[PATCH] D47194: [AMDGPU] Fixed non-uniform addr64 MUBUF in shader

[Tim Renouf via Phabricator via llvm-commits]

tpr created this revision.
Herald added subscribers: llvm-commits, t-tye, dstuttard, yaxunl, nhaehnle, wdng, kzhuravl, arsenm.

SIInstrInfo::legalizeOperands was assuming that it never sees an addr64
MUBUF in a shader. But it is generated from a global load, which LLPC
uses for a load from a read-only global variable.

Fixed to detect the addr64 case, and then legalize in the same way as in
compute.

Change-Id: I8aee2a0dc37d2061de445c30c7510e40b3e52f3c


Repository:
  rL LLVM

https://reviews.llvm.org/D47194

Files:
  lib/Target/AMDGPU/SIInstrInfo.cpp
  test/CodeGen/AMDGPU/shader-addr64-nonuniform.ll


Index: test/CodeGen/AMDGPU/shader-addr64-nonuniform.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/shader-addr64-nonuniform.ll
@@ -0,0 +1,17 @@
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx700 -verify-machineinstrs <%s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SICI %s
+
+; GCN-LABEL: {{^}}main:
+; GCN-NOT: readfirstlane
+; SICI: buffer_load_dwordx4 {{.*}} addr64
+
+ at indexable = internal unnamed_addr addrspace(1) constant [6 x <3 x float>] [<3 x float> <float 1.000000e+00, float 0.000000e+00, float 0.000000e+00>, <3 x float> <float 0.000000e+00, float 1.000000e+00, float 0.000000e+00>, <3 x float> <float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, <3 x float> <float 0.000000e+00, float 1.000000e+00, float 1.000000e+00>, <3 x float> <float 1.000000e+00, float 0.000000e+00, float 1.000000e+00>, <3 x float> <float 1.000000e+00, float 1.000000e+00, float 0.000000e+00>]
+
+define amdgpu_ps float @main(i32 %arg18) {
+.entry:
+  %tmp31 = sext i32 %arg18 to i64
+  %tmp32 = getelementptr [6 x <3 x float>], [6 x <3 x float>] addrspace(1)* @indexable, i64 0, i64 %tmp31
+  %tmp33 = load <3 x float>, <3 x float> addrspace(1)* %tmp32, align 16
+  %tmp34 = extractelement <3 x float> %tmp33, i32 0
+  ret float %tmp34
+}
+
Index: lib/Target/AMDGPU/SIInstrInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIInstrInfo.cpp
+++ lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -3660,12 +3660,15 @@
 
   // Legalize MIMG and MUBUF/MTBUF for shaders.
   //
-  // Shaders only generate MUBUF/MTBUF instructions via intrinsics or via
-  // scratch memory access. In both cases, the legalization never involves
-  // conversion to the addr64 form.
+  // Shaders only generate non-addr64 MUBUF/MTBUF instructions via intrinsics
+  // or via scratch memory access. In both cases, the legalization never
+  // involves conversion to the addr64 form.
+  // (Checking for a named operand "vaddr" is a way of checking whether the
+  // instruction is already addr64.)
   if (isMIMG(MI) ||
       (AMDGPU::isShader(MF.getFunction().getCallingConv()) &&
-       (isMUBUF(MI) || isMTBUF(MI)))) {
+       (isMUBUF(MI) || isMTBUF(MI))
+       && !getNamedOperand(MI, AMDGPU::OpName::vaddr))) {
     MachineOperand *SRsrc = getNamedOperand(MI, AMDGPU::OpName::srsrc);
     if (SRsrc && !RI.isSGPRClass(MRI.getRegClass(SRsrc->getReg()))) {
       unsigned SGPR = readlaneVGPRToSGPR(SRsrc->getReg(), MI, MRI);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D47194.147986.patch
Type: text/x-patch
Size: 2520 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180522/bec34b7f/attachment.bin>