[AMDGPU][GFX11] Add test coverage for 16-bit conversions, part 15.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D152908

diff --git a/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll b/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
index aeeb2fd..fedf13a 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
@@ -1,17 +1,64 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=GFX11 %s
 
-; GCN-LABEL: {{^}}sitofp_i16_to_f16
-; GCN: buffer_load_{{sshort|ushort}} v[[A_I16:[0-9]+]]
-
-; SI: v_cvt_f32_i32_e32 v[[A_F32:[0-9]+]], v[[A_I16]]
-; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]]
-
-; VI: v_cvt_f16_i16_e32 v[[R_F16:[0-9]+]], v[[A_I16]]
-
-; GCN: buffer_store_short v[[R_F16]]
-; GCN: s_endpgm
 define amdgpu_kernel void @sitofp_i16_to_f16(
+; SI-LABEL: sitofp_i16_to_f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_sshort v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: sitofp_i16_to_f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f16_i16_e32 v0, v0
+; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: sitofp_i16_to_f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f16_i16_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -21,13 +68,65 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}sitofp_i32_to_f16
-; GCN: buffer_load_dword v[[A_I32:[0-9]+]]
-; GCN: v_cvt_f32_i32_e32 v[[A_I16:[0-9]+]], v[[A_I32]]
-; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_I16]]
-; GCN: buffer_store_short v[[R_F16]]
-; GCN: s_endpgm
 define amdgpu_kernel void @sitofp_i32_to_f16(
+; SI-LABEL: sitofp_i32_to_f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: sitofp_i32_to_f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: sitofp_i32_to_f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -39,24 +138,74 @@ entry:
 
 ; f16 = sitofp i64 is in sint_to_fp.i64.ll
 
-; GCN-LABEL: {{^}}sitofp_v2i16_to_v2f16
-; GCN:     buffer_load_dword
-
-; SI: v_cvt_f32_i32_e32
-; SI: v_cvt_f32_i32_e32
-; SI: v_cvt_f16_f32_e32
-; SI: v_cvt_f16_f32_e32
-; SI-DAG: v_lshlrev_b32_e32
-; SI: v_or_b32_e32
-
-; VI-DAG: v_cvt_f16_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
-; VI-DAG: v_cvt_f16_i16_e32
-; VI:     v_or_b32_e32
-
-; GCN: buffer_store_dword
-; GCN: s_endpgm
-
 define amdgpu_kernel void @sitofp_v2i16_to_v2f16(
+; SI-LABEL: sitofp_v2i16_to_v2f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_bfe_i32 v1, v0, 0, 16
+; SI-NEXT:    v_ashrrev_i32_e32 v0, 16, v0
+; SI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; SI-NEXT:    v_cvt_f32_i32_e32 v1, v1
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT:    v_or_b32_e32 v0, v1, v0
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: sitofp_v2i16_to_v2f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f16_i16_sdwa v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
+; VI-NEXT:    v_cvt_f16_i16_e32 v0, v0
+; VI-NEXT:    v_or_b32_e32 v0, v0, v1
+; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: sitofp_v2i16_to_v2f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-NEXT:    v_cvt_f16_i16_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_i16_e32 v1, v1
+; GFX11-NEXT:    v_pack_b32_f16 v0, v0, v1
+; GFX11-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -66,25 +215,76 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}sitofp_v2i32_to_v2f16
-; GCN:    buffer_load_dwordx2
-
-; SI: v_cvt_f32_i32_e32
-; SI: v_cvt_f32_i32_e32
-; SI: v_cvt_f16_f32_e32
-; SI: v_cvt_f16_f32_e32
-; SI-DAG: v_lshlrev_b32_e32
-; SI: v_or_b32_e32
-
-; VI-DAG: v_cvt_f32_i32_e32
-; VI-DAG: v_cvt_f32_i32_e32
-; VI-DAG: v_cvt_f16_f32_e32
-; VI-DAG: v_cvt_f16_f32_sdwa
-; VI:     v_or_b32_e32
-
-; GCN: buffer_store_dword
-; GCN: s_endpgm
 define amdgpu_kernel void @sitofp_v2i32_to_v2f16(
+; SI-LABEL: sitofp_v2i32_to_v2f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_i32_e32 v1, v1
+; SI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT:    v_or_b32_e32 v0, v0, v1
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: sitofp_v2i32_to_v2f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f32_i32_e32 v1, v1
+; VI-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; VI-NEXT:    v_cvt_f16_f32_sdwa v1, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    v_or_b32_e32 v0, v0, v1
+; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: sitofp_v2i32_to_v2f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b64 v[0:1], off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f32_i32_e32 v1, v1
+; GFX11-NEXT:    v_cvt_f32_i32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_pack_b32_f16 v0, v0, v1
+; GFX11-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -94,15 +294,91 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}s_sint_to_fp_i1_to_f16:
-; GCN-DAG: v_cmp_le_f32_e32 [[CMP0:vcc]], 1.0, {{v[0-9]+}}
-; GCN-DAG: v_cmp_le_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]], 0, {{v[0-9]+}}
-; GCN: s_xor_b64 [[R_CMP:s\[[0-9]+:[0-9]+\]]], [[CMP1]], [[CMP0]]
-; GCN: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1.0, [[R_CMP]]
-; GCN-NEXT: v_cvt_f16_f32_e32 [[R_F16:v[0-9]+]], [[RESULT]]
-; GCN: buffer_store_short
-; GCN: s_endpgm
 define amdgpu_kernel void @s_sint_to_fp_i1_to_f16(ptr addrspace(1) %out, ptr addrspace(1) %in0, ptr addrspace(1) %in1) {
+; SI-LABEL: s_sint_to_fp_i1_to_f16:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0xd
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_mov_b32 s10, s2
+; SI-NEXT:    s_mov_b32 s11, s3
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s12, s6
+; SI-NEXT:    s_mov_b32 s13, s7
+; SI-NEXT:    s_mov_b32 s14, s2
+; SI-NEXT:    s_mov_b32 s15, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    buffer_load_dword v1, off, s[12:15], 0
+; SI-NEXT:    s_waitcnt vmcnt(1)
+; SI-NEXT:    v_cmp_le_f32_e32 vcc, 1.0, v0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_le_f32_e64 s[0:1], 0, v1
+; SI-NEXT:    s_xor_b64 s[0:1], s[0:1], vcc
+; SI-NEXT:    v_cndmask_b32_e64 v0, 0, -1.0, s[0:1]
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    s_mov_b32 s0, s4
+; SI-NEXT:    s_mov_b32 s1, s5
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: s_sint_to_fp_i1_to_f16:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x34
+; VI-NEXT:    s_mov_b32 s3, 0xf000
+; VI-NEXT:    s_mov_b32 s2, -1
+; VI-NEXT:    s_mov_b32 s10, s2
+; VI-NEXT:    s_mov_b32 s11, s3
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s12, s6
+; VI-NEXT:    s_mov_b32 s13, s7
+; VI-NEXT:    s_mov_b32 s14, s2
+; VI-NEXT:    s_mov_b32 s15, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    buffer_load_dword v1, off, s[12:15], 0
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    v_cmp_le_f32_e32 vcc, 1.0, v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cmp_le_f32_e64 s[0:1], 0, v1
+; VI-NEXT:    s_xor_b64 s[0:1], s[0:1], vcc
+; VI-NEXT:    v_cndmask_b32_e64 v0, 0, -1.0, s[0:1]
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    s_mov_b32 s0, s4
+; VI-NEXT:    s_mov_b32 s1, s5
+; VI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: s_sint_to_fp_i1_to_f16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-NEXT:    s_mov_b32 s10, -1
+; GFX11-NEXT:    s_mov_b32 s11, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s2, s10
+; GFX11-NEXT:    s_mov_b32 s3, s11
+; GFX11-NEXT:    s_mov_b32 s14, s10
+; GFX11-NEXT:    s_mov_b32 s15, s11
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s12, s6
+; GFX11-NEXT:    s_mov_b32 s13, s7
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[0:3], 0
+; GFX11-NEXT:    buffer_load_b32 v1, off, s[12:15], 0
+; GFX11-NEXT:    s_mov_b32 s8, s4
+; GFX11-NEXT:    s_mov_b32 s9, s5
+; GFX11-NEXT:    s_waitcnt vmcnt(1)
+; GFX11-NEXT:    v_cmp_le_f32_e32 vcc_lo, 1.0, v0
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cmp_le_f32_e64 s0, 0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_xor_b32 s0, s0, vcc_lo
+; GFX11-NEXT:    v_cndmask_b32_e64 v0, 0, -1.0, s0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   %a = load float, ptr addrspace(1) %in0
   %b = load float, ptr addrspace(1) %in1
   %acmp = fcmp oge float %a, 0.000000e+00

diff --git a/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll b/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
index fd99d80..22ebab0 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
@@ -1,16 +1,64 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefixes=GFX11 %s
 
-; GCN-LABEL: {{^}}uitofp_i16_to_f16
-; GCN: buffer_load_ushort v[[A_I16:[0-9]+]]
-; SI:  v_cvt_f32_u32_e32 v[[A_F32:[0-9]+]], v[[A_I16]]
-; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]]
-
-; VI:  v_cvt_f16_u16_e32 v[[R_F16:[0-9]+]], v[[A_I16]]
-
-; GCN: buffer_store_short v[[R_F16]]
-; GCN: s_endpgm
 define amdgpu_kernel void @uitofp_i16_to_f16(
+; SI-LABEL: uitofp_i16_to_f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: uitofp_i16_to_f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f16_u16_e32 v0, v0
+; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: uitofp_i16_to_f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_u16 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f16_u16_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -20,13 +68,65 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}uitofp_i32_to_f16
-; GCN: buffer_load_dword v[[A_I32:[0-9]+]]
-; GCN: v_cvt_f32_u32_e32 v[[A_I16:[0-9]+]], v[[A_I32]]
-; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_I16]]
-; GCN: buffer_store_short v[[R_F16]]
-; GCN: s_endpgm
 define amdgpu_kernel void @uitofp_i32_to_f16(
+; SI-LABEL: uitofp_i32_to_f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: uitofp_i32_to_f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: uitofp_i32_to_f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -38,24 +138,74 @@ entry:
 
 ; f16 = uitofp i64 is in uint_to_fp.i64.ll
 
-; GCN-LABEL: {{^}}uitofp_v2i16_to_v2f16
-; GCN:     buffer_load_dword
-
-; SI: v_cvt_f32_u32_e32
-; SI: v_cvt_f32_u32_e32
-; SI: v_cvt_f16_f32_e32
-; SI: v_cvt_f16_f32_e32
-; SI-DAG: v_lshlrev_b32_e32
-; SI: v_or_b32_e32
-
-
-; VI-DAG: v_cvt_f16_u16_e32
-; VI-DAG: v_cvt_f16_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
-; VI:     v_or_b32_e32
-
-; GCN: buffer_store_dword
-; GCN: s_endpgm
 define amdgpu_kernel void @uitofp_v2i16_to_v2f16(
+; SI-LABEL: uitofp_v2i16_to_v2f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_and_b32_e32 v1, 0xffff, v0
+; SI-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
+; SI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; SI-NEXT:    v_cvt_f32_u32_e32 v1, v1
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT:    v_or_b32_e32 v0, v1, v0
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: uitofp_v2i16_to_v2f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f16_u16_sdwa v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
+; VI-NEXT:    v_cvt_f16_u16_e32 v0, v0
+; VI-NEXT:    v_or_b32_e32 v0, v0, v1
+; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: uitofp_v2i16_to_v2f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-NEXT:    v_cvt_f16_u16_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_u16_e32 v1, v1
+; GFX11-NEXT:    v_pack_b32_f16 v0, v0, v1
+; GFX11-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -65,25 +215,76 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}uitofp_v2i32_to_v2f16
-; GCN:     buffer_load_dwordx2
-
-; SI: v_cvt_f32_u32_e32
-; SI: v_cvt_f32_u32_e32
-; SI: v_cvt_f16_f32_e32
-; SI: v_cvt_f16_f32_e32
-; SI-DAG: v_lshlrev_b32_e32
-; SI: v_or_b32_e32
-
-; VI-DAG: v_cvt_f32_u32_e32
-; VI-DAG: v_cvt_f32_u32_e32
-; VI-DAG: v_cvt_f16_f32_e32
-; VI-DAG: v_cvt_f16_f32_sdwa
-; VI:     v_or_b32_e32
-
-; GCN:     buffer_store_dword
-; GCN:     s_endpgm
 define amdgpu_kernel void @uitofp_v2i32_to_v2f16(
+; SI-LABEL: uitofp_v2i32_to_v2f16:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s10, s6
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s8, s2
+; SI-NEXT:    s_mov_b32 s9, s3
+; SI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cvt_f32_u32_e32 v1, v1
+; SI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; SI-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT:    v_or_b32_e32 v0, v0, v1
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: uitofp_v2i32_to_v2f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_mov_b32 s7, 0xf000
+; VI-NEXT:    s_mov_b32 s6, -1
+; VI-NEXT:    s_mov_b32 s10, s6
+; VI-NEXT:    s_mov_b32 s11, s7
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s8, s2
+; VI-NEXT:    s_mov_b32 s9, s3
+; VI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; VI-NEXT:    s_mov_b32 s4, s0
+; VI-NEXT:    s_mov_b32 s5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f32_u32_e32 v1, v1
+; VI-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; VI-NEXT:    v_cvt_f16_f32_sdwa v1, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    v_or_b32_e32 v0, v0, v1
+; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: uitofp_v2i32_to_v2f16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_mov_b32 s6, -1
+; GFX11-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s10, s6
+; GFX11-NEXT:    s_mov_b32 s11, s7
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s8, s2
+; GFX11-NEXT:    s_mov_b32 s9, s3
+; GFX11-NEXT:    s_mov_b32 s4, s0
+; GFX11-NEXT:    buffer_load_b64 v[0:1], off, s[8:11], 0
+; GFX11-NEXT:    s_mov_b32 s5, s1
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cvt_f32_u32_e32 v1, v1
+; GFX11-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_pack_b32_f16 v0, v0, v1
+; GFX11-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) {
 entry:
@@ -93,15 +294,91 @@ entry:
   ret void
 }
 
-; GCN-LABEL: {{^}}s_uint_to_fp_i1_to_f16:
-; GCN-DAG: v_cmp_le_f32_e32 [[CMP0:vcc]], 1.0, {{v[0-9]+}}
-; GCN-DAG: v_cmp_le_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]], 0, {{v[0-9]+}}
-; GCN: s_xor_b64 [[R_CMP:s\[[0-9]+:[0-9]+\]]], [[CMP1]], [[CMP0]]
-; GCN: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, 1.0, [[R_CMP]]
-; GCN-NEXT: v_cvt_f16_f32_e32 [[R_F16:v[0-9]+]], [[RESULT]]
-; GCN: buffer_store_short
-; GCN: s_endpgm
 define amdgpu_kernel void @s_uint_to_fp_i1_to_f16(ptr addrspace(1) %out, ptr addrspace(1) %in0, ptr addrspace(1) %in1) {
+; SI-LABEL: s_uint_to_fp_i1_to_f16:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0xd
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_mov_b32 s10, s2
+; SI-NEXT:    s_mov_b32 s11, s3
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s12, s6
+; SI-NEXT:    s_mov_b32 s13, s7
+; SI-NEXT:    s_mov_b32 s14, s2
+; SI-NEXT:    s_mov_b32 s15, s3
+; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NEXT:    buffer_load_dword v1, off, s[12:15], 0
+; SI-NEXT:    s_waitcnt vmcnt(1)
+; SI-NEXT:    v_cmp_le_f32_e32 vcc, 1.0, v0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_le_f32_e64 s[0:1], 0, v1
+; SI-NEXT:    s_xor_b64 s[0:1], s[0:1], vcc
+; SI-NEXT:    v_cndmask_b32_e64 v0, 0, 1.0, s[0:1]
+; SI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; SI-NEXT:    s_mov_b32 s0, s4
+; SI-NEXT:    s_mov_b32 s1, s5
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: s_uint_to_fp_i1_to_f16:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x34
+; VI-NEXT:    s_mov_b32 s3, 0xf000
+; VI-NEXT:    s_mov_b32 s2, -1
+; VI-NEXT:    s_mov_b32 s10, s2
+; VI-NEXT:    s_mov_b32 s11, s3
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_mov_b32 s12, s6
+; VI-NEXT:    s_mov_b32 s13, s7
+; VI-NEXT:    s_mov_b32 s14, s2
+; VI-NEXT:    s_mov_b32 s15, s3
+; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; VI-NEXT:    buffer_load_dword v1, off, s[12:15], 0
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    v_cmp_le_f32_e32 vcc, 1.0, v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cmp_le_f32_e64 s[0:1], 0, v1
+; VI-NEXT:    s_xor_b64 s[0:1], s[0:1], vcc
+; VI-NEXT:    v_cndmask_b32_e64 v0, 0, 1.0, s[0:1]
+; VI-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; VI-NEXT:    s_mov_b32 s0, s4
+; VI-NEXT:    s_mov_b32 s1, s5
+; VI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; VI-NEXT:    s_endpgm
+;
+; GFX11-LABEL: s_uint_to_fp_i1_to_f16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-NEXT:    s_mov_b32 s10, -1
+; GFX11-NEXT:    s_mov_b32 s11, 0x31016000
+; GFX11-NEXT:    s_mov_b32 s2, s10
+; GFX11-NEXT:    s_mov_b32 s3, s11
+; GFX11-NEXT:    s_mov_b32 s14, s10
+; GFX11-NEXT:    s_mov_b32 s15, s11
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s12, s6
+; GFX11-NEXT:    s_mov_b32 s13, s7
+; GFX11-NEXT:    buffer_load_b32 v0, off, s[0:3], 0
+; GFX11-NEXT:    buffer_load_b32 v1, off, s[12:15], 0
+; GFX11-NEXT:    s_mov_b32 s8, s4
+; GFX11-NEXT:    s_mov_b32 s9, s5
+; GFX11-NEXT:    s_waitcnt vmcnt(1)
+; GFX11-NEXT:    v_cmp_le_f32_e32 vcc_lo, 1.0, v0
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_cmp_le_f32_e64 s0, 0, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_xor_b32 s0, s0, vcc_lo
+; GFX11-NEXT:    v_cndmask_b32_e64 v0, 0, 1.0, s0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX11-NEXT:    buffer_store_b16 v0, off, s[8:11], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   %a = load float, ptr addrspace(1) %in0
   %b = load float, ptr addrspace(1) %in1
   %acmp = fcmp oge float %a, 0.000000e+00