Skip to content

AMDGPU: Fix no return atomicrmw fadd v2f16 selection for gfx908 #96948

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 27, 2024
Merged

AMDGPU: Fix no return atomicrmw fadd v2f16 selection for gfx908 #96948

merged 1 commit into from
Jun 27, 2024

Conversation

arsenm
Copy link
Contributor

@arsenm arsenm commented Jun 27, 2024

We previously would always expand this with a cmpxchg loop, while
it should be the same conditions as the f32 case (except for the
denormal concern).

@arsenm
AMDGPU: Fix no return atomicrmw fadd v2f16 selection for gfx908
e28f8d9 
We previously would always expand this with a cmpxchg loop, while
it should be the same conditions as the f32 case (except for the
denormal concern).
@arsenm Graphite App
Copy link
Contributor Author

arsenm commented Jun 27, 2024

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @arsenm and the rest of your teammates on Graphite Graphite

@llvmbot
Copy link
Member

llvmbot commented Jun 27, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

We previously would always expand this with a cmpxchg loop, while
it should be the same conditions as the f32 case (except for the
denormal concern).

Full diff: https://github.com/llvm/llvm-project/pull/96948.diff

5 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/FLATInstructions.td (+1)
  • (modified) llvm/lib/Target/AMDGPU/SIISelLowering.cpp (+14-7)
  • (modified) llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll (+2-36)
  • (modified) llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll (+5-70)
  • (modified) llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll (+1-13)
diff --git a/llvm/lib/Target/AMDGPU/FLATInstructions.td b/llvm/lib/Target/AMDGPU/FLATInstructions.td
index 98054dde398b3..4bf8f20269a15 100644
--- a/llvm/lib/Target/AMDGPU/FLATInstructions.td
+++ b/llvm/lib/Target/AMDGPU/FLATInstructions.td
@@ -1626,6 +1626,7 @@ defm : GlobalFLATAtomicPatsNoRtnWithAddrSpace <"GLOBAL_ATOMIC_ADD_F32", "int_amd
 }
 
 let OtherPredicates = [HasAtomicBufferGlobalPkAddF16NoRtnInsts] in {
+defm : GlobalFLATAtomicPatsNoRtn <"GLOBAL_ATOMIC_PK_ADD_F16", "atomic_load_fadd_global", v2f16>;
 defm : GlobalFLATAtomicPatsNoRtnWithAddrSpace <"GLOBAL_ATOMIC_PK_ADD_F16", "int_amdgcn_flat_atomic_fadd", "global_addrspace", v2f16>;
 defm : GlobalFLATAtomicPatsNoRtnWithAddrSpace <"GLOBAL_ATOMIC_PK_ADD_F16", "int_amdgcn_global_atomic_fadd", "global_addrspace", v2f16>;
 }
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 16fa7266a4b7d..12977af0d7e85 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -16219,13 +16219,20 @@ SITargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *RMW) const {
     if (Subtarget->hasGFX90AInsts() && Ty->isDoubleTy())
       return ReportUnsafeHWInst(AtomicExpansionKind::None);
 
-    if (AS != AMDGPUAS::FLAT_ADDRESS && Ty->isFloatTy()) {
-      // global/buffer atomic fadd f32 no-rtn: gfx908, gfx90a, gfx940, gfx11+.
-      if (RMW->use_empty() && Subtarget->hasAtomicFaddNoRtnInsts())
-        return ReportUnsafeHWInst(AtomicExpansionKind::None);
-      // global/buffer atomic fadd f32 rtn: gfx90a, gfx940, gfx11+.
-      if (!RMW->use_empty() && Subtarget->hasAtomicFaddRtnInsts())
-        return ReportUnsafeHWInst(AtomicExpansionKind::None);
+    if (AS != AMDGPUAS::FLAT_ADDRESS) {
+      if (Ty->isFloatTy()) {
+        // global/buffer atomic fadd f32 no-rtn: gfx908, gfx90a, gfx940, gfx11+.
+        if (RMW->use_empty() && Subtarget->hasAtomicFaddNoRtnInsts())
+          return ReportUnsafeHWInst(AtomicExpansionKind::None);
+        // global/buffer atomic fadd f32 rtn: gfx90a, gfx940, gfx11+.
+        if (!RMW->use_empty() && Subtarget->hasAtomicFaddRtnInsts())
+          return ReportUnsafeHWInst(AtomicExpansionKind::None);
+      } else {
+        // gfx908
+        if (RMW->use_empty() &&
+            Subtarget->hasAtomicBufferGlobalPkAddF16NoRtnInsts() && isHalf2(Ty))
+          return ReportUnsafeHWInst(AtomicExpansionKind::None);
+      }
     }
 
     // flat atomic fadd f32: gfx940, gfx11+.
diff --git a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
index 00a01e8d976bb..d50ba64ba5d47 100644
--- a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
@@ -6430,26 +6430,9 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_fin
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX908-NEXT:    v_mov_b32_e32 v1, s8
-; GFX908-NEXT:    buffer_load_dword v2, v1, s[4:7], 0 offen offset:1024
-; GFX908-NEXT:    s_add_i32 s10, s8, 0x400
-; GFX908-NEXT:    s_mov_b64 s[8:9], 0
-; GFX908-NEXT:    v_mov_b32_e32 v3, s10
-; GFX908-NEXT:  .LBB20_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v1, v2, v0
-; GFX908-NEXT:    v_mov_b32_e32 v5, v2
-; GFX908-NEXT:    v_mov_b32_e32 v4, v1
-; GFX908-NEXT:    buffer_atomic_cmpswap v[4:5], v3, s[4:7], 0 offen glc
+; GFX908-NEXT:    buffer_atomic_pk_add_f16 v0, v1, s[4:7], 0 offen offset:1024
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v2
-; GFX908-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
-; GFX908-NEXT:    v_mov_b32_e32 v2, v4
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; GFX908-NEXT:    s_cbranch_execnz .LBB20_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_fine_grained_memory:
@@ -7912,26 +7895,9 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX908-NEXT:    v_mov_b32_e32 v1, s8
-; GFX908-NEXT:    buffer_load_dword v2, v1, s[4:7], 0 offen offset:1024
-; GFX908-NEXT:    s_add_i32 s10, s8, 0x400
-; GFX908-NEXT:    s_mov_b64 s[8:9], 0
-; GFX908-NEXT:    v_mov_b32_e32 v3, s10
-; GFX908-NEXT:  .LBB25_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v1, v2, v0
-; GFX908-NEXT:    v_mov_b32_e32 v5, v2
-; GFX908-NEXT:    v_mov_b32_e32 v4, v1
-; GFX908-NEXT:    buffer_atomic_cmpswap v[4:5], v3, s[4:7], 0 offen glc
+; GFX908-NEXT:    buffer_atomic_pk_add_f16 v0, v1, s[4:7], 0 offen offset:1024
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v2
-; GFX908-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
-; GFX908-NEXT:    v_mov_b32_e32 v2, v4
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; GFX908-NEXT:    s_cbranch_execnz .LBB25_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_remote_memory:
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
index 62762008b0c23..e312b37b2e0bb 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
@@ -15550,22 +15550,9 @@ define void @global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory
 ; GFX908-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory:
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT:    global_load_dword v4, v[0:1], off
-; GFX908-NEXT:    s_mov_b64 s[4:5], 0
-; GFX908-NEXT:  .LBB67_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v3, v4, v2
-; GFX908-NEXT:    global_atomic_cmpswap v3, v[0:1], v[3:4], off glc
+; GFX908-NEXT:    global_atomic_pk_add_f16 v[0:1], v2, off
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX908-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT:    v_mov_b32_e32 v4, v3
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX908-NEXT:    s_cbranch_execnz .LBB67_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory:
@@ -15771,22 +15758,9 @@ define void @global_agent_atomic_fadd_noret_v2f16__offset12b_pos__amdgpu_no_fine
 ; GFX908-LABEL: global_agent_atomic_fadd_noret_v2f16__offset12b_pos__amdgpu_no_fine_grained_memory:
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT:    global_load_dword v4, v[0:1], off offset:2044
-; GFX908-NEXT:    s_mov_b64 s[4:5], 0
-; GFX908-NEXT:  .LBB68_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v3, v4, v2
-; GFX908-NEXT:    global_atomic_cmpswap v3, v[0:1], v[3:4], off offset:2044 glc
+; GFX908-NEXT:    global_atomic_pk_add_f16 v[0:1], v2, off offset:2044
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX908-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT:    v_mov_b32_e32 v4, v3
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX908-NEXT:    s_cbranch_execnz .LBB68_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: global_agent_atomic_fadd_noret_v2f16__offset12b_pos__amdgpu_no_fine_grained_memory:
@@ -15995,22 +15969,9 @@ define void @global_agent_atomic_fadd_noret_v2f16__offset12b_neg__amdgpu_no_fine
 ; GFX908-LABEL: global_agent_atomic_fadd_noret_v2f16__offset12b_neg__amdgpu_no_fine_grained_memory:
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT:    global_load_dword v4, v[0:1], off offset:-2048
-; GFX908-NEXT:    s_mov_b64 s[4:5], 0
-; GFX908-NEXT:  .LBB69_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v3, v4, v2
-; GFX908-NEXT:    global_atomic_cmpswap v3, v[0:1], v[3:4], off offset:-2048 glc
+; GFX908-NEXT:    global_atomic_pk_add_f16 v[0:1], v2, off offset:-2048
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX908-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT:    v_mov_b32_e32 v4, v3
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX908-NEXT:    s_cbranch_execnz .LBB69_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: global_agent_atomic_fadd_noret_v2f16__offset12b_neg__amdgpu_no_fine_grained_memory:
@@ -16917,22 +16878,9 @@ define void @global_agent_atomic_fadd_noret_v2f16__amdgpu_no_remote_memory(ptr a
 ; GFX908-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_remote_memory:
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT:    global_load_dword v4, v[0:1], off
-; GFX908-NEXT:    s_mov_b64 s[4:5], 0
-; GFX908-NEXT:  .LBB73_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v3, v4, v2
-; GFX908-NEXT:    global_atomic_cmpswap v3, v[0:1], v[3:4], off glc
+; GFX908-NEXT:    global_atomic_pk_add_f16 v[0:1], v2, off
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX908-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT:    v_mov_b32_e32 v4, v3
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX908-NEXT:    s_cbranch_execnz .LBB73_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_remote_memory:
@@ -17368,22 +17316,9 @@ define void @global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory
 ; GFX908-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory__amdgpu_no_remote_memory:
 ; GFX908:       ; %bb.0:
 ; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT:    global_load_dword v4, v[0:1], off
-; GFX908-NEXT:    s_mov_b64 s[4:5], 0
-; GFX908-NEXT:  .LBB75_1: ; %atomicrmw.start
-; GFX908-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT:    s_waitcnt vmcnt(0)
-; GFX908-NEXT:    v_pk_add_f16 v3, v4, v2
-; GFX908-NEXT:    global_atomic_cmpswap v3, v[0:1], v[3:4], off glc
+; GFX908-NEXT:    global_atomic_pk_add_f16 v[0:1], v2, off
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    buffer_wbinvl1
-; GFX908-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX908-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT:    v_mov_b32_e32 v4, v3
-; GFX908-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX908-NEXT:    s_cbranch_execnz .LBB75_1
-; GFX908-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX908-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX908-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: global_agent_atomic_fadd_noret_v2f16__amdgpu_no_fine_grained_memory__amdgpu_no_remote_memory:
diff --git a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
index 533f0f2691045..021a55f743f66 100644
--- a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
+++ b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
@@ -5750,19 +5750,7 @@ define void @test_atomicrmw_fadd_v2f16_global_agent_noret__unsafe(ptr addrspace(
 ; GFX9-NEXT:    ret void
 ;
 ; GFX908-LABEL: @test_atomicrmw_fadd_v2f16_global_agent_noret__unsafe(
-; GFX908-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(1) [[PTR:%.*]], align 4
-; GFX908-NEXT:    br label [[ATOMICRMW_START:%.*]]
-; GFX908:       atomicrmw.start:
-; GFX908-NEXT:    [[LOADED:%.*]] = phi <2 x half> [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
-; GFX908-NEXT:    [[NEW:%.*]] = fadd <2 x half> [[LOADED]], [[VALUE:%.*]]
-; GFX908-NEXT:    [[TMP2:%.*]] = bitcast <2 x half> [[NEW]] to i32
-; GFX908-NEXT:    [[TMP3:%.*]] = bitcast <2 x half> [[LOADED]] to i32
-; GFX908-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(1) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("agent") seq_cst seq_cst, align 4
-; GFX908-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
-; GFX908-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
-; GFX908-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to <2 x half>
-; GFX908-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
-; GFX908:       atomicrmw.end:
+; GFX908-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], <2 x half> [[VALUE:%.*]] syncscope("agent") seq_cst, align 4
 ; GFX908-NEXT:    ret void
 ;
 ; GFX90A-LABEL: @test_atomicrmw_fadd_v2f16_global_agent_noret__unsafe(

@arsenm arsenm marked this pull request as ready for review June 27, 2024 18:38
@arsenm arsenm merged commit a2a73d8 into main Jun 27, 2024
10 of 11 checks passed
@arsenm arsenm deleted the users/arsenm/amdgpu-select-nortn-atomicrmw-fadd-v2f16-gfx908 branch June 27, 2024 19:17
lravenclaw pushed a commit to lravenclaw/llvm-project that referenced this pull request Jul 3, 2024
@arsenm @lravenclaw
AMDGPU: Fix no return atomicrmw fadd v2f16 selection for gfx908 (llvm…
de3684b 
…#96948)

We previously would always expand this with a cmpxchg loop, while
it should be the same conditions as the f32 case (except for the
denormal concern).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@rampitec rampitec rampitec approved these changes

@cdevadas cdevadas Awaiting requested review from cdevadas

@jayfoad jayfoad Awaiting requested review from jayfoad

@mariusz-sikora-at-amd mariusz-sikora-at-amd Awaiting requested review from mariusz-sikora-at-amd

@Pierre-vh Pierre-vh Awaiting requested review from Pierre-vh

@shiltian shiltian Awaiting requested review from shiltian

@Sisyph Sisyph Awaiting requested review from Sisyph

@yxsamliu yxsamliu Awaiting requested review from yxsamliu

Assignees
No one assigned
Labels
backend:AMDGPU llvm:transforms
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@arsenm @llvmbot @rampitec