From ae4c9e7bc3cea05ed104e3532c4a9b5bc308401b Mon Sep 17 00:00:00 2001
From: Tom Stellard <thomas.stellard@amd.com>
Date: Fri, 20 Jun 2014 17:06:11 +0000
Subject: [PATCH] R600/SI: Add patterns for ctpop inside a branch

llvm-svn: 211378
---
 llvm/lib/Target/R600/SIInstructions.td | 50 ++++++++++++++++++++++++++--------
 llvm/test/CodeGen/R600/ctpop.ll        | 30 ++++++++++++++++++++
 llvm/test/CodeGen/R600/ctpop64.ll      | 31 +++++++++++++++++++++
 3 files changed, 99 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Target/R600/SIInstructions.td b/llvm/lib/Target/R600/SIInstructions.td
index d489f51..f888b8e 100644
--- a/llvm/lib/Target/R600/SIInstructions.td
+++ b/llvm/lib/Target/R600/SIInstructions.td
@@ -1742,6 +1742,24 @@ def : Pat <
   (S_BUFFER_LOAD_DWORD_SGPR $sbase, (S_MOV_B32 imm:$offset))
 >;
 
+} // Predicates = [isSI] in {
+
+//===----------------------------------------------------------------------===//
+// SOP1 Patterns
+//===----------------------------------------------------------------------===//
+
+let Predicates = [isSI, isCFDepth0] in {
+
+def : Pat <
+  (i64 (ctpop i64:$src)),
+  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+    (S_BCNT1_I32_B64 $src), sub0),
+    (S_MOV_B32 0), sub1)
+>;
+
+} // Predicates = [isSI, isCFDepth0]
+
+let  Predicates = [isSI] in {
 //===----------------------------------------------------------------------===//
 // SOP2 Patterns
 //===----------------------------------------------------------------------===//
@@ -1793,6 +1811,26 @@ class SextInReg <ValueType vt, int ShiftAmt> : Pat <
 def : SextInReg <i8, 24>;
 def : SextInReg <i16, 16>;
 
+def : Pat <
+  (i32 (add (i32 (ctpop i32:$popcnt)), i32:$val)),
+  (V_BCNT_U32_B32_e32 $popcnt, $val)
+>;
+
+def : Pat <
+   (i32 (ctpop i32:$popcnt)),
+   (V_BCNT_U32_B32_e64 $popcnt, 0, 0, 0)
+>;
+
+def : Pat <
+  (i64 (ctpop i64:$src)),
+  (INSERT_SUBREG
+    (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+      (V_BCNT_U32_B32_e32 (EXTRACT_SUBREG $src, sub1),
+        (V_BCNT_U32_B32_e64 (EXTRACT_SUBREG $src, sub0), 0, 0, 0)),
+      sub0),
+    (V_MOV_B32_e32 0), sub1)
+>;
+
 /********** ======================= **********/
 /********** Image sampling patterns **********/
 /********** ======================= **********/
@@ -2786,18 +2824,6 @@ def : Pat <
   (S_ADD_I32 $src0, $src1)
 >;
 
-def : Pat <
-  (i32 (add (i32 (ctpop i32:$popcnt)), i32:$val)),
-  (V_BCNT_U32_B32_e32 $popcnt, $val)
->;
-
-def : Pat <
-  (i64 (ctpop i64:$src)),
-  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
-    (S_BCNT1_I32_B64 $src), sub0),
-    (S_MOV_B32 0), sub1)
->;
-
 //============================================================================//
 // Miscellaneous Optimization Patterns
 //============================================================================//
diff --git a/llvm/test/CodeGen/R600/ctpop.ll b/llvm/test/CodeGen/R600/ctpop.ll
index e4d11e0..15be8e1 100644
--- a/llvm/test/CodeGen/R600/ctpop.ll
+++ b/llvm/test/CodeGen/R600/ctpop.ll
@@ -252,3 +252,33 @@ define void @v_ctpop_i32_add_vvar_inv(i32 addrspace(1)* noalias %out, i32 addrsp
   store i32 %add, i32 addrspace(1)* %out, align 4
   ret void
 }
+
+; FIXME: We currently disallow SALU instructions in all branches,
+; but there are some cases when the should be allowed.
+
+; FUNC-LABEL: @ctpop_i32_in_br
+; SI: BUFFER_LOAD_DWORD [[VAL:v[0-9]+]],
+; SI: V_BCNT_U32_B32_e64 [[RESULT:v[0-9]+]], [[VAL]], 0
+; SI: BUFFER_STORE_DWORD [[RESULT]],
+; SI: S_ENDPGM
+; EG: BCNT_INT
+define void @ctpop_i32_in_br(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond) {
+entry:
+  %0 = icmp eq i32 %cond, 0
+  br i1 %0, label %if, label %else
+
+if:
+  %1 = load i32 addrspace(1)* %in
+  %2 = call i32 @llvm.ctpop.i32(i32 %1)
+  br label %endif
+
+else:
+  %3 = getelementptr i32 addrspace(1)* %in, i32 1
+  %4 = load i32 addrspace(1)* %3
+  br label %endif
+
+endif:
+  %5 = phi i32 [%2, %if], [%4, %else]
+  store i32 %5, i32 addrspace(1)* %out
+  ret void
+}
diff --git a/llvm/test/CodeGen/R600/ctpop64.ll b/llvm/test/CodeGen/R600/ctpop64.ll
index 798d8f5..b36ecc6 100644
--- a/llvm/test/CodeGen/R600/ctpop64.ll
+++ b/llvm/test/CodeGen/R600/ctpop64.ll
@@ -89,3 +89,34 @@ define void @v_ctpop_v4i64(<4 x i32> addrspace(1)* noalias %out, <4 x i64> addrs
   store <4 x i32> %truncctpop, <4 x i32> addrspace(1)* %out, align 16
   ret void
 }
+
+; FIXME: We currently disallow SALU instructions in all branches,
+; but there are some cases when the should be allowed.
+
+; FUNC-LABEL: @ctpop_i64_in_br
+; SI: V_BCNT_U32_B32_e64 [[BCNT_LO:v[0-9]+]], v{{[0-9]+}}, 0
+; SI: V_BCNT_U32_B32_e32 v[[BCNT:[0-9]+]], v{{[0-9]+}}, [[BCNT_LO]]
+; SI: V_MOV_B32_e32 v[[ZERO:[0-9]+]], 0
+; SI: BUFFER_STORE_DWORDX2 v[
+; SI: [[BCNT]]:[[ZERO]]]
+; SI: S_ENDPGM
+define void @ctpop_i64_in_br(i64 addrspace(1)* %out, i64 addrspace(1)* %in, i32 %cond) {
+entry:
+  %0 = icmp eq i32 %cond, 0
+  br i1 %0, label %if, label %else
+
+if:
+  %1 = load i64 addrspace(1)* %in
+  %2 = call i64 @llvm.ctpop.i64(i64 %1)
+  br label %endif
+
+else:
+  %3 = getelementptr i64 addrspace(1)* %in, i32 1
+  %4 = load i64 addrspace(1)* %3
+  br label %endif
+
+endif:
+  %5 = phi i64 [%2, %if], [%4, %else]
+  store i64 %5, i64 addrspace(1)* %out
+  ret void
+}
-- 
2.7.4