AMDGPU: Make s34 the FP register
Make the FP register callee saved.
This is tricky because now the FP needs to be spilled in the prolog
relative to the incoming SP register, rather than the frame register
used throughout the rest of the function. I don't like how this
bypassess the standard mechanism for CSR spills just to get the
correct insert point. I may look for a better solution, since all CSR
VGPRs may also need to have all lanes activated. Another option might
be to make getFrameIndexReference change the base register if the
frame index is a CSR, and then try to figure out the right insertion
point in emitProlog.
If there is a free VGPR lane available for SGPR spilling, try to use
it for the FP. If that would require intrtoducing a new VGPR spill,
try to use a free call clobbered SGPR. Only fallback to introducing a
new VGPR spill as a last resort.
This also doesn't attempt to handle SGPR spilling with scalar stores.
llvm-svn: 365372
diff --git a/llvm/test/CodeGen/AMDGPU/byval-frame-setup.ll b/llvm/test/CodeGen/AMDGPU/byval-frame-setup.ll
index c272cba..2ab3327 100644
--- a/llvm/test/CodeGen/AMDGPU/byval-frame-setup.ll
+++ b/llvm/test/CodeGen/AMDGPU/byval-frame-setup.ll
@@ -69,28 +69,30 @@
}
; GCN-LABEL: {{^}}void_func_byval_struct_non_leaf:
-; GCN: s_mov_b32 s5, s32
-; GCN: s_add_u32 s32, s32, 0xc00{{$}}
-; GCN-DAG: buffer_store_dword v32
-; GCN-DAG: buffer_store_dword v33
+; GCN: buffer_store_dword v33, off, s[0:3], s32 offset:36
+; GCN-DAG: v_writelane_b32 v33, s34,
+; GCN: s_mov_b32 s34, s32
+; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s34{{$}}
+; GCN-DAG: s_add_u32 s32, s32, 0xc00{{$}}
+; GCN-DAG: buffer_store_dword v32, off, s[0:3], s34 offset:32
; GCN-NOT: v_writelane_b32 v{{[0-9]+}}, s32
-; GCN-DAG: v_writelane_b32
-; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s5{{$}}
-; GCN-DAG: v_add_{{[iu]}}32_e32 [[ADD0:v[0-9]+]], vcc, 1, [[LOAD0]]
-; GCN-DAG: buffer_store_dword [[ADD0]], off, s[0:3], s5{{$}}
-; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s5 offset:16{{$}}
+; GCN-DAG: v_add_{{[iu]}}32_e32 [[ADD0:v[0-9]+]], vcc, 1, [[LOAD0]]
+; GCN: buffer_store_dword [[ADD0]], off, s[0:3], s34{{$}}
+
+; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s34 offset:16{{$}}
; GCN-DAG: v_add_{{[iu]}}32_e32 [[ADD1:v[0-9]+]], vcc, 2, [[LOAD1]]
; GCN: s_swappc_b64
-; GCN: buffer_store_dword [[ADD1]], off, s[0:3], s5 offset:16{{$}}
+; GCN: buffer_store_dword [[ADD1]], off, s[0:3], s34 offset:16{{$}}
; GCN: v_readlane_b32
; GCN-NOT: v_readlane_b32 s32
-; GCN-DAG: buffer_load_dword v32,
-; GCN-DAG: buffer_load_dword v33,
+; GCN-DAG: buffer_load_dword v32, off, s[0:3], s34 offset:32
; GCN: s_sub_u32 s32, s32, 0xc00{{$}}
+; GCN: v_readlane_b32 s34, v33,
+; GCN-DAG: buffer_load_dword v33, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
; GCN: s_setpc_b64
define void @void_func_byval_struct_non_leaf(%struct.ByValStruct addrspace(5)* byval noalias nocapture align 4 %arg0, %struct.ByValStruct addrspace(5)* byval noalias nocapture align 4 %arg1) #1 {
entry:
@@ -108,20 +110,20 @@
}
; GCN-LABEL: {{^}}call_void_func_byval_struct_func:
-; GCN: s_mov_b32 s5, s32
+; GCN: s_mov_b32 s34, s32
; GCN-DAG: s_add_u32 s32, s32, 0xc00{{$}}
; GCN-DAG: v_writelane_b32
; GCN-DAG: v_mov_b32_e32 [[NINE:v[0-9]+]], 9
; GCN-DAG: v_mov_b32_e32 [[THIRTEEN:v[0-9]+]], 13
-; GCN-DAG: buffer_store_dword [[NINE]], off, s[0:3], s5{{$}}
-; GCN-DAG: buffer_store_dword [[THIRTEEN]], off, s[0:3], s5 offset:16
+; GCN-DAG: buffer_store_dword [[NINE]], off, s[0:3], s34{{$}}
+; GCN-DAG: buffer_store_dword [[THIRTEEN]], off, s[0:3], s34 offset:16
-; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s5{{$}}
-; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s5 offset:4
-; GCN-DAG: buffer_load_dword [[LOAD2:v[0-9]+]], off, s[0:3], s5 offset:8
-; GCN-DAG: buffer_load_dword [[LOAD3:v[0-9]+]], off, s[0:3], s5 offset:12
+; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s34{{$}}
+; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s34 offset:4
+; GCN-DAG: buffer_load_dword [[LOAD2:v[0-9]+]], off, s[0:3], s34 offset:8
+; GCN-DAG: buffer_load_dword [[LOAD3:v[0-9]+]], off, s[0:3], s34 offset:12
; GCN-NOT: s_add_u32 s32, s32, 0x800
@@ -131,10 +133,10 @@
; GCN-DAG: buffer_store_dword [[LOAD2]], off, s[0:3], s32 offset:8
; GCN-DAG: buffer_store_dword [[LOAD3]], off, s[0:3], s32 offset:12
-; GCN: buffer_load_dword [[LOAD4:v[0-9]+]], off, s[0:3], s5 offset:16
-; GCN: buffer_load_dword [[LOAD5:v[0-9]+]], off, s[0:3], s5 offset:20
-; GCN: buffer_load_dword [[LOAD6:v[0-9]+]], off, s[0:3], s5 offset:24
-; GCN: buffer_load_dword [[LOAD7:v[0-9]+]], off, s[0:3], s5 offset:28
+; GCN: buffer_load_dword [[LOAD4:v[0-9]+]], off, s[0:3], s34 offset:16
+; GCN: buffer_load_dword [[LOAD5:v[0-9]+]], off, s[0:3], s34 offset:20
+; GCN: buffer_load_dword [[LOAD6:v[0-9]+]], off, s[0:3], s34 offset:24
+; GCN: buffer_load_dword [[LOAD7:v[0-9]+]], off, s[0:3], s34 offset:28
; GCN-DAG: buffer_store_dword [[LOAD4]], off, s[0:3], s32 offset:16
; GCN-DAG: buffer_store_dword [[LOAD5]], off, s[0:3], s32 offset:20
@@ -149,8 +151,9 @@
; GCN-NOT: s_sub_u32 s32, s32, 0x800
; GCN: s_sub_u32 s32, s32, 0xc00{{$}}
-; GCN-NEXT: s_waitcnt
-; GCN-NEXT: s_setpc_b64
+; GCN: v_readlane_b32 s34, v
+; GCN: s_waitcnt
+; GCN: s_setpc_b64
define void @call_void_func_byval_struct_func() #1 {
entry:
%arg0 = alloca %struct.ByValStruct, align 4, addrspace(5)
@@ -306,20 +309,20 @@
}
; GCN-LABEL: {{^}}call_void_func_byval_struct_align8_func:
-; GCN: s_mov_b32 s5, s32
+; GCN: s_mov_b32 s34, s32
; GCN-DAG: s_add_u32 s32, s32, 0xc00{{$}}
; GCN-DAG: v_writelane_b32
; GCN-DAG: v_mov_b32_e32 [[NINE:v[0-9]+]], 9
; GCN-DAG: v_mov_b32_e32 [[THIRTEEN:v[0-9]+]], 13
-; GCN-DAG: buffer_store_dword [[NINE]], off, s[0:3], s5{{$}}
-; GCN-DAG: buffer_store_dword [[THIRTEEN]], off, s[0:3], s5 offset:16
+; GCN-DAG: buffer_store_dword [[NINE]], off, s[0:3], s34{{$}}
+; GCN-DAG: buffer_store_dword [[THIRTEEN]], off, s[0:3], s34 offset:16
-; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s5{{$}}
-; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s5 offset:4
-; GCN-DAG: buffer_load_dword [[LOAD2:v[0-9]+]], off, s[0:3], s5 offset:8
-; GCN-DAG: buffer_load_dword [[LOAD3:v[0-9]+]], off, s[0:3], s5 offset:12
+; GCN-DAG: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s34{{$}}
+; GCN-DAG: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s34 offset:4
+; GCN-DAG: buffer_load_dword [[LOAD2:v[0-9]+]], off, s[0:3], s34 offset:8
+; GCN-DAG: buffer_load_dword [[LOAD3:v[0-9]+]], off, s[0:3], s34 offset:12
; GCN-NOT: s_add_u32 s32, s32, 0x800
@@ -328,10 +331,10 @@
; GCN-DAG: buffer_store_dword [[LOAD2]], off, s[0:3], s32 offset:8
; GCN-DAG: buffer_store_dword [[LOAD3]], off, s[0:3], s32 offset:12
-; GCN: buffer_load_dword [[LOAD4:v[0-9]+]], off, s[0:3], s5 offset:16
-; GCN: buffer_load_dword [[LOAD5:v[0-9]+]], off, s[0:3], s5 offset:20
-; GCN: buffer_load_dword [[LOAD6:v[0-9]+]], off, s[0:3], s5 offset:24
-; GCN: buffer_load_dword [[LOAD7:v[0-9]+]], off, s[0:3], s5 offset:28
+; GCN: buffer_load_dword [[LOAD4:v[0-9]+]], off, s[0:3], s34 offset:16
+; GCN: buffer_load_dword [[LOAD5:v[0-9]+]], off, s[0:3], s34 offset:20
+; GCN: buffer_load_dword [[LOAD6:v[0-9]+]], off, s[0:3], s34 offset:24
+; GCN: buffer_load_dword [[LOAD7:v[0-9]+]], off, s[0:3], s34 offset:28
; GCN: s_waitcnt vmcnt(0)
; GCN-DAG: buffer_store_dword [[LOAD4]], off, s[0:3], s32 offset:16
; GCN-DAG: buffer_store_dword [[LOAD5]], off, s[0:3], s32 offset:20
@@ -346,7 +349,8 @@
; GCN-NOT: s_sub_u32 s32, s32, 0x800
; GCN: s_sub_u32 s32, s32, 0xc00{{$}}
-; GCN-NEXT: s_waitcnt
+; GCN: v_readlane_b32 s34, v
+; GCN: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @call_void_func_byval_struct_align8_func() #0 {
entry:
@@ -385,7 +389,7 @@
ret void
}
-declare void @external_void_func_void() #0
+declare hidden void @external_void_func_void() #0
declare void @llvm.lifetime.start.p5i8(i64, i8 addrspace(5)* nocapture) #3
declare void @llvm.lifetime.end.p5i8(i64, i8 addrspace(5)* nocapture) #3