diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll')
-rw-r--r-- | llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll | 34 |
1 files changed, 17 insertions, 17 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll b/llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll index 477ae3da507..9b713419e74 100644 --- a/llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll +++ b/llvm/test/CodeGen/AMDGPU/fmuladd.f16.ll @@ -16,7 +16,7 @@ declare half @llvm.fabs.f16(half) #1 ; VI-FLUSH: v_mac_f16_e32 {{v[0-9]+, v[0-9]+, v[0-9]+}} ; VI-DENORM: v_fma_f16 {{v[0-9]+, v[0-9]+, v[0-9]+}} -define void @fmuladd_f16(half addrspace(1)* %out, half addrspace(1)* %in1, +define amdgpu_kernel void @fmuladd_f16(half addrspace(1)* %out, half addrspace(1)* %in1, half addrspace(1)* %in2, half addrspace(1)* %in3) #0 { %r0 = load half, half addrspace(1)* %in1 %r1 = load half, half addrspace(1)* %in2 @@ -34,7 +34,7 @@ define void @fmuladd_f16(half addrspace(1)* %out, half addrspace(1)* %in1, ; VI-DENORM: v_fma_f16 [[RESULT:v[0-9]+]], [[R1]], 2.0, [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fmuladd_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -56,7 +56,7 @@ define void @fmuladd_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in ; VI-DENORM: v_fma_f16 [[RESULT:v[0-9]+]], [[R1]], 2.0, [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fmuladd_a_2.0_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_a_2.0_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -82,7 +82,7 @@ define void @fmuladd_a_2.0_b_f16(half addrspace(1)* %out, half addrspace(1)* %in ; VI-DENORM-STRICT: v_add_f16_e32 [[RESULT:v[0-9]+]], [[R2]], [[TMP]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fadd_a_a_b_f16(half addrspace(1)* %out, +define amdgpu_kernel void @fadd_a_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in1, half addrspace(1)* %in2) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() @@ -111,7 +111,7 @@ define void @fadd_a_a_b_f16(half addrspace(1)* %out, ; VI-DENORM-STRICT: v_add_f16_e32 [[RESULT:v[0-9]+]], [[TMP]], [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fadd_b_a_a_f16(half addrspace(1)* %out, +define amdgpu_kernel void @fadd_b_a_a_f16(half addrspace(1)* %out, half addrspace(1)* %in1, half addrspace(1)* %in2) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() @@ -134,7 +134,7 @@ define void @fadd_b_a_a_f16(half addrspace(1)* %out, ; VI-FLUSH: v_mac_f16_e32 [[R2]], -2.0, [[R1]] ; VI-DENORM: v_fma_f16 [[R2:v[0-9]+]], [[R1]], -2.0, [[R2]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[R2]] -define void @fmuladd_neg_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_neg_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -156,7 +156,7 @@ define void @fmuladd_neg_2.0_a_b_f16(half addrspace(1)* %out, half addrspace(1)* ; VI-DENORM: v_fma_f16 [[RESULT:v[0-9]+]], -[[R1]], -2.0, [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fmuladd_neg_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_neg_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -180,7 +180,7 @@ define void @fmuladd_neg_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace ; VI-DENORM: v_fma_f16 [[RESULT:v[0-9]+]], -[[R1]], 2.0, [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fmuladd_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -202,7 +202,7 @@ define void @fmuladd_2.0_neg_a_b_f16(half addrspace(1)* %out, half addrspace(1)* ; VI-FLUSH: v_mad_f16 [[RESULT:v[0-9]+]], [[R1]], 2.0, -[[R2]] ; VI-DENORM: v_fma_f16 [[RESULT:v[0-9]+]], [[R1]], 2.0, -[[R2]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fmuladd_2.0_a_neg_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { +define amdgpu_kernel void @fmuladd_2.0_a_neg_b_f16(half addrspace(1)* %out, half addrspace(1)* %in) #0 { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -231,7 +231,7 @@ define void @fmuladd_2.0_a_neg_b_f16(half addrspace(1)* %out, half addrspace(1)* ; VI-DENORM-STRICT: v_subrev_f16_e32 [[RESULT:v[0-9]+]], [[REGC]], [[TMP]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @mad_sub_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @mad_sub_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -261,7 +261,7 @@ define void @mad_sub_f16(half addrspace(1)* noalias nocapture %out, half addrspa ; VI-DENORM-STRICT: v_subrev_f16_e32 [[RESULT:v[0-9]+]], [[TMP]], [[REGC]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @mad_sub_inv_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @mad_sub_inv_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -291,7 +291,7 @@ define void @mad_sub_inv_f16(half addrspace(1)* noalias nocapture %out, half add ; VI-DENORM-STRICT: v_sub_f16_e64 [[RESULT:v[0-9]+]], [[TMP]], |[[REGC]]| ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @mad_sub_fabs_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @mad_sub_fabs_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -323,7 +323,7 @@ define void @mad_sub_fabs_f16(half addrspace(1)* noalias nocapture %out, half ad ; VI-DENORM-STRICT: v_sub_f16_e64 [[RESULT:v[0-9]+]], |[[REGC]]|, [[TMP]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @mad_sub_fabs_inv_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @mad_sub_fabs_inv_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -355,7 +355,7 @@ define void @mad_sub_fabs_inv_f16(half addrspace(1)* noalias nocapture %out, hal ; VI-DENORM-STRICT: v_mul_f16_e32 [[TMP:v[0-9]+]], [[REGB]], [[REGA]] ; VI-DENORM-STRICT: v_add_f16_e32 [[RESULT:v[0-9]+]], [[TMP]], [[REGC]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @neg_neg_mad_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @neg_neg_mad_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -388,7 +388,7 @@ define void @neg_neg_mad_f16(half addrspace(1)* noalias nocapture %out, half add ; VI-DENORM-STRICT: v_subrev_f16_e32 [[RESULT:v[0-9]+]], [[REGC]], [[TMP]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @mad_fabs_sub_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { +define amdgpu_kernel void @mad_fabs_sub_f16(half addrspace(1)* noalias nocapture %out, half addrspace(1)* noalias nocapture readonly %ptr) #1 { %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0 %tid.ext = sext i32 %tid to i64 %gep0 = getelementptr half, half addrspace(1)* %ptr, i64 %tid.ext @@ -419,7 +419,7 @@ define void @mad_fabs_sub_f16(half addrspace(1)* noalias nocapture %out, half ad ; VI-DENORM-STRICT: v_subrev_f16_e32 [[RESULT:v[0-9]+]], [[TMP]], [[R2]] ; VI-DENORM: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fsub_c_fadd_a_a_f16(half addrspace(1)* %out, half addrspace(1)* %in) { +define amdgpu_kernel void @fsub_c_fadd_a_a_f16(half addrspace(1)* %out, half addrspace(1)* %in) { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 @@ -447,7 +447,7 @@ define void @fsub_c_fadd_a_a_f16(half addrspace(1)* %out, half addrspace(1)* %in ; VI-DENORM-STRICT: v_subrev_f16_e32 [[RESULT:v[0-9]+]], [[R2]], [[TMP]] ; VI: flat_store_short v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]] -define void @fsub_fadd_a_a_c_f16(half addrspace(1)* %out, half addrspace(1)* %in) { +define amdgpu_kernel void @fsub_fadd_a_a_c_f16(half addrspace(1)* %out, half addrspace(1)* %in) { %tid = call i32 @llvm.amdgcn.workitem.id.x() %gep.0 = getelementptr half, half addrspace(1)* %out, i32 %tid %gep.1 = getelementptr half, half addrspace(1)* %gep.0, i32 1 |