From fa2ea1a8cf2e3fe4dae19032868010e917629d16 Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Tue, 4 Aug 2009 01:41:15 +0000
Subject: [PATCH] Emit sub r, #c instead of transforming it to add r, #-c if c
 fits in 8-bit. This is a bit of pre-mature optimization. 8-bit variant makes
 it likely it will be narrowed to a 16-bit instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78030 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/ARM/ARMInstrThumb2.td  | 11 +++++++++--
 test/CodeGen/Thumb2/thumb2-sub.ll | 20 +++++++++++++++++++-
 2 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/lib/Target/ARM/ARMInstrThumb2.td b/lib/Target/ARM/ARMInstrThumb2.td
index 8c562b04355..faa7dd5475d 100644
--- a/lib/Target/ARM/ARMInstrThumb2.td
+++ b/lib/Target/ARM/ARMInstrThumb2.td
@@ -83,6 +83,10 @@ def imm0_4095_neg : PatLeaf<(i32 imm), [{
  return (uint32_t)(-N->getZExtValue()) < 4096; 
 }], imm_neg_XFORM>; 
 
+def imm0_255_neg : PatLeaf<(i32 imm), [{
+  return (uint32_t)(-N->getZExtValue()) < 255;
+}], imm_neg_XFORM>; 
+
 /// imm0_65535 predicate - True if the 32-bit immediate is in the range 
 /// [0.65535].
 def imm0_65535 : PatLeaf<(i32 imm), [{
@@ -619,12 +623,12 @@ def t2STRB_POST : T2Iidxldst<(outs GPR:$base_wb),
 let mayLoad = 1 in
 def t2LDM : T2XI<(outs),
                  (ins addrmode4:$addr, pred:$p, reglist:$dst1, variable_ops),
-                 "ldm${addr:submode}${p}.w $addr, $dst1", []>;
+                 "ldm${addr:submode}${p} $addr, $dst1", []>;
 
 let mayStore = 1 in
 def t2STM : T2XI<(outs),
                  (ins addrmode4:$addr, pred:$p, reglist:$src1, variable_ops),
-                 "stm${addr:submode}${p}.w $addr, $src1", []>;
+                 "stm${addr:submode}${p} $addr, $src1", []>;
 
 //===----------------------------------------------------------------------===//
 //  Move Instructions.
@@ -704,6 +708,9 @@ defm t2RSB  : T2I_rbin_is   <"rsb", BinOpFrag<(sub  node:$LHS, node:$RHS)>>;
 defm t2RSBS : T2I_rbin_s_is <"rsb", BinOpFrag<(subc node:$LHS, node:$RHS)>>;
 
 // (sub X, imm) gets canonicalized to (add X, -imm).  Match this form.
+let AddedComplexity = 1 in
+def : T2Pat<(add       GPR:$src, imm0_255_neg:$imm),
+            (t2SUBri   GPR:$src, imm0_255_neg:$imm)>;
 def : T2Pat<(add       GPR:$src, t2_so_imm_neg:$imm),
             (t2SUBri   GPR:$src, t2_so_imm_neg:$imm)>;
 def : T2Pat<(add       GPR:$src, imm0_4095_neg:$imm),
diff --git a/test/CodeGen/Thumb2/thumb2-sub.ll b/test/CodeGen/Thumb2/thumb2-sub.ll
index a7c9676b978..74accdcb8ea 100644
--- a/test/CodeGen/Thumb2/thumb2-sub.ll
+++ b/test/CodeGen/Thumb2/thumb2-sub.ll
@@ -1,31 +1,49 @@
-; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {sub\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*#\[0-9\]*} | grep {#171\\|#1179666\\|#872428544\\|#1448498774\\|#510} | count 5
+; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
 
 ; 171 = 0x000000ab
 define i32 @f1(i32 %a) {
+; CHECK: f1:
+; CHECK: sub.w r0, r0, #171
     %tmp = sub i32 %a, 171
     ret i32 %tmp
 }
 
 ; 1179666 = 0x00120012
 define i32 @f2(i32 %a) {
+; CHECK: f2:
+; CHECK: sub.w r0, r0, #1179666
     %tmp = sub i32 %a, 1179666
     ret i32 %tmp
 }
 
 ; 872428544 = 0x34003400
 define i32 @f3(i32 %a) {
+; CHECK: f3:
+; CHECK: sub.w r0, r0, #872428544
     %tmp = sub i32 %a, 872428544
     ret i32 %tmp
 }
 
 ; 1448498774 = 0x56565656
 define i32 @f4(i32 %a) {
+; CHECK: f4:
+; CHECK: sub.w r0, r0, #1448498774
     %tmp = sub i32 %a, 1448498774
     ret i32 %tmp
 }
 
 ; 510 = 0x000001fe
 define i32 @f5(i32 %a) {
+; CHECK: f5:
+; CHECK: sub.w r0, r0, #510
     %tmp = sub i32 %a, 510
     ret i32 %tmp
 }
+
+; Don't change this to an add.
+define i32 @f6(i32 %a) {
+; CHECK: f6:
+; CHECK: sub.w r0, r0, #1
+    %tmp = sub i32 %a, 1
+    ret i32 %tmp
+}