Add intrinsic for s_bitreplicate. Lower to S_BITREPLICATE_B64_B32 machine instruction in both GISel and Selection DAG. Support VGPR arguments by inserting a `v_readfirstlane`.
46 lines
1.8 KiB
LLVM
46 lines
1.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
|
|
; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -global-isel=1 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
|
|
; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -global-isel=0 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
|
|
|
|
declare i64 @llvm.amdgcn.s.bitreplicate(i32)
|
|
|
|
define i64 @test_s_bitreplicate_constant() {
|
|
; GFX11-LABEL: test_s_bitreplicate_constant:
|
|
; GFX11: ; %bb.0: ; %entry
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: s_bitreplicate_b64_b32 s[0:1], 0x85fe3a92
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
entry:
|
|
%br = call i64 @llvm.amdgcn.s.bitreplicate(i32 u0x85FE3A92)
|
|
ret i64 %br
|
|
}
|
|
|
|
define amdgpu_cs void @test_s_bitreplicate_sgpr(i32 inreg %mask, ptr addrspace(1) %out) {
|
|
; GFX11-LABEL: test_s_bitreplicate_sgpr:
|
|
; GFX11: ; %bb.0: ; %entry
|
|
; GFX11-NEXT: s_bitreplicate_b64_b32 s[0:1], s0
|
|
; GFX11-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_nop 0
|
|
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
|
|
; GFX11-NEXT: s_endpgm
|
|
entry:
|
|
%br = call i64 @llvm.amdgcn.s.bitreplicate(i32 %mask)
|
|
store i64 %br, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define i64 @test_s_bitreplicate_vgpr(i32 %mask) {
|
|
; GFX11-LABEL: test_s_bitreplicate_vgpr:
|
|
; GFX11: ; %bb.0: ; %entry
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX11-NEXT: s_bitreplicate_b64_b32 s[0:1], s0
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
entry:
|
|
%br = call i64 @llvm.amdgcn.s.bitreplicate(i32 %mask)
|
|
ret i64 %br
|
|
}
|