R600/SI: Add global atomicrmw xor

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220109 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/R600/AMDGPUInstructions.td b/lib/Target/R600/AMDGPUInstructions.td
index 06171a0b40687a5fec45d1d37de926596664e169..1562c2a047534d2903413a15b83eb94376395fc8 100644 (file)
--- a/lib/Target/R600/AMDGPUInstructions.td
+++ b/lib/Target/R600/AMDGPUInstructions.td
@@ -394,6 +394,7 @@ def atomic_or_global : global_binary_atomic_op<atomic_load_or>;
 def atomic_sub_global : global_binary_atomic_op<atomic_load_sub>;
 def atomic_umax_global : global_binary_atomic_op<atomic_load_umax>;
 def atomic_umin_global : global_binary_atomic_op<atomic_load_umin>;
+def atomic_xor_global : global_binary_atomic_op<atomic_load_xor>;
 
 //===----------------------------------------------------------------------===//
 // Misc Pattern Fragments

diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index 504b086ea7ae2ec1438fc3c6868ea9cbd2970de8..0a8bb460e757d3dde8e185c38746c8fdaf333fbd 100644 (file)
--- a/lib/Target/R600/SIInstructions.td
+++ b/lib/Target/R600/SIInstructions.td
@@ -922,7 +922,9 @@ defm BUFFER_ATOMIC_AND : MUBUF_Atomic <
 defm BUFFER_ATOMIC_OR : MUBUF_Atomic <
   0x0000003a, "BUFFER_ATOMIC_OR", VReg_32, i32, atomic_or_global
 >;
-//def BUFFER_ATOMIC_XOR : MUBUF_ <0x0000003b, "BUFFER_ATOMIC_XOR", []>;
+defm BUFFER_ATOMIC_XOR : MUBUF_Atomic <
+  0x0000003b, "BUFFER_ATOMIC_XOR", VReg_32, i32, atomic_xor_global
+>;
 //def BUFFER_ATOMIC_INC : MUBUF_ <0x0000003c, "BUFFER_ATOMIC_INC", []>;
 //def BUFFER_ATOMIC_DEC : MUBUF_ <0x0000003d, "BUFFER_ATOMIC_DEC", []>;
 //def BUFFER_ATOMIC_FCMPSWAP : MUBUF_ <0x0000003e, "BUFFER_ATOMIC_FCMPSWAP", []>;

diff --git a/test/CodeGen/R600/global_atomics.ll b/test/CodeGen/R600/global_atomics.ll
index c46cee3d9d9e2116627eb6f3300963fa0b31b935..8d21dbdc7939d84e11863ef9b58e666a813c8f57 100644 (file)
--- a/test/CodeGen/R600/global_atomics.ll
+++ b/test/CodeGen/R600/global_atomics.ll
@@ -639,3 +639,83 @@ entry:
   store i32 %0, i32 addrspace(1)* %out2
   ret void
 }
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_offset:
+; SI: BUFFER_ATOMIC_XOR v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:0x10{{$}}
+define void @atomic_xor_i32_offset(i32 addrspace(1)* %out, i32 %in) {
+entry:
+  %gep = getelementptr i32 addrspace(1)* %out, i32 4
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_ret_offset:
+; SI: BUFFER_ATOMIC_XOR [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:0x10 glc {{$}}
+; SI: BUFFER_STORE_DWORD [[RET]]
+define void @atomic_xor_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
+entry:
+  %gep = getelementptr i32 addrspace(1)* %out, i32 4
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %0, i32 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_addr64_offset:
+; SI: BUFFER_ATOMIC_XOR v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:0x10{{$}}
+define void @atomic_xor_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
+entry:
+  %ptr = getelementptr i32 addrspace(1)* %out, i64 %index
+  %gep = getelementptr i32 addrspace(1)* %ptr, i32 4
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_ret_addr64_offset:
+; SI: BUFFER_ATOMIC_XOR [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:0x10 glc{{$}}
+; SI: BUFFER_STORE_DWORD [[RET]]
+define void @atomic_xor_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
+entry:
+  %ptr = getelementptr i32 addrspace(1)* %out, i64 %index
+  %gep = getelementptr i32 addrspace(1)* %ptr, i32 4
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %0, i32 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32:
+; SI: BUFFER_ATOMIC_XOR v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+define void @atomic_xor_i32(i32 addrspace(1)* %out, i32 %in) {
+entry:
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_ret:
+; SI: BUFFER_ATOMIC_XOR [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; SI: BUFFER_STORE_DWORD [[RET]]
+define void @atomic_xor_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
+entry:
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %0, i32 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_addr64:
+; SI: BUFFER_ATOMIC_XOR v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
+define void @atomic_xor_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
+entry:
+  %ptr = getelementptr i32 addrspace(1)* %out, i64 %index
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_xor_i32_ret_addr64:
+; SI: BUFFER_ATOMIC_XOR [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
+; SI: BUFFER_STORE_DWORD [[RET]]
+define void @atomic_xor_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
+entry:
+  %ptr = getelementptr i32 addrspace(1)* %out, i64 %index
+  %0  = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %0, i32 addrspace(1)* %out2
+  ret void
+}