECK-NEXT:        shrq    $63, {{.*}}
-; CHECK-NEXT:        leaq    ({{.*}},{{.*}}), {{.*}}
-
+; CHECK-LABEL:       lshift1:
+; CHECK:             shrq    $63, %rsi
+; CHECK-NEXT:        leaq    (%rsi,%rdi,2), %rax
 
 define i64 @lshift1(i64 %a, i64 %b) nounwind readnone uwtable {
 entry:
@@ -27,10 +25,9 @@ entry:
 ;    return (a << 2) | (b >> 62);
 ;}
 
-; CHECK:             lshift2:
-; CHECK:             shlq    $2, {{.*}}
-; CHECK-NEXT:        shrq    $62, {{.*}}
-; CHECK-NEXT:        leaq    ({{.*}},{{.*}}), {{.*}}
+; CHECK-LABEL:       lshift2:
+; CHECK:             shrq    $62, %rsi
+; CHECK-NEXT:        leaq    (%rsi,%rdi,4), %rax
 
 define i64 @lshift2(i64 %a, i64 %b) nounwind readnone uwtable {
 entry:
diff --git a/llvm/test/CodeGen/X86/x86-64-double-precision-shift-right.ll b/llvm/test/CodeGen/X86/x86-64-double-precision-shift-right.ll
index 5edaad89df4..bc2f39ee666 100644
--- a/llvm/test/CodeGen/X86/x86-64-double-precision-shift-right.ll
+++ b/llvm/test/CodeGen/X86/x86-64-double-precision-shift-right.ll
@@ -61,10 +61,9 @@ define i64 @rshift7(i64 %a, i64 %b) nounwind readnone uwtable {
 ;    return (a >> 63) | (b << 1);
 ;}
 
-; CHECK:             rshift63:
-; CHECK:             shrq    $63, {{.*}}
-; CHECK-NEXT:        leaq    ({{.*}},{{.*}}), {{.*}}
-; CHECK-NEXT:        orq     {{.*}}, {{.*}}
+; CHECK-LABEL:       rshift63:
+; CHECK:             shrq    $63, %rdi
+; CHECK-NEXT:        leaq    (%rdi,%rsi,2), %rax
 
 define i64 @rshift63(i64 %a, i64 %b) nounwind readnone uwtable {
   %1 = lshr i64 %a, 63