[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-06-05 Thread Matt Arsenault via llvm-branch-commits 

arsenm wrote:

> This is only required for the isa.bc file right.

BC input is "not offload" since there's only one compile for it 

https://github.com/llvm/llvm-project/pull/196586
___
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-06-05 Thread Joseph Huber via llvm-branch-commits 

https://github.com/jhuber6 approved this pull request.

This is only required for the isa.bc file right.

https://github.com/llvm/llvm-project/pull/196586
___
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-05-09 Thread Matt Arsenault via llvm-branch-commits 

https://github.com/arsenm created 
https://github.com/llvm/llvm-project/pull/196586

Pre-work to consolidate target identification for future target
option bug fixes. Also requires updating flang to match recent
clang changes.

Co-authored-by: Claude Sonnet 4 

>From 89264ffcfcbae79d28a3ef8d5dc096e1c4012c37 Mon Sep 17 00:00:00 2001
From: Matt Arsenault 
Date: Fri, 8 May 2026 11:19:48 +0100
Subject: [PATCH] clang/AMDGPU: Pass BoundArch through device libs handling

Pre-work to consolidate target identification for future target
option bug fixes. Also requires updating flang to match recent
clang changes.

Co-authored-by: Claude Sonnet 4 
---
 clang/include/clang/Driver/ToolChain.h   |  2 +-
 clang/lib/Driver/ToolChain.cpp   |  2 +-
 clang/lib/Driver/ToolChains/AMDGPU.cpp   | 11 ++---
 clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp | 11 -
 clang/lib/Driver/ToolChains/AMDGPUOpenMP.h   |  2 +-
 clang/lib/Driver/ToolChains/Flang.cpp| 24 
 clang/lib/Driver/ToolChains/Flang.h  | 16 ++---
 clang/lib/Driver/ToolChains/HIPAMD.cpp   | 15 
 clang/lib/Driver/ToolChains/HIPAMD.h |  2 +-
 clang/lib/Driver/ToolChains/HIPSPV.cpp   |  4 ++--
 clang/lib/Driver/ToolChains/HIPSPV.h |  2 +-
 11 files changed, 57 insertions(+), 34 deletions(-)

diff --git a/clang/include/clang/Driver/ToolChain.h 
b/clang/include/clang/Driver/ToolChain.h
index 117563ef5c950..e5daadba00e04 100644
--- a/clang/include/clang/Driver/ToolChain.h
+++ b/clang/include/clang/Driver/ToolChain.h
@@ -834,7 +834,7 @@ class ToolChain {
 
   /// Get paths for device libraries.
   virtual llvm::SmallVector
-  getDeviceLibs(const llvm::opt::ArgList &Args,
+  getDeviceLibs(const llvm::opt::ArgList &Args, llvm::StringRef BoundArch,
 const Action::OffloadKind DeviceOffloadingKind) const;
 
   /// Add the system specific libraries for the active offload kinds.
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index a68d39deda915..f920c2014c4a4 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -1799,7 +1799,7 @@ void ToolChain::addSYCLIncludeArgs(const ArgList 
&DriverArgs,
ArgStringList &CC1Args) const {}
 
 llvm::SmallVector
-ToolChain::getDeviceLibs(const ArgList &DriverArgs,
+ToolChain::getDeviceLibs(const ArgList &DriverArgs, StringRef BoundArch,
  const Action::OffloadKind DeviceOffloadingKind) const 
{
   return {};
 }
diff --git a/clang/lib/Driver/ToolChains/AMDGPU.cpp 
b/clang/lib/Driver/ToolChains/AMDGPU.cpp
index 0f87043cf801e..06f9c6ac97047 100644
--- a/clang/lib/Driver/ToolChains/AMDGPU.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPU.cpp
@@ -982,9 +982,14 @@ void ROCMToolChain::addClangTargetOptions(
   if (TT.getEnvironment() == llvm::Triple::LLVM)
 return;
 
-  // Get the device name and canonicalize it
-  const StringRef GpuArch = getGPUArch(DriverArgs);
-  auto Kind = llvm::AMDGPU::parseArchAMDGCN(GpuArch);
+  // Get the device name and canonicalize it. For offload compilation,
+  // BoundArch contains the full target ID. For non-offload (OpenCL),
+  // fall back to -mcpu.
+  StringRef GpuArch = BoundArch.empty() ? getGPUArch(DriverArgs) : BoundArch;
+
+  // Extract processor name for canonical arch lookup
+  StringRef Processor = getProcessorFromTargetID(getTriple(), GpuArch);
+  auto Kind = llvm::AMDGPU::parseArchAMDGCN(Processor);
   const StringRef CanonArch = llvm::AMDGPU::getArchNameAMDGCN(Kind);
   StringRef LibDeviceFile = RocmInstallation->getLibDeviceFile(CanonArch);
   auto ABIVer = DeviceLibABIVersion::fromCodeObjectVersion(
diff --git a/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp 
b/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
index 1f8b078dc4f44..c9965e8e6014b 100644
--- a/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
@@ -43,7 +43,8 @@ void AMDGPUOpenMPToolChain::addClangTargetOptions(
   true))
 return;
 
-  for (auto BCFile : getDeviceLibs(DriverArgs, DeviceOffloadingKind)) {
+  for (auto BCFile :
+   getDeviceLibs(DriverArgs, BoundArch, DeviceOffloadingKind)) {
 CC1Args.push_back(BCFile.ShouldInternalize ? "-mlink-builtin-bitcode"
: "-mlink-bitcode-file");
 CC1Args.push_back(DriverArgs.MakeArgString(BCFile.Path));
@@ -126,17 +127,15 @@ AMDGPUOpenMPToolChain::computeMSVCVersion(const Driver *D,
 
 llvm::SmallVector
 AMDGPUOpenMPToolChain::getDeviceLibs(
-const llvm::opt::ArgList &Args,
+const llvm::opt::ArgList &Args, llvm::StringRef BoundArch,
 const Action::OffloadKind DeviceOffloadingKind) const {
   if (!Args.hasFlag(options::OPT_offloadlib, options::OPT_no_offloadlib, true))
 return {};
 
-  StringRef GpuArch = getProcessorFromTargetID(
-  getTriple(), Args.getLastArgValue(options::OPT_march_EQ));
-
+  StringRef GpuArch = getProces

[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-05-09 Thread via llvm-branch-commits 

llvmorg-github-actions[bot] wrote:




@llvm/pr-subscribers-clang-driver

Author: Matt Arsenault (arsenm)


Changes

Pre-work to consolidate target identification for future target
option bug fixes. Also requires updating flang to match recent
clang changes.

Co-authored-by: Claude Sonnet 4 

---
Full diff: https://github.com/llvm/llvm-project/pull/196586.diff


11 Files Affected:

- (modified) clang/include/clang/Driver/ToolChain.h (+1-1) 
- (modified) clang/lib/Driver/ToolChain.cpp (+1-1) 
- (modified) clang/lib/Driver/ToolChains/AMDGPU.cpp (+8-3) 
- (modified) clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp (+5-6) 
- (modified) clang/lib/Driver/ToolChains/AMDGPUOpenMP.h (+1-1) 
- (modified) clang/lib/Driver/ToolChains/Flang.cpp (+14-10) 
- (modified) clang/lib/Driver/ToolChains/Flang.h (+13-3) 
- (modified) clang/lib/Driver/ToolChains/HIPAMD.cpp (+10-5) 
- (modified) clang/lib/Driver/ToolChains/HIPAMD.h (+1-1) 
- (modified) clang/lib/Driver/ToolChains/HIPSPV.cpp (+2-2) 
- (modified) clang/lib/Driver/ToolChains/HIPSPV.h (+1-1) 


``diff
diff --git a/clang/include/clang/Driver/ToolChain.h 
b/clang/include/clang/Driver/ToolChain.h
index 117563ef5c950..e5daadba00e04 100644
--- a/clang/include/clang/Driver/ToolChain.h
+++ b/clang/include/clang/Driver/ToolChain.h
@@ -834,7 +834,7 @@ class ToolChain {
 
   /// Get paths for device libraries.
   virtual llvm::SmallVector
-  getDeviceLibs(const llvm::opt::ArgList &Args,
+  getDeviceLibs(const llvm::opt::ArgList &Args, llvm::StringRef BoundArch,
 const Action::OffloadKind DeviceOffloadingKind) const;
 
   /// Add the system specific libraries for the active offload kinds.
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index a68d39deda915..f920c2014c4a4 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -1799,7 +1799,7 @@ void ToolChain::addSYCLIncludeArgs(const ArgList 
&DriverArgs,
ArgStringList &CC1Args) const {}
 
 llvm::SmallVector
-ToolChain::getDeviceLibs(const ArgList &DriverArgs,
+ToolChain::getDeviceLibs(const ArgList &DriverArgs, StringRef BoundArch,
  const Action::OffloadKind DeviceOffloadingKind) const 
{
   return {};
 }
diff --git a/clang/lib/Driver/ToolChains/AMDGPU.cpp 
b/clang/lib/Driver/ToolChains/AMDGPU.cpp
index 0f87043cf801e..06f9c6ac97047 100644
--- a/clang/lib/Driver/ToolChains/AMDGPU.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPU.cpp
@@ -982,9 +982,14 @@ void ROCMToolChain::addClangTargetOptions(
   if (TT.getEnvironment() == llvm::Triple::LLVM)
 return;
 
-  // Get the device name and canonicalize it
-  const StringRef GpuArch = getGPUArch(DriverArgs);
-  auto Kind = llvm::AMDGPU::parseArchAMDGCN(GpuArch);
+  // Get the device name and canonicalize it. For offload compilation,
+  // BoundArch contains the full target ID. For non-offload (OpenCL),
+  // fall back to -mcpu.
+  StringRef GpuArch = BoundArch.empty() ? getGPUArch(DriverArgs) : BoundArch;
+
+  // Extract processor name for canonical arch lookup
+  StringRef Processor = getProcessorFromTargetID(getTriple(), GpuArch);
+  auto Kind = llvm::AMDGPU::parseArchAMDGCN(Processor);
   const StringRef CanonArch = llvm::AMDGPU::getArchNameAMDGCN(Kind);
   StringRef LibDeviceFile = RocmInstallation->getLibDeviceFile(CanonArch);
   auto ABIVer = DeviceLibABIVersion::fromCodeObjectVersion(
diff --git a/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp 
b/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
index 1f8b078dc4f44..c9965e8e6014b 100644
--- a/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
@@ -43,7 +43,8 @@ void AMDGPUOpenMPToolChain::addClangTargetOptions(
   true))
 return;
 
-  for (auto BCFile : getDeviceLibs(DriverArgs, DeviceOffloadingKind)) {
+  for (auto BCFile :
+   getDeviceLibs(DriverArgs, BoundArch, DeviceOffloadingKind)) {
 CC1Args.push_back(BCFile.ShouldInternalize ? "-mlink-builtin-bitcode"
: "-mlink-bitcode-file");
 CC1Args.push_back(DriverArgs.MakeArgString(BCFile.Path));
@@ -126,17 +127,15 @@ AMDGPUOpenMPToolChain::computeMSVCVersion(const Driver *D,
 
 llvm::SmallVector
 AMDGPUOpenMPToolChain::getDeviceLibs(
-const llvm::opt::ArgList &Args,
+const llvm::opt::ArgList &Args, llvm::StringRef BoundArch,
 const Action::OffloadKind DeviceOffloadingKind) const {
   if (!Args.hasFlag(options::OPT_offloadlib, options::OPT_no_offloadlib, true))
 return {};
 
-  StringRef GpuArch = getProcessorFromTargetID(
-  getTriple(), Args.getLastArgValue(options::OPT_march_EQ));
-
+  StringRef GpuArch = getProcessorFromTargetID(getTriple(), BoundArch);
   SmallVector BCLibs;
   for (auto BCLib :
-   getCommonDeviceLibNames(Args, GpuArch.str(), DeviceOffloadingKind))
+   getCommonDeviceLibNames(Args, GpuArch, DeviceOffloadingKind))
 BCLibs.emplace_back(BCLib);

[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-05-09 Thread Matt Arsenault via llvm-branch-commits 

https://github.com/arsenm ready_for_review 
https://github.com/llvm/llvm-project/pull/196586
___
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

[llvm-branch-commits] [clang] clang/AMDGPU: Pass BoundArch through device libs handling (PR #196586)

2026-05-09 Thread Matt Arsenault via llvm-branch-commits 

arsenm wrote:

> [!WARNING]
> This pull request is not mergeable via GitHub because a downstack PR is 
> open. Once all requirements are satisfied, merge this PR as a stack  href="https://app.graphite.com/github/pr/llvm/llvm-project/196586?utm_source=stack-comment-downstack-mergeability-warning";
>  >on Graphite.
> https://graphite.dev/docs/merge-pull-requests";>Learn more

* **#196586** https://app.graphite.com/github/pr/llvm/llvm-project/196586?utm_source=stack-comment-icon";
 target="_blank">https://static.graphite.dev/graphite-32x32-black.png"; alt="Graphite" 
width="10px" height="10px"/> 👈 https://app.graphite.com/github/pr/llvm/llvm-project/196586?utm_source=stack-comment-view-in-graphite";
 target="_blank">(View in Graphite)
* **#196504** https://app.graphite.com/github/pr/llvm/llvm-project/196504?utm_source=stack-comment-icon";
 target="_blank">https://static.graphite.dev/graphite-32x32-black.png"; alt="Graphite" 
width="10px" height="10px"/>
* `main`




This stack of pull requests is managed by https://graphite.dev?utm-source=stack-comment";>Graphite. Learn 
more about https://stacking.dev/?utm_source=stack-comment";>stacking.


https://github.com/llvm/llvm-project/pull/196586
___
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits