AMDGPU/SI: Better handle s_wait insertion

We can wait on either VM, EXP or LGKM.
The waits are independent.

Without this patch, a wait inserted because of one of them
would also wait for all the previous others.
This patch makes s_wait only wait for the ones we need for the next
instruction.

Here's an example of subtle perf reduction this patch solves:

This is without the patch:

buffer_load_format_xyzw v[8:11], v0, s[44:47], 0 idxen
buffer_load_format_xyzw v[12:15], v0, s[48:51], 0 idxen
s_load_dwordx4 s[44:47], s[8:9], 0xc
s_waitcnt lgkmcnt(0)
buffer_load_format_xyzw v[16:19], v0, s[52:55], 0 idxen
s_load_dwordx4 s[48:51], s[8:9], 0x10
s_waitcnt vmcnt(1)
buffer_load_format_xyzw v[20:23], v0, s[44:47], 0 idxen

The s_waitcnt vmcnt(1) is useless.
The reason it is added is because the last
buffer_load_format_xyzw needs s[44:47], which was issued
by the first s_load_dwordx4. It waits for all VM
before that call to have finished.

Internally after every instruction, 3 counters (for VM, EXP and LGTM)
are updated after every instruction. For example buffer_load_format_xyzw
will
increase the VM counter, and s_load_dwordx4 the LGKM one.

Without the patch, for every defined register,
the current 3 counters are stored, and are used to know
how long to wait when an instruction needs the register.

Because of that, the s[44:47] counter includes that to use the register
you need to wait for the previous buffer_load_format_xyzw.

Instead this patch stores only the counters that matter for the
register,
and puts zero for the other ones, since we don't need any wait for them.

Patch by: Axel Davy

Differential Revision: http://reviews.llvm.org/D11883

llvm-svn: 245755

1 files changed, 49 insertions, 8 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/wait.ll b/llvm/test/CodeGen/AMDGPU/wait.ll
index 5cc7577cad3..5aaf0028320 100644
--- a/llvm/test/CodeGen/AMDGPU/wait.ll
+++ b/llvm/test/CodeGen/AMDGPU/wait.ll
@@ -1,11 +1,16 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -strict-whitespace %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -strict-whitespace %s
-
-; CHECK-LABEL: {{^}}main:
-; CHECK: s_load_dwordx4
-; CHECK: s_load_dwordx4
-; CHECK: s_waitcnt vmcnt(0) lgkmcnt(0){{$}}
-; CHECK: s_endpgm
+; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=DEFAULT
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=DEFAULT
+; RUN: llc -march=amdgcn --misched=ilpmax -mcpu=SI -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=ILPMAX
+; RUN: llc -march=amdgcn --misched=ilpmax -mcpu=tonga -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=ILPMAX
+; The ilpmax scheduler is used for the second test to get the ordering we want for the test.
+
+; DEFAULT-LABEL: {{^}}main:
+; DEFAULT: s_load_dwordx4
+; DEFAULT: s_load_dwordx4
+; DEFAULT: s_waitcnt vmcnt(0)
+; DEFAULT: exp
+; DEFAULT: s_waitcnt lgkmcnt(0)
+; DEFAULT: s_endpgm
 define void @main(<16 x i8> addrspace(2)* inreg %arg, <16 x i8> addrspace(2)* inreg %arg1, <32 x i8> addrspace(2)* inreg %arg2, <16 x i8> addrspace(2)* inreg %arg3, <16 x i8> addrspace(2)* inreg %arg4, i32 inreg %arg5, i32 %arg6, i32 %arg7, i32 %arg8, i32 %arg9, float addrspace(2)* inreg %constptr) #0 {
 main_body:
   %tmp = getelementptr <16 x i8>, <16 x i8> addrspace(2)* %arg3, i32 0
@@ -29,6 +34,42 @@ main_body:
   ret void
 }
 
+; ILPMAX-LABEL: {{^}}main2:
+; ILPMAX: s_load_dwordx4
+; ILPMAX: s_waitcnt lgkmcnt(0)
+; ILPMAX: buffer_load
+; ILPMAX: s_load_dwordx4
+; ILPMAX: s_waitcnt lgkmcnt(0)
+; ILPMAX: buffer_load
+; ILPMAX: s_waitcnt vmcnt(1)
+; ILPMAX: s_waitcnt vmcnt(0)
+; ILPMAX: s_endpgm
+
+define void @main2([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, [16 x <16 x i8>] addrspace(2)*
+byval, i32 inreg, i32 inreg, i32, i32, i32, i32) #0 {
+main_body:
+  %11 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 0
+  %12 = load <16 x i8>, <16 x i8> addrspace(2)* %11, align 16, !tbaa !0
+  %13 = add i32 %5, %7
+  %14 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %12, i32 0, i32 %13)
+  %15 = extractelement <4 x float> %14, i32 0
+  %16 = extractelement <4 x float> %14, i32 1
+  %17 = extractelement <4 x float> %14, i32 2
+  %18 = extractelement <4 x float> %14, i32 3
+  %19 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %4, i64 0, i64 1
+  %20 = load <16 x i8>, <16 x i8> addrspace(2)* %19, align 16, !tbaa !0
+  %21 = add i32 %5, %7
+  %22 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %20, i32 0, i32 %21)
+  %23 = extractelement <4 x float> %22, i32 0
+  %24 = extractelement <4 x float> %22, i32 1
+  %25 = extractelement <4 x float> %22, i32 2
+  %26 = extractelement <4 x float> %22, i32 3
+  call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %15, float %16, float %17, float %18)
+  call void @llvm.SI.export(i32 15, i32 0, i32 0, i32 32, i32 0, float %23, float %24, float %25, float %26)
+  ret void
+}
+
+
 ; Function Attrs: noduplicate nounwind
 declare void @llvm.AMDGPU.barrier.global() #1