[PATCH] D31728: [AMDGPU] Translate reqd_work_group_size into amdgpu_flat_work_group_size

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
Herald added subscribers: t-tye, tpr, dstuttard, nhaehnle, wdng, kzhuravl.

These two attributes specify the same info in a different way.
AMGPU BE only checks the latter as a target specific attribute
as opposed to language specific reqd_work_group_size.

This change produces amdgpu_flat_work_group_size out of
reqd_work_group_size if specified.


https://reviews.llvm.org/D31728

Files:
  lib/CodeGen/TargetInfo.cpp
  test/CodeGenOpenCL/amdgpu-attrs.cl


Index: test/CodeGenOpenCL/amdgpu-attrs.cl
===================================================================
--- test/CodeGenOpenCL/amdgpu-attrs.cl
+++ test/CodeGenOpenCL/amdgpu-attrs.cl
@@ -129,6 +129,16 @@
 // CHECK: define amdgpu_kernel void @flat_work_group_size_32_64_waves_per_eu_2_4_num_sgpr_32_num_vgpr_64() [[FLAT_WORK_GROUP_SIZE_32_64_WAVES_PER_EU_2_4_NUM_SGPR_32_NUM_VGPR_64:#[0-9]+]]
 }
 
+__attribute__((reqd_work_group_size(32, 2, 1))) // expected-no-diagnostics
+kernel void reqd_work_group_size_32_2_1() {
+// CHECK: define amdgpu_kernel void @reqd_work_group_size_32_2_1() [[FLAT_WORK_GROUP_SIZE_64_64:#[0-9]+]]
+}
+__attribute__((reqd_work_group_size(32, 2, 1), amdgpu_flat_work_group_size(16, 128))) // expected-no-diagnostics
+kernel void reqd_work_group_size_32_2_1_flat_work_group_size_16_128() {
+// CHECK: define amdgpu_kernel void @reqd_work_group_size_32_2_1_flat_work_group_size_16_128() [[FLAT_WORK_GROUP_SIZE_64_64:#[0-9]+]]
+}
+
+
 // Make sure this is silently accepted on other targets.
 // X86-NOT: "amdgpu-flat-work-group-size"
 // X86-NOT: "amdgpu-waves-per-eu"
@@ -142,6 +152,7 @@
 // CHECK-NOT: "amdgpu-num-vgpr"="0"
 
 // CHECK-DAG: attributes [[FLAT_WORK_GROUP_SIZE_32_64]] = { noinline nounwind "amdgpu-flat-work-group-size"="32,64"
+// CHECK-DAG: attributes [[FLAT_WORK_GROUP_SIZE_64_64]] = { noinline nounwind "amdgpu-flat-work-group-size"="64,64"
 // CHECK-DAG: attributes [[WAVES_PER_EU_2]] = { noinline nounwind "amdgpu-waves-per-eu"="2"
 // CHECK-DAG: attributes [[WAVES_PER_EU_2_4]] = { noinline nounwind "amdgpu-waves-per-eu"="2,4"
 // CHECK-DAG: attributes [[NUM_SGPR_32]] = { noinline nounwind "amdgpu-num-sgpr"="32"
Index: lib/CodeGen/TargetInfo.cpp
===================================================================
--- lib/CodeGen/TargetInfo.cpp
+++ lib/CodeGen/TargetInfo.cpp
@@ -7302,9 +7302,18 @@
 
   llvm::Function *F = cast<llvm::Function>(GV);
 
-  if (const auto *Attr = FD->getAttr<AMDGPUFlatWorkGroupSizeAttr>()) {
-    unsigned Min = Attr->getMin();
-    unsigned Max = Attr->getMax();
+  const auto *ReqdWGS = M.getLangOpts().OpenCL ?
+    FD->getAttr<ReqdWorkGroupSizeAttr>() : nullptr;
+  const auto *FlatWGS = FD->getAttr<AMDGPUFlatWorkGroupSizeAttr>();
+  if (ReqdWGS || FlatWGS) {
+    unsigned Min = FlatWGS ? FlatWGS->getMin() : 0;
+    unsigned Max = FlatWGS ? FlatWGS->getMax() : 0;
+    if (ReqdWGS) {
+      unsigned Req = ReqdWGS->getXDim() * ReqdWGS->getYDim() *
+                     ReqdWGS->getZDim();
+      assert((Min == 0 || Min <= Req) && (Max == 0 || Max >= Req));
+      Min = Max = Req;
+    }
 
     if (Min != 0) {
       assert(Min <= Max && "Min must be less than or equal Max");


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D31728.94290.patch
Type: text/x-patch
Size: 2675 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170405/e046bbd7/attachment.bin>