[llvm] 33b45ee - AMDGPU: Handle addrspacecast of constant 32-bit to flat

[Matt Arsenault via llvm-commits]

Author: Matt Arsenault
Date: 2022-01-27T11:01:44-05:00
New Revision: 33b45ee44b1f32ffdbc995e6fec806271b4b3ba4

URL: https://github.com/llvm/llvm-project/commit/33b45ee44b1f32ffdbc995e6fec806271b4b3ba4
DIFF: https://github.com/llvm/llvm-project/commit/33b45ee44b1f32ffdbc995e6fec806271b4b3ba4.diff

LOG: AMDGPU: Handle addrspacecast of constant 32-bit to flat

I accidentally made this work on the GlobalISel path, and there's no
real reason not to handle this.

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    llvm/test/CodeGen/AMDGPU/addrspacecast.ll
    llvm/test/CodeGen/AMDGPU/invalid-addrspacecast.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 47c9db6627e7..561866b5a398 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -5632,6 +5632,15 @@ SDValue SITargetLowering::lowerADDRSPACECAST(SDValue Op,
     }
   }
 
+  if (SrcAS == AMDGPUAS::CONSTANT_ADDRESS_32BIT &&
+      Op.getValueType() == MVT::i64) {
+    const SIMachineFunctionInfo *Info =
+        DAG.getMachineFunction().getInfo<SIMachineFunctionInfo>();
+    SDValue Hi = DAG.getConstant(Info->get32BitAddressHighBits(), SL, MVT::i32);
+    SDValue Vec = DAG.getNode(ISD::BUILD_VECTOR, SL, MVT::v2i32, Src, Hi);
+    return DAG.getNode(ISD::BITCAST, SL, MVT::i64, Vec);
+  }
+
   if (ASC->getDestAddressSpace() == AMDGPUAS::CONSTANT_ADDRESS_32BIT &&
       Src.getValueType() == MVT::i64)
     return DAG.getNode(ISD::TRUNCATE, SL, MVT::i32, Src);

diff  --git a/llvm/test/CodeGen/AMDGPU/addrspacecast.ll b/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
index 355158783b52..b9095de7d116 100644
--- a/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
+++ b/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
@@ -405,9 +405,32 @@ define amdgpu_kernel void @use_global_to_constant32_addrspacecast(i8 addrspace(1
   ret void
 }
 
+; GCN-LABEL: {{^}}use_constant32bit_to_flat_addrspacecast_0:
+; GCN: s_load_dword [[PTR:s[0-9]+]],
+; GCN: v_mov_b32_e32 v[[HI:[0-9]+]], 0
+; GCN: v_mov_b32_e32 v[[LO:[0-9]+]], [[PTR]]
+; GCN: flat_load_dword v{{[0-9]+}}, v{{\[}}[[LO]]:[[HI]]{{\]}}
+define amdgpu_kernel void @use_constant32bit_to_flat_addrspacecast_0(i32 addrspace(6)* %ptr) #0 {
+  %stof = addrspacecast i32 addrspace(6)* %ptr to i32*
+  %load = load volatile i32, i32* %stof
+  ret void
+}
+
+; GCN-LABEL: {{^}}use_constant32bit_to_flat_addrspacecast_1:
+; GCN: s_load_dword [[PTR:s[0-9]+]],
+; GCN: v_mov_b32_e32 v[[HI:[0-9]+]], 0xffff8000
+; GCN: v_mov_b32_e32 v[[LO:[0-9]+]], [[PTR]]
+; GCN: flat_load_dword v{{[0-9]+}}, v{{\[}}[[LO]]:[[HI]]{{\]}}
+define amdgpu_kernel void @use_constant32bit_to_flat_addrspacecast_1(i32 addrspace(6)* %ptr) #3 {
+  %stof = addrspacecast i32 addrspace(6)* %ptr to i32*
+  %load = load volatile i32, i32* %stof
+  ret void
+}
+
 declare void @llvm.amdgcn.s.barrier() #1
 declare i32 @llvm.amdgcn.workitem.id.x() #2
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind convergent }
 attributes #2 = { nounwind readnone }
+attributes #3 = { nounwind "amdgpu-32bit-address-high-bits"="0xffff8000" }

diff  --git a/llvm/test/CodeGen/AMDGPU/invalid-addrspacecast.ll b/llvm/test/CodeGen/AMDGPU/invalid-addrspacecast.ll
index cec4eac48154..7f1124221796 100644
--- a/llvm/test/CodeGen/AMDGPU/invalid-addrspacecast.ll
+++ b/llvm/test/CodeGen/AMDGPU/invalid-addrspacecast.ll
@@ -7,13 +7,6 @@ define amdgpu_kernel void @use_group_to_global_addrspacecast(i32 addrspace(3)* %
   ret void
 }
 
-; ERROR: error: <unknown>:0:0: in function use_constant32bit_to_flat_addrspacecast void (i32 addrspace(6)*): invalid addrspacecast
-define amdgpu_kernel void @use_constant32bit_to_flat_addrspacecast(i32 addrspace(6)* %ptr) #0 {
-  %stof = addrspacecast i32 addrspace(6)* %ptr to i32*
-  store volatile i32 7, i32* %stof
-  ret void
-}
-
 ; ERROR: error: <unknown>:0:0: in function use_local_to_constant32bit_addrspacecast void (i32 addrspace(3)*): invalid addrspacecast
 define amdgpu_kernel void @use_local_to_constant32bit_addrspacecast(i32 addrspace(3)* %ptr) #0 {
   %stof = addrspacecast i32 addrspace(3)* %ptr to i32 addrspace(6)*