Skip to content

AMDGPU: Remove global_atomic_ordered_add_b64 from gfx1250 support #145781

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Jun 25, 2025
Merged

AMDGPU: Remove global_atomic_ordered_add_b64 from gfx1250 support #145781

merged 1 commit into from
Jun 25, 2025

Conversation

@changpeng
Copy link
Contributor

@changpeng changpeng commented Jun 25, 2025

No description provided.

@llvmbot llvmbot added backend:AMDGPU llvm:mc Machine (object) code labels Jun 25, 2025
@llvmbot
Copy link
Member

llvmbot commented Jun 25, 2025

@llvm/pr-subscribers-mc

Author: Changpeng Fang (changpeng)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/145781.diff

2 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/FLATInstructions.td (+14-12)
  • (modified) llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s (+3)
diff --git a/llvm/lib/Target/AMDGPU/FLATInstructions.td b/llvm/lib/Target/AMDGPU/FLATInstructions.td
index 88b694862d376..5f575fc9fd588 100644
--- a/llvm/lib/Target/AMDGPU/FLATInstructions.td
+++ b/llvm/lib/Target/AMDGPU/FLATInstructions.td
@@ -1036,9 +1036,11 @@ defm GLOBAL_LOAD_LDS_DWORDX3 : FLAT_Global_Load_LDS_Pseudo <"global_load_lds_dwo
 defm GLOBAL_LOAD_LDS_DWORDX4 : FLAT_Global_Load_LDS_Pseudo <"global_load_lds_dwordx4">;
 }
 
+let SubtargetPredicate = isGFX12PlusNot12_50 in
+  defm GLOBAL_ATOMIC_ORDERED_ADD_B64 : FLAT_Global_Atomic_Pseudo <"global_atomic_ordered_add_b64", VReg_64, i64>;
+
 let SubtargetPredicate = isGFX12Plus in {
   defm GLOBAL_ATOMIC_COND_SUB_U32    : FLAT_Global_Atomic_Pseudo <"global_atomic_cond_sub_u32", VGPR_32, i32>;
-  defm GLOBAL_ATOMIC_ORDERED_ADD_B64 : FLAT_Global_Atomic_Pseudo <"global_atomic_ordered_add_b64", VReg_64, i64>;
 
   def GLOBAL_INV    : FLAT_Global_Invalidate_Writeback<"global_inv">;
   def GLOBAL_WB     : FLAT_Global_Invalidate_Writeback<"global_wb">;
@@ -1827,19 +1829,19 @@ let SubtargetPredicate = isGFX12Plus in {
     defm : GlobalFLATAtomicPatsNoRtnWithAddrSpace  <"GLOBAL_ATOMIC_COND_SUB_U32", "int_amdgcn_atomic_cond_sub_u32", "global_addrspace",  i32>;
 }
 
-let OtherPredicates = [isGFX12Plus] in {
+let OtherPredicates = [isGFX12PlusNot12_50] in
   defm : GlobalFLATAtomicPatsRtn <"GLOBAL_ATOMIC_ORDERED_ADD_B64", "int_amdgcn_global_atomic_ordered_add_b64", i64, i64, /* isIntr */ 1>;
 
-  let WaveSizePredicate = isWave32 in {
-    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w32, int_amdgcn_global_load_tr_b64, v2i32>;
-    foreach vt = [v8i16, v8f16, v8bf16] in
-      defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w32, int_amdgcn_global_load_tr_b128, vt>;
-  }
-  let WaveSizePredicate = isWave64 in {
-    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w64, int_amdgcn_global_load_tr_b64, i32>;
-    foreach vt = [v4i16, v4f16, v4bf16] in
-      defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w64, int_amdgcn_global_load_tr_b128, vt>;
-  }
+let WaveSizePredicate = isWave32, OtherPredicates = [isGFX12Plus] in {
+  defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w32, int_amdgcn_global_load_tr_b64, v2i32>;
+  foreach vt = [v8i16, v8f16, v8bf16] in
+    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w32, int_amdgcn_global_load_tr_b128, vt>;
+}
+
+let WaveSizePredicate = isWave64, OtherPredicates = [isGFX12PlusNot12_50] in {
+  defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w64, int_amdgcn_global_load_tr_b64, i32>;
+  foreach vt = [v4i16, v4f16, v4bf16] in
+    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w64, int_amdgcn_global_load_tr_b128, vt>;
 }
 
 let SubtargetPredicate = HasAtomicFMinFMaxF32GlobalInsts, OtherPredicates = [HasFlatGlobalInsts] in {
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s b/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
index 2111e91cd5ef2..89bd507942a22 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
@@ -1,5 +1,8 @@
 ; RUN: not llvm-mc -triple=amdgcn -mcpu=gfx1250 -show-encoding %s 2>&1 | FileCheck --check-prefix=GFX1250-ERR --implicit-check-not=error: --strict-whitespace %s
 
+global_atomic_ordered_add_b64 v0, v[2:3], s[0:1] offset:-64
+// GFX1250-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: instruction not supported on this GPU
+
 ;; DOT4_F32_*, DOT2_F32_*, DOT2_F16 and DOT2_BF16
 
 v_dot4_f32_fp8_fp8 v0, v1, v2, v3

@llvmbot
Copy link
Member

llvmbot commented Jun 25, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: Changpeng Fang (changpeng)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/145781.diff

2 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/FLATInstructions.td (+14-12)
  • (modified) llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s (+3)
diff --git a/llvm/lib/Target/AMDGPU/FLATInstructions.td b/llvm/lib/Target/AMDGPU/FLATInstructions.td
index 88b694862d376..5f575fc9fd588 100644
--- a/llvm/lib/Target/AMDGPU/FLATInstructions.td
+++ b/llvm/lib/Target/AMDGPU/FLATInstructions.td
@@ -1036,9 +1036,11 @@ defm GLOBAL_LOAD_LDS_DWORDX3 : FLAT_Global_Load_LDS_Pseudo <"global_load_lds_dwo
 defm GLOBAL_LOAD_LDS_DWORDX4 : FLAT_Global_Load_LDS_Pseudo <"global_load_lds_dwordx4">;
 }
 
+let SubtargetPredicate = isGFX12PlusNot12_50 in
+  defm GLOBAL_ATOMIC_ORDERED_ADD_B64 : FLAT_Global_Atomic_Pseudo <"global_atomic_ordered_add_b64", VReg_64, i64>;
+
 let SubtargetPredicate = isGFX12Plus in {
   defm GLOBAL_ATOMIC_COND_SUB_U32    : FLAT_Global_Atomic_Pseudo <"global_atomic_cond_sub_u32", VGPR_32, i32>;
-  defm GLOBAL_ATOMIC_ORDERED_ADD_B64 : FLAT_Global_Atomic_Pseudo <"global_atomic_ordered_add_b64", VReg_64, i64>;
 
   def GLOBAL_INV    : FLAT_Global_Invalidate_Writeback<"global_inv">;
   def GLOBAL_WB     : FLAT_Global_Invalidate_Writeback<"global_wb">;
@@ -1827,19 +1829,19 @@ let SubtargetPredicate = isGFX12Plus in {
     defm : GlobalFLATAtomicPatsNoRtnWithAddrSpace  <"GLOBAL_ATOMIC_COND_SUB_U32", "int_amdgcn_atomic_cond_sub_u32", "global_addrspace",  i32>;
 }
 
-let OtherPredicates = [isGFX12Plus] in {
+let OtherPredicates = [isGFX12PlusNot12_50] in
   defm : GlobalFLATAtomicPatsRtn <"GLOBAL_ATOMIC_ORDERED_ADD_B64", "int_amdgcn_global_atomic_ordered_add_b64", i64, i64, /* isIntr */ 1>;
 
-  let WaveSizePredicate = isWave32 in {
-    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w32, int_amdgcn_global_load_tr_b64, v2i32>;
-    foreach vt = [v8i16, v8f16, v8bf16] in
-      defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w32, int_amdgcn_global_load_tr_b128, vt>;
-  }
-  let WaveSizePredicate = isWave64 in {
-    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w64, int_amdgcn_global_load_tr_b64, i32>;
-    foreach vt = [v4i16, v4f16, v4bf16] in
-      defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w64, int_amdgcn_global_load_tr_b128, vt>;
-  }
+let WaveSizePredicate = isWave32, OtherPredicates = [isGFX12Plus] in {
+  defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w32, int_amdgcn_global_load_tr_b64, v2i32>;
+  foreach vt = [v8i16, v8f16, v8bf16] in
+    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w32, int_amdgcn_global_load_tr_b128, vt>;
+}
+
+let WaveSizePredicate = isWave64, OtherPredicates = [isGFX12PlusNot12_50] in {
+  defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B64_w64, int_amdgcn_global_load_tr_b64, i32>;
+  foreach vt = [v4i16, v4f16, v4bf16] in
+    defm : GlobalFLATLoadPats <GLOBAL_LOAD_TR_B128_w64, int_amdgcn_global_load_tr_b128, vt>;
 }
 
 let SubtargetPredicate = HasAtomicFMinFMaxF32GlobalInsts, OtherPredicates = [HasFlatGlobalInsts] in {
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s b/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
index 2111e91cd5ef2..89bd507942a22 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_unsupported.s
@@ -1,5 +1,8 @@
 ; RUN: not llvm-mc -triple=amdgcn -mcpu=gfx1250 -show-encoding %s 2>&1 | FileCheck --check-prefix=GFX1250-ERR --implicit-check-not=error: --strict-whitespace %s
 
+global_atomic_ordered_add_b64 v0, v[2:3], s[0:1] offset:-64
+// GFX1250-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: instruction not supported on this GPU
+
 ;; DOT4_F32_*, DOT2_F32_*, DOT2_F16 and DOT2_BF16
 
 v_dot4_f32_fp8_fp8 v0, v1, v2, v3

@changpeng changpeng requested review from rampitec and shiltian June 25, 2025 21:05
@changpeng
Copy link
Contributor Author

changpeng commented Jun 25, 2025

Why "CI Checks Build and Test Linux" always fails? @shiltian

@shiltian
Copy link
Contributor

shiltian commented Jun 25, 2025

I have no idea…I did break it though in one of my previous PRs.

@changpeng
Copy link
Contributor Author

changpeng commented Jun 25, 2025

I have no idea…I did break it though in one of my previous PRs.

I meant I could not find what is the error, just see in the log:

ninja: build stopped: cannot make progress due to previous errors.

@changpeng
Copy link
Contributor Author

changpeng commented Jun 25, 2025

Could be from others' change:

2025-06-25T18:50:23.0053861Z FAILED: tools/bolt/lib/Core/CMakeFiles/LLVMBOLTCore.dir/BinarySection.cpp.o
2025-06-25T18:50:23.0064673Z CCACHE_CPP2=yes CCACHE_HASHDIR=yes CCACHE_SLOPPINESS=pch_defines,time_macros /usr/bin/ccache /opt/llvm/bin/clang++ -DCMAKE_INSTALL_FULL_LIBDIR="/home/gha/actions-runner/_work/llvm-project/llvm-project/build/install/lib" -DGTEST_HAS_RTTI=0 -DLLVM_BUILD_STATIC -D_DEBUG -D_GLIBCXX_ASSERTIONS -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -I/home/gha/actions-runner/_work/llvm-project/llvm-project/build/tools/bolt/lib/Core -I/home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/lib/Core -I/home/gha/actions-runner/_work/llvm-project/llvm-project/build/include -I/home/gha/actions-runner/_work/llvm-project/llvm-project/llvm/include -I/home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/include -I/home/gha/actions-runner/_work/llvm-project/llvm-project/build/tools/bolt/include -gmlt -fPIC -fno-semantic-interposition -fvisibility-inlines-hidden -Werror=date-time -Werror=unguarded-availability-new -Wall -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers -pedantic -Wno-long-long -Wc++98-compat-extra-semi -Wimplicit-fallthrough -Wcovered-switch-default -Wno-noexcept-type -Wnon-virtual-dtor -Wdelete-non-virtual-dtor -Wsuggest-override -Wstring-conversion -Wmisleading-indentation -Wctad-maybe-unsupported -fdiagnostics-color -ffunction-sections -fdata-sections -O3 -DNDEBUG -std=c++17 -fno-exceptions -funwind-tables -fno-rtti -UNDEBUG -MD -MT tools/bolt/lib/Core/CMakeFiles/LLVMBOLTCore.dir/BinarySection.cpp.o -MF tools/bolt/lib/Core/CMakeFiles/LLVMBOLTCore.dir/BinarySection.cpp.o.d -o tools/bolt/lib/Core/CMakeFiles/LLVMBOLTCore.dir/BinarySection.cpp.o -c /home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/lib/Core/BinarySection.cpp
2025-06-25T18:50:23.0077017Z /home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/lib/Core/BinarySection.cpp:298:44: error: call to 'copyByteArray' is ambiguous
2025-06-25T18:50:23.0078434Z 298 | auto *NewData = reinterpret_cast<char *>(copyByteArray(OS.str()));
2025-06-25T18:50:23.0079197Z | ^~~~~~~~~~~~~
2025-06-25T18:50:23.0080455Z /home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/include/bolt/Core/BinarySection.h:526:17: note: candidate function
2025-06-25T18:50:23.0081623Z 526 | inline uint8_t *copyByteArray(StringRef Buffer) {
2025-06-25T18:50:23.0082217Z | ^
2025-06-25T18:50:23.0083216Z /home/gha/actions-runner/_work/llvm-project/llvm-project/bolt/include/bolt/Core/BinarySection.h:531:17: note: candidate function
2025-06-25T18:50:23.0084346Z 531 | inline uint8_t *copyByteArray(ArrayRef Buffer) {
2025-06-25T18:50:23.0084918Z | ^
2025-06-25T18:50:23.0085253Z 1 error generated.

@changpeng changpeng merged commit ac29858 into llvm:main Jun 25, 2025
8 of 10 checks passed
@changpeng changpeng deleted the atom branch June 25, 2025 22:09
anthonyhatran pushed a commit to anthonyhatran/llvm-project that referenced this pull request Jun 26, 2025
@changpeng
AMDGPU: Remove global_atomic_ordered_add_b64 from gfx1250 support (ll…
a00c9a9 
…vm#145781)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@shiltian shiltian shiltian approved these changes

@rampitec rampitec Awaiting requested review from rampitec

Assignees

No one assigned

Labels

backend:AMDGPU llvm:mc Machine (object) code

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@changpeng @llvmbot @shiltian