[AMDGPU] SDWA peephole: enable by default
Reviewers: vpykhtin, rampitec, arsenm
Subscribers: qcolombet, kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Differential Revision: https://reviews.llvm.org/D31671
llvm-svn: 299536
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
index 4d0580a..eb1f32c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
@@ -26,29 +26,32 @@
; GCN-LABEL: {{^}}sin_v2f16
; GCN: buffer_load_dword v[[A_V2_F16:[0-9]+]]
; SI: v_mov_b32_e32 v[[HALF_PIE:[0-9]+]], 0x3e22f983{{$}}
-; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
+; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
; SI-DAG: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
; SI-DAG: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
+; SI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], v[[HALF_PIE]], v[[A_F32_0]]
+; SI-DAG: v_fract_f32_e32 v[[F_F32_0:[0-9]+]], v[[M_F32_0]]
+; SI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], v[[HALF_PIE]], v[[A_F32_1]]
+; SI-DAG: v_fract_f32_e32 v[[F_F32_1:[0-9]+]], v[[M_F32_1]]
-; VI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
; VI-DAG: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
-; VI-DAG: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
-
-
-; SI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], v[[HALF_PIE]], v[[A_F32_0]]
-; VI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], 0.15915494, v[[A_F32_0]]
-; GCN-DAG: v_fract_f32_e32 v[[F_F32_0:[0-9]+]], v[[M_F32_0]]
-
-; SI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], v[[HALF_PIE]], v[[A_F32_1]]
-; VI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], 0.15915494, v[[A_F32_1]]
-; GCN-DAG: v_fract_f32_e32 v[[F_F32_1:[0-9]+]], v[[M_F32_1]]
+; VI-DAG: v_cvt_f32_f16_sdwa v[[A_F32_1:[0-9]+]], v[[A_V2_F16]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
+; VI-DAG: v_mul_f32_e32 v[[M_F32_0:[0-9]+]], 0.15915494, v[[A_F32_0]]
+; VI-DAG: v_mul_f32_e32 v[[M_F32_1:[0-9]+]], 0.15915494, v[[A_F32_1]]
+; VI-DAG: v_fract_f32_e32 v[[F_F32_0:[0-9]+]], v[[M_F32_0]]
+; VI-DAG: v_fract_f32_e32 v[[F_F32_1:[0-9]+]], v[[M_F32_1]]
; GCN-DAG: v_sin_f32_e32 v[[R_F32_0:[0-9]+]], v[[F_F32_0]]
; GCN-DAG: v_sin_f32_e32 v[[R_F32_1:[0-9]+]], v[[F_F32_1]]
; GCN-DAG: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
-; GCN-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
-; GCN: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
-; GCN: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
+
+; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
+; SI: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
+; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
+
+; VI-DAG: v_cvt_f16_f32_sdwa v[[R_F16_1:[0-9]+]], v[[R_F32_1]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_1]], v[[R_F16_0]]
+
; GCN: buffer_store_dword v[[R_V2_F16]]
; GCN: s_endpgm
define amdgpu_kernel void @sin_v2f16(