[llvm] 093aa37 - [AMDGPU] Autogenerate min.ll/max.ll tests. NFC. (#65786)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Sep 11 10:29:58 PDT 2023
Author: Stanislav Mekhanoshin
Date: 2023-09-11T10:29:53-07:00
New Revision: 093aa37744462971a71dafba4fe308a5929f26a0
URL: https://github.com/llvm/llvm-project/commit/093aa37744462971a71dafba4fe308a5929f26a0
DIFF: https://github.com/llvm/llvm-project/commit/093aa37744462971a71dafba4fe308a5929f26a0.diff
LOG: [AMDGPU] Autogenerate min.ll/max.ll tests. NFC. (#65786)
Added:
Modified:
llvm/test/CodeGen/AMDGPU/max.ll
llvm/test/CodeGen/AMDGPU/min.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AMDGPU/max.ll b/llvm/test/CodeGen/AMDGPU/max.ll
index 52da8fe00168e16..ad8ef8bc90c05d8 100644
--- a/llvm/test/CodeGen/AMDGPU/max.ll
+++ b/llvm/test/CodeGen/AMDGPU/max.ll
@@ -1,12 +1,49 @@
-; RUN: llc -march=amdgcn -mcpu=pitcairn < %s | FileCheck -enable-var-scope -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -enable-var-scope -check-prefix=EG -check-prefix=FUNC %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -march=amdgcn -mcpu=pitcairn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=SI %s
+; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=EG %s
-
-; FUNC-LABEL: {{^}}v_test_imax_sge_i32:
-; SI: v_max_i32_e32
-
-; EG: MAX_INT
define amdgpu_kernel void @v_test_imax_sge_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_imax_sge_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xd
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s6, 0
+; SI-NEXT: s_mov_b32 s7, s3
+; SI-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; SI-NEXT: v_mov_b32_e32 v1, 0
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_load_dword s6, s[6:7], 0x0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; SI-NEXT: v_max_i32_e32 v0, s6, v0
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_imax_sge_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 2, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T1.X, KC0[2].Z,
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MAX_INT T0.X, T1.X, T0.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.in = getelementptr inbounds i32, ptr addrspace(1) %bptr, i32 %tid
%a = load i32, ptr addrspace(1) %aptr, align 4
@@ -17,18 +54,55 @@ define amdgpu_kernel void @v_test_imax_sge_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: {{^}}v_test_imax_sge_v4i32:
-; SI: v_max_i32_e32
-; SI: v_max_i32_e32
-; SI: v_max_i32_e32
-; SI: v_max_i32_e32
-
; These could be merged into one
-; EG: MAX_INT
-; EG: MAX_INT
-; EG: MAX_INT
-; EG: MAX_INT
define amdgpu_kernel void @v_test_imax_sge_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_imax_sge_v4i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xd
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s6, 0
+; SI-NEXT: s_mov_b32 s7, s3
+; SI-NEXT: v_lshlrev_b32_e32 v0, 4, v0
+; SI-NEXT: v_mov_b32_e32 v1, 0
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_load_dwordx4 s[8:11], s[6:7], 0x0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; SI-NEXT: v_max_i32_e32 v3, s11, v3
+; SI-NEXT: v_max_i32_e32 v2, s10, v2
+; SI-NEXT: v_max_i32_e32 v1, s9, v1
+; SI-NEXT: v_max_i32_e32 v0, s8, v0
+; SI-NEXT: buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_imax_sge_v4i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 5, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT: VTX_READ_128 T1.XYZW, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T1.X, KC0[2].Z,
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MAX_INT * T0.W, T1.W, T0.W,
+; EG-NEXT: MAX_INT * T0.Z, T1.Z, T0.Z,
+; EG-NEXT: MAX_INT * T0.Y, T1.Y, T0.Y,
+; EG-NEXT: MAX_INT T0.X, T1.X, T0.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.in = getelementptr inbounds <4 x i32>, ptr addrspace(1) %bptr, i32 %tid
%a = load <4 x i32>, ptr addrspace(1) %aptr, align 4
@@ -39,35 +113,122 @@ define amdgpu_kernel void @v_test_imax_sge_v4i32(ptr addrspace(1) %out, ptr addr
ret void
}
-; FUNC-LABEL: @s_test_imax_sge_i32
-; SI: s_max_i32
-
-; EG: MAX_INT
define amdgpu_kernel void @s_test_imax_sge_i32(ptr addrspace(1) %out, i32 %a, i32 %b) nounwind {
+; SI-LABEL: s_test_imax_sge_i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_max_i32 s0, s2, s3
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sge_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_INT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp sge i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_imax_sge_imm_i32:
-; SI: s_max_i32 {{s[0-9]+}}, {{s[0-9]+}}, 9
-
-; EG: MAX_INT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_imax_sge_imm_i32(ptr addrspace(1) %out, i32 %a) nounwind {
+; SI-LABEL: s_test_imax_sge_imm_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dword s4, s[0:1], 0xb
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_max_i32 s4, s4, 9
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sge_imm_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_INT * T1.X, KC0[2].Z, literal.y,
+; EG-NEXT: 2(2.802597e-45), 9(1.261169e-44)
%cmp = icmp sge i32 %a, 9
%val = select i1 %cmp, i32 %a, i32 9
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}v_test_imax_sge_i8:
-; SI: buffer_load_sbyte
-; SI: buffer_load_sbyte
-; SI: v_max_i32_e32
-
-; EG: MAX_INT
define amdgpu_kernel void @v_test_imax_sge_i8(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_imax_sge_i8:
+; 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 s14, s2
+; 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 s15, s3
+; SI-NEXT: s_mov_b32 s10, s2
+; SI-NEXT: s_mov_b32 s11, s3
+; SI-NEXT: buffer_load_sbyte v0, off, s[12:15], 0
+; SI-NEXT: buffer_load_sbyte v1, off, s[8:11], 0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_max_i32_e32 v0, v0, v1
+; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_imax_sge_i8:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @8
+; EG-NEXT: ALU 0, @13, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @10
+; EG-NEXT: ALU 14, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 8:
+; EG-NEXT: VTX_READ_8 T0.X, T0.X, 0, #1
+; EG-NEXT: Fetch clause starting at 10:
+; EG-NEXT: VTX_READ_8 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: MOV * T0.X, KC0[2].W,
+; EG-NEXT: ALU clause starting at 13:
+; EG-NEXT: MOV * T1.X, KC0[2].Z,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: BFE_INT T0.Z, T0.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.W, T1.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: AND_INT * T1.W, KC0[2].Y, literal.y,
+; EG-NEXT: 8(1.121039e-44), 3(4.203895e-45)
+; EG-NEXT: MAX_INT * T0.W, PV.W, PV.Z,
+; EG-NEXT: AND_INT T0.W, PV.W, literal.x,
+; EG-NEXT: LSHL * T1.W, T1.W, literal.y,
+; EG-NEXT: 255(3.573311e-43), 3(4.203895e-45)
+; EG-NEXT: LSHL T0.X, PV.W, PS,
+; EG-NEXT: LSHL * T0.W, literal.x, PS,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%a = load i8, ptr addrspace(1) %aptr, align 1
%b = load i8, ptr addrspace(1) %bptr, align 1
%cmp = icmp sge i8 %a, %b
@@ -76,35 +237,111 @@ define amdgpu_kernel void @v_test_imax_sge_i8(ptr addrspace(1) %out, ptr addrspa
ret void
}
-; FUNC-LABEL: {{^}}s_test_imax_sgt_imm_i32:
-; SI: s_max_i32 {{s[0-9]+}}, {{s[0-9]+}}, 9
-
-; EG: MAX_INT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_imax_sgt_imm_i32(ptr addrspace(1) %out, i32 %a) nounwind {
+; SI-LABEL: s_test_imax_sgt_imm_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dword s4, s[0:1], 0xb
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_max_i32 s4, s4, 9
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sgt_imm_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_INT * T1.X, KC0[2].Z, literal.y,
+; EG-NEXT: 2(2.802597e-45), 9(1.261169e-44)
%cmp = icmp sgt i32 %a, 9
%val = select i1 %cmp, i32 %a, i32 9
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_imax_sgt_imm_v2i32:
-; SI: s_max_i32 {{s[0-9]+}}, {{s[0-9]+}}, 9
-; SI: s_max_i32 {{s[0-9]+}}, {{s[0-9]+}}, 9
-
-; EG: MAX_INT {{.*}}literal.{{[xyzw]}}
-; EG: MAX_INT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_imax_sgt_imm_v2i32(ptr addrspace(1) %out, <2 x i32> %a) nounwind {
+; SI-LABEL: s_test_imax_sgt_imm_v2i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: s_max_i32 s0, s3, 9
+; SI-NEXT: s_max_i32 s1, s2, 9
+; SI-NEXT: v_mov_b32_e32 v0, s1
+; SI-NEXT: v_mov_b32_e32 v1, s0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sgt_imm_v2i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 4, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MAX_INT * T0.Y, KC0[3].X, literal.x,
+; EG-NEXT: 9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT: MAX_INT T0.X, KC0[2].W, literal.x,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT: 9(1.261169e-44), 2(2.802597e-45)
%cmp = icmp sgt <2 x i32> %a, <i32 9, i32 9>
%val = select <2 x i1> %cmp, <2 x i32> %a, <2 x i32> <i32 9, i32 9>
store <2 x i32> %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: @v_test_imax_sgt_i32
-; SI: v_max_i32_e32
-
-; EG: MAX_INT
define amdgpu_kernel void @v_test_imax_sgt_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_imax_sgt_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xd
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s6, 0
+; SI-NEXT: s_mov_b32 s7, s3
+; SI-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; SI-NEXT: v_mov_b32_e32 v1, 0
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_load_dword s6, s[6:7], 0x0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; SI-NEXT: v_max_i32_e32 v0, s6, v0
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_imax_sgt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 2, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T1.X, KC0[2].Z,
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MAX_INT T0.X, T1.X, T0.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.in = getelementptr inbounds i32, ptr addrspace(1) %bptr, i32 %tid
%a = load i32, ptr addrspace(1) %aptr, align 4
@@ -115,22 +352,78 @@ define amdgpu_kernel void @v_test_imax_sgt_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: @s_test_imax_sgt_i32
-; SI: s_max_i32
-
-; EG: MAX_INT
define amdgpu_kernel void @s_test_imax_sgt_i32(ptr addrspace(1) %out, i32 %a, i32 %b) nounwind {
+; SI-LABEL: s_test_imax_sgt_i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_max_i32 s0, s2, s3
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sgt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_INT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp sgt i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: @v_test_umax_uge_i32
-; SI: v_max_u32_e32
-
-; EG: MAX_UINT
define amdgpu_kernel void @v_test_umax_uge_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_umax_uge_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xd
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s6, 0
+; SI-NEXT: s_mov_b32 s7, s3
+; SI-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; SI-NEXT: v_mov_b32_e32 v1, 0
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_load_dword s6, s[6:7], 0x0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; SI-NEXT: v_max_u32_e32 v0, s6, v0
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_umax_uge_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 2, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T1.X, KC0[2].Z,
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MAX_UINT T0.X, T1.X, T0.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.in = getelementptr inbounds i32, ptr addrspace(1) %bptr, i32 %tid
%a = load i32, ptr addrspace(1) %aptr, align 4
@@ -141,42 +434,126 @@ define amdgpu_kernel void @v_test_umax_uge_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: @s_test_umax_uge_i32
-; SI: s_max_u32
-
-; EG: MAX_UINT
define amdgpu_kernel void @s_test_umax_uge_i32(ptr addrspace(1) %out, i32 %a, i32 %b) nounwind {
+; SI-LABEL: s_test_umax_uge_i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_max_u32 s0, s2, s3
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_umax_uge_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_UINT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp uge i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_umax_uge_v3i32:
-; SI: s_max_u32
-; SI: s_max_u32
-; SI: s_max_u32
-; SI-NOT: s_max_u32
-; SI: s_endpgm
-
-; EG: MAX_UINT
-; EG: MAX_UINT
-; EG: MAX_UINT
-; EG-NOT: MAX_UINT
define amdgpu_kernel void @s_test_umax_uge_v3i32(ptr addrspace(1) %out, <3 x i32> %a, <3 x i32> %b) nounwind {
+; SI-LABEL: s_test_umax_uge_v3i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx8 s[4:11], s[0:1], 0xd
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_max_u32 s6, s6, s10
+; SI-NEXT: s_max_u32 s5, s5, s9
+; SI-NEXT: s_max_u32 s4, s4, s8
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0 offset:8
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: v_mov_b32_e32 v1, s5
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_umax_uge_v3i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MAX_UINT * T0.Y, KC0[3].Z, KC0[4].Z,
+; EG-NEXT: MAX_UINT * T0.X, KC0[3].Y, KC0[4].Y,
+; EG-NEXT: LSHR T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, literal.y,
+; EG-NEXT: 2(2.802597e-45), 8(1.121039e-44)
+; EG-NEXT: LSHR T2.X, PV.W, literal.x,
+; EG-NEXT: MAX_UINT * T3.X, KC0[3].W, KC0[4].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp uge <3 x i32> %a, %b
%val = select <3 x i1> %cmp, <3 x i32> %a, <3 x i32> %b
store <3 x i32> %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}v_test_umax_uge_i8:
-; SI: buffer_load_ubyte
-; SI: buffer_load_ubyte
-; SI: v_max_u32_e32
-
-; EG: MAX_UINT
define amdgpu_kernel void @v_test_umax_uge_i8(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_umax_uge_i8:
+; 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 s14, s2
+; 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 s15, s3
+; SI-NEXT: s_mov_b32 s10, s2
+; SI-NEXT: s_mov_b32 s11, s3
+; SI-NEXT: buffer_load_ubyte v0, off, s[12:15], 0
+; SI-NEXT: buffer_load_ubyte v1, off, s[8:11], 0
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_max_u32_e32 v0, v0, v1
+; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_umax_uge_i8:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 11, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_8 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_8 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T0.X, KC0[2].Z,
+; EG-NEXT: MOV * T1.X, KC0[2].W,
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: AND_INT T0.W, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_UINT * T1.W, T0.X, T1.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T0.X, T1.W, PV.W,
+; EG-NEXT: LSHL * T0.W, literal.x, PV.W,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%a = load i8, ptr addrspace(1) %aptr, align 1
%b = load i8, ptr addrspace(1) %bptr, align 1
%cmp = icmp uge i8 %a, %b
@@ -185,11 +562,45 @@ define amdgpu_kernel void @v_test_umax_uge_i8(ptr addrspace(1) %out, ptr addrspa
ret void
}
-; FUNC-LABEL: @v_test_umax_ugt_i32
-; SI: v_max_u32_e32
-
-; EG: MAX_UINT
define amdgpu_kernel void @v_test_umax_ugt_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
+; SI-LABEL: v_test_umax_ugt_i32:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xd
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s6, 0
+; SI-NEXT: s_mov_b32 s7, s3
+; SI-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; SI-NEXT: v_mov_b32_e32 v1, 0
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: s_load_dword s4, s[4:5], 0x0
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; SI-NEXT: v_max_u32_e32 v0, s4, v0
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: v_test_umax_ugt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 2, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T1.X, KC0[2].W,
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MAX_UINT T0.X, T0.X, T1.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.in = getelementptr inbounds i32, ptr addrspace(1) %bptr, i32 %tid
%a = load i32, ptr addrspace(1) %gep.in, align 4
@@ -200,24 +611,64 @@ define amdgpu_kernel void @v_test_umax_ugt_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: {{^}}s_test_umax_ugt_i32:
-; SI: s_max_u32
-
-; EG: MAX_UINT
define amdgpu_kernel void @s_test_umax_ugt_i32(ptr addrspace(1) %out, i32 %a, i32 %b) nounwind {
+; SI-LABEL: s_test_umax_ugt_i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_max_u32 s0, s2, s3
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_umax_ugt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MAX_UINT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp ugt i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_umax_ugt_imm_v2i32:
-; SI-DAG: s_max_u32 {{s[0-9]+}}, {{s[0-9]+}}, 15
-; SI-DAG: s_max_u32 {{s[0-9]+}}, {{s[0-9]+}}, 23
-
-; EG: MAX_UINT {{.*}}literal.{{[xyzw]}}
-; EG: MAX_UINT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_umax_ugt_imm_v2i32(ptr addrspace(1) %out, <2 x i32> %a) nounwind {
+; SI-LABEL: s_test_umax_ugt_imm_v2i32:
+; SI: ; %bb.0:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s4, s0
+; SI-NEXT: s_mov_b32 s5, s1
+; SI-NEXT: s_max_u32 s0, s3, 23
+; SI-NEXT: s_max_u32 s1, s2, 15
+; SI-NEXT: v_mov_b32_e32 v0, s1
+; SI-NEXT: v_mov_b32_e32 v1, s0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_umax_ugt_imm_v2i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 4, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MAX_UINT * T0.Y, KC0[3].X, literal.x,
+; EG-NEXT: 23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT: MAX_UINT T0.X, KC0[2].W, literal.x,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT: 15(2.101948e-44), 2(2.802597e-45)
%cmp = icmp ugt <2 x i32> %a, <i32 15, i32 23>
%val = select <2 x i1> %cmp, <2 x i32> %a, <2 x i32> <i32 15, i32 23>
store <2 x i32> %val, ptr addrspace(1) %out, align 4
@@ -225,17 +676,43 @@ define amdgpu_kernel void @s_test_umax_ugt_imm_v2i32(ptr addrspace(1) %out, <2 x
}
; Make sure redundant and removed
-; FUNC-LABEL: {{^}}simplify_demanded_bits_test_umax_ugt_i16:
-; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0x13
-; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0x1c
-; SI-DAG: s_and_b32 [[A16:s[0-9]+]], [[A]], 0xffff
-; SI-DAG: s_and_b32 [[B16:s[0-9]+]], [[B]], 0xffff
-; SI: s_max_u32 [[MAX:s[0-9]+]], [[A16]], [[B16]]
-; SI: v_mov_b32_e32 [[VMAX:v[0-9]+]], [[MAX]]
-; SI: buffer_store_dword [[VMAX]]
-; EG: MAX_UINT
define amdgpu_kernel void @simplify_demanded_bits_test_umax_ugt_i16(ptr addrspace(1) %out, [8 x i32], i16 zeroext %a, [8 x i32], i16 zeroext %b) nounwind {
+; SI-LABEL: simplify_demanded_bits_test_umax_ugt_i16:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dword s4, s[0:1], 0x13
+; SI-NEXT: s_load_dword s5, s[0:1], 0x1c
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_and_b32 s5, s5, 0xffff
+; SI-NEXT: s_max_u32 s4, s4, s5
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: simplify_demanded_bits_test_umax_ugt_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 5, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MAX_UINT T0.X, PV.Z, PV.W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%a.ext = zext i16 %a to i32
%b.ext = zext i16 %b to i32
%cmp = icmp ugt i32 %a.ext, %b.ext
@@ -247,18 +724,42 @@ define amdgpu_kernel void @simplify_demanded_bits_test_umax_ugt_i16(ptr addrspac
; Make sure redundant sign_extend_inreg removed.
-; FUNC-LABEL: {{^}}simplify_demanded_bits_test_max_slt_i16:
-; SI-DAG: s_load_dword [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0x13
-; SI-DAG: s_load_dword [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, 0x1c
-; SI-DAG: s_sext_i32_i16 [[EXT_A:s[0-9]+]], [[A]]
-; SI-DAG: s_sext_i32_i16 [[EXT_B:s[0-9]+]], [[B]]
-
-; SI: s_max_i32 [[MAX:s[0-9]+]], [[EXT_A]], [[EXT_B]]
-; SI: v_mov_b32_e32 [[VMAX:v[0-9]+]], [[MAX]]
-; SI: buffer_store_dword [[VMAX]]
-
-; EG: MAX_INT
define amdgpu_kernel void @simplify_demanded_bits_test_max_slt_i16(ptr addrspace(1) %out, [8 x i32], i16 signext %a, [8 x i32], i16 signext %b) nounwind {
+; SI-LABEL: simplify_demanded_bits_test_max_slt_i16:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dword s4, s[0:1], 0x13
+; SI-NEXT: s_load_dword s5, s[0:1], 0x1c
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_sext_i32_i16 s4, s4
+; SI-NEXT: s_sext_i32_i16 s5, s5
+; SI-NEXT: s_max_i32 s4, s4, s5
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: simplify_demanded_bits_test_max_slt_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 5, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MAX_INT T0.X, PV.Z, PV.W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%a.ext = sext i16 %a to i32
%b.ext = sext i16 %b to i32
%cmp = icmp sgt i32 %a.ext, %b.ext
@@ -269,15 +770,51 @@ define amdgpu_kernel void @simplify_demanded_bits_test_max_slt_i16(ptr addrspace
ret void
}
-; FUNC-LABEL: {{^}}s_test_imax_sge_i16:
-; SI: s_load_dword
-; SI: s_load_dword
-; SI: s_sext_i32_i16
-; SI: s_sext_i32_i16
-; SI: s_max_i32
-
-; EG: MAX_INT
define amdgpu_kernel void @s_test_imax_sge_i16(ptr addrspace(1) %out, [8 x i32], i16 %a, [8 x i32], i16 %b) nounwind {
+; SI-LABEL: s_test_imax_sge_i16:
+; SI: ; %bb.0:
+; SI-NEXT: s_load_dword s4, s[0:1], 0x13
+; SI-NEXT: s_load_dword s5, s[0:1], 0x1c
+; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_sext_i32_i16 s4, s4
+; SI-NEXT: s_sext_i32_i16 s5, s5
+; SI-NEXT: s_max_i32 s4, s4, s5
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: s_test_imax_sge_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 14, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: AND_INT * T1.W, KC0[2].Y, literal.y,
+; EG-NEXT: 16(2.242078e-44), 3(4.203895e-45)
+; EG-NEXT: MAX_INT * T0.W, PV.Z, PV.W,
+; EG-NEXT: AND_INT T0.W, PV.W, literal.x,
+; EG-NEXT: LSHL * T1.W, T1.W, literal.y,
+; EG-NEXT: 65535(9.183409e-41), 3(4.203895e-45)
+; EG-NEXT: LSHL T0.X, PV.W, PS,
+; EG-NEXT: LSHL * T0.W, literal.x, PS,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%cmp = icmp sge i16 %a, %b
%val = select i1 %cmp, i16 %a, i16 %b
store i16 %val, ptr addrspace(1) %out
@@ -285,71 +822,175 @@ define amdgpu_kernel void @s_test_imax_sge_i16(ptr addrspace(1) %out, [8 x i32],
}
; 64 bit
-; FUNC-LABEL: {{^}}test_umax_ugt_i64
-; SI: s_endpgm
-; EG: SETE_INT
-; EG: SETGT_UINT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_umax_ugt_i64(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind {
+; SI-LABEL: test_umax_ugt_i64:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
+; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1]
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_cselect_b32 s4, s7, s9
+; SI-NEXT: s_cselect_b32 s5, s6, s8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: v_mov_b32_e32 v1, s4
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: test_umax_ugt_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T0.W, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[2].W, KC0[3].Y,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tmp = icmp ugt i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_umax_uge_i64
-; SI: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_UINT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_umax_uge_i64(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind {
+; SI-LABEL: test_umax_uge_i64:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
+; SI-NEXT: v_cmp_ge_u64_e32 vcc, s[6:7], v[0:1]
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_cselect_b32 s4, s7, s9
+; SI-NEXT: s_cselect_b32 s5, s6, s8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: v_mov_b32_e32 v1, s4
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: test_umax_uge_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T0.W, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[2].W, KC0[3].Y,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tmp = icmp uge i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_imax_sgt_i64
-; SI: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_INT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_imax_sgt_i64(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind {
+; SI-LABEL: test_imax_sgt_i64:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
+; SI-NEXT: v_cmp_gt_i64_e32 vcc, s[6:7], v[0:1]
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_cselect_b32 s4, s7, s9
+; SI-NEXT: s_cselect_b32 s5, s6, s8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: v_mov_b32_e32 v1, s4
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: test_imax_sgt_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_INT * T0.W, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[2].W, KC0[3].Y,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tmp = icmp sgt i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_imax_sge_i64
-; SI: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_INT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_imax_sge_i64(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind {
+; SI-LABEL: test_imax_sge_i64:
+; 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_waitcnt lgkmcnt(0)
+; SI-NEXT: s_mov_b32 s0, s4
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[6:7], v[0:1]
+; SI-NEXT: s_mov_b32 s1, s5
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_cselect_b32 s4, s7, s9
+; SI-NEXT: s_cselect_b32 s5, s6, s8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: v_mov_b32_e32 v1, s4
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; EG-LABEL: test_imax_sge_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_INT * T0.W, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[2].W, KC0[3].Y,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%tmp = icmp sge i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-
declare i32 @llvm.amdgcn.workitem.id.x() #0
attributes #0 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/min.ll b/llvm/test/CodeGen/AMDGPU/min.ll
index efe8e419e0af0ef..9dafa27ece86f65 100644
--- a/llvm/test/CodeGen/AMDGPU/min.ll
+++ b/llvm/test/CodeGen/AMDGPU/min.ll
@@ -1,15 +1,122 @@
-; RUN: llc -mtriple=r600-- -mcpu=cypress -verify-machineinstrs < %s | FileCheck --check-prefixes=EG,FUNC %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,CI,FUNC %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=tonga -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,VI,GFX8_9_10,FUNC %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX9_10,GFX8_9_10,FUNC %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10,GFX9_10,GFX8_9_10,FUNC %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10,GFX9_10,GFX8_9_10,FUNC %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=r600-- -mcpu=cypress -verify-machineinstrs < %s | FileCheck --check-prefix=EG %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck --check-prefix=CI %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=tonga -verify-machineinstrs < %s | FileCheck --check-prefix=VI %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck --check-prefixes=GFX11 %s
-; FUNC-LABEL: {{^}}v_test_imin_sle_i32:
-; GCN: v_min_i32_e32
-
-; EG: MIN_INT
define amdgpu_kernel void @v_test_imin_sle_i32(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_imin_sle_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 3, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T1.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MIN_INT T0.X, T0.X, T1.X,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_imin_sle_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v5, v[0:1]
+; CI-NEXT: flat_load_dword v2, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_i32_e32 v2, v5, v2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_imin_sle_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_i32_e32 v2, v5, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_imin_sle_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_imin_sle_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_imin_sle_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds i32, ptr addrspace(1) %b.ptr, i32 %tid
@@ -22,52 +129,368 @@ define amdgpu_kernel void @v_test_imin_sle_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_i32:
-; GCN: s_min_i32
-
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_i32(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; EG-LABEL: s_test_imin_sle_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_INT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_v1i32:
-; GCN: s_min_i32
-
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_v1i32(ptr addrspace(1) %out, <1 x i32> %a, <1 x i32> %b) #0 {
+; EG-LABEL: s_test_imin_sle_v1i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_INT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_v1i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_v1i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_v1i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_v1i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_v1i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle <1 x i32> %a, %b
%val = select <1 x i1> %cmp, <1 x i32> %a, <1 x i32> %b
store <1 x i32> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_v4i32:
-; GCN: s_min_i32
-; GCN: s_min_i32
-; GCN: s_min_i32
-; GCN: s_min_i32
-
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_v4i32(ptr addrspace(1) %out, <4 x i32> %a, <4 x i32> %b) #0 {
+; EG-LABEL: s_test_imin_sle_v4i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MIN_INT * T0.W, KC0[4].X, KC0[5].X,
+; EG-NEXT: MIN_INT * T0.Z, KC0[3].W, KC0[4].W,
+; EG-NEXT: MIN_INT * T0.Y, KC0[3].Z, KC0[4].Z,
+; EG-NEXT: MIN_INT * T0.X, KC0[3].Y, KC0[4].Y,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_v4i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x4
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s11, s15
+; CI-NEXT: s_min_i32 s3, s10, s14
+; CI-NEXT: s_min_i32 s4, s9, s13
+; CI-NEXT: s_min_i32 s5, s8, s12
+; CI-NEXT: v_mov_b32_e32 v5, s1
+; CI-NEXT: v_mov_b32_e32 v0, s5
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s3
+; CI-NEXT: v_mov_b32_e32 v3, s2
+; CI-NEXT: v_mov_b32_e32 v4, s0
+; CI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_v4i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s11, s15
+; VI-NEXT: s_min_i32 s3, s10, s14
+; VI-NEXT: s_min_i32 s4, s9, s13
+; VI-NEXT: s_min_i32 s5, s8, s12
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: v_mov_b32_e32 v0, s5
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: v_mov_b32_e32 v3, s2
+; VI-NEXT: v_mov_b32_e32 v4, s0
+; VI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_v4i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s11, s15
+; GFX9-NEXT: s_min_i32 s3, s10, s14
+; GFX9-NEXT: s_min_i32 s4, s9, s13
+; GFX9-NEXT: s_min_i32 s5, s8, s12
+; GFX9-NEXT: v_mov_b32_e32 v0, s5
+; GFX9-NEXT: v_mov_b32_e32 v1, s4
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: v_mov_b32_e32 v3, s2
+; GFX9-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_v4i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v4, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s11, s15
+; GFX10-NEXT: s_min_i32 s3, s10, s14
+; GFX10-NEXT: s_min_i32 s4, s8, s12
+; GFX10-NEXT: s_min_i32 s5, s9, s13
+; GFX10-NEXT: v_mov_b32_e32 v0, s4
+; GFX10-NEXT: v_mov_b32_e32 v1, s5
+; GFX10-NEXT: v_mov_b32_e32 v2, s3
+; GFX10-NEXT: v_mov_b32_e32 v3, s2
+; GFX10-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_v4i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b256 s[4:11], s[0:1], 0x10
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v4, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s7, s11
+; GFX11-NEXT: s_min_i32 s3, s6, s10
+; GFX11-NEXT: s_min_i32 s4, s4, s8
+; GFX11-NEXT: s_min_i32 s5, s5, s9
+; GFX11-NEXT: v_mov_b32_e32 v0, s4
+; GFX11-NEXT: v_mov_b32_e32 v1, s5
+; GFX11-NEXT: v_mov_b32_e32 v2, s3
+; GFX11-NEXT: v_mov_b32_e32 v3, s2
+; GFX11-NEXT: global_store_b128 v4, v[0:3], s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle <4 x i32> %a, %b
%val = select <4 x i1> %cmp, <4 x i32> %a, <4 x i32> %b
store <4 x i32> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_i8:
-; GCN: s_load_{{dword|b32}}
-; GCN: s_load_{{dword|b32}}
-; GCN: s_sext_i32_i8
-; GCN: s_sext_i32_i8
-; GCN: s_min_i32
define amdgpu_kernel void @s_test_imin_sle_i8(ptr addrspace(1) %out, [8 x i32], i8 %a, [8 x i32], i8 %b) #0 {
+; EG-LABEL: s_test_imin_sle_i8:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 14, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_8 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_8 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: AND_INT * T1.W, KC0[2].Y, literal.y,
+; EG-NEXT: 8(1.121039e-44), 3(4.203895e-45)
+; EG-NEXT: MIN_INT * T0.W, PV.Z, PV.W,
+; EG-NEXT: AND_INT T0.W, PV.W, literal.x,
+; EG-NEXT: LSHL * T1.W, T1.W, literal.y,
+; EG-NEXT: 255(3.573311e-43), 3(4.203895e-45)
+; EG-NEXT: LSHL T0.X, PV.W, PS,
+; EG-NEXT: LSHL * T0.W, literal.x, PS,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0xa
+; CI-NEXT: s_load_dword s3, s[4:5], 0x13
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_sext_i32_i8 s2, s2
+; CI-NEXT: s_sext_i32_i8 s3, s3
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_byte v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_i8:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x28
+; VI-NEXT: s_load_dword s3, s[4:5], 0x4c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_sext_i32_i8 s2, s2
+; VI-NEXT: s_sext_i32_i8 s3, s3
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_byte v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX9-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_sext_i32_i8 s2, s2
+; GFX9-NEXT: s_sext_i32_i8 s3, s3
+; GFX9-NEXT: s_min_i32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_byte v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x2
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX10-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_sext_i32_i8 s2, s2
+; GFX10-NEXT: s_sext_i32_i8 s3, s3
+; GFX10-NEXT: s_min_i32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_byte v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x2
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x28
+; GFX11-NEXT: s_load_b32 s3, s[0:1], 0x4c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_sext_i32_i8 s2, s2
+; GFX11-NEXT: s_sext_i32_i8 s3, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT: s_min_i32 s2, s2, s3
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b8 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle i8 %a, %b
%val = select i1 %cmp, i8 %a, i8 %b
store i8 %val, ptr addrspace(1) %out
@@ -75,95 +498,624 @@ define amdgpu_kernel void @s_test_imin_sle_i8(ptr addrspace(1) %out, [8 x i32],
}
; FIXME: Why vector and sdwa for last element?
-; FUNC-LABEL: {{^}}s_test_imin_sle_v4i8:
-; GCN-DAG: s_load_{{dwordx2|b64}}
-; GCN-DAG: s_load_{{dword|b32}} s
-; GCN-DAG: s_load_{{dword|b32}} s
-; GCN-NOT: _load_
-
-; CI: s_min_i32
-; CI: s_min_i32
-; CI: s_min_i32
-; CI: s_min_i32
-
-; VI-DAG: s_min_i32
-; VI-DAG: s_min_i32
-; VI-DAG: s_min_i32
-; VI-DAG: v_min_i32_sdwa
-
-; GFX9_10: v_min_i16
-; GFX9_10: v_min_i16
-; GFX9_10: v_min_i16
-; GFX9_10: v_min_i16
-
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
+
define amdgpu_kernel void @s_test_imin_sle_v4i8(ptr addrspace(1) %out, [8 x i32], <4 x i8> %a, [8 x i32], <4 x i8> %b) #0 {
+; EG-LABEL: s_test_imin_sle_v4i8:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @22, KC0[], KC1[]
+; EG-NEXT: TEX 7 @6
+; EG-NEXT: ALU 30, @23, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T4.X, T5.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_8 T5.X, T4.X, 74, #3
+; EG-NEXT: VTX_READ_8 T6.X, T4.X, 108, #3
+; EG-NEXT: VTX_READ_8 T7.X, T4.X, 72, #3
+; EG-NEXT: VTX_READ_8 T8.X, T4.X, 111, #3
+; EG-NEXT: VTX_READ_8 T9.X, T4.X, 75, #3
+; EG-NEXT: VTX_READ_8 T10.X, T4.X, 109, #3
+; EG-NEXT: VTX_READ_8 T11.X, T4.X, 73, #3
+; EG-NEXT: VTX_READ_8 T4.X, T4.X, 110, #3
+; EG-NEXT: ALU clause starting at 22:
+; EG-NEXT: MOV * T4.X, 0.0,
+; EG-NEXT: ALU clause starting at 23:
+; EG-NEXT: BFE_INT T0.Z, T5.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T4.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT: BFE_INT T4.X, T11.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.Y, T10.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: BFE_INT * T1.Z, T9.X, 0.0, literal.x, BS:VEC_201
+; EG-NEXT: 8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT: BFE_INT T1.W, T8.X, 0.0, literal.x,
+; EG-NEXT: MIN_INT * T0.W, T0.Z, T0.W,
+; EG-NEXT: 8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T0.Z, T1.Z, PV.W,
+; EG-NEXT: AND_INT T0.W, PS, literal.x,
+; EG-NEXT: MIN_INT * T1.W, T4.X, T0.Y,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: AND_INT T4.X, PS, literal.x,
+; EG-NEXT: LSHL T0.Y, PV.W, literal.y,
+; EG-NEXT: BFE_INT T1.Z, T7.X, 0.0, literal.z,
+; EG-NEXT: BFE_INT T0.W, T6.X, 0.0, literal.z, BS:VEC_120/SCL_212
+; EG-NEXT: LSHL * T1.W, PV.Z, literal.w,
+; EG-NEXT: 255(3.573311e-43), 16(2.242078e-44)
+; EG-NEXT: 8(1.121039e-44), 24(3.363116e-44)
+; EG-NEXT: MIN_INT T0.Z, PV.Z, PV.W,
+; EG-NEXT: OR_INT T0.W, PS, PV.Y,
+; EG-NEXT: LSHL * T1.W, PV.X, literal.x,
+; EG-NEXT: 8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT T0.W, PV.W, PS,
+; EG-NEXT: AND_INT * T1.W, PV.Z, literal.x,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: OR_INT T4.X, PV.W, PS,
+; EG-NEXT: LSHR * T5.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_v4i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0xa
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_load_dword s3, s[4:5], 0x13
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_ashr_i32 s4, s2, 24
+; CI-NEXT: s_sext_i32_i8 s5, s2
+; CI-NEXT: s_bfe_i32 s6, s2, 0x80008
+; CI-NEXT: s_bfe_i32 s2, s2, 0x80010
+; CI-NEXT: s_ashr_i32 s7, s3, 24
+; CI-NEXT: s_sext_i32_i8 s8, s3
+; CI-NEXT: s_bfe_i32 s9, s3, 0x80008
+; CI-NEXT: s_bfe_i32 s3, s3, 0x80010
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: s_min_i32 s4, s4, s7
+; CI-NEXT: s_and_b32 s2, s2, 0xff
+; CI-NEXT: s_lshl_b32 s4, s4, 24
+; CI-NEXT: s_lshl_b32 s2, s2, 16
+; CI-NEXT: s_or_b32 s2, s4, s2
+; CI-NEXT: s_min_i32 s3, s6, s9
+; CI-NEXT: s_min_i32 s4, s5, s8
+; CI-NEXT: s_lshl_b32 s3, s3, 8
+; CI-NEXT: s_and_b32 s4, s4, 0xff
+; CI-NEXT: s_or_b32 s3, s4, s3
+; CI-NEXT: s_and_b32 s3, s3, 0xffff
+; CI-NEXT: s_or_b32 s2, s3, s2
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_v4i8:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x28
+; VI-NEXT: s_load_dword s3, s[4:5], 0x4c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_lshrrev_b16_e64 v0, 8, s2
+; VI-NEXT: v_lshrrev_b16_e64 v1, 8, s3
+; VI-NEXT: s_ashr_i32 s4, s2, 24
+; VI-NEXT: s_bfe_i32 s5, s2, 0x80010
+; VI-NEXT: s_sext_i32_i8 s2, s2
+; VI-NEXT: s_ashr_i32 s6, s3, 24
+; VI-NEXT: s_bfe_i32 s7, s3, 0x80010
+; VI-NEXT: s_sext_i32_i8 s3, s3
+; VI-NEXT: s_min_i32 s4, s4, s6
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: s_min_i32 s3, s5, s7
+; VI-NEXT: v_min_i32_sdwa v0, sext(v0), sext(v1) dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0
+; VI-NEXT: v_lshlrev_b16_e64 v1, 8, s4
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_v4i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_lshr_b32 s4, s2, 16
+; GFX9-NEXT: s_bfe_i32 s6, s4, 0x80000
+; GFX9-NEXT: s_lshr_b32 s7, s3, 16
+; GFX9-NEXT: s_bfe_i32 s8, s7, 0x80000
+; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v2, s8
+; GFX9-NEXT: s_bfe_i32 s9, s3, 0x80000
+; GFX9-NEXT: v_min_i16_sdwa v1, sext(s4), sext(v1) dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT: v_min_i16_e32 v2, s6, v2
+; GFX9-NEXT: s_bfe_i32 s5, s2, 0x80000
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: v_mov_b32_e32 v3, s9
+; GFX9-NEXT: v_min_i16_sdwa v2, sext(s2), sext(v2) dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT: v_min_i16_e32 v3, s5, v3
+; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_v4i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x2
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX10-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_lshr_b32 s4, s2, 16
+; GFX10-NEXT: s_lshr_b32 s5, s3, 16
+; GFX10-NEXT: v_ashrrev_i16 v0, 8, s2
+; GFX10-NEXT: v_ashrrev_i16 v1, 8, s4
+; GFX10-NEXT: v_ashrrev_i16 v2, 8, s5
+; GFX10-NEXT: v_ashrrev_i16 v3, 8, s3
+; GFX10-NEXT: s_bfe_i32 s2, s2, 0x80000
+; GFX10-NEXT: s_bfe_i32 s3, s3, 0x80000
+; GFX10-NEXT: s_bfe_i32 s4, s4, 0x80000
+; GFX10-NEXT: v_min_i16 v1, v1, v2
+; GFX10-NEXT: v_min_i16 v0, v0, v3
+; GFX10-NEXT: s_bfe_i32 s5, s5, 0x80000
+; GFX10-NEXT: v_min_i16 v2, s2, s3
+; GFX10-NEXT: v_min_i16 v3, s4, s5
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX10-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_dword v2, v0, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_v4i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x2
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x28
+; GFX11-NEXT: s_load_b32 s3, s[0:1], 0x4c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-NEXT: v_ashrrev_i16 v0, 8, s2
+; GFX11-NEXT: v_ashrrev_i16 v1, 8, s3
+; GFX11-NEXT: v_ashrrev_i16 v2, 8, s4
+; GFX11-NEXT: v_ashrrev_i16 v3, 8, s5
+; GFX11-NEXT: s_bfe_i32 s2, s2, 0x80000
+; GFX11-NEXT: s_bfe_i32 s3, s3, 0x80000
+; GFX11-NEXT: s_bfe_i32 s4, s4, 0x80000
+; GFX11-NEXT: s_bfe_i32 s5, s5, 0x80000
+; GFX11-NEXT: v_min_i16 v4, s2, s3
+; GFX11-NEXT: v_min_i16 v5, s4, s5
+; GFX11-NEXT: v_min_i16 v2, v2, v3
+; GFX11-NEXT: v_min_i16 v0, v0, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v4
+; GFX11-NEXT: v_and_b32_e32 v3, 0xff, v5
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_lshlrev_b16 v2, 8, v2
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v2, v3, v2
+; GFX11-NEXT: v_or_b32_e32 v0, v1, v0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_lshlrev_b32_e32 v1, 16, v2
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: global_store_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle <4 x i8> %a, %b
%val = select <4 x i1> %cmp, <4 x i8> %a, <4 x i8> %b
store <4 x i8> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_v2i16:
-; GCN: s_load_{{dwordx4|b128}} s
-
-; CI: s_ashr_i32
-; CI: s_sext_i32_i16
-; CI: s_ashr_i32
-; CI: s_sext_i32_i16
-; CI: s_min_i32
-; CI: s_min_i32
-
-; VI: s_sext_i32_i16
-; VI: s_sext_i32_i16
-; VI: s_min_i32
-; VI: s_min_i32
-
-; GFX9_10: v_pk_min_i16
-
-; EG: MIN_INT
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_v2i16(ptr addrspace(1) %out, <2 x i16> %a, <2 x i16> %b) #0 {
+; EG-LABEL: s_test_imin_sle_v2i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @14, KC0[], KC1[]
+; EG-NEXT: TEX 3 @6
+; EG-NEXT: ALU 13, @15, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T4.X, T5.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T5.X, T4.X, 42, #3
+; EG-NEXT: VTX_READ_16 T6.X, T4.X, 44, #3
+; EG-NEXT: VTX_READ_16 T7.X, T4.X, 40, #3
+; EG-NEXT: VTX_READ_16 T4.X, T4.X, 46, #3
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MOV * T4.X, 0.0,
+; EG-NEXT: ALU clause starting at 15:
+; EG-NEXT: BFE_INT T5.X, T5.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.Y, T4.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: BFE_INT * T0.Z, T7.X, 0.0, literal.x, BS:VEC_201
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: BFE_INT * T0.W, T6.X, 0.0, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T0.W, T0.Z, PV.W,
+; EG-NEXT: MIN_INT * T1.W, T5.X, T0.Y,
+; EG-NEXT: LSHL T1.W, PS, literal.x,
+; EG-NEXT: AND_INT * T0.W, PV.W, literal.y,
+; EG-NEXT: 16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT: OR_INT T4.X, PV.W, PS,
+; EG-NEXT: LSHR * T5.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_v2i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_ashr_i32 s4, s2, 16
+; CI-NEXT: s_sext_i32_i16 s2, s2
+; CI-NEXT: s_ashr_i32 s5, s3, 16
+; CI-NEXT: s_sext_i32_i16 s3, s3
+; CI-NEXT: s_min_i32 s4, s4, s5
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: s_lshl_b32 s3, s4, 16
+; CI-NEXT: s_and_b32 s2, s2, 0xffff
+; CI-NEXT: s_or_b32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_v2i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_ashr_i32 s4, s2, 16
+; VI-NEXT: s_sext_i32_i16 s2, s2
+; VI-NEXT: s_ashr_i32 s5, s3, 16
+; VI-NEXT: s_sext_i32_i16 s3, s3
+; VI-NEXT: s_min_i32 s4, s4, s5
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: s_lshl_b32 s3, s4, 16
+; VI-NEXT: s_and_b32 s2, s2, 0xffff
+; VI-NEXT: s_or_b32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_v2i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: v_pk_min_i16 v1, s2, v1
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_v2i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_pk_min_i16 v1, s2, s3
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_v2i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_pk_min_i16 v1, s2, s3
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle <2 x i16> %a, %b
%val = select <2 x i1> %cmp, <2 x i16> %a, <2 x i16> %b
store <2 x i16> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_v4i16:
-; CI-NOT: buffer_load
-; CI: s_min_i32
-; CI: s_min_i32
-; CI: s_min_i32
-; CI: s_min_i32
-
-; VI: s_min_i32
-; VI: s_min_i32
-; VI: s_min_i32
-; VI: s_min_i32
-
-; GFX9_10: v_pk_min_i16
-; GFX9_10: v_pk_min_i16
-
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_v4i16(ptr addrspace(1) %out, <4 x i16> %a, <4 x i16> %b) #0 {
+; EG-LABEL: s_test_imin_sle_v4i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @28, KC0[], KC1[]
+; EG-NEXT: TEX 1 @12
+; EG-NEXT: ALU 9, @30, KC0[], KC1[]
+; EG-NEXT: TEX 1 @16
+; EG-NEXT: ALU 10, @40, KC0[], KC1[]
+; EG-NEXT: TEX 1 @20
+; EG-NEXT: ALU 10, @51, KC0[], KC1[]
+; EG-NEXT: TEX 1 @24
+; EG-NEXT: ALU 11, @62, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T6.XY, T5.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 12:
+; EG-NEXT: VTX_READ_16 T6.X, T5.X, 50, #3
+; EG-NEXT: VTX_READ_16 T7.X, T5.X, 58, #3
+; EG-NEXT: Fetch clause starting at 16:
+; EG-NEXT: VTX_READ_16 T6.X, T5.X, 48, #3
+; EG-NEXT: VTX_READ_16 T7.X, T5.X, 56, #3
+; EG-NEXT: Fetch clause starting at 20:
+; EG-NEXT: VTX_READ_16 T6.X, T5.X, 46, #3
+; EG-NEXT: VTX_READ_16 T7.X, T5.X, 54, #3
+; EG-NEXT: Fetch clause starting at 24:
+; EG-NEXT: VTX_READ_16 T6.X, T5.X, 44, #3
+; EG-NEXT: VTX_READ_16 T5.X, T5.X, 52, #3
+; EG-NEXT: ALU clause starting at 28:
+; EG-NEXT: MOV * T0.Y, T3.X,
+; EG-NEXT: MOV * T5.X, 0.0,
+; EG-NEXT: ALU clause starting at 30:
+; EG-NEXT: BFE_INT T0.Z, T6.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T7.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT * T0.W, PV.Z, PV.W,
+; EG-NEXT: LSHL T0.W, PV.W, literal.x,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT: 16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT: OR_INT * T0.W, PS, PV.W,
+; EG-NEXT: MOV * T3.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 40:
+; EG-NEXT: BFE_INT T0.Z, T6.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T7.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T0.W, PV.Z, PV.W,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.x,
+; EG-NEXT: -65536(nan), 0(0.000000e+00)
+; EG-NEXT: AND_INT * T0.W, PV.W, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT: MOV T3.X, PV.W,
+; EG-NEXT: MOV * T0.Y, T2.X,
+; EG-NEXT: ALU clause starting at 51:
+; EG-NEXT: BFE_INT T0.Z, T6.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T7.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T0.W, PV.Z, PV.W,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT: MOV * T2.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 62:
+; EG-NEXT: BFE_INT T0.Z, T6.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T5.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT * T0.W, PV.Z, PV.W,
+; EG-NEXT: LSHR T5.X, KC0[2].Y, literal.x,
+; EG-NEXT: AND_INT T1.W, T0.Y, literal.y,
+; EG-NEXT: AND_INT * T0.W, PV.W, literal.z,
+; EG-NEXT: 2(2.802597e-45), -65536(nan)
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T6.X, PV.W, PS,
+; EG-NEXT: MOV T2.X, PV.X,
+; EG-NEXT: MOV * T6.Y, T3.X,
+;
+; CI-LABEL: s_test_imin_sle_v4i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x2
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_ashr_i32 s6, s0, 16
+; CI-NEXT: s_ashr_i32 s7, s1, 16
+; CI-NEXT: s_sext_i32_i16 s0, s0
+; CI-NEXT: s_sext_i32_i16 s1, s1
+; CI-NEXT: s_ashr_i32 s8, s2, 16
+; CI-NEXT: s_ashr_i32 s9, s3, 16
+; CI-NEXT: s_sext_i32_i16 s2, s2
+; CI-NEXT: s_sext_i32_i16 s3, s3
+; CI-NEXT: s_min_i32 s7, s7, s9
+; CI-NEXT: s_min_i32 s1, s1, s3
+; CI-NEXT: s_min_i32 s3, s6, s8
+; CI-NEXT: s_min_i32 s0, s0, s2
+; CI-NEXT: s_lshl_b32 s7, s7, 16
+; CI-NEXT: s_and_b32 s1, s1, 0xffff
+; CI-NEXT: s_lshl_b32 s3, s3, 16
+; CI-NEXT: s_and_b32 s0, s0, 0xffff
+; CI-NEXT: s_or_b32 s1, s1, s7
+; CI-NEXT: s_or_b32 s0, s0, s3
+; CI-NEXT: v_mov_b32_e32 v2, s4
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_v4i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_ashr_i32 s6, s1, 16
+; VI-NEXT: s_sext_i32_i16 s1, s1
+; VI-NEXT: s_ashr_i32 s8, s3, 16
+; VI-NEXT: s_sext_i32_i16 s3, s3
+; VI-NEXT: s_ashr_i32 s7, s0, 16
+; VI-NEXT: s_sext_i32_i16 s0, s0
+; VI-NEXT: s_ashr_i32 s9, s2, 16
+; VI-NEXT: s_sext_i32_i16 s2, s2
+; VI-NEXT: s_min_i32 s6, s6, s8
+; VI-NEXT: s_min_i32 s1, s1, s3
+; VI-NEXT: s_min_i32 s7, s7, s9
+; VI-NEXT: s_min_i32 s0, s0, s2
+; VI-NEXT: s_lshl_b32 s2, s6, 16
+; VI-NEXT: s_and_b32 s1, s1, 0xffff
+; VI-NEXT: s_or_b32 s1, s1, s2
+; VI-NEXT: s_lshl_b32 s2, s7, 16
+; VI-NEXT: s_and_b32 s0, s0, 0xffff
+; VI-NEXT: s_or_b32 s0, s0, s2
+; VI-NEXT: v_mov_b32_e32 v2, s4
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_v4i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s3
+; GFX9-NEXT: v_mov_b32_e32 v3, s2
+; GFX9-NEXT: v_pk_min_i16 v1, s1, v0
+; GFX9-NEXT: v_pk_min_i16 v0, s0, v3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[6:7]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_v4i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_pk_min_i16 v1, s1, s3
+; GFX10-NEXT: v_pk_min_i16 v0, s0, s2
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[6:7]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_v4i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x8
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_pk_min_i16 v1, s5, s7
+; GFX11-NEXT: v_pk_min_i16 v0, s4, s6
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle <4 x i16> %a, %b
%val = select <4 x i1> %cmp, <4 x i16> %a, <4 x i16> %b
store <4 x i16> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: @v_test_imin_slt_i32
-; GCN: v_min_i32_e32
-
-; EG: MIN_INT
define amdgpu_kernel void @v_test_imin_slt_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) #0 {
+; EG-LABEL: v_test_imin_slt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 3, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T1.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MIN_INT T0.X, T0.X, T1.X,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_imin_slt_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v5, v[0:1]
+; CI-NEXT: flat_load_dword v2, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_i32_e32 v2, v5, v2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_imin_slt_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_i32_e32 v2, v5, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_imin_slt_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_imin_slt_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_imin_slt_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_i32_e32 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i32, ptr addrspace(1) %aptr, i32 %tid
%b.gep = getelementptr inbounds i32, ptr addrspace(1) %bptr, i32 %tid
@@ -176,14 +1128,133 @@ define amdgpu_kernel void @v_test_imin_slt_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: @v_test_imin_slt_i16
-; CI: v_min_i32_e32
-
-; GFX8_9: v_min_i16_e32
-; GFX10: v_min_i16
-
-; EG: MIN_INT
define amdgpu_kernel void @v_test_imin_slt_i16(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) #0 {
+; EG-LABEL: v_test_imin_slt_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @8
+; EG-NEXT: ALU 0, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @10
+; EG-NEXT: ALU 16, @15, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T1.XW, T0.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 8:
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 0, #1
+; EG-NEXT: Fetch clause starting at 10:
+; EG-NEXT: VTX_READ_16 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: LSHL * T0.W, T0.X, 1,
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: ADD_INT * T1.X, KC0[2].Z, T0.W,
+; EG-NEXT: ALU clause starting at 15:
+; EG-NEXT: BFE_INT T0.Z, T0.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T1.W, T1.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: AND_INT T2.W, PS, literal.x,
+; EG-NEXT: MIN_INT * T1.W, PV.W, PV.Z,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: AND_INT T1.W, PS, literal.x,
+; EG-NEXT: LSHL * T2.W, PV.W, literal.y,
+; EG-NEXT: 65535(9.183409e-41), 3(4.203895e-45)
+; EG-NEXT: LSHL T1.X, PV.W, PS,
+; EG-NEXT: LSHL * T1.W, literal.x, PS,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MOV T1.Y, 0.0,
+; EG-NEXT: MOV * T1.Z, 0.0,
+; EG-NEXT: LSHR * T0.X, T0.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_imin_slt_i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 1, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_sshort v5, v[0:1]
+; CI-NEXT: flat_load_sshort v2, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_i32_e32 v2, v5, v2
+; CI-NEXT: flat_store_short v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_imin_slt_i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 1, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_ushort v5, v[0:1]
+; VI-NEXT: flat_load_ushort v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_i16_e32 v2, v5, v2
+; VI-NEXT: flat_store_short v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_imin_slt_i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 1, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_ushort v1, v0, s[2:3]
+; GFX9-NEXT: global_load_ushort v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_i16_e32 v1, v1, v2
+; GFX9-NEXT: global_store_short v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_imin_slt_i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_ushort v1, v0, s[2:3]
+; GFX10-NEXT: global_load_ushort v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_i16 v1, v1, v2
+; GFX10-NEXT: global_store_short v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_imin_slt_i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 1, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_u16 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_u16 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_i16 v1, v1, v2
+; GFX11-NEXT: global_store_b16 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i16, ptr addrspace(1) %aptr, i32 %tid
%b.gep = getelementptr inbounds i16, ptr addrspace(1) %bptr, i32 %tid
@@ -197,57 +1268,436 @@ define amdgpu_kernel void @v_test_imin_slt_i16(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: @s_test_imin_slt_i32
-; GCN: s_min_i32
-
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_slt_i32(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; EG-LABEL: s_test_imin_slt_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_INT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_slt_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_slt_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_slt_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_slt_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_slt_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp slt i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_slt_v2i32:
-; GCN: s_min_i32
-; GCN: s_min_i32
-
-; EG: MIN_INT
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_slt_v2i32(ptr addrspace(1) %out, <2 x i32> %a, <2 x i32> %b) #0 {
+; EG-LABEL: s_test_imin_slt_v2i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MIN_INT * T0.Y, KC0[3].X, KC0[3].Z,
+; EG-NEXT: MIN_INT * T0.X, KC0[2].W, KC0[3].Y,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_slt_v2i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x2
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s1, s1, s3
+; CI-NEXT: s_min_i32 s0, s0, s2
+; CI-NEXT: v_mov_b32_e32 v2, s4
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_slt_v2i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s1, s1, s3
+; VI-NEXT: s_min_i32 s0, s0, s2
+; VI-NEXT: v_mov_b32_e32 v2, s4
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_slt_v2i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s1, s1, s3
+; GFX9-NEXT: s_min_i32 s0, s0, s2
+; GFX9-NEXT: v_mov_b32_e32 v0, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[6:7]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_slt_v2i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s0, s0, s2
+; GFX10-NEXT: s_min_i32 s1, s1, s3
+; GFX10-NEXT: v_mov_b32_e32 v0, s0
+; GFX10-NEXT: v_mov_b32_e32 v1, s1
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[6:7]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_slt_v2i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x8
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s4, s6
+; GFX11-NEXT: s_min_i32 s3, s5, s7
+; GFX11-NEXT: v_mov_b32_e32 v0, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s3
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp slt <2 x i32> %a, %b
%val = select <2 x i1> %cmp, <2 x i32> %a, <2 x i32> %b
store <2 x i32> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_slt_imm_i32:
-; GCN: s_min_i32 {{s[0-9]+}}, {{s[0-9]+}}, 8
-
-; EG: MIN_INT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_imin_slt_imm_i32(ptr addrspace(1) %out, i32 %a) #0 {
+; EG-LABEL: s_test_imin_slt_imm_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_INT * T1.X, KC0[2].Z, literal.y,
+; EG-NEXT: 2(2.802597e-45), 8(1.121039e-44)
+;
+; CI-LABEL: s_test_imin_slt_imm_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0x2
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s2, 8
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_slt_imm_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x8
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s2, 8
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_slt_imm_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s2, 8
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_slt_imm_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s2, 8
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_slt_imm_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x8
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s2, 8
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp slt i32 %a, 8
%val = select i1 %cmp, i32 %a, i32 8
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_imm_i32:
-; GCN: s_min_i32 {{s[0-9]+}}, {{s[0-9]+}}, 8
-
-; EG: MIN_INT {{.*}}literal.{{[xyzw]}}
define amdgpu_kernel void @s_test_imin_sle_imm_i32(ptr addrspace(1) %out, i32 %a) #0 {
+; EG-LABEL: s_test_imin_sle_imm_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_INT * T1.X, KC0[2].Z, literal.y,
+; EG-NEXT: 2(2.802597e-45), 8(1.121039e-44)
+;
+; CI-LABEL: s_test_imin_sle_imm_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0x2
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_i32 s2, s2, 8
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_imm_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x8
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_i32 s2, s2, 8
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_imm_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_i32 s2, s2, 8
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_imm_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_i32 s2, s2, 8
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_imm_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x8
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_i32 s2, s2, 8
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle i32 %a, 8
%val = select i1 %cmp, i32 %a, i32 8
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: @v_test_umin_ule_i32
-; GCN: v_min_u32_e32
-
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_umin_ule_i32(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_umin_ule_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 3, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T1.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MIN_UINT T0.X, T0.X, T1.X,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ule_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v5, v[0:1]
+; CI-NEXT: flat_load_dword v2, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_u32_e32 v2, v5, v2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ule_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u32_e32 v2, v5, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ule_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ule_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ule_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds i32, ptr addrspace(1) %b.ptr, i32 %tid
@@ -260,17 +1710,133 @@ define amdgpu_kernel void @v_test_umin_ule_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: @v_test_umin_ule_v3i32
-; GCN: v_min_u32_e32
-; GCN: v_min_u32_e32
-; GCN: v_min_u32_e32
-; GCN-NOT: v_min_u32_e32
-; GCN: s_endpgm
-
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_umin_ule_v3i32(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_umin_ule_v3i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 9, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T2.X, T3.X, 0
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_128 T1.XYZW, T1.X, 0, #1
+; EG-NEXT: VTX_READ_128 T2.XYZW, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T1.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MIN_UINT * T0.Y, T2.Y, T1.Y,
+; EG-NEXT: MIN_UINT T0.X, T2.X, T1.X,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR T1.X, PV.W, literal.x,
+; EG-NEXT: MIN_UINT * T2.X, T2.Z, T1.Z,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.W, T0.W, literal.x,
+; EG-NEXT: 8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT: LSHR * T3.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ule_v3i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v6, 4, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v6
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_add_i32_e32 v3, vcc, s4, v6
+; CI-NEXT: v_addc_u32_e32 v4, vcc, 0, v2, vcc
+; CI-NEXT: flat_load_dwordx3 v[0:2], v[0:1]
+; CI-NEXT: flat_load_dwordx3 v[3:5], v[3:4]
+; CI-NEXT: v_mov_b32_e32 v7, s1
+; CI-NEXT: v_add_i32_e32 v6, vcc, s0, v6
+; CI-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_u32_e32 v2, v2, v5
+; CI-NEXT: v_min_u32_e32 v1, v1, v4
+; CI-NEXT: v_min_u32_e32 v0, v0, v3
+; CI-NEXT: flat_store_dwordx3 v[6:7], v[0:2]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ule_v3i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v6, 4, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v6
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_add_u32_e32 v3, vcc, s4, v6
+; VI-NEXT: v_addc_u32_e32 v4, vcc, 0, v2, vcc
+; VI-NEXT: flat_load_dwordx3 v[0:2], v[0:1]
+; VI-NEXT: flat_load_dwordx3 v[3:5], v[3:4]
+; VI-NEXT: v_mov_b32_e32 v7, s1
+; VI-NEXT: v_add_u32_e32 v6, vcc, s0, v6
+; VI-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u32_e32 v2, v2, v5
+; VI-NEXT: v_min_u32_e32 v1, v1, v4
+; VI-NEXT: v_min_u32_e32 v0, v0, v3
+; VI-NEXT: flat_store_dwordx3 v[6:7], v[0:2]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ule_v3i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v6, 4, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dwordx3 v[0:2], v6, s[2:3]
+; GFX9-NEXT: global_load_dwordx3 v[3:5], v6, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_u32_e32 v2, v2, v5
+; GFX9-NEXT: v_min_u32_e32 v1, v1, v4
+; GFX9-NEXT: v_min_u32_e32 v0, v0, v3
+; GFX9-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ule_v3i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v6, 4, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dwordx3 v[0:2], v6, s[2:3]
+; GFX10-NEXT: global_load_dwordx3 v[3:5], v6, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_u32_e32 v2, v2, v5
+; GFX10-NEXT: v_min_u32_e32 v1, v1, v4
+; GFX10-NEXT: v_min_u32_e32 v0, v0, v3
+; GFX10-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ule_v3i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v6, 4, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b96 v[0:2], v6, s[6:7]
+; GFX11-NEXT: global_load_b96 v[3:5], v6, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_u32_e32 v2, v2, v5
+; GFX11-NEXT: v_min_u32_e32 v1, v1, v4
+; GFX11-NEXT: v_min_u32_e32 v0, v0, v3
+; GFX11-NEXT: global_store_b96 v6, v[0:2], s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds <3 x i32>, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds <3 x i32>, ptr addrspace(1) %b.ptr, i32 %tid
@@ -285,26 +1851,170 @@ define amdgpu_kernel void @v_test_umin_ule_v3i32(ptr addrspace(1) %out, ptr addr
}
; FIXME: Reduce unused packed component to scalar
-; FUNC-LABEL: @v_test_umin_ule_v3i16{{$}}
-; CI: v_min_u32_e32
-; CI: v_min_u32_e32
-; CI: v_min_u32_e32
-; CI-NOT: v_min_u32_e32
-
-; VI: v_min_u16_e32
-; VI: v_min_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI: v_min_u16_e32
-; VI-NOT: v_min_u16
-
-; GFX9_10: v_pk_min_u16
-; GFX9_10: v_pk_min_u16
-; GCN: s_endpgm
-
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_umin_ule_v3i16(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_umin_ule_v3i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @20, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @8
+; EG-NEXT: ALU 11, @24, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 3 @12
+; EG-NEXT: ALU 8, @36, KC0[], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T6.X, T8.X, 0
+; EG-NEXT: MEM_RAT MSKOR T7.XW, T0.X
+; EG-NEXT: CF_END
+; EG-NEXT: Fetch clause starting at 8:
+; EG-NEXT: VTX_READ_16 T7.X, T6.X, 4, #1
+; EG-NEXT: VTX_READ_16 T8.X, T0.X, 4, #1
+; EG-NEXT: Fetch clause starting at 12:
+; EG-NEXT: VTX_READ_16 T8.X, T6.X, 0, #1
+; EG-NEXT: VTX_READ_16 T9.X, T0.X, 0, #1
+; EG-NEXT: VTX_READ_16 T6.X, T6.X, 2, #1
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 2, #1
+; EG-NEXT: ALU clause starting at 20:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T6.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 24:
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: ADD_INT * T1.W, PV.W, literal.x,
+; EG-NEXT: 4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT: AND_INT * T2.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T2.W, PV.W, literal.x,
+; EG-NEXT: MIN_UINT * T3.W, T8.X, T7.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T7.X, PS, PV.W,
+; EG-NEXT: LSHL * T7.W, literal.x, PV.W,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MOV * T7.Y, 0.0,
+; EG-NEXT: ALU clause starting at 36:
+; EG-NEXT: MOV T7.Z, 0.0,
+; EG-NEXT: MIN_UINT * T2.W, T0.X, T6.X,
+; EG-NEXT: LSHR T0.X, T1.W, literal.x,
+; EG-NEXT: LSHL T1.W, PV.W, literal.y,
+; EG-NEXT: MIN_UINT * T2.W, T9.X, T8.X,
+; EG-NEXT: 2(2.802597e-45), 16(2.242078e-44)
+; EG-NEXT: OR_INT T6.X, PV.W, PS,
+; EG-NEXT: LSHR * T8.X, T0.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ule_v3i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 3, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
+; CI-NEXT: flat_load_dwordx2 v[2:3], v[2:3]
+; CI-NEXT: v_mov_b32_e32 v5, s1
+; CI-NEXT: v_add_i32_e32 v4, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc
+; CI-NEXT: v_add_i32_e32 v6, vcc, 4, v4
+; CI-NEXT: v_addc_u32_e32 v7, vcc, 0, v5, vcc
+; CI-NEXT: s_waitcnt vmcnt(1)
+; CI-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_lshrrev_b32_e32 v9, 16, v2
+; CI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; CI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; CI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; CI-NEXT: v_min_u32_e32 v0, v0, v2
+; CI-NEXT: v_min_u32_e32 v2, v8, v9
+; CI-NEXT: v_min_u32_e32 v1, v1, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v0, v0, v2
+; CI-NEXT: flat_store_short v[6:7], v1
+; CI-NEXT: flat_store_dword v[4:5], v0
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ule_v3i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 3, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT: flat_load_dwordx2 v[2:3], v[2:3]
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: v_add_u32_e32 v4, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc
+; VI-NEXT: v_add_u32_e32 v6, vcc, 4, v4
+; VI-NEXT: v_addc_u32_e32 v7, vcc, 0, v5, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u16_e32 v8, v0, v2
+; VI-NEXT: v_min_u16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT: v_min_u16_e32 v1, v1, v3
+; VI-NEXT: v_or_b32_e32 v0, v8, v0
+; VI-NEXT: flat_store_short v[6:7], v1
+; VI-NEXT: flat_store_dword v[4:5], v0
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ule_v3i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v4, 3, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v4, s[2:3]
+; GFX9-NEXT: global_load_dwordx2 v[2:3], v4, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_pk_min_u16 v1, v1, v3
+; GFX9-NEXT: v_pk_min_u16 v0, v0, v2
+; GFX9-NEXT: global_store_short v4, v1, s[0:1] offset:4
+; GFX9-NEXT: global_store_dword v4, v0, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ule_v3i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[2:3]
+; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_pk_min_u16 v1, v1, v3
+; GFX10-NEXT: v_pk_min_u16 v0, v0, v2
+; GFX10-NEXT: global_store_short v4, v1, s[0:1] offset:4
+; GFX10-NEXT: global_store_dword v4, v0, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ule_v3i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v4, 3, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b64 v[0:1], v4, s[6:7]
+; GFX11-NEXT: global_load_b64 v[2:3], v4, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_pk_min_u16 v1, v1, v3
+; GFX11-NEXT: v_pk_min_u16 v0, v0, v2
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b16 v4, v1, s[4:5] offset:4
+; GFX11-NEXT: global_store_b32 v4, v0, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds <3 x i16>, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds <3 x i16>, ptr addrspace(1) %b.ptr, i32 %tid
@@ -318,22 +2028,189 @@ define amdgpu_kernel void @v_test_umin_ule_v3i16(ptr addrspace(1) %out, ptr addr
ret void
}
-; FUNC-LABEL: @s_test_umin_ule_i32
-; GCN: s_min_u32
-
-; EG: MIN_UINT
define amdgpu_kernel void @s_test_umin_ule_i32(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; EG-LABEL: s_test_umin_ule_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_UINT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_umin_ule_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_u32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_umin_ule_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_u32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_umin_ule_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_u32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_umin_ule_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_u32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_umin_ule_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_u32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp ule i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: @v_test_umin_ult_i32
-; GCN: v_min_u32_e32
-
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_umin_ult_i32(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_umin_ult_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 3, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 3, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ADD_INT * T1.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 14:
+; EG-NEXT: MIN_UINT T0.X, T0.X, T1.X,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ult_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
+; CI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v5, v[0:1]
+; CI-NEXT: flat_load_dword v2, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v4
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_u32_e32 v2, v5, v2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ult_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u32_e32 v2, v5, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ult_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ult_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ult_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_u32_e32 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds i32, ptr addrspace(1) %b.ptr, i32 %tid
@@ -346,18 +2223,119 @@ define amdgpu_kernel void @v_test_umin_ult_i32(ptr addrspace(1) %out, ptr addrsp
ret void
}
-; FUNC-LABEL: {{^}}v_test_umin_ult_i8:
-; CI: {{buffer|flat|global}}_load_ubyte
-; CI: {{buffer|flat|global}}_load_ubyte
-; CI: v_min_u32_e32
-
-; GFX8_9_10: {{flat|global}}_load_{{ubyte|u8}}
-; GFX8_9_10: {{flat|global}}_load_{{ubyte|u8}}
-; GFX8_9: v_min_u16_e32
-; GFX10: v_min_u16
-
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_umin_ult_i8(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_umin_ult_i8:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 12, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T1.XW, T0.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_8 T2.X, T2.X, 0, #1
+; EG-NEXT: VTX_READ_8 T1.X, T1.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: ADD_INT T1.X, KC0[2].Z, T0.X,
+; EG-NEXT: ADD_INT * T2.X, KC0[2].W, T0.X,
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.X,
+; EG-NEXT: AND_INT T1.W, PV.W, literal.x,
+; EG-NEXT: MIN_UINT * T2.W, T1.X, T2.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T1.X, T2.W, PV.W,
+; EG-NEXT: LSHL * T1.W, literal.x, PV.W,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T1.Y, 0.0,
+; EG-NEXT: MOV * T1.Z, 0.0,
+; EG-NEXT: LSHR * T0.X, T0.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ult_i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v2, s3
+; CI-NEXT: v_add_i32_e32 v1, vcc, s2, v0
+; CI-NEXT: v_addc_u32_e32 v2, vcc, 0, v2, vcc
+; CI-NEXT: v_mov_b32_e32 v4, s5
+; CI-NEXT: v_add_i32_e32 v3, vcc, s4, v0
+; CI-NEXT: v_addc_u32_e32 v4, vcc, 0, v4, vcc
+; CI-NEXT: flat_load_ubyte v2, v[1:2]
+; CI-NEXT: flat_load_ubyte v3, v[3:4]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v0
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_min_u32_e32 v2, v2, v3
+; CI-NEXT: flat_store_byte v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ult_i8:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: v_add_u32_e32 v1, vcc, s2, v0
+; VI-NEXT: v_addc_u32_e32 v2, vcc, 0, v2, vcc
+; VI-NEXT: v_mov_b32_e32 v4, s5
+; VI-NEXT: v_add_u32_e32 v3, vcc, s4, v0
+; VI-NEXT: v_addc_u32_e32 v4, vcc, 0, v4, vcc
+; VI-NEXT: flat_load_ubyte v2, v[1:2]
+; VI-NEXT: flat_load_ubyte v3, v[3:4]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v0
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u16_e32 v2, v2, v3
+; VI-NEXT: flat_store_byte v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ult_i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_ubyte v1, v0, s[2:3]
+; GFX9-NEXT: global_load_ubyte v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_min_u16_e32 v1, v1, v2
+; GFX9-NEXT: global_store_byte v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ult_i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_ubyte v1, v0, s[2:3]
+; GFX10-NEXT: global_load_ubyte v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_min_u16 v1, v1, v2
+; GFX10-NEXT: global_store_byte v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ult_i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_u8 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_u8 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_min_u16 v1, v1, v2
+; GFX11-NEXT: global_store_b8 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds i8, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds i8, ptr addrspace(1) %b.ptr, i32 %tid
@@ -371,27 +2349,211 @@ define amdgpu_kernel void @v_test_umin_ult_i8(ptr addrspace(1) %out, ptr addrspa
ret void
}
-; FUNC-LABEL: @s_test_umin_ult_i32
-; GCN: s_min_u32
-
-; EG: MIN_UINT
define amdgpu_kernel void @s_test_umin_ult_i32(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; EG-LABEL: s_test_umin_ult_i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_UINT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_umin_ult_i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_u32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_umin_ult_i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_u32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_umin_ult_i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_u32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_umin_ult_i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_u32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_umin_ult_i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_u32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp ult i32 %a, %b
%val = select i1 %cmp, i32 %a, i32 %b
store i32 %val, ptr addrspace(1) %out, align 4
ret void
}
-; FUNC-LABEL: @v_test_umin_ult_i32_multi_use
-; CI-NOT: v_min
-; GCN: s_cmp_lt_u32
-; CI-NOT: v_min
-; CI: v_cndmask_b32
-; CI-NOT: v_min
-; GCN: s_endpgm
-
-; EG-NOT: MIN_UINT
define amdgpu_kernel void @v_test_umin_ult_i32_multi_use(ptr addrspace(1) %out0, ptr addrspace(1) %out1, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) #0 {
+; EG-LABEL: v_test_umin_ult_i32_multi_use:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 16, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 0
+; EG-NEXT: MEM_RAT MSKOR T2.XW, T3.X
+; EG-NEXT: CF_END
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_32 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T0.X, KC0[2].W,
+; EG-NEXT: MOV * T1.X, KC0[3].X,
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: AND_INT T0.W, KC0[2].Z, literal.x,
+; EG-NEXT: SETGT_UINT * T1.W, T1.X, T0.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: AND_INT T1.W, PS, 1,
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T2.X, PV.W, PS,
+; EG-NEXT: LSHL * T2.W, literal.x, PS,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T2.Y, 0.0,
+; EG-NEXT: MOV * T2.Z, 0.0,
+; EG-NEXT: LSHR T3.X, KC0[2].Z, literal.x,
+; EG-NEXT: SETGE_UINT * T0.W, T0.X, T1.X,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: CNDE_INT T0.X, PV.W, T0.X, T1.X,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ult_i32_multi_use:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_load_dword s4, s[4:5], 0x0
+; CI-NEXT: s_load_dword s5, s[6:7], 0x0
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: v_mov_b32_e32 v3, s3
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_cmp_lt_u32 s4, s5
+; CI-NEXT: s_cselect_b64 s[0:1], -1, 0
+; CI-NEXT: v_cndmask_b32_e64 v4, 0, 1, s[0:1]
+; CI-NEXT: s_and_b64 s[0:1], s[0:1], exec
+; CI-NEXT: s_cselect_b32 s0, s4, s5
+; CI-NEXT: v_mov_b32_e32 v5, s0
+; CI-NEXT: flat_store_dword v[0:1], v5
+; CI-NEXT: flat_store_byte v[2:3], v4
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ult_i32_multi_use:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_load_dword s4, s[4:5], 0x0
+; VI-NEXT: s_load_dword s5, s[6:7], 0x0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_cmp_lt_u32 s4, s5
+; VI-NEXT: s_cselect_b64 s[0:1], -1, 0
+; VI-NEXT: v_cndmask_b32_e64 v4, 0, 1, s[0:1]
+; VI-NEXT: s_and_b64 s[0:1], s[0:1], exec
+; VI-NEXT: s_cselect_b32 s0, s4, s5
+; VI-NEXT: v_mov_b32_e32 v5, s0
+; VI-NEXT: flat_store_dword v[0:1], v5
+; VI-NEXT: flat_store_byte v[2:3], v4
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ult_i32_multi_use:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_load_dword s8, s[4:5], 0x0
+; GFX9-NEXT: s_load_dword s9, s[6:7], 0x0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_cmp_lt_u32 s8, s9
+; GFX9-NEXT: s_cselect_b64 s[4:5], -1, 0
+; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[4:5]
+; GFX9-NEXT: s_and_b64 s[4:5], s[4:5], exec
+; GFX9-NEXT: s_cselect_b32 s4, s8, s9
+; GFX9-NEXT: v_mov_b32_e32 v2, s4
+; GFX9-NEXT: global_store_dword v0, v2, s[0:1]
+; GFX9-NEXT: global_store_byte v0, v1, s[2:3]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ult_i32_multi_use:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_load_dword s8, s[4:5], 0x0
+; GFX10-NEXT: s_load_dword s9, s[6:7], 0x0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_cmp_lt_u32 s8, s9
+; GFX10-NEXT: s_cselect_b32 s4, -1, 0
+; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s4
+; GFX10-NEXT: s_and_b32 s4, s4, exec_lo
+; GFX10-NEXT: s_cselect_b32 s4, s8, s9
+; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: global_store_dword v1, v2, s[0:1]
+; GFX10-NEXT: global_store_byte v1, v0, s[2:3]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ult_i32_multi_use:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b256 s[0:7], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_load_b32 s4, s[4:5], 0x0
+; GFX11-NEXT: s_load_b32 s5, s[6:7], 0x0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_cmp_lt_u32 s4, s5
+; GFX11-NEXT: s_cselect_b32 s6, -1, 0
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s6
+; GFX11-NEXT: s_and_b32 s6, s6, exec_lo
+; GFX11-NEXT: s_cselect_b32 s4, s4, s5
+; GFX11-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b32 v1, v2, s[0:1]
+; GFX11-NEXT: global_store_b8 v1, v0, s[2:3]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%a = load i32, ptr addrspace(1) %aptr, align 4
%b = load i32, ptr addrspace(1) %bptr, align 4
%cmp = icmp ult i32 %a, %b
@@ -401,15 +2563,141 @@ define amdgpu_kernel void @v_test_umin_ult_i32_multi_use(ptr addrspace(1) %out0,
ret void
}
-; FUNC-LABEL: @v_test_umin_ult_i16_multi_use
-; GCN-NOT: v_min
-; GCN: v_cmp_lt_u32
-; GCN: v_cndmask_b32
-; GCN-NOT: v_min
-; GCN: s_endpgm
-
-; EG-NOT: MIN_UINT
define amdgpu_kernel void @v_test_umin_ult_i16_multi_use(ptr addrspace(1) %out0, ptr addrspace(1) %out1, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) #0 {
+; EG-LABEL: v_test_umin_ult_i16_multi_use:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 24, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T2.XW, T3.X
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T1.X, 0, #1
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 0, #1
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV T0.X, KC0[2].W,
+; EG-NEXT: MOV * T1.X, KC0[3].X,
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: AND_INT T0.W, KC0[2].Y, literal.x,
+; EG-NEXT: SETGE_UINT * T1.W, T0.X, T1.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: CNDE_INT T1.W, PS, T0.X, T1.X,
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T2.X, PV.W, PS,
+; EG-NEXT: LSHL * T2.W, literal.x, PS,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MOV T2.Y, 0.0,
+; EG-NEXT: AND_INT T0.W, KC0[2].Z, literal.x,
+; EG-NEXT: SETGT_UINT * T1.W, T1.X, T0.X,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: AND_INT T1.W, PS, 1,
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT: LSHL T0.X, PV.W, PS,
+; EG-NEXT: LSHL * T0.W, literal.x, PS,
+; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV T2.Z, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR T1.X, KC0[2].Z, literal.x,
+; EG-NEXT: LSHR * T3.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_umin_ult_i16_multi_use:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v0, s4
+; CI-NEXT: v_mov_b32_e32 v1, s5
+; CI-NEXT: v_mov_b32_e32 v2, s6
+; CI-NEXT: v_mov_b32_e32 v3, s7
+; CI-NEXT: flat_load_ushort v4, v[0:1]
+; CI-NEXT: flat_load_ushort v5, v[2:3]
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: v_mov_b32_e32 v3, s3
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_cmp_lt_u32_e32 vcc, v4, v5
+; CI-NEXT: v_cndmask_b32_e32 v4, v5, v4, vcc
+; CI-NEXT: flat_store_short v[0:1], v4
+; CI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc
+; CI-NEXT: flat_store_byte v[2:3], v0
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_umin_ult_i16_multi_use:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s4
+; VI-NEXT: v_mov_b32_e32 v1, s5
+; VI-NEXT: v_mov_b32_e32 v2, s6
+; VI-NEXT: v_mov_b32_e32 v3, s7
+; VI-NEXT: flat_load_ushort v4, v[0:1]
+; VI-NEXT: flat_load_ushort v5, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_cmp_lt_u32_e32 vcc, v4, v5
+; VI-NEXT: v_cndmask_b32_e32 v4, v5, v4, vcc
+; VI-NEXT: flat_store_short v[0:1], v4
+; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc
+; VI-NEXT: flat_store_byte v[2:3], v0
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_umin_ult_i16_multi_use:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_ushort v1, v0, s[4:5]
+; GFX9-NEXT: global_load_ushort v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_cmp_lt_u32_e32 vcc, v1, v2
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v1, vcc
+; GFX9-NEXT: global_store_short v0, v1, s[0:1]
+; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc
+; GFX9-NEXT: global_store_byte v0, v1, s[2:3]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_umin_ult_i16_multi_use:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_ushort v1, v0, s[4:5]
+; GFX10-NEXT: global_load_ushort v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_cmp_lt_u32_e32 vcc_lo, v1, v2
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v1, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
+; GFX10-NEXT: global_store_short v0, v1, s[0:1]
+; GFX10-NEXT: global_store_byte v0, v2, s[2:3]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_umin_ult_i16_multi_use:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b256 s[0:7], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_u16 v1, v0, s[4:5]
+; GFX11-NEXT: global_load_u16 v2, v0, s[6:7]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_cmp_lt_u32_e32 vcc_lo, v1, v2
+; GFX11-NEXT: v_cndmask_b32_e32 v1, v2, v1, vcc_lo
+; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT: global_store_b8 v0, v2, s[2:3]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%a = load i16, ptr addrspace(1) %aptr, align 2
%b = load i16, ptr addrspace(1) %bptr, align 2
%cmp = icmp ult i16 %a, %b
@@ -419,72 +2707,530 @@ define amdgpu_kernel void @v_test_umin_ult_i16_multi_use(ptr addrspace(1) %out0,
ret void
}
-
-; FUNC-LABEL: @s_test_umin_ult_v1i32
-; GCN: s_min_u32
-
-; EG: MIN_UINT
define amdgpu_kernel void @s_test_umin_ult_v1i32(ptr addrspace(1) %out, <1 x i32> %a, <1 x i32> %b) #0 {
+; EG-LABEL: s_test_umin_ult_v1i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT: MIN_UINT * T1.X, KC0[2].Z, KC0[2].W,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_umin_ult_v1i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_u32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_umin_ult_v1i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_u32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_umin_ult_v1i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_u32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_umin_ult_v1i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_u32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_umin_ult_v1i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_u32 s2, s2, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp ult <1 x i32> %a, %b
%val = select <1 x i1> %cmp, <1 x i32> %a, <1 x i32> %b
store <1 x i32> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_umin_ult_v8i32:
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-; GCN: s_min_u32
-
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
define amdgpu_kernel void @s_test_umin_ult_v8i32(ptr addrspace(1) %out, <8 x i32> %a, <8 x i32> %b) #0 {
+; EG-LABEL: s_test_umin_ult_v8i32:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 13, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T2.XYZW, T3.X, 0
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: MIN_UINT * T0.W, KC0[5].X, KC0[7].X,
+; EG-NEXT: MIN_UINT * T0.Z, KC0[4].W, KC0[6].W,
+; EG-NEXT: MIN_UINT * T0.Y, KC0[4].Z, KC0[6].Z,
+; EG-NEXT: MIN_UINT * T0.X, KC0[4].Y, KC0[6].Y,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT * T2.W, KC0[6].X, KC0[8].X,
+; EG-NEXT: MIN_UINT * T2.Z, KC0[5].W, KC0[7].W,
+; EG-NEXT: MIN_UINT * T2.Y, KC0[5].Z, KC0[7].Z,
+; EG-NEXT: MIN_UINT * T2.X, KC0[5].Y, KC0[7].Y,
+; EG-NEXT: ADD_INT * T1.W, KC0[2].Y, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: LSHR * T3.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_umin_ult_v8i32:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x8
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_min_u32 s4, s11, s19
+; CI-NEXT: s_min_u32 s5, s10, s18
+; CI-NEXT: s_min_u32 s6, s9, s17
+; CI-NEXT: s_min_u32 s7, s8, s16
+; CI-NEXT: s_min_u32 s2, s15, s23
+; CI-NEXT: s_min_u32 s3, s14, s22
+; CI-NEXT: s_min_u32 s8, s13, s21
+; CI-NEXT: s_min_u32 s9, s12, s20
+; CI-NEXT: v_mov_b32_e32 v3, s2
+; CI-NEXT: s_add_u32 s2, s0, 16
+; CI-NEXT: v_mov_b32_e32 v2, s3
+; CI-NEXT: s_addc_u32 s3, s1, 0
+; CI-NEXT: v_mov_b32_e32 v5, s3
+; CI-NEXT: v_mov_b32_e32 v0, s9
+; CI-NEXT: v_mov_b32_e32 v1, s8
+; CI-NEXT: v_mov_b32_e32 v4, s2
+; CI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; CI-NEXT: v_mov_b32_e32 v5, s1
+; CI-NEXT: v_mov_b32_e32 v0, s7
+; CI-NEXT: v_mov_b32_e32 v1, s6
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_mov_b32_e32 v3, s4
+; CI-NEXT: v_mov_b32_e32 v4, s0
+; CI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_umin_ult_v8i32:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x20
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_min_u32 s4, s11, s19
+; VI-NEXT: s_min_u32 s5, s10, s18
+; VI-NEXT: s_min_u32 s6, s9, s17
+; VI-NEXT: s_min_u32 s7, s8, s16
+; VI-NEXT: s_min_u32 s2, s15, s23
+; VI-NEXT: s_min_u32 s3, s14, s22
+; VI-NEXT: s_min_u32 s8, s13, s21
+; VI-NEXT: s_min_u32 s9, s12, s20
+; VI-NEXT: v_mov_b32_e32 v3, s2
+; VI-NEXT: s_add_u32 s2, s0, 16
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: s_addc_u32 s3, s1, 0
+; VI-NEXT: v_mov_b32_e32 v5, s3
+; VI-NEXT: v_mov_b32_e32 v0, s9
+; VI-NEXT: v_mov_b32_e32 v1, s8
+; VI-NEXT: v_mov_b32_e32 v4, s2
+; VI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: v_mov_b32_e32 v0, s7
+; VI-NEXT: v_mov_b32_e32 v1, s6
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: v_mov_b32_e32 v4, s0
+; VI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_umin_ult_v8i32:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x20
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_min_u32 s4, s9, s17
+; GFX9-NEXT: s_min_u32 s5, s8, s16
+; GFX9-NEXT: s_min_u32 s6, s15, s23
+; GFX9-NEXT: s_min_u32 s7, s14, s22
+; GFX9-NEXT: s_min_u32 s8, s13, s21
+; GFX9-NEXT: s_min_u32 s9, s12, s20
+; GFX9-NEXT: s_min_u32 s2, s11, s19
+; GFX9-NEXT: s_min_u32 s3, s10, s18
+; GFX9-NEXT: v_mov_b32_e32 v0, s9
+; GFX9-NEXT: v_mov_b32_e32 v1, s8
+; GFX9-NEXT: v_mov_b32_e32 v2, s7
+; GFX9-NEXT: v_mov_b32_e32 v3, s6
+; GFX9-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: v_mov_b32_e32 v0, s5
+; GFX9-NEXT: v_mov_b32_e32 v1, s4
+; GFX9-NEXT: v_mov_b32_e32 v2, s3
+; GFX9-NEXT: v_mov_b32_e32 v3, s2
+; GFX9-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_umin_ult_v8i32:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x20
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v8, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_min_u32 s4, s9, s17
+; GFX10-NEXT: s_min_u32 s5, s8, s16
+; GFX10-NEXT: s_min_u32 s6, s15, s23
+; GFX10-NEXT: s_min_u32 s7, s14, s22
+; GFX10-NEXT: s_min_u32 s8, s12, s20
+; GFX10-NEXT: s_min_u32 s9, s13, s21
+; GFX10-NEXT: s_min_u32 s2, s11, s19
+; GFX10-NEXT: s_min_u32 s3, s10, s18
+; GFX10-NEXT: v_mov_b32_e32 v0, s8
+; GFX10-NEXT: v_mov_b32_e32 v1, s9
+; GFX10-NEXT: v_mov_b32_e32 v2, s7
+; GFX10-NEXT: v_mov_b32_e32 v3, s6
+; GFX10-NEXT: v_mov_b32_e32 v4, s5
+; GFX10-NEXT: v_mov_b32_e32 v5, s4
+; GFX10-NEXT: v_mov_b32_e32 v6, s3
+; GFX10-NEXT: v_mov_b32_e32 v7, s2
+; GFX10-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16
+; GFX10-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_umin_ult_v8i32:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b512 s[4:19], s[0:1], 0x20
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v8, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_min_u32 s2, s7, s15
+; GFX11-NEXT: s_min_u32 s3, s6, s14
+; GFX11-NEXT: s_min_u32 s6, s11, s19
+; GFX11-NEXT: s_min_u32 s7, s10, s18
+; GFX11-NEXT: s_min_u32 s8, s8, s16
+; GFX11-NEXT: s_min_u32 s9, s9, s17
+; GFX11-NEXT: s_min_u32 s5, s5, s13
+; GFX11-NEXT: s_min_u32 s4, s4, s12
+; GFX11-NEXT: v_mov_b32_e32 v0, s8
+; GFX11-NEXT: v_mov_b32_e32 v1, s9
+; GFX11-NEXT: v_mov_b32_e32 v2, s7
+; GFX11-NEXT: v_mov_b32_e32 v3, s6
+; GFX11-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-NEXT: v_mov_b32_e32 v5, s5
+; GFX11-NEXT: v_mov_b32_e32 v6, s3
+; GFX11-NEXT: v_mov_b32_e32 v7, s2
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b128 v8, v[0:3], s[0:1] offset:16
+; GFX11-NEXT: global_store_b128 v8, v[4:7], s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp ult <8 x i32> %a, %b
%val = select <8 x i1> %cmp, <8 x i32> %a, <8 x i32> %b
store <8 x i32> %val, ptr addrspace(1) %out
ret void
}
-; FUNC-LABEL: {{^}}s_test_umin_ult_v8i16:
-; GCN-NOT: {{buffer|flat|global}}_load
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-; CI: s_min_u32
-
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-; VI: s_min_u32
-
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
-; EG: MIN_UINT
define amdgpu_kernel void @s_test_umin_ult_v8i16(ptr addrspace(1) %out, <8 x i16> %a, <8 x i16> %b) #0 {
+; EG-LABEL: s_test_umin_ult_v8i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 1, @52, KC0[], KC1[]
+; EG-NEXT: TEX 1 @20
+; EG-NEXT: ALU 9, @54, KC0[], KC1[]
+; EG-NEXT: TEX 1 @24
+; EG-NEXT: ALU 8, @64, KC0[], KC1[]
+; EG-NEXT: TEX 1 @28
+; EG-NEXT: ALU 10, @73, KC0[], KC1[]
+; EG-NEXT: TEX 1 @32
+; EG-NEXT: ALU 8, @84, KC0[], KC1[]
+; EG-NEXT: TEX 1 @36
+; EG-NEXT: ALU 10, @93, KC0[], KC1[]
+; EG-NEXT: TEX 1 @40
+; EG-NEXT: ALU 8, @104, KC0[], KC1[]
+; EG-NEXT: TEX 1 @44
+; EG-NEXT: ALU 10, @113, KC0[], KC1[]
+; EG-NEXT: TEX 1 @48
+; EG-NEXT: ALU 10, @124, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T7.XYZW, T8.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 20:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 66, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 82, #3
+; EG-NEXT: Fetch clause starting at 24:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 64, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 80, #3
+; EG-NEXT: Fetch clause starting at 28:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 62, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 78, #3
+; EG-NEXT: Fetch clause starting at 32:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 60, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 76, #3
+; EG-NEXT: Fetch clause starting at 36:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 58, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 74, #3
+; EG-NEXT: Fetch clause starting at 40:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 56, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 72, #3
+; EG-NEXT: Fetch clause starting at 44:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 54, #3
+; EG-NEXT: VTX_READ_16 T9.X, T7.X, 70, #3
+; EG-NEXT: Fetch clause starting at 48:
+; EG-NEXT: VTX_READ_16 T8.X, T7.X, 52, #3
+; EG-NEXT: VTX_READ_16 T7.X, T7.X, 68, #3
+; EG-NEXT: ALU clause starting at 52:
+; EG-NEXT: MOV * T0.Y, T3.X,
+; EG-NEXT: MOV * T7.X, 0.0,
+; EG-NEXT: ALU clause starting at 54:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT * T0.W, PV.W, PS,
+; EG-NEXT: LSHL T0.W, PV.W, literal.x,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT: 16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT: OR_INT * T0.W, PS, PV.W,
+; EG-NEXT: MOV * T3.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 64:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: AND_INT T2.W, T0.Y, literal.x,
+; EG-NEXT: MIN_UINT * T0.W, PV.W, PS,
+; EG-NEXT: -65536(nan), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, PV.W, PS,
+; EG-NEXT: MOV T3.X, PV.W,
+; EG-NEXT: MOV * T0.Y, T2.X,
+; EG-NEXT: ALU clause starting at 73:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT T0.W, PV.W, PS,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT: MOV * T2.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 84:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: AND_INT T2.W, T0.Y, literal.x,
+; EG-NEXT: MIN_UINT * T0.W, PV.W, PS,
+; EG-NEXT: -65536(nan), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T7.Z, PV.W, PS,
+; EG-NEXT: MOV T2.X, PV.Z,
+; EG-NEXT: MOV * T0.Y, T5.X,
+; EG-NEXT: ALU clause starting at 93:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT T0.W, PV.W, PS,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT: MOV * T5.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 104:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: AND_INT T2.W, T0.Y, literal.x,
+; EG-NEXT: MIN_UINT * T0.W, PV.W, PS,
+; EG-NEXT: -65536(nan), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, PV.W, PS,
+; EG-NEXT: MOV T5.X, PV.W,
+; EG-NEXT: MOV * T0.Y, T4.X,
+; EG-NEXT: ALU clause starting at 113:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T9.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT T0.W, PV.W, PS,
+; EG-NEXT: AND_INT * T1.W, T0.Y, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT: MOV * T4.X, PV.W,
+; EG-NEXT: MOV * T0.Y, PV.X,
+; EG-NEXT: ALU clause starting at 124:
+; EG-NEXT: AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT: AND_INT * T1.W, T7.X, literal.x,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHR T8.X, KC0[2].Y, literal.x,
+; EG-NEXT: AND_INT T2.W, T0.Y, literal.y,
+; EG-NEXT: MIN_UINT * T0.W, PV.W, PS,
+; EG-NEXT: 2(2.802597e-45), -65536(nan)
+; EG-NEXT: OR_INT * T7.X, PV.W, PS,
+; EG-NEXT: MOV T4.X, PV.X,
+; EG-NEXT: MOV * T7.W, T3.X,
+; EG-NEXT: MOV * T7.Y, T5.X,
+;
+; CI-LABEL: s_test_umin_ult_v8i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x4
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_lshr_b32 s2, s8, 16
+; CI-NEXT: s_and_b32 s3, s8, 0xffff
+; CI-NEXT: s_lshr_b32 s4, s9, 16
+; CI-NEXT: s_and_b32 s5, s9, 0xffff
+; CI-NEXT: s_lshr_b32 s6, s10, 16
+; CI-NEXT: s_and_b32 s7, s10, 0xffff
+; CI-NEXT: s_lshr_b32 s8, s11, 16
+; CI-NEXT: s_and_b32 s9, s11, 0xffff
+; CI-NEXT: s_lshr_b32 s10, s12, 16
+; CI-NEXT: s_and_b32 s11, s12, 0xffff
+; CI-NEXT: s_lshr_b32 s12, s13, 16
+; CI-NEXT: s_lshr_b32 s16, s14, 16
+; CI-NEXT: s_lshr_b32 s17, s15, 16
+; CI-NEXT: s_and_b32 s13, s13, 0xffff
+; CI-NEXT: s_and_b32 s14, s14, 0xffff
+; CI-NEXT: s_and_b32 s15, s15, 0xffff
+; CI-NEXT: s_min_u32 s8, s8, s17
+; CI-NEXT: s_min_u32 s6, s6, s16
+; CI-NEXT: s_min_u32 s4, s4, s12
+; CI-NEXT: s_min_u32 s2, s2, s10
+; CI-NEXT: s_min_u32 s9, s9, s15
+; CI-NEXT: s_lshl_b32 s8, s8, 16
+; CI-NEXT: s_min_u32 s7, s7, s14
+; CI-NEXT: s_lshl_b32 s6, s6, 16
+; CI-NEXT: s_min_u32 s5, s5, s13
+; CI-NEXT: s_lshl_b32 s4, s4, 16
+; CI-NEXT: s_min_u32 s3, s3, s11
+; CI-NEXT: s_lshl_b32 s2, s2, 16
+; CI-NEXT: s_or_b32 s8, s9, s8
+; CI-NEXT: s_or_b32 s6, s7, s6
+; CI-NEXT: s_or_b32 s4, s5, s4
+; CI-NEXT: s_or_b32 s2, s3, s2
+; CI-NEXT: v_mov_b32_e32 v5, s1
+; CI-NEXT: v_mov_b32_e32 v0, s2
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s6
+; CI-NEXT: v_mov_b32_e32 v3, s8
+; CI-NEXT: v_mov_b32_e32 v4, s0
+; CI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_umin_ult_v8i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s2, s11, 16
+; VI-NEXT: s_lshr_b32 s4, s10, 16
+; VI-NEXT: s_and_b32 s5, s10, 0xffff
+; VI-NEXT: s_lshr_b32 s10, s15, 16
+; VI-NEXT: s_and_b32 s3, s11, 0xffff
+; VI-NEXT: s_and_b32 s11, s15, 0xffff
+; VI-NEXT: s_lshr_b32 s15, s14, 16
+; VI-NEXT: s_min_u32 s2, s2, s10
+; VI-NEXT: s_lshr_b32 s6, s9, 16
+; VI-NEXT: s_and_b32 s7, s9, 0xffff
+; VI-NEXT: s_lshr_b32 s9, s8, 16
+; VI-NEXT: s_and_b32 s14, s14, 0xffff
+; VI-NEXT: s_lshr_b32 s16, s13, 16
+; VI-NEXT: s_lshr_b32 s17, s12, 16
+; VI-NEXT: s_min_u32 s4, s4, s15
+; VI-NEXT: s_min_u32 s3, s3, s11
+; VI-NEXT: s_lshl_b32 s2, s2, 16
+; VI-NEXT: s_and_b32 s8, s8, 0xffff
+; VI-NEXT: s_and_b32 s13, s13, 0xffff
+; VI-NEXT: s_and_b32 s12, s12, 0xffff
+; VI-NEXT: s_min_u32 s9, s9, s17
+; VI-NEXT: s_min_u32 s6, s6, s16
+; VI-NEXT: s_min_u32 s5, s5, s14
+; VI-NEXT: s_or_b32 s2, s3, s2
+; VI-NEXT: s_lshl_b32 s3, s4, 16
+; VI-NEXT: s_min_u32 s8, s8, s12
+; VI-NEXT: s_min_u32 s7, s7, s13
+; VI-NEXT: s_or_b32 s3, s5, s3
+; VI-NEXT: s_lshl_b32 s4, s6, 16
+; VI-NEXT: s_lshl_b32 s5, s9, 16
+; VI-NEXT: s_or_b32 s4, s7, s4
+; VI-NEXT: s_or_b32 s5, s8, s5
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: v_mov_b32_e32 v0, s5
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: v_mov_b32_e32 v3, s2
+; VI-NEXT: v_mov_b32_e32 v4, s0
+; VI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_umin_ult_v8i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s15
+; GFX9-NEXT: v_mov_b32_e32 v1, s14
+; GFX9-NEXT: v_pk_min_u16 v3, s11, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s13
+; GFX9-NEXT: v_pk_min_u16 v2, s10, v1
+; GFX9-NEXT: v_pk_min_u16 v1, s9, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, s12
+; GFX9-NEXT: v_pk_min_u16 v0, s8, v0
+; GFX9-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_umin_ult_v8i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x10
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v4, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_pk_min_u16 v3, s11, s15
+; GFX10-NEXT: v_pk_min_u16 v2, s10, s14
+; GFX10-NEXT: v_pk_min_u16 v1, s9, s13
+; GFX10-NEXT: v_pk_min_u16 v0, s8, s12
+; GFX10-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_umin_ult_v8i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b256 s[4:11], s[0:1], 0x10
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v4, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_pk_min_u16 v3, s7, s11
+; GFX11-NEXT: v_pk_min_u16 v2, s6, s10
+; GFX11-NEXT: v_pk_min_u16 v1, s5, s9
+; GFX11-NEXT: v_pk_min_u16 v0, s4, s8
+; GFX11-NEXT: global_store_b128 v4, v[0:3], s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp ult <8 x i16> %a, %b
%val = select <8 x i1> %cmp, <8 x i16> %a, <8 x i16> %b
store <8 x i16> %val, ptr addrspace(1) %out
@@ -492,15 +3238,105 @@ define amdgpu_kernel void @s_test_umin_ult_v8i16(ptr addrspace(1) %out, <8 x i16
}
; Make sure redundant and removed
-; FUNC-LABEL: {{^}}simplify_demanded_bits_test_umin_ult_i16:
-; GCN-DAG: s_load_{{dword|b32}} [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, {{0xa|0x28}}
-; GCN-DAG: s_load_{{dword|b32}} [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, {{0x13|0x4c}}
-; GCN: s_min_u32 [[MIN:s[0-9]+]], s{{[0-9]}}, s{{[0-9]}}
-; GCN: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
-; GCN: {{flat|global}}_store_{{dword|b32}} v{{.+}}, [[VMIN]]
-
-; EG: MIN_UINT
+
define amdgpu_kernel void @simplify_demanded_bits_test_umin_ult_i16(ptr addrspace(1) %out, [8 x i32], i16 zeroext %a, [8 x i32], i16 zeroext %b) #0 {
+; EG-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 5, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_UINT T0.X, PV.Z, PV.W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0xa
+; CI-NEXT: s_load_dword s3, s[4:5], 0x13
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_and_b32 s2, s2, 0xffff
+; CI-NEXT: s_and_b32 s3, s3, 0xffff
+; CI-NEXT: s_min_u32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x28
+; VI-NEXT: s_load_dword s3, s[4:5], 0x4c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_and_b32 s2, s2, 0xffff
+; VI-NEXT: s_and_b32 s3, s3, 0xffff
+; VI-NEXT: s_min_u32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX9-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX9-NEXT: s_and_b32 s3, s3, 0xffff
+; GFX9-NEXT: s_min_u32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x2
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX10-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX10-NEXT: s_and_b32 s3, s3, 0xffff
+; GFX10-NEXT: s_min_u32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: simplify_demanded_bits_test_umin_ult_i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x2
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x28
+; GFX11-NEXT: s_load_b32 s3, s[0:1], 0x4c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-NEXT: s_and_b32 s3, s3, 0xffff
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT: s_min_u32 s2, s2, s3
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%a.ext = zext i16 %a to i32
%b.ext = zext i16 %b to i32
%cmp = icmp ult i32 %a.ext, %b.ext
@@ -512,18 +3348,104 @@ define amdgpu_kernel void @simplify_demanded_bits_test_umin_ult_i16(ptr addrspac
; Make sure redundant sign_extend_inreg removed.
-; FUNC-LABEL: {{^}}simplify_demanded_bits_test_min_slt_i16:
-; GCN-DAG: s_load_{{dword|b32}} [[A:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, {{0xa|0x28}}
-; GCN-DAG: s_load_{{dword|b32}} [[B:s[0-9]+]], {{s\[[0-9]+:[0-9]+\]}}, {{0x13|0x4c}}
-; GCN-DAG: s_sext_i32_i16 [[EXT_A:s[0-9]+]], [[A]]
-; GCN-DAG: s_sext_i32_i16 [[EXT_B:s[0-9]+]], [[B]]
-
-; GCN: s_min_i32 [[MIN:s[0-9]+]], [[EXT_A]], [[EXT_B]]
-; GCN: v_mov_b32_e32 [[VMIN:v[0-9]+]], [[MIN]]
-; GCN: {{flat|global}}_store_{{dword|b32}} v{{.+}}, [[VMIN]]
-
-; EG: MIN_INT
define amdgpu_kernel void @simplify_demanded_bits_test_min_slt_i16(ptr addrspace(1) %out, [8 x i32], i16 signext %a, [8 x i32], i16 signext %b) #0 {
+; EG-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 5, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 72, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 108, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T0.X, PV.Z, PV.W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0xa
+; CI-NEXT: s_load_dword s3, s[4:5], 0x13
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_sext_i32_i16 s2, s2
+; CI-NEXT: s_sext_i32_i16 s3, s3
+; CI-NEXT: s_min_i32 s2, s2, s3
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x28
+; VI-NEXT: s_load_dword s3, s[4:5], 0x4c
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_sext_i32_i16 s2, s2
+; VI-NEXT: s_sext_i32_i16 s3, s3
+; VI-NEXT: s_min_i32 s2, s2, s3
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX9-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_sext_i32_i16 s2, s2
+; GFX9-NEXT: s_sext_i32_i16 s3, s3
+; GFX9-NEXT: s_min_i32 s2, s2, s3
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x2
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x28
+; GFX10-NEXT: s_load_dword s3, s[4:5], 0x4c
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_sext_i32_i16 s2, s2
+; GFX10-NEXT: s_sext_i32_i16 s3, s3
+; GFX10-NEXT: s_min_i32 s2, s2, s3
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: simplify_demanded_bits_test_min_slt_i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x2
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x28
+; GFX11-NEXT: s_load_b32 s3, s[0:1], 0x4c
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_sext_i32_i16 s2, s2
+; GFX11-NEXT: s_sext_i32_i16 s3, s3
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT: s_min_i32 s2, s2, s3
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%a.ext = sext i16 %a to i32
%b.ext = sext i16 %b to i32
%cmp = icmp slt i32 %a.ext, %b.ext
@@ -534,11 +3456,108 @@ define amdgpu_kernel void @simplify_demanded_bits_test_min_slt_i16(ptr addrspace
ret void
}
-; FUNC-LABEL: {{^}}s_test_imin_sle_i16:
-; GCN: s_min_i32
-
-; EG: MIN_INT
define amdgpu_kernel void @s_test_imin_sle_i16(ptr addrspace(1) %out, i16 %a, i16 %b) #0 {
+; EG-LABEL: s_test_imin_sle_i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 0, @10, KC0[], KC1[]
+; EG-NEXT: TEX 1 @6
+; EG-NEXT: ALU 14, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 6:
+; EG-NEXT: VTX_READ_16 T1.X, T0.X, 40, #3
+; EG-NEXT: VTX_READ_16 T0.X, T0.X, 42, #3
+; EG-NEXT: ALU clause starting at 10:
+; EG-NEXT: MOV * T0.X, 0.0,
+; EG-NEXT: ALU clause starting at 11:
+; EG-NEXT: BFE_INT T0.Z, T1.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: AND_INT * T1.W, KC0[2].Y, literal.y,
+; EG-NEXT: 16(2.242078e-44), 3(4.203895e-45)
+; EG-NEXT: MIN_INT * T0.W, PV.Z, PV.W,
+; EG-NEXT: AND_INT T0.W, PV.W, literal.x,
+; EG-NEXT: LSHL * T1.W, T1.W, literal.y,
+; EG-NEXT: 65535(9.183409e-41), 3(4.203895e-45)
+; EG-NEXT: LSHL T0.X, PV.W, PS,
+; EG-NEXT: LSHL * T0.W, literal.x, PS,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: MOV T0.Y, 0.0,
+; EG-NEXT: MOV * T0.Z, 0.0,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: s_test_imin_sle_i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dword s2, s[4:5], 0x2
+; CI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: s_sext_i32_i16 s3, s2
+; CI-NEXT: s_ashr_i32 s2, s2, 16
+; CI-NEXT: s_min_i32 s2, s3, s2
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_mov_b32_e32 v2, s2
+; CI-NEXT: flat_store_short v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: s_test_imin_sle_i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dword s2, s[4:5], 0x8
+; VI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_sext_i32_i16 s3, s2
+; VI-NEXT: s_ashr_i32 s2, s2, 16
+; VI-NEXT: s_min_i32 s2, s3, s2
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: flat_store_short v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: s_test_imin_sle_i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: s_sext_i32_i16 s3, s2
+; GFX9-NEXT: s_ashr_i32 s2, s2, 16
+; GFX9-NEXT: s_min_i32 s2, s3, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s2
+; GFX9-NEXT: global_store_short v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: s_test_imin_sle_i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dword s2, s[4:5], 0x8
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_sext_i32_i16 s3, s2
+; GFX10-NEXT: s_ashr_i32 s2, s2, 16
+; GFX10-NEXT: s_min_i32 s2, s3, s2
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: global_store_short v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: s_test_imin_sle_i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x8
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_sext_i32_i16 s3, s2
+; GFX11-NEXT: s_ashr_i32 s2, s2, 16
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT: s_min_i32 s2, s3, s2
+; GFX11-NEXT: v_mov_b32_e32 v1, s2
+; GFX11-NEXT: global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%cmp = icmp sle i16 %a, %b
%val = select i1 %cmp, i16 %a, i16 %b
store i16 %val, ptr addrspace(1) %out
@@ -546,82 +3565,586 @@ define amdgpu_kernel void @s_test_imin_sle_i16(ptr addrspace(1) %out, i16 %a, i1
}
; 64 bit
-; FUNC-LABEL: {{^}}test_umin_ult_i64
-; GCN: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_UINT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
+
define amdgpu_kernel void @test_umin_ult_i64(ptr addrspace(1) %out, i64 %a, i64 %b) #0 {
+; EG-LABEL: test_umin_ult_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T0.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: test_umin_ult_i64:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[1:2]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: s_and_b64 s[0:1], vcc, exec
+; CI-NEXT: s_cselect_b32 s0, s3, s5
+; CI-NEXT: s_cselect_b32 s1, s2, s4
+; CI-NEXT: v_mov_b32_e32 v2, s1
+; CI-NEXT: v_mov_b32_e32 v3, s0
+; CI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: test_umin_ult_i64:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[1:2]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_and_b64 s[0:1], vcc, exec
+; VI-NEXT: s_cselect_b32 s0, s3, s5
+; VI-NEXT: s_cselect_b32 s1, s2, s4
+; VI-NEXT: v_mov_b32_e32 v2, s1
+; VI-NEXT: v_mov_b32_e32 v3, s0
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: test_umin_ult_i64:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s6
+; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1]
+; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s3, s3, s7
+; GFX9-NEXT: s_cselect_b32 s2, s2, s6
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: test_umin_ult_i64:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_cmp_lt_u64_e64 s4, s[2:3], s[6:7]
+; GFX10-NEXT: s_and_b32 s4, s4, exec_lo
+; GFX10-NEXT: s_cselect_b32 s2, s2, s6
+; GFX10-NEXT: s_cselect_b32 s3, s3, s7
+; GFX10-NEXT: v_mov_b32_e32 v0, s2
+; GFX10-NEXT: v_mov_b32_e32 v1, s3
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: test_umin_ult_i64:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_lt_u64_e64 s2, s[6:7], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: s_and_b32 s2, s2, exec_lo
+; GFX11-NEXT: s_cselect_b32 s0, s6, s0
+; GFX11-NEXT: s_cselect_b32 s1, s7, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, s0
+; GFX11-NEXT: v_mov_b32_e32 v1, s1
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tmp = icmp ult i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_umin_ule_i64
-; GCN: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_UINT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_umin_ule_i64(ptr addrspace(1) %out, i64 %a, i64 %b) #0 {
+; EG-LABEL: test_umin_ule_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_UINT * T0.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: test_umin_ule_i64:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_cmp_le_u64_e32 vcc, s[2:3], v[1:2]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: s_and_b64 s[0:1], vcc, exec
+; CI-NEXT: s_cselect_b32 s0, s3, s5
+; CI-NEXT: s_cselect_b32 s1, s2, s4
+; CI-NEXT: v_mov_b32_e32 v2, s1
+; CI-NEXT: v_mov_b32_e32 v3, s0
+; CI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: test_umin_ule_i64:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_cmp_le_u64_e32 vcc, s[2:3], v[1:2]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_and_b64 s[0:1], vcc, exec
+; VI-NEXT: s_cselect_b32 s0, s3, s5
+; VI-NEXT: s_cselect_b32 s1, s2, s4
+; VI-NEXT: v_mov_b32_e32 v2, s1
+; VI-NEXT: v_mov_b32_e32 v3, s0
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: test_umin_ule_i64:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s6
+; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_cmp_le_u64_e32 vcc, s[2:3], v[0:1]
+; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s3, s3, s7
+; GFX9-NEXT: s_cselect_b32 s2, s2, s6
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: test_umin_ule_i64:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_cmp_le_u64_e64 s4, s[2:3], s[6:7]
+; GFX10-NEXT: s_and_b32 s4, s4, exec_lo
+; GFX10-NEXT: s_cselect_b32 s2, s2, s6
+; GFX10-NEXT: s_cselect_b32 s3, s3, s7
+; GFX10-NEXT: v_mov_b32_e32 v0, s2
+; GFX10-NEXT: v_mov_b32_e32 v1, s3
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: test_umin_ule_i64:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_le_u64_e64 s2, s[6:7], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: s_and_b32 s2, s2, exec_lo
+; GFX11-NEXT: s_cselect_b32 s0, s6, s0
+; GFX11-NEXT: s_cselect_b32 s1, s7, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, s0
+; GFX11-NEXT: v_mov_b32_e32 v1, s1
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tmp = icmp ule i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_imin_slt_i64
-; GCN: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_INT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_imin_slt_i64(ptr addrspace(1) %out, i64 %a, i64 %b) #0 {
+; EG-LABEL: test_imin_slt_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_INT * T0.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: test_imin_slt_i64:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[1:2]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: s_and_b64 s[0:1], vcc, exec
+; CI-NEXT: s_cselect_b32 s0, s3, s5
+; CI-NEXT: s_cselect_b32 s1, s2, s4
+; CI-NEXT: v_mov_b32_e32 v2, s1
+; CI-NEXT: v_mov_b32_e32 v3, s0
+; CI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: test_imin_slt_i64:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[1:2]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_and_b64 s[0:1], vcc, exec
+; VI-NEXT: s_cselect_b32 s0, s3, s5
+; VI-NEXT: s_cselect_b32 s1, s2, s4
+; VI-NEXT: v_mov_b32_e32 v2, s1
+; VI-NEXT: v_mov_b32_e32 v3, s0
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: test_imin_slt_i64:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s6
+; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[0:1]
+; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s3, s3, s7
+; GFX9-NEXT: s_cselect_b32 s2, s2, s6
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: test_imin_slt_i64:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_cmp_lt_i64_e64 s4, s[2:3], s[6:7]
+; GFX10-NEXT: s_and_b32 s4, s4, exec_lo
+; GFX10-NEXT: s_cselect_b32 s2, s2, s6
+; GFX10-NEXT: s_cselect_b32 s3, s3, s7
+; GFX10-NEXT: v_mov_b32_e32 v0, s2
+; GFX10-NEXT: v_mov_b32_e32 v1, s3
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: test_imin_slt_i64:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_lt_i64_e64 s2, s[6:7], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: s_and_b32 s2, s2, exec_lo
+; GFX11-NEXT: s_cselect_b32 s0, s6, s0
+; GFX11-NEXT: s_cselect_b32 s1, s7, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, s0
+; GFX11-NEXT: v_mov_b32_e32 v1, s1
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tmp = icmp slt i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}test_imin_sle_i64
-; GCN: s_endpgm
-
-; EG: SETE_INT
-; EG: SETGT_INT
-; EG: SETGT_UINT
-; EG: CNDE_INT
-; EG: CNDE_INT
-; EG: CNDE_INT
define amdgpu_kernel void @test_imin_sle_i64(ptr addrspace(1) %out, i64 %a, i64 %b) #0 {
+; EG-LABEL: test_imin_sle_i64:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: ALU clause starting at 4:
+; EG-NEXT: SETE_INT T0.Z, KC0[3].X, KC0[3].Z,
+; EG-NEXT: SETGT_INT * T0.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: SETGT_UINT * T1.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: CNDE_INT * T0.W, T0.Z, T0.W, PV.W,
+; EG-NEXT: CNDE_INT * T0.Y, PV.W, KC0[3].Z, KC0[3].X,
+; EG-NEXT: CNDE_INT * T0.X, T0.W, KC0[3].Y, KC0[2].W,
+; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: test_imin_sle_i64:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v0, s0
+; CI-NEXT: v_mov_b32_e32 v1, s4
+; CI-NEXT: v_mov_b32_e32 v2, s5
+; CI-NEXT: v_cmp_le_i64_e32 vcc, s[2:3], v[1:2]
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: s_and_b64 s[0:1], vcc, exec
+; CI-NEXT: s_cselect_b32 s0, s3, s5
+; CI-NEXT: s_cselect_b32 s1, s2, s4
+; CI-NEXT: v_mov_b32_e32 v2, s1
+; CI-NEXT: v_mov_b32_e32 v3, s0
+; CI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: test_imin_sle_i64:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: v_mov_b32_e32 v1, s4
+; VI-NEXT: v_mov_b32_e32 v2, s5
+; VI-NEXT: v_cmp_le_i64_e32 vcc, s[2:3], v[1:2]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_and_b64 s[0:1], vcc, exec
+; VI-NEXT: s_cselect_b32 s0, s3, s5
+; VI-NEXT: s_cselect_b32 s1, s2, s4
+; VI-NEXT: v_mov_b32_e32 v2, s1
+; VI-NEXT: v_mov_b32_e32 v3, s0
+; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: test_imin_sle_i64:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: v_mov_b32_e32 v2, 0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: v_mov_b32_e32 v0, s6
+; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_cmp_le_i64_e32 vcc, s[2:3], v[0:1]
+; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s3, s3, s7
+; GFX9-NEXT: s_cselect_b32 s2, s2, s6
+; GFX9-NEXT: v_mov_b32_e32 v0, s2
+; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: test_imin_sle_i64:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: v_cmp_le_i64_e64 s4, s[2:3], s[6:7]
+; GFX10-NEXT: s_and_b32 s4, s4, exec_lo
+; GFX10-NEXT: s_cselect_b32 s2, s2, s6
+; GFX10-NEXT: s_cselect_b32 s3, s3, s7
+; GFX10-NEXT: v_mov_b32_e32 v0, s2
+; GFX10-NEXT: v_mov_b32_e32 v1, s3
+; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: test_imin_sle_i64:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_cmp_le_i64_e64 s2, s[6:7], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: s_and_b32 s2, s2, exec_lo
+; GFX11-NEXT: s_cselect_b32 s0, s6, s0
+; GFX11-NEXT: s_cselect_b32 s1, s7, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, s0
+; GFX11-NEXT: v_mov_b32_e32 v1, s1
+; GFX11-NEXT: global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tmp = icmp sle i64 %a, %b
%val = select i1 %tmp, i64 %a, i64 %b
store i64 %val, ptr addrspace(1) %out, align 8
ret void
}
-; FUNC-LABEL: {{^}}v_test_imin_sle_v2i16:
-; CI: v_min_i32
-; CI: v_min_i32
-
-; VI: v_min_i16
-; VI: v_min_i16
-
-; GFX9_10: v_pk_min_i16
-
-; EG: MIN_INT
-; EG: MIN_INT
define amdgpu_kernel void @v_test_imin_sle_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_imin_sle_v2i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @8
+; EG-NEXT: ALU 0, @15, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @10
+; EG-NEXT: ALU 16, @16, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T7.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 8:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: Fetch clause starting at 10:
+; EG-NEXT: VTX_READ_32 T7.X, T7.X, 0, #1
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].Z, PV.W,
+; EG-NEXT: ALU clause starting at 15:
+; EG-NEXT: ADD_INT * T7.X, KC0[2].W, T0.W,
+; EG-NEXT: ALU clause starting at 16:
+; EG-NEXT: LSHR T1.W, T0.X, literal.x,
+; EG-NEXT: LSHR * T2.W, T7.X, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: BFE_INT T8.X, PS, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.Y, PV.W, 0.0, literal.x,
+; EG-NEXT: BFE_INT T0.Z, T7.X, 0.0, literal.x,
+; EG-NEXT: BFE_INT * T1.W, T0.X, 0.0, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: MIN_INT T1.W, PV.W, PV.Z,
+; EG-NEXT: MIN_INT * T2.W, PV.Y, PV.X,
+; EG-NEXT: LSHL T2.W, PS, literal.x,
+; EG-NEXT: AND_INT * T1.W, PV.W, literal.y,
+; EG-NEXT: 16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT: OR_INT T0.X, PS, PV.W,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T7.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_imin_sle_v2i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: flat_load_dword v4, v[0:1]
+; CI-NEXT: v_add_i32_e32 v0, vcc, s4, v2
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v3, v[0:1]
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(1)
+; CI-NEXT: v_bfe_i32 v2, v4, 0, 16
+; CI-NEXT: v_ashrrev_i32_e32 v4, 16, v4
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_bfe_i32 v5, v3, 0, 16
+; CI-NEXT: v_ashrrev_i32_e32 v3, 16, v3
+; CI-NEXT: v_min_i32_e32 v3, v4, v3
+; CI-NEXT: v_min_i32_e32 v2, v2, v5
+; CI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; CI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; CI-NEXT: v_or_b32_e32 v2, v2, v3
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_imin_sle_v2i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_i16_e32 v3, v5, v2
+; VI-NEXT: v_min_i16_sdwa v2, v5, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT: v_or_b32_e32 v2, v3, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_imin_sle_v2i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_pk_min_i16 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_imin_sle_v2i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_pk_min_i16 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_imin_sle_v2i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_pk_min_i16 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds <2 x i16>, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds <2 x i16>, ptr addrspace(1) %b.ptr, i32 %tid
@@ -635,18 +4158,142 @@ define amdgpu_kernel void @v_test_imin_sle_v2i16(ptr addrspace(1) %out, ptr addr
}
; FIXME: i16 min
-; FUNC-LABEL: {{^}}v_test_imin_ule_v2i16:
-; CI: v_min_u32
-; CI: v_min_u32
-
-; VI: v_min_u16
-; VI: v_min_u16
-
-; GFX9_10: v_pk_min_u16
-; EG: MIN_UINT
-; EG: MIN_UINT
define amdgpu_kernel void @v_test_imin_ule_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %a.ptr, ptr addrspace(1) %b.ptr) #0 {
+; EG-LABEL: v_test_imin_ule_v2i16:
+; EG: ; %bb.0:
+; EG-NEXT: ALU 2, @12, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @8
+; EG-NEXT: ALU 0, @15, KC0[CB0:0-32], KC1[]
+; EG-NEXT: TEX 0 @10
+; EG-NEXT: ALU 13, @16, KC0[CB0:0-32], KC1[]
+; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T7.X, 1
+; EG-NEXT: CF_END
+; EG-NEXT: PAD
+; EG-NEXT: Fetch clause starting at 8:
+; EG-NEXT: VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT: Fetch clause starting at 10:
+; EG-NEXT: VTX_READ_32 T7.X, T7.X, 0, #1
+; EG-NEXT: ALU clause starting at 12:
+; EG-NEXT: LSHL * T0.W, T0.X, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT: ADD_INT * T0.X, KC0[2].W, PV.W,
+; EG-NEXT: ALU clause starting at 15:
+; EG-NEXT: ADD_INT * T7.X, KC0[2].Z, T0.W,
+; EG-NEXT: ALU clause starting at 16:
+; EG-NEXT: LSHR T1.W, T0.X, literal.x,
+; EG-NEXT: LSHR * T2.W, T7.X, literal.x,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: AND_INT T0.Z, T0.X, literal.x,
+; EG-NEXT: AND_INT T3.W, T7.X, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT: MIN_UINT * T1.W, PS, PV.W,
+; EG-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT: LSHL T1.W, PS, literal.x,
+; EG-NEXT: MIN_UINT * T2.W, PV.W, PV.Z,
+; EG-NEXT: 16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT: OR_INT T0.X, PS, PV.W,
+; EG-NEXT: ADD_INT * T0.W, KC0[2].Y, T0.W,
+; EG-NEXT: LSHR * T7.X, PV.W, literal.x,
+; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+;
+; CI-LABEL: v_test_imin_ule_v2i16:
+; CI: ; %bb.0:
+; CI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x4
+; CI-NEXT: v_lshlrev_b32_e32 v2, 2, v0
+; CI-NEXT: s_waitcnt lgkmcnt(0)
+; CI-NEXT: v_mov_b32_e32 v1, s3
+; CI-NEXT: v_add_i32_e32 v0, vcc, s2, v2
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: v_mov_b32_e32 v3, s5
+; CI-NEXT: flat_load_dword v4, v[0:1]
+; CI-NEXT: v_add_i32_e32 v0, vcc, s4, v2
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v3, vcc
+; CI-NEXT: flat_load_dword v3, v[0:1]
+; CI-NEXT: v_add_i32_e32 v0, vcc, s0, v2
+; CI-NEXT: v_mov_b32_e32 v1, s1
+; CI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CI-NEXT: s_waitcnt vmcnt(1)
+; CI-NEXT: v_lshrrev_b32_e32 v2, 16, v4
+; CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_lshrrev_b32_e32 v5, 16, v3
+; CI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; CI-NEXT: v_min_u32_e32 v2, v2, v5
+; CI-NEXT: v_min_u32_e32 v3, v4, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: flat_store_dword v[0:1], v2
+; CI-NEXT: s_endpgm
+;
+; VI-LABEL: v_test_imin_ule_v2i16:
+; VI: ; %bb.0:
+; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x10
+; VI-NEXT: v_lshlrev_b32_e32 v4, 2, v0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v1, s3
+; VI-NEXT: v_add_u32_e32 v0, vcc, s2, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
+; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT: flat_load_dword v5, v[0:1]
+; VI-NEXT: flat_load_dword v2, v[2:3]
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: v_add_u32_e32 v0, vcc, s0, v4
+; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_min_u16_e32 v3, v5, v2
+; VI-NEXT: v_min_u16_sdwa v2, v5, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT: v_or_b32_e32 v2, v3, v2
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX9-LABEL: v_test_imin_ule_v2i16:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_pk_min_u16 v1, v1, v2
+; GFX9-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT: s_endpgm
+;
+; GFX10-LABEL: v_test_imin_ule_v2i16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT: global_load_dword v2, v0, s[6:7]
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_pk_min_u16 v1, v1, v2
+; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT: s_endpgm
+;
+; GFX11-LABEL: v_test_imin_ule_v2i16:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: s_load_b128 s[4:7], s[0:1], 0x0
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x10
+; GFX11-NEXT: v_lshlrev_b32_e32 v0, 2, v0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_load_b32 v1, v0, s[6:7]
+; GFX11-NEXT: global_load_b32 v2, v0, s[0:1]
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_pk_min_u16 v1, v1, v2
+; GFX11-NEXT: global_store_b32 v0, v1, s[4:5]
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%a.gep = getelementptr inbounds <2 x i16>, ptr addrspace(1) %a.ptr, i32 %tid
%b.gep = getelementptr inbounds <2 x i16>, ptr addrspace(1) %b.ptr, i32 %tid
More information about the llvm-commits
mailing list