Fun x86 encoding tricks: when adding an immediate value of 128,

use a SUB instruction instead of an ADD, because -128 can be
encoded in an 8-bit signed immediate field, while +128 can't be.
This avoids the need for a 32-bit immediate field in this case.

A similar optimization applies to 64-bit adds with 0x80000000,
with the 32-bit signed immediate field.

To support this, teach tablegen how to handle 64-bit constants.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@57663 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 26 insertions, 0 deletions

diff --git a/test/CodeGen/X86/add-trick32.ll b/test/CodeGen/X86/add-trick32.ll
new file mode 100644
index 0000000..42909b4
--- /dev/null
+++ b/test/CodeGen/X86/add-trick32.ll
@@ -0,0 +1,11 @@
+; RUN: llvm-as < %s | llc -march=x86 > %t
+; RUN: not grep add %t
+; RUN: grep subl %t | count 1
+
+; The immediate can be encoded in a smaller way if the
+; instruction is a sub instead of an add.
+
+define i32 @foo(i32 inreg %a) nounwind {
+  %b = add i32 %a, 128
+  ret i32 %b
+}
diff --git a/test/CodeGen/X86/add-trick64.ll b/test/CodeGen/X86/add-trick64.ll
new file mode 100644
index 0000000..5466d9d
--- /dev/null
+++ b/test/CodeGen/X86/add-trick64.ll
@@ -0,0 +1,15 @@
+; RUN: llvm-as < %s | llc -march=x86-64 > %t
+; RUN: not grep add %t
+; RUN: grep subq %t | count 2
+
+; The immediate can be encoded in a smaller way if the
+; instruction is a sub instead of an add.
+
+define i64 @foo(i64 inreg %a) nounwind {
+  %b = add i64 %a, 2147483648
+  ret i64 %b
+}
+define i64 @bar(i64 inreg %a) nounwind {
+  %b = add i64 %a, 128
+  ret i64 %b
+}