[AMDGPU] Add MSG_RTN_GET_CLUSTER_BARRIER_STATE (#157549)

Author: rampitec

llvm/lib/Target/AMDGPU/SIDefines.h

@@ -458,6 +458,8 @@ enum Id { // Message ID, width(4) [3:0].
   ID_RTN_GET_TBA_TO_PC = 134,
   ID_RTN_GET_SE_AID_ID = 135,
 
+  ID_RTN_GET_CLUSTER_BARRIER_STATE = 136, // added in GFX1250
+
   ID_MASK_PreGFX11_ = 0xF,
   ID_MASK_GFX11Plus_ = 0xFF
 };

llvm/lib/Target/AMDGPU/Utils/AMDGPUAsmUtils.cpp

@@ -116,6 +116,8 @@ static constexpr CustomOperand MsgOperands[] = {
   {{"MSG_RTN_GET_TBA"},         ID_RTN_GET_TBA,             isGFX11Plus},
   {{"MSG_RTN_GET_TBA_TO_PC"},   ID_RTN_GET_TBA_TO_PC,       isGFX11Plus},
   {{"MSG_RTN_GET_SE_AID_ID"},   ID_RTN_GET_SE_AID_ID,       isGFX12Plus},
+  {{"MSG_RTN_GET_CLUSTER_BARRIER_STATE"}, ID_RTN_GET_CLUSTER_BARRIER_STATE,
+                                                            isGFX1250},
 };
 
 static constexpr CustomOperand SysMsgOperands[] = {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll

@@ -268,5 +268,30 @@ define amdgpu_kernel void @test_get_99999_i64(ptr addrspace(1) %out) {
   ret void
 }
 
+define amdgpu_kernel void @test_get_136_i64(ptr addrspace(1) %out) {
+; GFX11-LABEL: test_get_136_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
+; GFX11-NEXT:    s_sendmsg_rtn_b64 s[2:3], sendmsg(136, 0, 0)
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: test_get_136_i64:
+; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
+; GFX1250-NEXT:    s_sendmsg_rtn_b64 s[2:3], sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE)
+; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX1250-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1250-NEXT:    s_endpgm
+  %ret = call i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32 136)
+  store i64 %ret, ptr addrspace(1) %out
+  ret void
+}
+
 declare i32 @llvm.amdgcn.s.sendmsg.rtn.i32(i32)
 declare i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32)

llvm/test/MC/AMDGPU/gfx1250_asm_sop1.s

@@ -45,6 +45,14 @@ s_rfe_i64 s[2:3]
 s_rfe_b64 s[2:3]
 // GFX1250: s_rfe_i64 s[2:3]                        ; encoding: [0x02,0x4a,0x80,0xbe]
 
+s_sendmsg_rtn_b32 s2, sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE)
+// GFX1250: s_sendmsg_rtn_b32 s2, sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE) ; encoding: [0x88,0x4c,0x82,0xbe]
+// GFX12-ERR: :[[@LINE-2]]:{{[0-9]+}}: error: specified message id is not supported on this GPU
+
+s_sendmsg_rtn_b64 s[2:3], sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE)
+// GFX1250: s_sendmsg_rtn_b64 s[2:3], sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE) ; encoding: [0x88,0x4d,0x82,0xbe]
+// GFX12-ERR: :[[@LINE-2]]:{{[0-9]+}}: error: specified message id is not supported on this GPU
+
 s_get_shader_cycles_u64 s[2:3]
 // GFX1250: s_get_shader_cycles_u64 s[2:3]          ; encoding: [0x00,0x06,0x82,0xbe]
 // GFX12-ERR: :[[@LINE-2]]:{{[0-9]+}}: error: instruction not supported on this GPU

llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_sop1.txt

@@ -12,6 +12,12 @@
 # GFX1250: s_add_pc_i64 s[2:3]                     ; encoding: [0x02,0x4b,0x80,0xbe]
 0x02,0x4b,0x80,0xbe
 
+# GFX1250: s_sendmsg_rtn_b32 s2, sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE) ; encoding: [0x88,0x4c,0x82,0xbe]
+0x88,0x4c,0x82,0xbe
+
+# GFX1250: s_sendmsg_rtn_b64 s[2:3], sendmsg(MSG_RTN_GET_CLUSTER_BARRIER_STATE) ; encoding: [0x88,0x4d,0x82,0xbe]
+0x88,0x4d,0x82,0xbe
+
 # GFX1250: s_get_shader_cycles_u64 s[2:3]          ; encoding: [0x00,0x06,0x82,0xbe]
 0x00,0x06,0x82,0xbe