HIP: Link correct denormal mode library

[arsenm]

Summary

This wasn't respecting the flush mode based on the default, and also
wasn't correctly handling the explicit
-fno-cuda-flush-denormals-to-zero overriding the mode.

[arsenm]

ping

[yaxunl]

LGTM. Thanks!

clang/test/Driver/hip-device-libs.hip

5

dup if

[arsenm]

4ea3650c212ae471657d3a253cd424ce9d1316ac

clang/lib/Driver/ToolChains/HIP.cpp

void HIPToolChain::addClangTargetOptions(
     Action::OffloadKind DeviceOffloadingKind) const {
   HostTC.addClangTargetOptions(DriverArgs, CC1Args, DeviceOffloadingKind);
 
   StringRef GpuArch = DriverArgs.getLastArgValue(options::OPT_march_EQ);
   assert(!GpuArch.empty() && "Must have an explicit GPU arch.");
   (void) GpuArch;
   assert(DeviceOffloadingKind == Action::OFK_HIP &&
          "Only HIP offloading kinds are supported for GPUs.");
+  auto Kind = llvm::AMDGPU::parseArchAMDGCN(GpuArch);
 
   CC1Args.push_back("-target-cpu");
   CC1Args.push_back(DriverArgs.MakeArgStringRef(GpuArch));
   CC1Args.push_back("-fcuda-is-device");
 
   if (DriverArgs.hasFlag(options::OPT_fcuda_approx_transcendentals,
                          options::OPT_fno_cuda_approx_transcendentals, false))
     CC1Args.push_back("-fcuda-approx-transcendentals");
void HIPToolChain::addClangTargetOptions(
 
   // If --hip-device-lib is not set, add the default bitcode libraries.
   if (BCLibs.empty()) {
     // Get the bc lib file name for ISA version. For example,
     // gfx803 => oclc_isa_version_803.amdgcn.bc.
     std::string GFXVersion = GpuArch.drop_front(3).str();
     std::string ISAVerBC = "oclc_isa_version_" + GFXVersion + ".amdgcn.bc";
 
-    llvm::StringRef FlushDenormalControlBC;
-    if (DriverArgs.hasArg(options::OPT_fcuda_flush_denormals_to_zero))
-      FlushDenormalControlBC = "oclc_daz_opt_on.amdgcn.bc";
-    else
-      FlushDenormalControlBC = "oclc_daz_opt_off.amdgcn.bc";
+    bool FTZDAZ = DriverArgs.hasFlag(
+      options::OPT_fcuda_flush_denormals_to_zero,
+      options::OPT_fno_cuda_flush_denormals_to_zero,
+      getDefaultDenormsAreZeroForTarget(Kind));
+
+    std::string FlushDenormalControlBC = FTZDAZ ?
+      "oclc_daz_opt_on.amdgcn.bc" :
+      "oclc_daz_opt_off.amdgcn.bc";
 
     llvm::StringRef WaveFrontSizeBC;
     if (stoi(GFXVersion) < 1000)
       WaveFrontSizeBC = "oclc_wavefrontsize64_on.amdgcn.bc";
     else
       WaveFrontSizeBC = "oclc_wavefrontsize64_off.amdgcn.bc";
 
     BCLibs.append({"hip.amdgcn.bc", "ocml.amdgcn.bc", "ockl.amdgcn.bc",
                    "oclc_finite_only_off.amdgcn.bc",
-                   std::string(FlushDenormalControlBC),
+                   FlushDenormalControlBC,
                    "oclc_correctly_rounded_sqrt_on.amdgcn.bc",
                    "oclc_unsafe_math_off.amdgcn.bc", ISAVerBC,
                    std::string(WaveFrontSizeBC)});
   }
   for (auto Lib : BCLibs)
     addBCLib(getDriver(), DriverArgs, CC1Args, LibraryPaths, Lib);
 }
 

clang/test/Driver/hip-device-libs.hip

 // REQUIRES: clang-driver
 // REQUIRES: x86-registered-target
 // REQUIRES: amdgpu-registered-target
 
-// Test flush-denormals-to-zero enabled uses oclc_daz_opt_on
+// Test if if oclc_daz_opt_on or if oclc_daz_opt_off is linked depending on
+// expected denormal mode.
 
+// Test subtarget with flushing on by default.
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx803 \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib   \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,FLUSHD
+
+
+// Test subtarget with flushing off by ddefault.
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx900 \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,NOFLUSHD
+
+
+// Test explicit flag, opposite of target default.
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx900 \
+// RUN:   -fcuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,FLUSHD
+
+
+// Test explicit flag, opposite of target default.
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx803 \
+// RUN:   -fno-cuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,NOFLUSHD
+
+
+// Test explicit flag, same as target default.
 // RUN: %clang -### -target x86_64-linux-gnu \
 // RUN:   --cuda-gpu-arch=gfx900 \
+// RUN:   -fno-cuda-flush-denormals-to-zero \
 // RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,NOFLUSHD
+
+
+// Test explicit flag, same as target default.
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx803 \
 // RUN:   -fcuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
 // RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
 // RUN: 2>&1 | FileCheck %s --check-prefixes=COM,FLUSHD
 
-// Test flush-denormals-to-zero disabled uses oclc_daz_opt_off
+
+// Test last flag wins, not flushing
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx803 \
+// RUN:   -fcuda-flush-denormals-to-zero -fno-cuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,NOFLUSHD
+
 
 // RUN: %clang -### -target x86_64-linux-gnu \
 // RUN:   --cuda-gpu-arch=gfx900 \
-// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   -fcuda-flush-denormals-to-zero -fno-cuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib   \
 // RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
 // RUN: 2>&1 | FileCheck %s --check-prefixes=COM,NOFLUSHD
 
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx900 \
+// RUN:   -fno-cuda-flush-denormals-to-zero -fcuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib   \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,FLUSHD
+
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   --cuda-gpu-arch=gfx803 \
+// RUN:   -fno-cuda-flush-denormals-to-zero -fcuda-flush-denormals-to-zero \
+// RUN:   --hip-device-lib-path=%S/Inputs/hip_dev_lib \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck %s --check-prefixes=COM,FLUSHD
+
+
 // Test environment variable HIP_DEVICE_LIB_PATH
 
 // RUN: env HIP_DEVICE_LIB_PATH=%S/Inputs/hip_dev_lib \
 // RUN:   %clang -### -target x86_64-linux-gnu \
 // RUN:   --cuda-gpu-arch=gfx900 \
 // RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
 // RUN: 2>&1 | FileCheck %s --check-prefixes=COM
 
 // COM: {{"[^"]*clang[^"]*"}}
 // COM-SAME: "-mlink-builtin-bitcode" "{{.*}}hip.amdgcn.bc"
 // COM-SAME: "-mlink-builtin-bitcode" "{{.*}}ocml.amdgcn.bc"
 // COM-SAME: "-mlink-builtin-bitcode" "{{.*}}ockl.amdgcn.bc"
 // FLUSHD-SAME: "-mlink-builtin-bitcode" "{{.*}}oclc_daz_opt_on.amdgcn.bc"
 // NOFLUSHD-SAME: "-mlink-builtin-bitcode" "{{.*}}oclc_daz_opt_off.amdgcn.bc"
-