AMDGPU: Force skip over SMRD, VMEM and s_waitcnt instructions

[pendingchaos]

Summary

This fixes a large Dawn of War 3 performance regression with RADV from Mesa 19.0 to master which was caused by creating less code in some branches.

[pendingchaos]

Seems I forgot to update a few test cases. I'll add a updated revision soon.

[pendingchaos]

This second version updates various tests.

[nhaehnle]

LGTM

llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp

for (MachineBasicBlock::const_iterator I = MBB.begin(), E = MBB.end();
       // when EXEC = 0. We should skip the loop lest it becomes infinite.
       if (I->getOpcode() == AMDGPU::S_CBRANCH_VCCNZ ||
           I->getOpcode() == AMDGPU::S_CBRANCH_VCCZ)
         return true;
 
       if (TII->hasUnwantedEffectsWhenEXECEmpty(*I))
         return true;
 
+      // These instructions are potentially expensive even if EXEC = 0.
+      if (TII->isSMRD(*I) || TII->isVMEM(*I) || I->getOpcode() == AMDGPU::S_WAITCNT)
+        return true;
+
       ++NumInstr;
       if (NumInstr >= SkipThreshold)
         return true;
     }
   }
 
   return false;
 }

llvm/trunk/test/CodeGen/AMDGPU/collapse-endcf.ll

 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,ALL %s
 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -amdgpu-opt-exec-mask-pre-ra=0 < %s | FileCheck -enable-var-scope -check-prefixes=DISABLED,ALL %s
 
 ; ALL-LABEL: {{^}}simple_nested_if:
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[ENDIF:BB[0-9_]+]]
 ; GCN-NEXT: s_cbranch_execz [[ENDIF]]
 ; GCN:      s_and_b64 exec, exec, vcc
 ; GCN-NEXT: ; mask branch [[ENDIF]]
+; GCN-NEXT: s_cbranch_execz [[ENDIF]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[ENDIF]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC]]
 ; GCN: ds_write_b32
 ; GCN: s_endpgm
 
 

 }
 
 ; ALL-LABEL: {{^}}uncollapsable_nested_if:
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC_OUTER:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[ENDIF_OUTER:BB[0-9_]+]]
 ; GCN-NEXT: s_cbranch_execz [[ENDIF_OUTER]]
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC_INNER:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[ENDIF_INNER:BB[0-9_]+]]
+; GCN-NEXT: s_cbranch_execz [[ENDIF_INNER]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[ENDIF_INNER]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC_INNER]]
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[ENDIF_OUTER]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC_OUTER]]
 ; GCN: ds_write_b32

 
 ; ALL-LABEL: {{^}}nested_if_if_else:
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC_OUTER:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[ENDIF_OUTER:BB[0-9_]+]]
 ; GCN-NEXT: s_cbranch_execz [[ENDIF_OUTER]]
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC_INNER:s\[[0-9:]+\]]]
 ; GCN-NEXT: s_xor_b64 [[SAVEEXEC_INNER2:s\[[0-9:]+\]]], exec, [[SAVEEXEC_INNER]]
 ; GCN-NEXT: ; mask branch [[THEN_INNER:BB[0-9_]+]]
+; GCN-NEXT: s_cbranch_execz [[THEN_INNER]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[THEN_INNER]]:
 ; GCN-NEXT: s_or_saveexec_b64 [[SAVEEXEC_INNER3:s\[[0-9:]+\]]], [[SAVEEXEC_INNER2]]
 ; GCN-NEXT: s_xor_b64 exec, exec, [[SAVEEXEC_INNER3]]
 ; GCN-NEXT: ; mask branch [[ENDIF_OUTER]]
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[ENDIF_OUTER]]:

 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC_OUTER:s\[[0-9:]+\]]]
 ; GCN-NEXT: s_xor_b64 [[SAVEEXEC_OUTER2:s\[[0-9:]+\]]], exec, [[SAVEEXEC_OUTER]]
 ; GCN-NEXT: ; mask branch [[THEN_OUTER:BB[0-9_]+]]
 ; GCN-NEXT: s_cbranch_execz [[THEN_OUTER]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: s_and_saveexec_b64 [[SAVEEXEC_INNER_IF_OUTER_ELSE:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[THEN_OUTER_FLOW:BB[0-9_]+]]
+; GCN-NEXT: s_cbranch_execz [[THEN_OUTER_FLOW]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[THEN_OUTER_FLOW]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC_INNER_IF_OUTER_ELSE]]
 ; GCN-NEXT: {{^}}[[THEN_OUTER]]:
 ; GCN-NEXT: s_or_saveexec_b64 [[SAVEEXEC_OUTER3:s\[[0-9:]+\]]], [[SAVEEXEC_OUTER2]]
 ; GCN-NEXT: s_xor_b64 exec, exec, [[SAVEEXEC_OUTER3]]
 ; GCN-NEXT: ; mask branch [[ENDIF_OUTER:BB[0-9_]+]]
 ; GCN-NEXT: s_cbranch_execz [[ENDIF_OUTER]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: s_and_saveexec_b64 [[SAVEEXEC_INNER_IF_OUTER_THEN:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[FLOW1:BB[0-9_]+]]
+; GCN-NEXT: s_cbranch_execz [[FLOW1]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: [[FLOW1]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC_INNER_IF_OUTER_THEN]]
 ; GCN-NEXT: {{^}}[[ENDIF_OUTER]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC_OUTER]]
 ; GCN: ds_write_b32
 ; GCN: s_endpgm

 bb.outer.end:
   store i32 3, i32 addrspace(3)* null
   ret void
 }
 
 ; ALL-LABEL: {{^}}s_endpgm_unsafe_barrier:
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC:s\[[0-9:]+\]]]
 ; GCN-NEXT: ; mask branch [[ENDIF:BB[0-9_]+]]
+; GCN-NEXT: s_cbranch_execz [[ENDIF]]
 ; GCN-NEXT: {{^BB[0-9_]+}}:
 ; GCN:      store_dword
 ; GCN-NEXT: {{^}}[[ENDIF]]:
 ; GCN-NEXT: s_or_b64 exec, exec, [[SAVEEXEC]]
 ; GCN:      s_barrier
 ; GCN-NEXT: s_endpgm
 define amdgpu_kernel void @s_endpgm_unsafe_barrier(i32 addrspace(1)* nocapture %arg) {
 bb:

llvm/trunk/test/CodeGen/AMDGPU/skip-if-dead.ll

 ; CHECK: v_cmpx_gt_f32_e32 vcc, 0,
 ; CHECK: [[BB4]]:
 ; CHECK: s_or_b64 exec, exec
 ; CHECK: image_sample_c
 
 ; CHECK: v_cmp_neq_f32_e32 vcc, 0,
 ; CHECK: s_and_saveexec_b64 s{{\[[0-9]+:[0-9]+\]}}, vcc
 ; CHECK: mask branch [[END:BB[0-9]+_[0-9]+]]
+; CHECK-NEXT: s_cbranch_execz [[END]]
 ; CHECK-NOT: branch
 
 ; CHECK: BB{{[0-9]+_[0-9]+}}: ; %bb8
 ; CHECK: buffer_store_dword
 
 ; CHECK: [[END]]:
 ; CHECK: s_endpgm
 define amdgpu_ps void @if_after_kill_block(float %arg, float %arg1, float %arg2, float %arg3) #0 {

llvm/trunk/test/CodeGen/AMDGPU/valu-i1.ll

 end:
   ret void
 }
 
 ; SI-LABEL: {{^}}simple_test_v_if:
 ; SI: v_cmp_ne_u32_e32 vcc, 0, v{{[0-9]+}}
 ; SI: s_and_saveexec_b64 [[BR_SREG:s\[[0-9]+:[0-9]+\]]], vcc
 ; SI-NEXT: ; mask branch [[EXIT:BB[0-9]+_[0-9]+]]
+; SI-NEXT: s_cbranch_execz [[EXIT]]
 
 ; SI-NEXT: BB{{[0-9]+_[0-9]+}}:
 ; SI: buffer_store_dword
 
 ; SI-NEXT: {{^}}[[EXIT]]:
 ; SI: s_endpgm
 define amdgpu_kernel void @simple_test_v_if(i32 addrspace(1)* %dst, i32 addrspace(1)* %src) #1 {
   %tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone

 }
 
 ; FIXME: It would be better to endpgm in the then block.
 
 ; SI-LABEL: {{^}}simple_test_v_if_ret_else_ret:
 ; SI: v_cmp_ne_u32_e32 vcc, 0, v{{[0-9]+}}
 ; SI: s_and_saveexec_b64 [[BR_SREG:s\[[0-9]+:[0-9]+\]]], vcc
 ; SI-NEXT: ; mask branch [[EXIT:BB[0-9]+_[0-9]+]]
+; SI-NEXT: s_cbranch_execz [[EXIT]]
 
 ; SI-NEXT: BB{{[0-9]+_[0-9]+}}:
 ; SI: buffer_store_dword
 
 ; SI-NEXT: {{^}}[[EXIT]]:
 ; SI: s_endpgm
 define amdgpu_kernel void @simple_test_v_if_ret_else_ret(i32 addrspace(1)* %dst, i32 addrspace(1)* %src) #1 {
   %tid = call i32 @llvm.amdgcn.workitem.id.x()

 
 ; SI-NEXT: {{^BB[0-9]+_[0-9]+}}: ; %exit
 ; SI: ds_write_b32
 
 ; SI-NEXT: {{^}}[[FLOW]]:
 ; SI-NEXT: s_or_saveexec_b64
 ; SI-NEXT: s_xor_b64 exec, exec
 ; SI-NEXT: ; mask branch [[UNIFIED_RETURN:BB[0-9]+_[0-9]+]]
+; SI-NEXT: s_cbranch_execz [[UNIFIED_RETURN]]
 
 ; SI-NEXT: {{^BB[0-9]+_[0-9]+}}: ; %then
 ; SI: s_waitcnt
 ; SI-NEXT: buffer_store_dword
 
 ; SI-NEXT: {{^}}[[UNIFIED_RETURN]]: ; %UnifiedReturnBlock
 ; SI: s_endpgm
 define amdgpu_kernel void @simple_test_v_if_ret_else_code_ret(i32 addrspace(1)* %dst, i32 addrspace(1)* %src) #1 {

llvm/trunk/test/CodeGen/AMDGPU/wqm.ll

 ;CHECK: image_sample
 ;CHECK: image_sample
 ;CHECK: %Flow
 ;CHECK-NEXT: s_or_saveexec_b64 [[SAVED:s\[[0-9]+:[0-9]+\]]],
 ;CHECK-NEXT: s_and_b64 exec, exec, [[ORIG]]
 ;CHECK-NEXT: s_and_b64 [[SAVED]], exec, [[SAVED]]
 ;CHECK-NEXT: s_xor_b64 exec, exec, [[SAVED]]
 ;CHECK-NEXT: mask branch [[END_BB:BB[0-9]+_[0-9]+]]
+;CHECK-NEXT: s_cbranch_execz [[END_BB]]
 ;CHECK-NEXT: BB{{[0-9]+_[0-9]+}}: ; %ELSE
 ;CHECK: store_dword
 ;CHECK: [[END_BB]]: ; %END
 ;CHECK: s_or_b64 exec, exec,
 ;CHECK: v_mov_b32_e32 v0
 ;CHECK: ; return
 define amdgpu_ps float @test_control_flow_1(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %c, i32 %z, float %data) {
 main_body: