AMDGPU: Assume SMRD access for constant address space

Since r243294 these are selected to SMRD and
moved later if required.

llvm-svn: 244354

1 files changed, 224 insertions, 32 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
index a68d110fdc9..5cb6cc3ff67 100644
--- a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
+++ b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
@@ -1,5 +1,7 @@
+; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tahiti < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-SI %s
 ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI %s
 ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
 ; RUN: llc -march=amdgcn -mcpu=bonaire -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
 
@@ -115,35 +117,6 @@ done:
   ret void
 }
 
-; OPT-LABEL: @test_no_sink_flat_small_offset_i32(
-; OPT: getelementptr i32, i32 addrspace(4)* %in
-; OPT: br i1
-; OPT-NOT: ptrtoint
-
-; GCN-LABEL: {{^}}test_no_sink_flat_small_offset_i32:
-; GCN: flat_load_dword
-; GCN: {{^}}BB4_2:
-
-define void @test_no_sink_flat_small_offset_i32(i32 addrspace(4)* %out, i32 addrspace(4)* %in, i32 %cond) {
-entry:
-  %out.gep = getelementptr i32, i32 addrspace(4)* %out, i64 999999
-  %in.gep = getelementptr i32, i32 addrspace(4)* %in, i64 7
-  %tmp0 = icmp eq i32 %cond, 0
-  br i1 %tmp0, label %endif, label %if
-
-if:
-  %tmp1 = load i32, i32 addrspace(4)* %in.gep
-  br label %endif
-
-endif:
-  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
-  store i32 %x, i32 addrspace(4)* %out.gep
-  br label %done
-
-done:
-  ret void
-}
-
 ; OPT-LABEL: @test_sink_scratch_small_offset_i32(
 ; OPT-NOT:  getelementptr [512 x i32]
 ; OPT: br i1
@@ -153,7 +126,7 @@ done:
 ; GCN: s_and_saveexec_b64
 ; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
 ; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
-; GCN: {{^}}BB5_2:
+; GCN: {{^}}BB4_2:
 define void @test_sink_scratch_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
 entry:
   %alloca = alloca [512 x i32], align 4
@@ -189,7 +162,7 @@ done:
 ; GCN: s_and_saveexec_b64
 ; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
 ; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
-; GCN: {{^}}BB6_2:
+; GCN: {{^}}BB5_2:
 define void @test_no_sink_scratch_large_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
 entry:
   %alloca = alloca [512 x i32], align 4
@@ -222,7 +195,7 @@ done:
 ; GCN: s_and_saveexec_b64
 ; CI: buffer_load_dword {{v[0-9]+}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
 ; VI: flat_load_dword v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}]
-; GCN: {{^}}BB7_2:
+; GCN: {{^}}BB6_2:
 define void @test_sink_global_vreg_sreg_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %offset, i32 %cond) {
 entry:
   %offset.ext = zext i32 %offset to i64
@@ -246,3 +219,222 @@ done:
 
 attributes #0 = { nounwind readnone }
 attributes #1 = { nounwind }
+
+
+
+; OPT-LABEL: @test_sink_constant_small_offset_i32
+; OPT-NOT:  getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_small_offset_i32:
+; GCN: s_and_saveexec_b64
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0x7{{$}}
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 7
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; OPT-LABEL: @test_sink_constant_max_8_bit_offset_i32
+; OPT-NOT:  getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_max_8_bit_offset_i32:
+; GCN: s_and_saveexec_b64
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0xff{{$}}
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_8_bit_offset_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 255
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; OPT-LABEL: @test_sink_constant_max_8_bit_offset_p1_i32
+; OPT-SI:  getelementptr i32, i32 addrspace(2)*
+; OPT-CI-NOT:  getelementptr i32, i32 addrspace(2)*
+; OPT-VI-NOT:  getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_max_8_bit_offset_p1_i32:
+; GCN: s_and_saveexec_b64
+; SI: s_movk_i32 [[OFFSET:s[0-9]+]], 0x400
+
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, [[OFFSET]]{{$}}
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_8_bit_offset_p1_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 256
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; OPT-LABEL: @test_sink_constant_max_32_bit_offset_i32
+; OPT-SI: getelementptr i32, i32 addrspace(2)*
+; OPT-CI-NOT: getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_max_32_bit_offset_i32:
+; GCN: s_and_saveexec_b64
+; GCN-DAG: s_mov_b32 s{{[0-9]+}}, 3{{$}}
+; GCN-DAG: s_mov_b32 s{{[0-9]+}}, -4{{$}}
+; GCN: s_add_u32
+; GCN: s_addc_u32
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0x0{{$}}
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_32_bit_offset_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 4294967295
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; OPT-LABEL: @test_sink_constant_max_32_bit_offset_p1_i32
+; OPT: getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_max_32_bit_offset_p1_i32:
+; GCN: s_and_saveexec_b64
+; GCN: s_add_u32
+; GCN: s_addc_u32
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0x0{{$}}
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_32_bit_offset_p1_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 17179869181
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; GCN-LABEL: {{^}}test_sink_constant_max_20_bit_byte_offset_i32:
+; GCN: s_and_saveexec_b64
+; SI: s_mov_b32 [[OFFSET:s[0-9]+]], 0xffffc{{$}}
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, [[OFFSET]]{{$}}
+
+; CI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0x3ffff{{$}}
+; VI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0xffffc{{$}}
+
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_20_bit_byte_offset_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 262143
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}
+
+; OPT-LABEL: @test_sink_constant_max_20_bit_byte_offset_p1_i32
+; OPT-SI: getelementptr i32, i32 addrspace(2)*
+; OPT-CI-NOT: getelementptr i32, i32 addrspace(2)*
+; OPT-VI: getelementptr i32, i32 addrspace(2)*
+; OPT: br i1
+
+; GCN-LABEL: {{^}}test_sink_constant_max_20_bit_byte_offset_p1_i32:
+; GCN: s_and_saveexec_b64
+; SI: s_mov_b32 [[OFFSET:s[0-9]+]], 0x100000{{$}}
+; SI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, [[OFFSET]]{{$}}
+
+; CI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0x40000{{$}}
+
+; VI: s_mov_b32 [[OFFSET:s[0-9]+]], 0x100000{{$}}
+; VI: s_load_dword s{{[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, [[OFFSET]]{{$}}
+
+; GCN: s_or_b64 exec, exec
+define void @test_sink_constant_max_20_bit_byte_offset_p1_i32(i32 addrspace(1)* %out, i32 addrspace(2)* %in, i32 %cond) {
+entry:
+  %out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
+  %in.gep = getelementptr i32, i32 addrspace(2)* %in, i64 262144
+  %tmp0 = icmp eq i32 %cond, 0
+  br i1 %tmp0, label %endif, label %if
+
+if:
+  %tmp1 = load i32, i32 addrspace(2)* %in.gep
+  br label %endif
+
+endif:
+  %x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
+  store i32 %x, i32 addrspace(1)* %out.gep
+  br label %done
+
+done:
+  ret void
+}