[llvm] [AMDGPU] Remove calling conv check on entry function (PR #162080)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Nov 10 04:33:58 PST 2025
https://github.com/jofrn updated https://github.com/llvm/llvm-project/pull/162080
>From a0f0b23893fea1edd52a515c5d9a3d7326efec35 Mon Sep 17 00:00:00 2001
From: jofrn <jofernau at amd.com>
Date: Mon, 6 Oct 2025 08:42:09 -0400
Subject: [PATCH 1/3] [AMDGPU] Remove calling conv check on entry function
It is undefined behavior to call a function with a
mismatched calling convention. Rather than crash on
this behavior, it should compile.
---
llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp | 7 -------
1 file changed, 7 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp b/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
index 8101c68986241..2c1729f107321 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
@@ -256,13 +256,6 @@ AMDGPUResourceUsageAnalysisImpl::analyzeResourceUsage(
const Function *Callee = getCalleeFunction(*CalleeOp);
- // Avoid crashing on undefined behavior with an illegal call to a
- // kernel. If a callsite's calling convention doesn't match the
- // function's, it's undefined behavior. If the callsite calling
- // convention does match, that would have errored earlier.
- if (Callee && AMDGPU::isEntryFunctionCC(Callee->getCallingConv()))
- report_fatal_error("invalid call to entry function");
-
auto isSameFunction = [](const MachineFunction &MF, const Function *F) {
return F == &MF.getFunction();
};
>From 2f3d478476279ef1a4cdaeadfe10011db5d0223d Mon Sep 17 00:00:00 2001
From: jofrn <jofernau at amd.com>
Date: Mon, 13 Oct 2025 08:42:44 -0400
Subject: [PATCH 2/3] Add test
---
llvm/test/CodeGen/AMDGPU/cc-entry.ll | 8 ++++++++
1 file changed, 8 insertions(+)
create mode 100644 llvm/test/CodeGen/AMDGPU/cc-entry.ll
diff --git a/llvm/test/CodeGen/AMDGPU/cc-entry.ll b/llvm/test/CodeGen/AMDGPU/cc-entry.ll
new file mode 100644
index 0000000000000..c4fbc6958894c
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/cc-entry.ll
@@ -0,0 +1,8 @@
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 < %s 2>&1 | FileCheck %s
+
+define amdgpu_kernel void @entry_fn() {
+; CHECK-NOT: LLVM ERROR: invalid call to entry function
+entry:
+ call void @entry_fn()
+ ret void
+}
>From 78863591a44a5dd8398835f7324b99b18179d5e9 Mon Sep 17 00:00:00 2001
From: jofrn <jofernau at amd.com>
Date: Mon, 10 Nov 2025 07:33:30 -0500
Subject: [PATCH 3/3] Use autogenerate, use stdout, add call to entry from
another fn
---
llvm/test/CodeGen/AMDGPU/cc-entry.ll | 65 +++++++++++++++++++++++++++-
1 file changed, 63 insertions(+), 2 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/cc-entry.ll b/llvm/test/CodeGen/AMDGPU/cc-entry.ll
index c4fbc6958894c..d807f321c4009 100644
--- a/llvm/test/CodeGen/AMDGPU/cc-entry.ll
+++ b/llvm/test/CodeGen/AMDGPU/cc-entry.ll
@@ -1,7 +1,68 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 < %s 2>&1 | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck %s
define amdgpu_kernel void @entry_fn() {
-; CHECK-NOT: LLVM ERROR: invalid call to entry function
+; CHECK-LABEL: entry_fn:
+; CHECK: ; %bb.0: ; %entry
+; CHECK-NEXT: s_getpc_b64 s[4:5]
+; CHECK-NEXT: s_sext_i32_i16 s5, s5
+; CHECK-NEXT: s_add_co_u32 s4, s4, entry_fn at gotpcrel32@lo+8
+; CHECK-NEXT: s_add_co_ci_u32 s5, s5, entry_fn at gotpcrel32@hi+16
+; CHECK-NEXT: s_mov_b32 s32, 0
+; CHECK-NEXT: s_load_b64 s[4:5], s[4:5], 0x0
+; CHECK-NEXT: s_wait_kmcnt 0x0
+; CHECK-NEXT: s_swappc_b64 s[30:31], s[4:5]
+; CHECK-NEXT: s_endpgm
+entry:
+ call void @entry_fn()
+ ret void
+}
+
+define void @caller() {
+; CHECK-LABEL: caller:
+; CHECK: ; %bb.0: ; %entry
+; CHECK-NEXT: s_wait_loadcnt_dscnt 0x0
+; CHECK-NEXT: s_wait_expcnt 0x0
+; CHECK-NEXT: s_wait_samplecnt 0x0
+; CHECK-NEXT: s_wait_bvhcnt 0x0
+; CHECK-NEXT: s_wait_kmcnt 0x0
+; CHECK-NEXT: s_mov_b32 s0, s33
+; CHECK-NEXT: s_mov_b32 s33, s32
+; CHECK-NEXT: s_or_saveexec_b32 s1, -1
+; CHECK-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_mov_b32 exec_lo, s1
+; CHECK-NEXT: s_add_co_i32 s32, s32, 16
+; CHECK-NEXT: v_writelane_b32 v40, s0, 2
+; CHECK-NEXT: s_mov_b64 s[0:1], s[4:5]
+; CHECK-NEXT: s_getpc_b64 s[4:5]
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_sext_i32_i16 s5, s5
+; CHECK-NEXT: s_add_co_u32 s4, s4, entry_fn at gotpcrel32@lo+12
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_add_co_ci_u32 s5, s5, entry_fn at gotpcrel32@hi+24
+; CHECK-NEXT: v_mov_b32_e32 v0, v31
+; CHECK-NEXT: s_load_b64 s[4:5], s[4:5], 0x0
+; CHECK-NEXT: v_writelane_b32 v40, s30, 0
+; CHECK-NEXT: s_mov_b64 s[2:3], s[6:7]
+; CHECK-NEXT: s_mov_b64 s[6:7], s[10:11]
+; CHECK-NEXT: v_writelane_b32 v40, s31, 1
+; CHECK-NEXT: s_wait_kmcnt 0x0
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_swappc_b64 s[30:31], s[4:5]
+; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; CHECK-NEXT: v_readlane_b32 s31, v40, 1
+; CHECK-NEXT: v_readlane_b32 s30, v40, 0
+; CHECK-NEXT: s_mov_b32 s32, s33
+; CHECK-NEXT: v_readlane_b32 s0, v40, 2
+; CHECK-NEXT: s_or_saveexec_b32 s1, -1
+; CHECK-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_mov_b32 exec_lo, s1
+; CHECK-NEXT: s_mov_b32 s33, s0
+; CHECK-NEXT: s_wait_loadcnt 0x0
+; CHECK-NEXT: s_wait_alu 0xfffe
+; CHECK-NEXT: s_setpc_b64 s[30:31]
entry:
call void @entry_fn()
ret void
More information about the llvm-commits
mailing list