[AMDGPU][GlobalISel] Legalize G_VECREDUCE_FADD (#223909)

llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 4ef314f..bf165db 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -2310,9 +2310,9 @@
 
   getActionDefinitionsBuilder(
       {G_VECREDUCE_SMIN, G_VECREDUCE_SMAX, G_VECREDUCE_UMIN, G_VECREDUCE_UMAX,
-       G_VECREDUCE_ADD, G_VECREDUCE_MUL, G_VECREDUCE_FMUL, G_VECREDUCE_FMIN,
-       G_VECREDUCE_FMAX, G_VECREDUCE_FMINIMUM, G_VECREDUCE_FMAXIMUM,
-       G_VECREDUCE_OR, G_VECREDUCE_AND, G_VECREDUCE_XOR})
+       G_VECREDUCE_ADD, G_VECREDUCE_MUL, G_VECREDUCE_FADD, G_VECREDUCE_FMUL,
+       G_VECREDUCE_FMIN, G_VECREDUCE_FMAX, G_VECREDUCE_FMINIMUM,
+       G_VECREDUCE_FMAXIMUM, G_VECREDUCE_OR, G_VECREDUCE_AND, G_VECREDUCE_XOR})
       .legalFor(AllVectors)
       .scalarize(1)
       .lower();

llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll

diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
index 71baa8a..7cd101b1 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
@@ -1409,6 +1409,58 @@
   ret float %res
 }
 
+define float @test_vector_reduce_fadd_reassoc_v3float(<3 x float> %v) {
+; GFX7-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX7:       ; %bb.0:
+; GFX7-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX7-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX7-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX8:       ; %bb.0:
+; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX8-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX8-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX9-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX10-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: test_vector_reduce_fadd_reassoc_v3float:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_add_f32_e32 v0, v0, v1
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_add_f32_e32 v0, v0, v2
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %res = call reassoc float @llvm.vector.reduce.fadd.v3float(float -0.0, <3 x float> %v)
+  ret float %res
+}
+
 define float @test_vector_reduce_fadd_v4float(float %sp, <4 x float> %v) {
 ; GFX7-LABEL: test_vector_reduce_fadd_v4float:
 ; GFX7:       ; %bb.0: ; %entry