[PATCH] R600/SI: Extend private extload pattern to include zext loads

[Tom Stellard]

---
 lib/Target/R600/AMDGPUInstructions.td | 10 +++++---
 test/CodeGen/R600/extload-private.ll  | 46 +++++++++++++++++++++++++++++++++++
 2 files changed, 52 insertions(+), 4 deletions(-)
 create mode 100644 test/CodeGen/R600/extload-private.ll

diff --git a/lib/Target/R600/AMDGPUInstructions.td b/lib/Target/R600/AMDGPUInstructions.td
index 12b494e..aa61d73 100644
--- a/lib/Target/R600/AMDGPUInstructions.td
+++ b/lib/Target/R600/AMDGPUInstructions.td
@@ -164,10 +164,6 @@ class PrivateStore <SDPatternOperator op> : PrivateMemOp <
   (ops node:$value, node:$ptr), (op node:$value, node:$ptr)
 >;
 
-def extloadi8_private : PrivateLoad <extloadi8>;
-def sextloadi8_private : PrivateLoad <sextloadi8>;
-def extloadi16_private : PrivateLoad <extloadi16>;
-def sextloadi16_private : PrivateLoad <sextloadi16>;
 def load_private : PrivateLoad <load>;
 
 def truncstorei8_private : PrivateStore <truncstorei8>;
@@ -231,6 +227,9 @@ def sextloadi8_local : PatFrag<(ops node:$ptr), (sextloadi8 node:$ptr), [{
     return isLocalLoad(dyn_cast<LoadSDNode>(N));
 }]>;
 
+def extloadi8_private : PrivateLoad <az_extloadi8>;
+def sextloadi8_private : PrivateLoad <sextloadi8>;
+
 def az_extloadi16 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
   return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
 }]>;
@@ -267,6 +266,9 @@ def sextloadi16_local : PatFrag<(ops node:$ptr), (sextloadi16 node:$ptr), [{
     return isLocalLoad(dyn_cast<LoadSDNode>(N));
 }]>;
 
+def extloadi16_private : PrivateLoad <az_extloadi16>;
+def sextloadi16_private : PrivateLoad <sextloadi16>;
+
 def az_extloadi32 : PatFrag<(ops node:$ptr), (az_extload node:$ptr), [{
   return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
 }]>;
diff --git a/test/CodeGen/R600/extload-private.ll b/test/CodeGen/R600/extload-private.ll
new file mode 100644
index 0000000..4991eab
--- /dev/null
+++ b/test/CodeGen/R600/extload-private.ll
@@ -0,0 +1,46 @@
+; RUN: llc < %s -march=amdgcn -mcpu=SI -mattr=-promote-alloca -verify-machineinstrs | FileCheck --check-prefix=SI --check-prefix=FUNC %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-promote-alloca -verify-machineinstrs | FileCheck --check-prefix=SI --check-prefix=FUNC %s
+
+; FUNC-Label: {{^}}load_i8_sext_private:
+; SI: buffer_load_sbyte v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen
+define void @load_i8_sext_private(i32 addrspace(1)* %out) {
+entry:
+  %tmp0 = alloca i8
+  %tmp1 = load i8* %tmp0
+  %tmp2 = sext i8 %tmp1 to i32
+  store i32 %tmp2, i32 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-Label: {{^}}load_i8_zext_private:
+; SI: buffer_load_ubyte v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen
+define void @load_i8_zext_private(i32 addrspace(1)* %out) {
+entry:
+  %tmp0 = alloca i8
+  %tmp1 = load i8* %tmp0
+  %tmp2 = zext i8 %tmp1 to i32
+  store i32 %tmp2, i32 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-Label: {{^}}load_i16_sext_private:
+; SI: buffer_load_sshort v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen
+define void @load_i16_sext_private(i32 addrspace(1)* %out) {
+entry:
+  %tmp0 = alloca i16
+  %tmp1 = load i16* %tmp0
+  %tmp2 = sext i16 %tmp1 to i32
+  store i32 %tmp2, i32 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-Label: {{^}}load_i16_zext_private:
+; SI: buffer_load_ushort v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen
+define void @load_i16_zext_private(i32 addrspace(1)* %out) {
+entry:
+  %tmp0 = alloca i16
+  %tmp1 = load i16* %tmp0
+  %tmp2 = zext i16 %tmp1 to i32
+  store i32 %tmp2, i32 addrspace(1)* %out
+  ret void
+}
-- 
2.0.4