[llvm] [AMDGPU] Skip the iterative atomic scan on native LDS atomics (PR #227222)
Dmitry Sidorov via llvm-commits
llvm-commits at lists.llvm.org
Fri Oct 2 06:21:15 PDT 2026
https://github.com/MrSidims updated https://github.com/llvm/llvm-project/pull/227222
>From 998066d42c7b5d4a36948b10d43be1d9ff14e4d7 Mon Sep 17 00:00:00 2001
From: Dmitry Sidorov <Dmitry.Sidorov at amd.com>
Date: Tue, 29 Sep 2026 00:14:48 +0200
Subject: [PATCH] [AMDGPU] Skip the iterative atomic scan on native LDS atomics
The iterative scan runs a serial loop with one iteration per active lane.
For an LDS atomic that the target executes natively, this costs more than
the hardware serialization it replaces. Leave such atomics alone when the
value is divergent.
---
.../Target/AMDGPU/AMDGPUAtomicOptimizer.cpp | 7 +
.../atomic_optimizations_local_pointer.ll | 6170 ++---------------
.../CodeGen/AMDGPU/local-atomicrmw-fadd.ll | 547 +-
3 files changed, 800 insertions(+), 5924 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
index aa822c13dd8fe1..5c7eeb8da6cfa0 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
@@ -247,6 +247,13 @@ void AMDGPUAtomicOptimizerImpl::visitAtomicRMWInst(AtomicRMWInst &I) {
const bool IsLDS = I.getPointerAddressSpace() == AMDGPUAS::LOCAL_ADDRESS;
+ // The iterative scan runs once per active lane and costs more than the
+ // hardware serialization of a native LDS atomic.
+ if (IsLDS && ValDivergent && ScanImpl == ScanOptions::Iterative &&
+ ST.getTargetLowering()->shouldExpandAtomicRMWInIR(&I) ==
+ TargetLowering::AtomicExpansionKind::None)
+ return;
+
// If we get here, we can optimize the atomic using a single wavefront-wide
// atomic operation to do the calculation for the entire wavefront, so
// remember the instruction so we can come back to it.
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
index 4b16a4a371c183..58281f6ee42f08 100644
--- a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
@@ -571,302 +571,97 @@ entry:
}
define amdgpu_kernel void @add_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: add_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB2_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB2_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_add_i32_e32 v0, vcc, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: add_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: add_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_add_i32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB2_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_add_u32_e32 v0, vcc, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: add_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_add_i32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB2_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_add_u32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: add_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB2_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_add_nc_u32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: add_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB2_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_add_nc_u32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: add_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_add_rtn_u32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB2_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_add_nc_u32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: add_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB2_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB2_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB2_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_add_rtn_u32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB2_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_add_nc_u32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_add_rtn_u32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: add_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_add_rtn_u32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: add_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v3, 0
@@ -1278,198 +1073,62 @@ entry:
}
define amdgpu_kernel void @add_i32_varying_nouse() {
-; GFX7LESS_ITERATIVE-LABEL: add_i32_varying_nouse:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v1, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v1, exec_hi, v1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX7LESS_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 s[4:5], s[0:1], s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2:
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_add_u32 v1, v0
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB3_4:
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: add_i32_varying_nouse:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: ds_add_u32 v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX8_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX8_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX8_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2:
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
; GFX8_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB3_4:
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX9_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX9_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX9_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2:
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB3_4:
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1064_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1064_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX1064_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX1064_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2:
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB3_4:
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: s_bitset0_b32 s1, s2
-; GFX1032_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_cmp_lg_u32 s1, 0
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2:
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB3_4:
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1164_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX1164_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX1164_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2:
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v1
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1164_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB3_4:
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: add_i32_varying_nouse:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1132_ITERATIVE-NEXT: s_bitset0_b32 s1, s2
-; GFX1132_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_cmp_lg_u32 s1, 0
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB3_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2:
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB3_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, 0
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: ds_add_u32 v1, v0
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB3_4:
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: add_i32_varying_nouse:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: ds_add_u32 v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: add_i32_varying_nouse:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
@@ -2427,357 +2086,97 @@ entry:
}
define amdgpu_kernel void @add_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: add_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB6_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_add_rtn_u64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB6_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX7LESS_ITERATIVE-NEXT: v_add_i32_e32 v0, vcc, s5, v1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_addc_u32_e32 v1, vcc, v3, v2, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: add_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: add_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_add_rtn_u64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB6_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX8_ITERATIVE-NEXT: v_add_u32_e32 v0, vcc, s5, v1
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_addc_u32_e32 v1, vcc, v3, v2, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: add_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_add_rtn_u64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB6_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX9_ITERATIVE-NEXT: v_add_co_u32_e32 v0, vcc, s5, v1
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_addc_co_u32_e32 v1, vcc, v3, v2, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: add_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v0, s6
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v3, s6
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s6
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_add_u32 s0, s0, s7
-; GFX1064_ITERATIVE-NEXT: s_addc_u32 s1, s1, s8
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_add_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB6_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_add_co_u32 v0, vcc, s2, v1
-; GFX1064_ITERATIVE-NEXT: v_add_co_ci_u32_e32 v1, vcc, s3, v2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: add_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_add_u32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_add_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB6_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_add_co_u32 v0, vcc_lo, s2, v1
-; GFX1032_ITERATIVE-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, s3, v2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: add_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s6, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v2, s6
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v3, s6
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s6
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_add_u32 s0, s0, s7
-; GFX1164_ITERATIVE-NEXT: s_addc_u32 s1, s1, s8
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_add_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB6_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_add_co_u32 v0, vcc, s2, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1164_ITERATIVE-NEXT: v_add_co_ci_u32_e64 v1, null, s3, v1, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
+; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: add_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .LBB6_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_add_u32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB6_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB6_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_add_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB6_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_add_co_u32 v0, vcc_lo, s2, v0
-; GFX1132_ITERATIVE-NEXT: v_add_co_ci_u32_e64 v1, null, s3, v1, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: add_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_add_rtn_u64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: add_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -3483,225 +2882,62 @@ entry:
}
define amdgpu_kernel void @add_i64_varying_nouse() {
-; GFX7LESS_ITERATIVE-LABEL: add_i64_varying_nouse:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v1, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v1, exec_hi, v1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX7LESS_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s4, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s5, v0, s4
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s4
-; GFX7LESS_ITERATIVE-NEXT: s_add_u32 s0, s0, s5
-; GFX7LESS_ITERATIVE-NEXT: s_bitset0_b64 s[2:3], s4
-; GFX7LESS_ITERATIVE-NEXT: s_addc_u32 s1, s1, s6
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 s[4:5], s[2:3], s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2:
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB7_4:
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: add_i64_varying_nouse:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: ds_add_u64 v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX8_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s4, s[2:3]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s5, v0, s4
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s4
-; GFX8_ITERATIVE-NEXT: s_add_u32 s0, s0, s5
-; GFX8_ITERATIVE-NEXT: s_bitset0_b64 s[2:3], s4
-; GFX8_ITERATIVE-NEXT: s_addc_u32 s1, s1, s6
-; GFX8_ITERATIVE-NEXT: s_cmp_lg_u64 s[2:3], 0
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2:
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX8_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB7_4:
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX9_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s4, s[2:3]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s5, v0, s4
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s4
-; GFX9_ITERATIVE-NEXT: s_add_u32 s0, s0, s5
-; GFX9_ITERATIVE-NEXT: s_bitset0_b64 s[2:3], s4
-; GFX9_ITERATIVE-NEXT: s_addc_u32 s1, s1, s6
-; GFX9_ITERATIVE-NEXT: s_cmp_lg_u64 s[2:3], 0
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2:
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX9_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX9_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB7_4:
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1064_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s4, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s5, v0, s4
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s4
-; GFX1064_ITERATIVE-NEXT: s_bitset0_b64 s[2:3], s4
-; GFX1064_ITERATIVE-NEXT: s_add_u32 s0, s0, s5
-; GFX1064_ITERATIVE-NEXT: s_addc_u32 s1, s1, s6
-; GFX1064_ITERATIVE-NEXT: s_cmp_lg_u64 s[2:3], 0
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2:
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1064_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB7_4:
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s5, v2, s3
-; GFX1032_ITERATIVE-NEXT: s_bitset0_b32 s2, s3
-; GFX1032_ITERATIVE-NEXT: s_add_u32 s0, s0, s4
-; GFX1032_ITERATIVE-NEXT: s_addc_u32 s1, s1, s5
-; GFX1032_ITERATIVE-NEXT: s_cmp_lg_u32 s2, 0
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2:
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1032_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1032_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB7_4:
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1164_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s4, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s5, v1, s4
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s4
-; GFX1164_ITERATIVE-NEXT: s_bitset0_b64 s[2:3], s4
-; GFX1164_ITERATIVE-NEXT: s_add_u32 s0, s0, s5
-; GFX1164_ITERATIVE-NEXT: s_addc_u32 s1, s1, s6
-; GFX1164_ITERATIVE-NEXT: s_cmp_lg_u64 s[2:3], 0
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2:
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v0
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1164_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
+; GFX1164_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB7_4:
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: add_i64_varying_nouse:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s5, v2, s3
-; GFX1132_ITERATIVE-NEXT: s_bitset0_b32 s2, s3
-; GFX1132_ITERATIVE-NEXT: s_add_u32 s0, s0, s4
-; GFX1132_ITERATIVE-NEXT: s_addc_u32 s1, s1, s5
-; GFX1132_ITERATIVE-NEXT: s_cmp_lg_u32 s2, 0
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB7_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2:
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB7_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, 0
-; GFX1132_ITERATIVE-NEXT: ds_add_u64 v2, v[0:1]
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
+; GFX1132_ITERATIVE-NEXT: ds_add_u64 v1, v[0:1]
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB7_4:
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: add_i64_varying_nouse:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: ds_add_u64 v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: add_i64_varying_nouse:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v5, 0
@@ -4817,302 +4053,97 @@ entry:
}
define amdgpu_kernel void @sub_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: sub_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB10_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB10_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_sub_i32_e32 v0, vcc, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: sub_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: sub_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_add_i32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB10_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_sub_u32_e32 v0, vcc, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: sub_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_add_i32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB10_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_sub_u32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: sub_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB10_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_sub_nc_u32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: sub_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB10_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_sub_nc_u32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: sub_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_add_i32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_sub_rtn_u32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB10_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_sub_nc_u32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: sub_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB10_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB10_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB10_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_sub_rtn_u32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB10_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_sub_nc_u32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_sub_rtn_u32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: sub_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_sub_rtn_u32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: sub_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v3, 0
@@ -5524,198 +4555,62 @@ entry:
}
define amdgpu_kernel void @sub_i32_varying_nouse() {
-; GFX7LESS_ITERATIVE-LABEL: sub_i32_varying_nouse:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v1, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v1, exec_hi, v1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX7LESS_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 s[4:5], s[0:1], s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2:
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_sub_u32 v1, v0
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB11_4:
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: sub_i32_varying_nouse:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: ds_sub_u32 v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX8_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX8_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX8_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2:
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
; GFX8_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB11_4:
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX9_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX9_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX9_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2:
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB11_4:
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1064_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1064_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX1064_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX1064_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2:
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB11_4:
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: s_bitset0_b32 s1, s2
-; GFX1032_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_cmp_lg_u32 s1, 0
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2:
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, s0
; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB11_4:
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s4, v0, s3
-; GFX1164_ITERATIVE-NEXT: s_bitset0_b64 s[0:1], s3
-; GFX1164_ITERATIVE-NEXT: s_add_i32 s2, s2, s4
-; GFX1164_ITERATIVE-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2:
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v1
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v0, s2
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1164_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB11_4:
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: sub_i32_varying_nouse:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: .LBB11_1: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1132_ITERATIVE-NEXT: s_bitset0_b32 s1, s2
-; GFX1132_ITERATIVE-NEXT: s_add_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_cmp_lg_u32 s1, 0
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB11_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2:
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: v_cmpx_eq_u32_e32 0, v1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB11_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, 0
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: ds_sub_u32 v1, v0
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB11_4:
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: sub_i32_varying_nouse:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: ds_sub_u32 v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: sub_i32_varying_nouse:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
@@ -6704,357 +5599,97 @@ entry:
}
define amdgpu_kernel void @sub_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: sub_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB14_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_sub_rtn_u64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB14_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX7LESS_ITERATIVE-NEXT: v_sub_i32_e32 v0, vcc, s5, v1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_subb_u32_e32 v1, vcc, v3, v2, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: sub_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: sub_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_sub_rtn_u64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB14_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX8_ITERATIVE-NEXT: v_sub_u32_e32 v0, vcc, s5, v1
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_subb_u32_e32 v1, vcc, v3, v2, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: sub_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s6
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s6
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s6
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_add_u32 s0, s0, s8
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_sub_rtn_u64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB14_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s4
-; GFX9_ITERATIVE-NEXT: v_sub_co_u32_e32 v0, vcc, s5, v1
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v2, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: sub_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s6, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v0, s6
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v3, s6
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s6
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_add_u32 s0, s0, s7
-; GFX1064_ITERATIVE-NEXT: s_addc_u32 s1, s1, s8
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_sub_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB14_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_sub_co_u32 v0, vcc, s2, v1
-; GFX1064_ITERATIVE-NEXT: v_sub_co_ci_u32_e32 v1, vcc, s3, v2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: sub_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_add_u32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_sub_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB14_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_sub_co_u32 v0, vcc_lo, s2, v1
-; GFX1032_ITERATIVE-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, s3, v2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: sub_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s6, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v2, s6
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v3, s6
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s6
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_add_u32 s0, s0, s7
-; GFX1164_ITERATIVE-NEXT: s_addc_u32 s1, s1, s8
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s6
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_sub_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB14_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_sub_co_u32 v0, vcc, s2, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1164_ITERATIVE-NEXT: v_sub_co_ci_u32_e64 v1, null, s3, v1, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
+; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: sub_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .LBB14_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_add_u32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_addc_u32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB14_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB14_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_sub_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB14_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_sub_co_u32 v0, vcc_lo, s2, v0
-; GFX1132_ITERATIVE-NEXT: v_sub_co_ci_u32_e64 v1, null, s3, v1, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: sub_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_sub_rtn_u64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: sub_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -7760,302 +6395,97 @@ entry:
}
define amdgpu_kernel void @and_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: and_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB15_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB15_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_and_b32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: and_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: and_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_and_b32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB15_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_and_b32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: and_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_and_b32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB15_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_and_b32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: and_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_and_b32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB15_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_and_b32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: and_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_and_b32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB15_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_and_b32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: and_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_and_b32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_and_rtn_b32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB15_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: and_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB15_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_and_b32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB15_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB15_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_and_rtn_b32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB15_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_and_rtn_b32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: and_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_and_rtn_b32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: and_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v3, exec_lo, 0
@@ -8495,345 +6925,97 @@ entry:
}
define amdgpu_kernel void @and_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: and_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[8:9], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB16_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_and_rtn_b64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB16_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_and_b32_e32 v2, s4, v2
-; GFX7LESS_ITERATIVE-NEXT: v_and_b32_e32 v1, s5, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: and_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: and_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_and_rtn_b64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB16_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_and_b32_e32 v2, s4, v2
-; GFX8_ITERATIVE-NEXT: v_and_b32_e32 v1, s5, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: and_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_and_rtn_b64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB16_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_and_b32_e32 v2, s4, v2
-; GFX9_ITERATIVE-NEXT: v_and_b32_e32 v1, s5, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: and_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s8
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1064_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_and_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB16_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1064_ITERATIVE-NEXT: v_and_b32_e32 v2, s2, v2
-; GFX1064_ITERATIVE-NEXT: v_and_b32_e32 v1, s3, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: and_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_and_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB16_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1032_ITERATIVE-NEXT: v_and_b32_e32 v2, s2, v2
-; GFX1032_ITERATIVE-NEXT: v_and_b32_e32 v1, s3, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: and_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s8, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s8
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1164_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[8:9]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_and_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB16_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, s2, v1
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, s3, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: and_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .LBB16_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_and_b64 s[0:1], s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB16_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB16_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_and_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB16_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, s2, v1
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v0, s3, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: and_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_and_rtn_b64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: and_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -9398,302 +7580,97 @@ entry:
}
define amdgpu_kernel void @or_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: or_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_or_b32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB17_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB17_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_or_b32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: or_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: or_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_or_b32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB17_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_or_b32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: or_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_or_b32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB17_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_or_b32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: or_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_or_b32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB17_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_or_b32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: or_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_or_b32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB17_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_or_b32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: or_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_or_b32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_or_rtn_b32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB17_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_or_b32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: or_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB17_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_or_b32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB17_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB17_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_or_rtn_b32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB17_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_or_b32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_or_rtn_b32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: or_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_or_rtn_b32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: or_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v3, 0
@@ -10105,345 +8082,97 @@ entry:
}
define amdgpu_kernel void @or_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: or_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[8:9], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB18_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_or_rtn_b64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB18_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_or_b32_e32 v2, s4, v2
-; GFX7LESS_ITERATIVE-NEXT: v_or_b32_e32 v1, s5, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: or_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: or_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_or_rtn_b64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB18_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_or_b32_e32 v2, s4, v2
-; GFX8_ITERATIVE-NEXT: v_or_b32_e32 v1, s5, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: or_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_or_rtn_b64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB18_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_or_b32_e32 v2, s4, v2
-; GFX9_ITERATIVE-NEXT: v_or_b32_e32 v1, s5, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: or_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s8
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1064_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_or_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB18_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1064_ITERATIVE-NEXT: v_or_b32_e32 v2, s2, v2
-; GFX1064_ITERATIVE-NEXT: v_or_b32_e32 v1, s3, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: or_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_or_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB18_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1032_ITERATIVE-NEXT: v_or_b32_e32 v2, s2, v2
-; GFX1032_ITERATIVE-NEXT: v_or_b32_e32 v1, s3, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: or_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s8, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s8
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1164_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[8:9]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_or_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB18_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_or_b32_e32 v1, s2, v1
-; GFX1164_ITERATIVE-NEXT: v_or_b32_e32 v0, s3, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: or_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .LBB18_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_or_b64 s[0:1], s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB18_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB18_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_or_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB18_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132_ITERATIVE-NEXT: v_or_b32_e32 v1, s2, v1
-; GFX1132_ITERATIVE-NEXT: v_or_b32_e32 v0, s3, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: or_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_or_rtn_b64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: or_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -11008,302 +8737,97 @@ entry:
}
define amdgpu_kernel void @xor_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: xor_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB19_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB19_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_xor_b32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: xor_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: xor_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_xor_b32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB19_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_xor_b32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: xor_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_xor_b32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB19_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_xor_b32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: xor_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_xor_b32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB19_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_xor_b32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: xor_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB19_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_xor_b32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: xor_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_xor_b32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_xor_rtn_b32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB19_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_xor_b32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: xor_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB19_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB19_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB19_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_xor_rtn_b32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB19_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_xor_b32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_xor_rtn_b32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: xor_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_xor_rtn_b32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: xor_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v3, 0
@@ -11715,345 +9239,97 @@ entry:
}
define amdgpu_kernel void @xor_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: xor_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[8:9], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[8:9]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB20_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_xor_rtn_b64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB20_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_xor_b32_e32 v2, s4, v2
-; GFX7LESS_ITERATIVE-NEXT: v_xor_b32_e32 v1, s5, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: xor_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: xor_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_xor_rtn_b64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB20_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_xor_b32_e32 v2, s4, v2
-; GFX8_ITERATIVE-NEXT: v_xor_b32_e32 v1, s5, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: xor_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_xor_rtn_b64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB20_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v3
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_xor_b32_e32 v2, s4, v2
-; GFX9_ITERATIVE-NEXT: v_xor_b32_e32 v1, s5, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: xor_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s8
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s8
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[8:9]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_xor_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB20_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1064_ITERATIVE-NEXT: v_xor_b32_e32 v2, s2, v2
-; GFX1064_ITERATIVE-NEXT: v_xor_b32_e32 v1, s3, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: xor_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_xor_rtn_b64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB20_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1032_ITERATIVE-NEXT: v_xor_b32_e32 v2, s2, v2
-; GFX1032_ITERATIVE-NEXT: v_xor_b32_e32 v1, s3, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: xor_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s8, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s8
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s8
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s8
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[8:9], 1, s8
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[8:9]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_xor_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB20_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164_ITERATIVE-NEXT: v_xor_b32_e32 v1, s2, v1
-; GFX1164_ITERATIVE-NEXT: v_xor_b32_e32 v0, s3, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: xor_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .LBB20_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_xor_b64 s[0:1], s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB20_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB20_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_xor_rtn_b64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB20_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132_ITERATIVE-NEXT: v_xor_b32_e32 v1, s2, v1
-; GFX1132_ITERATIVE-NEXT: v_xor_b32_e32 v0, s3, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: xor_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_xor_rtn_b64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: xor_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -12618,302 +9894,97 @@ entry:
}
define amdgpu_kernel void @max_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: max_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_brev_b32 s2, 1
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_max_i32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB21_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB21_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_max_i32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: max_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: max_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_brev_b32 s2, 1
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_max_i32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB21_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_max_i32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: max_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_brev_b32 s2, 1
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_max_i32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB21_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_max_i32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: max_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_brev_b32 s2, 1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_max_i32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB21_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_max_i32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: max_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_brev_b32 s0, 1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_max_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB21_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_max_i32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: max_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_brev_b32 s2, 1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_max_i32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_max_rtn_i32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB21_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_max_i32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: max_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_brev_b32 s0, 1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB21_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_max_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB21_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB21_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_max_rtn_i32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB21_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_max_i32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_max_rtn_i32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: max_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_max_rtn_i32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: max_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v3, exec_lo, 0
@@ -13626,395 +10697,97 @@ entry:
}
define amdgpu_kernel void @max_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: max_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB23_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_max_rtn_i64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB23_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: max_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: max_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX8_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_max_rtn_i64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB23_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX8_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: max_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX9_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_max_rtn_i64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB23_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX9_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: max_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s10, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s10
-; GFX1064_ITERATIVE-NEXT: v_cmp_gt_i64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_max_rtn_i64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB23_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[2:3], v[1:2]
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: max_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: v_cmp_gt_i64_e64 s8, s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_max_rtn_i64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB23_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[1:2]
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc_lo
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: max_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .p2align 6
-; GFX1164_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s10, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s10
-; GFX1164_ITERATIVE-NEXT: v_cmp_gt_i64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_max_rtn_i64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB23_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1]
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: max_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_brev_b32 s1, 1
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .p2align 6
-; GFX1132_ITERATIVE-NEXT: .LBB23_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: v_cmp_gt_i64_e64 s8, s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB23_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB23_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_max_rtn_i64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB23_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[0:1]
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: max_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_max_rtn_i64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: max_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v7, exec_lo, 0
@@ -14823,302 +11596,97 @@ entry:
}
define amdgpu_kernel void @min_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: min_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_brev_b32 s2, -2
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_min_i32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB24_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB24_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_min_i32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: min_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: min_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_brev_b32 s2, -2
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_min_i32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB24_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_min_i32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: min_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_brev_b32 s2, -2
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_min_i32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB24_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_min_i32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: min_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_brev_b32 s2, -2
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_min_i32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB24_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_min_i32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: min_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_brev_b32 s0, -2
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_min_i32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB24_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_min_i32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: min_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_brev_b32 s2, -2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_min_i32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_min_rtn_i32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB24_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_min_i32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: min_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_brev_b32 s0, -2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB24_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_min_i32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB24_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB24_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_min_rtn_i32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB24_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_min_i32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_min_rtn_i32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: min_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_min_rtn_i32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: min_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v3, exec_lo, 0
@@ -15831,395 +12399,97 @@ entry:
}
define amdgpu_kernel void @min_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: min_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB26_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_min_rtn_i64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB26_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: min_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: min_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX8_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_min_rtn_i64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB26_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX8_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: min_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX9_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[4:5]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_min_rtn_i64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB26_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX9_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[1:2]
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: min_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s10, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s10
-; GFX1064_ITERATIVE-NEXT: v_cmp_lt_i64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_min_rtn_i64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB26_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[1:2]
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: min_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: v_cmp_lt_i64_e64 s8, s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_min_rtn_i64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB26_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc_lo, s[2:3], v[1:2]
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc_lo
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: min_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .p2align 6
-; GFX1164_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s10, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s10
-; GFX1164_ITERATIVE-NEXT: v_cmp_lt_i64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_min_rtn_i64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB26_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[0:1]
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: min_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_brev_b32 s1, -2
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .p2align 6
-; GFX1132_ITERATIVE-NEXT: .LBB26_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: v_cmp_lt_i64_e64 s8, s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB26_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB26_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_min_rtn_i64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB26_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_lt_i64_e32 vcc_lo, s[2:3], v[0:1]
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: min_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_min_rtn_i64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: min_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v7, exec_lo, 0
@@ -17028,302 +13298,97 @@ entry:
}
define amdgpu_kernel void @umax_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: umax_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_max_u32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB27_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB27_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_max_u32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: umax_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: umax_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_max_u32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB27_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_max_u32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: umax_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_max_u32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB27_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_max_u32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: umax_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_max_u32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB27_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_max_u32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: umax_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_max_u32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB27_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_max_u32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: umax_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_max_u32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_max_rtn_u32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB27_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_max_u32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: umax_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB27_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_max_u32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB27_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB27_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_max_rtn_u32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB27_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_max_u32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_max_rtn_u32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: umax_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_max_rtn_u32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: umax_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mov_b32_e32 v3, 0
@@ -18031,388 +14096,97 @@ entry:
}
define amdgpu_kernel void @umax_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: umax_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB29_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_max_rtn_u64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB29_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: umax_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: umax_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX8_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_max_rtn_u64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB29_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX8_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: umax_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX9_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_max_rtn_u64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB29_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX9_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: umax_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s10, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s10
-; GFX1064_ITERATIVE-NEXT: v_cmp_gt_u64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_max_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB29_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[1:2]
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: umax_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: v_cmp_gt_u64_e64 s8, s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_max_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB29_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[1:2]
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc_lo
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: umax_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .p2align 6
-; GFX1164_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s10, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s10
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_cmp_gt_u64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_max_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB29_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1]
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: umax_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], 0
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .p2align 6
-; GFX1132_ITERATIVE-NEXT: .LBB29_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_gt_u64_e64 s8, s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB29_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB29_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_max_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB29_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[0:1]
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: umax_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_max_rtn_u64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: umax_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v5, exec_lo, 0
@@ -19215,302 +14989,97 @@ entry:
}
define amdgpu_kernel void @umin_i32_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: umin_i32_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX7LESS_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[0:1], 0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX7LESS_ITERATIVE-NEXT: s_min_u32 s2, s2, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB30_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v0, v2
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB30_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_min_u32_e32 v0, s4, v1
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: umin_i32_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: umin_i32_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX8_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX8_ITERATIVE-NEXT: s_min_u32 s2, s2, s6
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v0, v2
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB30_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_min_u32_e32 v0, s4, v1
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: umin_i32_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX9_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s3
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s2, m0
-; GFX9_ITERATIVE-NEXT: s_min_u32 s2, s2, s6
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX9_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v0, v2
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB30_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_min_u32_e32 v0, s4, v1
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: umin_i32_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1064_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s8, v0, s3
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s2, s3
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1064_ITERATIVE-NEXT: s_min_u32 s2, s2, s8
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1064_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v0, v2
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB30_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1064_ITERATIVE-NEXT: v_min_u32_e32 v0, s2, v1
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1064_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: umin_i32_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1032_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s2, s1
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s3, v0, s2
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s2
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1032_ITERATIVE-NEXT: s_min_u32 s0, s0, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s1, s1, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1032_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v0, v2
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB30_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v0
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1032_ITERATIVE-NEXT: v_min_u32_e32 v0, s2, v1
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1032_ITERATIVE-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: umin_i32_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1164_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s3, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s8, v1, s3
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s2, s3
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX1164_ITERATIVE-NEXT: s_min_u32 s2, s2, s8
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[0:1], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v1, exec_hi, v1
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s2
-; GFX1164_ITERATIVE-NEXT: ds_min_rtn_u32 v1, v1, v2
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB30_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[0:1]
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1164_ITERATIVE-NEXT: v_min_u32_e32 v0, s2, v0
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: umin_i32_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_and_b32_e32 v1, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s1, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s0, -1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0
-; GFX1132_ITERATIVE-NEXT: .LBB30_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s2, s1
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s3, v1, s2
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s2
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s2, 1, s2
-; GFX1132_ITERATIVE-NEXT: s_min_u32 s0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s1, s1, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB30_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr1
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB30_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_min_rtn_u32 v1, v1, v2
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB30_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s1
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v1
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
-; GFX1132_ITERATIVE-NEXT: v_min_u32_e32 v0, s2, v0
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_min_rtn_u32 v0, v1, v0
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: umin_i32_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_min_rtn_u32 v0, v1, v0
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dword v0, off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: umin_i32_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v3, exec_lo, 0
@@ -20219,388 +15788,97 @@ entry:
}
define amdgpu_kernel void @umin_i64_varying(ptr addrspace(1) %out) {
-; GFX7LESS_ITERATIVE-LABEL: umin_i64_varying:
-; GFX7LESS_ITERATIVE: ; %bb.0: ; %entry
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX7LESS_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX7LESS_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX7LESS_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX7LESS_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX7LESS_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX7LESS_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX7LESS_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_ne_u64_e64 s[6:7], s[2:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_and_b64 vcc, exec, s[6:7]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_vccnz .LBB32_1
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX7LESS_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX7LESS_ITERATIVE-NEXT: ; %bb.3:
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_ITERATIVE-NEXT: ds_min_rtn_u64 v[3:4], v0, v[3:4]
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: .LBB32_4:
-; GFX7LESS_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX7LESS_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX7LESS_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX7LESS_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
-; GFX7LESS_ITERATIVE-NEXT: s_endpgm
+; GFX7LESS-LABEL: umin_i64_varying:
+; GFX7LESS: ; %bb.0: ; %entry
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0
+; GFX7LESS-NEXT: s_mov_b32 m0, -1
+; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT: s_mov_b32 s2, -1
+; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX7LESS-NEXT: s_endpgm
;
; GFX8_ITERATIVE-LABEL: umin_i64_varying:
; GFX8_ITERATIVE: ; %bb.0: ; %entry
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX8_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX8_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX8_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX8_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX8_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX8_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX8_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX8_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX8_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX8_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX8_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX8_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX8_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX8_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX8_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX8_ITERATIVE-NEXT: ; %bb.3:
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
+; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX8_ITERATIVE-NEXT: s_mov_b32 m0, -1
-; GFX8_ITERATIVE-NEXT: ds_min_rtn_u64 v[3:4], v0, v[3:4]
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: .LBB32_4:
-; GFX8_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX8_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX8_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX8_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX8_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX8_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX8_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX8_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX8_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX8_ITERATIVE-NEXT: s_endpgm
;
; GFX9_ITERATIVE-LABEL: umin_i64_varying:
; GFX9_ITERATIVE: ; %bb.0: ; %entry
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX9_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX9_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX9_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9_ITERATIVE-NEXT: s_ff1_i32_b64 s8, s[2:3]
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s9, v3, s8
-; GFX9_ITERATIVE-NEXT: v_readlane_b32 s10, v0, s8
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s10
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v5, s9
-; GFX9_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[4:5]
-; GFX9_ITERATIVE-NEXT: s_mov_b32 m0, s8
-; GFX9_ITERATIVE-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v2, s1, m0
-; GFX9_ITERATIVE-NEXT: v_writelane_b32 v1, s0, m0
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s9
-; GFX9_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s10
-; GFX9_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s8
-; GFX9_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX9_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX9_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX9_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX9_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX9_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX9_ITERATIVE-NEXT: ; %bb.3:
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX9_ITERATIVE-NEXT: ds_min_rtn_u64 v[3:4], v0, v[3:4]
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: .LBB32_4:
-; GFX9_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX9_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s5, v4
-; GFX9_ITERATIVE-NEXT: v_readfirstlane_b32 s4, v3
-; GFX9_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[1:2]
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s5
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
-; GFX9_ITERATIVE-NEXT: v_mov_b32_e32 v0, s4
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
; GFX9_ITERATIVE-NEXT: s_mov_b32 s3, 0xf000
; GFX9_ITERATIVE-NEXT: s_mov_b32 s2, -1
-; GFX9_ITERATIVE-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc
-; GFX9_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX9_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX9_ITERATIVE-NEXT: s_endpgm
;
; GFX1064_ITERATIVE-LABEL: umin_i64_varying:
; GFX1064_ITERATIVE: ; %bb.0: ; %entry
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1064_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1064_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX1064_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064_ITERATIVE-NEXT: s_ff1_i32_b64 s10, s[2:3]
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1064_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s10
-; GFX1064_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s10
-; GFX1064_ITERATIVE-NEXT: v_cmp_lt_u64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1064_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1064_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1064_ITERATIVE-NEXT: s_andn2_b64 s[2:3], s[2:3], s[6:7]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX1064_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1064_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1064_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX1064_ITERATIVE-NEXT: ; %bb.3:
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1064_ITERATIVE-NEXT: ds_min_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1064_ITERATIVE-NEXT: .LBB32_4:
-; GFX1064_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1064_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1064_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1064_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1064_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[1:2]
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc
-; GFX1064_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1064_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1064_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1064_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1064_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1064_ITERATIVE-NEXT: s_endpgm
;
; GFX1032_ITERATIVE-LABEL: umin_i64_varying:
; GFX1032_ITERATIVE: ; %bb.0: ; %entry
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr1_vgpr2
-; GFX1032_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX1032_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032_ITERATIVE-NEXT: s_ff1_i32_b32 s3, s2
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1032_ITERATIVE-NEXT: v_readlane_b32 s6, v0, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v2, s1, s3
-; GFX1032_ITERATIVE-NEXT: v_writelane_b32 v1, s0, s3
-; GFX1032_ITERATIVE-NEXT: v_cmp_lt_u64_e64 s8, s[0:1], s[6:7]
-; GFX1032_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1032_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1032_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1032_ITERATIVE-NEXT: s_andn2_b32 s2, s2, s3
-; GFX1032_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX1032_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1032_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032_ITERATIVE-NEXT: ; implicit-def: $vgpr3_vgpr4
-; GFX1032_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1032_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1032_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX1032_ITERATIVE-NEXT: ; %bb.3:
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v0, 0
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v4, s1
-; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v3, s0
-; GFX1032_ITERATIVE-NEXT: ds_min_rtn_u64 v[3:4], v0, v[3:4]
-; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1032_ITERATIVE-NEXT: .LBB32_4:
-; GFX1032_ITERATIVE-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX1032_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1032_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
; GFX1032_ITERATIVE-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v4
-; GFX1032_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v3
-; GFX1032_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[2:3], v[1:2]
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v2, v2, s3, vcc_lo
-; GFX1032_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[1:2], off, s[0:3], 0
+; GFX1032_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1032_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1032_ITERATIVE-NEXT: buffer_gl0_inv
+; GFX1032_ITERATIVE-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; GFX1032_ITERATIVE-NEXT: s_endpgm
;
; GFX1164_ITERATIVE-LABEL: umin_i64_varying:
; GFX1164_ITERATIVE: ; %bb.0: ; %entry
-; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v2, 0x3ff, v0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, 0
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[2:3], exec
-; GFX1164_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1164_ITERATIVE-NEXT: .p2align 6
-; GFX1164_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX1164_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164_ITERATIVE-NEXT: s_ctz_i32_b64 s10, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s10
-; GFX1164_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s10
-; GFX1164_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s10
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: v_cmp_lt_u64_e64 s[8:9], s[0:1], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_and_b64 s[8:9], s[8:9], exec
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1164_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1164_ITERATIVE-NEXT: s_lshl_b64 s[6:7], 1, s10
-; GFX1164_ITERATIVE-NEXT: s_and_not1_b64 s[2:3], s[2:3], s[6:7]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX1164_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164_ITERATIVE-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX1164_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX1164_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1164_ITERATIVE-NEXT: s_and_saveexec_b64 s[2:3], vcc
-; GFX1164_ITERATIVE-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164_ITERATIVE-NEXT: s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX1164_ITERATIVE-NEXT: ; %bb.3:
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v4, 0
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v3, s1
-; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1164_ITERATIVE-NEXT: ds_min_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1164_ITERATIVE-NEXT: .LBB32_4:
-; GFX1164_ITERATIVE-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1164_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
+; GFX1164_ITERATIVE-NEXT: v_mov_b32_e32 v1, 0
+; GFX1164_ITERATIVE-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX1164_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1164_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1]
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc
-; GFX1164_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1164_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1164_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1164_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1164_ITERATIVE-NEXT: s_endpgm
;
; GFX1132_ITERATIVE-LABEL: umin_i64_varying:
; GFX1132_ITERATIVE: ; %bb.0: ; %entry
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v2, 0x3ff, v0
-; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_mov_b64 s[0:1], -1
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr0_vgpr1
-; GFX1132_ITERATIVE-NEXT: .p2align 6
-; GFX1132_ITERATIVE-NEXT: .LBB32_1: ; %ComputeLoop
-; GFX1132_ITERATIVE-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132_ITERATIVE-NEXT: s_ctz_i32_b32 s3, s2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s7, v3, s3
-; GFX1132_ITERATIVE-NEXT: v_readlane_b32 s6, v2, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v1, s1, s3
-; GFX1132_ITERATIVE-NEXT: v_writelane_b32 v0, s0, s3
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_lt_u64_e64 s8, s[0:1], s[6:7]
-; GFX1132_ITERATIVE-NEXT: s_and_b32 s8, s8, exec_lo
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s1, s1, s7
-; GFX1132_ITERATIVE-NEXT: s_cselect_b32 s0, s0, s6
-; GFX1132_ITERATIVE-NEXT: s_lshl_b32 s3, 1, s3
-; GFX1132_ITERATIVE-NEXT: s_and_not1_b32 s2, s2, s3
-; GFX1132_ITERATIVE-NEXT: s_cbranch_scc1 .LBB32_1
-; GFX1132_ITERATIVE-NEXT: ; %bb.2: ; %ComputeEnd
-; GFX1132_ITERATIVE-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX1132_ITERATIVE-NEXT: ; implicit-def: $vgpr2_vgpr3
-; GFX1132_ITERATIVE-NEXT: s_and_saveexec_b32 s2, vcc_lo
-; GFX1132_ITERATIVE-NEXT: s_xor_b32 s2, exec_lo, s2
-; GFX1132_ITERATIVE-NEXT: s_cbranch_execz .LBB32_4
-; GFX1132_ITERATIVE-NEXT: ; %bb.3:
-; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s1
-; GFX1132_ITERATIVE-NEXT: v_mov_b32_e32 v2, s0
-; GFX1132_ITERATIVE-NEXT: ds_min_rtn_u64 v[2:3], v4, v[2:3]
-; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
-; GFX1132_ITERATIVE-NEXT: .LBB32_4:
-; GFX1132_ITERATIVE-NEXT: s_or_b32 exec_lo, exec_lo, s2
+; GFX1132_ITERATIVE-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX1132_ITERATIVE-NEXT: s_load_b64 s[0:1], s[4:5], 0x24
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s3, v3
-; GFX1132_ITERATIVE-NEXT: v_readfirstlane_b32 s2, v2
-; GFX1132_ITERATIVE-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132_ITERATIVE-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[2:3], v[0:1]
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
-; GFX1132_ITERATIVE-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132_ITERATIVE-NEXT: s_mov_b32 s2, -1
; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
+; GFX1132_ITERATIVE-NEXT: s_waitcnt lgkmcnt(0)
+; GFX1132_ITERATIVE-NEXT: buffer_gl0_inv
; GFX1132_ITERATIVE-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
; GFX1132_ITERATIVE-NEXT: s_endpgm
;
-; GFX7LESS_DPP-LABEL: umin_i64_varying:
-; GFX7LESS_DPP: ; %bb.0: ; %entry
-; GFX7LESS_DPP-NEXT: v_mov_b32_e32 v1, 0
-; GFX7LESS_DPP-NEXT: s_mov_b32 m0, -1
-; GFX7LESS_DPP-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: ds_min_rtn_u64 v[0:1], v1, v[0:1]
-; GFX7LESS_DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS_DPP-NEXT: s_mov_b32 s3, 0xf000
-; GFX7LESS_DPP-NEXT: s_mov_b32 s2, -1
-; GFX7LESS_DPP-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
-; GFX7LESS_DPP-NEXT: s_endpgm
-;
; GFX8_DPP-LABEL: umin_i64_varying:
; GFX8_DPP: ; %bb.0: ; %entry
; GFX8_DPP-NEXT: v_mbcnt_lo_u32_b32 v6, exec_lo, 0
@@ -21402,3 +16680,5 @@ entry:
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX1332_DPP: {{.*}}
; GFX1364_DPP: {{.*}}
+; GFX7LESS_DPP: {{.*}}
+; GFX7LESS_ITERATIVE: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/local-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/local-atomicrmw-fadd.ll
index e5e07ff1a94eb1..f87274ae4bf77b 100644
--- a/llvm/test/CodeGen/AMDGPU/local-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/local-atomicrmw-fadd.ll
@@ -7747,51 +7747,17 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX12-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX12-NEXT: s_mov_b32 s1, exec_lo
-; GFX12-NEXT: s_brev_b32 s0, 1
+; GFX12-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
+; GFX12-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mul_f32 v0, 0x42280000, v0
; GFX12-NEXT: v_add_f32_e32 v0, s3, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_cndmask_b32_e64 v1, v0, s3, vcc_lo
-; GFX12-NEXT: ; implicit-def: $vgpr0
-; GFX12-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_ctz_i32_b32 s3, s1
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_readlane_b32 s6, v1, s3
-; GFX12-NEXT: v_writelane_b32 v0, s0, s3
-; GFX12-NEXT: s_lshl_b32 s3, 1, s3
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_and_not1_b32 s1, s1, s3
-; GFX12-NEXT: s_add_f32 s0, s0, s6
-; GFX12-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX12-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX12-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX12-NEXT: ; implicit-def: $vgpr1
-; GFX12-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX12-NEXT: s_cbranch_execz .LBB28_8
-; GFX12-NEXT: ; %bb.7:
-; GFX12-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s0
+; GFX12-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: ds_add_rtn_f32 v1, v1, v2
+; GFX12-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX12-NEXT: s_wait_dscnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_SE
-; GFX12-NEXT: .LBB28_8:
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
-; GFX12-NEXT: v_readfirstlane_b32 s2, v1
-; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_add_f32 v0, s2, v0
-; GFX12-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
+; GFX12-NEXT: v_mov_b32_e32 v1, 0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: global_store_b32 v1, v0, s[0:1]
; GFX12-NEXT: s_endpgm
@@ -7843,43 +7809,13 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX942-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX942-NEXT: v_add_f32_e32 v0, s8, v0
; GFX942-NEXT: v_mov_b32_e32 v1, s8
-; GFX942-NEXT: s_mov_b64 s[0:1], exec
-; GFX942-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX942-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX942-NEXT: ; implicit-def: $vgpr0
-; GFX942-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX942-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX942-NEXT: v_readfirstlane_b32 s6, v1
-; GFX942-NEXT: s_mov_b32 m0, s3
-; GFX942-NEXT: v_readlane_b32 s8, v2, s3
-; GFX942-NEXT: v_writelane_b32 v0, s6, m0
-; GFX942-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX942-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX942-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX942-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX942-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX942-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX942-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX942-NEXT: ; implicit-def: $vgpr2
-; GFX942-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX942-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX942-NEXT: s_cbranch_execz .LBB28_8
-; GFX942-NEXT: ; %bb.7:
-; GFX942-NEXT: v_mov_b32_e32 v2, s2
-; GFX942-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX942-NEXT: s_waitcnt lgkmcnt(0)
-; GFX942-NEXT: .LBB28_8:
-; GFX942-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX942-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX942-NEXT: v_mov_b32_e32 v1, s2
; GFX942-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX942-NEXT: v_readfirstlane_b32 s2, v2
-; GFX942-NEXT: v_mov_b32_e32 v1, 0
-; GFX942-NEXT: s_nop 0
-; GFX942-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX942-NEXT: v_mov_b32_e32 v2, s2
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX942-NEXT: s_waitcnt lgkmcnt(0)
+; GFX942-NEXT: ds_add_rtn_f32 v0, v1, v0
+; GFX942-NEXT: s_waitcnt lgkmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v1, 0
; GFX942-NEXT: global_store_dword v1, v0, s[0:1]
; GFX942-NEXT: s_endpgm
;
@@ -7932,47 +7868,17 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX11-NEXT: .LBB28_4:
; GFX11-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX11-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX11-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX11-NEXT: s_mov_b32 s0, exec_lo
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
+; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mul_f32 v0, 0x42280000, v0
; GFX11-NEXT: v_add_f32_e32 v0, s3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_cndmask_b32_e64 v2, v0, s3, vcc_lo
-; GFX11-NEXT: ; implicit-def: $vgpr0
-; GFX11-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX11-NEXT: s_ctz_i32_b32 s1, s0
-; GFX11-NEXT: v_readfirstlane_b32 s3, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_readlane_b32 s6, v2, s1
-; GFX11-NEXT: v_writelane_b32 v0, s3, s1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: v_add_f32_e32 v1, s6, v1
-; GFX11-NEXT: s_lshl_b32 s1, 1, s1
-; GFX11-NEXT: s_and_not1_b32 s0, s0, s1
-; GFX11-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX11-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX11-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX11-NEXT: ; implicit-def: $vgpr2
-; GFX11-NEXT: s_and_saveexec_b32 s0, vcc_lo
-; GFX11-NEXT: s_xor_b32 s0, exec_lo, s0
-; GFX11-NEXT: s_cbranch_execz .LBB28_8
-; GFX11-NEXT: ; %bb.7:
-; GFX11-NEXT: v_mov_b32_e32 v2, s2
-; GFX11-NEXT: ds_add_rtn_f32 v2, v2, v1
+; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: .LBB28_8:
-; GFX11-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
-; GFX11-NEXT: v_readfirstlane_b32 s2, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_add_f32 v0, s2, v0
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
+; GFX11-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: buffer_gl0_inv
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
; GFX11-NEXT: global_store_b32 v1, v0, s[0:1]
; GFX11-NEXT: s_endpgm
;
@@ -8023,44 +7929,17 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX10-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX10-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX10-NEXT: s_mov_b32 s0, exec_lo
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX10-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX10-NEXT: v_add_f32_e32 v0, s3, v0
-; GFX10-NEXT: v_cndmask_b32_e64 v2, v0, s3, vcc_lo
-; GFX10-NEXT: ; implicit-def: $vgpr0
-; GFX10-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX10-NEXT: s_ff1_i32_b32 s1, s0
-; GFX10-NEXT: v_readfirstlane_b32 s3, v1
-; GFX10-NEXT: v_readlane_b32 s6, v2, s1
-; GFX10-NEXT: v_writelane_b32 v0, s3, s1
-; GFX10-NEXT: v_add_f32_e32 v1, s6, v1
-; GFX10-NEXT: s_lshl_b32 s1, 1, s1
-; GFX10-NEXT: s_andn2_b32 s0, s0, s1
-; GFX10-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX10-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX10-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX10-NEXT: ; implicit-def: $vgpr2
-; GFX10-NEXT: s_and_saveexec_b32 s0, vcc_lo
-; GFX10-NEXT: s_xor_b32 s0, exec_lo, s0
-; GFX10-NEXT: s_cbranch_execz .LBB28_8
-; GFX10-NEXT: ; %bb.7:
-; GFX10-NEXT: v_mov_b32_e32 v2, s2
+; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: ds_add_rtn_f32 v2, v2, v1
+; GFX10-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: .LBB28_8:
-; GFX10-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX10-NEXT: v_readfirstlane_b32 s2, v2
; GFX10-NEXT: v_mov_b32_e32 v1, 0
-; GFX10-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: global_store_dword v1, v0, s[0:1]
; GFX10-NEXT: s_endpgm
;
@@ -8111,42 +7990,13 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX90A-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX90A-NEXT: v_add_f32_e32 v0, s8, v0
; GFX90A-NEXT: v_mov_b32_e32 v1, s8
-; GFX90A-NEXT: s_mov_b64 s[0:1], exec
-; GFX90A-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX90A-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX90A-NEXT: ; implicit-def: $vgpr0
-; GFX90A-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX90A-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX90A-NEXT: v_readfirstlane_b32 s6, v1
-; GFX90A-NEXT: s_mov_b32 m0, s3
-; GFX90A-NEXT: v_readlane_b32 s8, v2, s3
-; GFX90A-NEXT: v_writelane_b32 v0, s6, m0
-; GFX90A-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX90A-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX90A-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX90A-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX90A-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX90A-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX90A-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX90A-NEXT: ; implicit-def: $vgpr2
-; GFX90A-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX90A-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX90A-NEXT: s_cbranch_execz .LBB28_8
-; GFX90A-NEXT: ; %bb.7:
-; GFX90A-NEXT: v_mov_b32_e32 v2, s2
-; GFX90A-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX90A-NEXT: s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT: .LBB28_8:
-; GFX90A-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX90A-NEXT: v_mov_b32_e32 v1, s2
; GFX90A-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX90A-NEXT: v_readfirstlane_b32 s2, v2
-; GFX90A-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX90A-NEXT: v_mov_b32_e32 v2, s2
-; GFX90A-NEXT: v_mov_b32_e32 v1, 0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX90A-NEXT: s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT: ds_add_rtn_f32 v0, v1, v0
+; GFX90A-NEXT: s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v1, 0
; GFX90A-NEXT: global_store_dword v1, v0, s[0:1]
; GFX90A-NEXT: s_endpgm
;
@@ -8197,42 +8047,13 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX908-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX908-NEXT: v_add_f32_e32 v0, s8, v0
; GFX908-NEXT: v_mov_b32_e32 v1, s8
-; GFX908-NEXT: s_mov_b64 s[0:1], exec
-; GFX908-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX908-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX908-NEXT: ; implicit-def: $vgpr0
-; GFX908-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX908-NEXT: v_readfirstlane_b32 s6, v1
-; GFX908-NEXT: s_mov_b32 m0, s3
-; GFX908-NEXT: v_readlane_b32 s8, v2, s3
-; GFX908-NEXT: v_writelane_b32 v0, s6, m0
-; GFX908-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX908-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX908-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX908-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX908-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX908-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX908-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX908-NEXT: ; implicit-def: $vgpr2
-; GFX908-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX908-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX908-NEXT: s_cbranch_execz .LBB28_8
-; GFX908-NEXT: ; %bb.7:
-; GFX908-NEXT: v_mov_b32_e32 v2, s2
-; GFX908-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX908-NEXT: s_waitcnt lgkmcnt(0)
-; GFX908-NEXT: .LBB28_8:
-; GFX908-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX908-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX908-NEXT: v_mov_b32_e32 v1, s2
; GFX908-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX908-NEXT: v_readfirstlane_b32 s2, v2
-; GFX908-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX908-NEXT: v_mov_b32_e32 v2, s2
-; GFX908-NEXT: v_mov_b32_e32 v1, 0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX908-NEXT: s_waitcnt lgkmcnt(0)
+; GFX908-NEXT: ds_add_rtn_f32 v0, v1, v0
+; GFX908-NEXT: s_waitcnt lgkmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v1, 0
; GFX908-NEXT: global_store_dword v1, v0, s[0:1]
; GFX908-NEXT: s_endpgm
;
@@ -8249,8 +8070,8 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: s_add_i32 s3, s3, 4
; GFX8-NEXT: v_readfirstlane_b32 s6, v1
-; GFX8-NEXT: ; implicit-def: $vgpr1
; GFX8-NEXT: s_mov_b32 m0, -1
+; GFX8-NEXT: ; implicit-def: $vgpr1
; GFX8-NEXT: s_and_saveexec_b64 s[0:1], vcc
; GFX8-NEXT: s_cbranch_execz .LBB28_2
; GFX8-NEXT: ; %bb.1:
@@ -8284,41 +8105,11 @@ define amdgpu_kernel void @local_ds_fadd(ptr addrspace(1) %out, ptr addrspace(3)
; GFX8-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX8-NEXT: v_add_f32_e32 v0, s8, v0
; GFX8-NEXT: v_mov_b32_e32 v1, s8
-; GFX8-NEXT: s_mov_b64 s[0:1], exec
-; GFX8-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX8-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX8-NEXT: ; implicit-def: $vgpr0
-; GFX8-NEXT: .LBB28_5: ; %ComputeLoop
-; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8-NEXT: v_readfirstlane_b32 s6, v1
-; GFX8-NEXT: s_mov_b32 m0, s3
-; GFX8-NEXT: v_readlane_b32 s8, v2, s3
-; GFX8-NEXT: v_writelane_b32 v0, s6, m0
-; GFX8-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX8-NEXT: s_cbranch_scc1 .LBB28_5
-; GFX8-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX8-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX8-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX8-NEXT: ; implicit-def: $vgpr2
-; GFX8-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8-NEXT: s_cbranch_execz .LBB28_8
-; GFX8-NEXT: ; %bb.7:
-; GFX8-NEXT: v_mov_b32_e32 v2, s2
-; GFX8-NEXT: s_mov_b32 m0, -1
-; GFX8-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: .LBB28_8:
-; GFX8-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX8-NEXT: v_readfirstlane_b32 s2, v2
-; GFX8-NEXT: v_add_f32_e32 v0, s2, v0
; GFX8-NEXT: v_mov_b32_e32 v1, s2
-; GFX8-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
+; GFX8-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8-NEXT: ds_add_rtn_f32 v2, v1, v0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, s0
; GFX8-NEXT: v_mov_b32_e32 v1, s1
@@ -8637,48 +8428,15 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX12-NEXT: ; %bb.4:
; GFX12-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX12-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX12-NEXT: s_mov_b32 s1, exec_lo
-; GFX12-NEXT: s_brev_b32 s0, 1
+; GFX12-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
+; GFX12-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mul_f32 v0, 0x42280000, v0
; GFX12-NEXT: v_add_f32_e32 v0, s3, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_cndmask_b32_e64 v1, v0, s3, vcc_lo
-; GFX12-NEXT: ; implicit-def: $vgpr0
-; GFX12-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_ctz_i32_b32 s3, s1
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_readlane_b32 s6, v1, s3
-; GFX12-NEXT: v_writelane_b32 v0, s0, s3
-; GFX12-NEXT: s_lshl_b32 s3, 1, s3
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_and_not1_b32 s1, s1, s3
-; GFX12-NEXT: s_add_f32 s0, s0, s6
-; GFX12-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX12-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX12-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX12-NEXT: ; implicit-def: $vgpr1
-; GFX12-NEXT: s_and_saveexec_b32 s1, vcc_lo
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_xor_b32 s1, exec_lo, s1
-; GFX12-NEXT: ; %bb.7:
-; GFX12-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s0
-; GFX12-NEXT: ds_add_rtn_f32 v1, v1, v2
-; GFX12-NEXT: ; %bb.8:
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
+; GFX12-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
+; GFX12-NEXT: ds_add_rtn_f32 v0, v1, v0
+; GFX12-NEXT: v_mov_b32_e32 v1, 0
; GFX12-NEXT: s_wait_dscnt 0x0
-; GFX12-NEXT: v_readfirstlane_b32 s2, v1
-; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_add_f32 v0, s2, v0
-; GFX12-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: global_store_b32 v1, v0, s[0:1]
; GFX12-NEXT: s_endpgm
@@ -8727,41 +8485,12 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX942-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX942-NEXT: v_add_f32_e32 v0, s8, v0
; GFX942-NEXT: v_mov_b32_e32 v1, s8
-; GFX942-NEXT: s_mov_b64 s[0:1], exec
-; GFX942-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX942-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX942-NEXT: ; implicit-def: $vgpr0
-; GFX942-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX942-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX942-NEXT: v_readfirstlane_b32 s6, v1
-; GFX942-NEXT: s_mov_b32 m0, s3
-; GFX942-NEXT: v_readlane_b32 s8, v2, s3
-; GFX942-NEXT: v_writelane_b32 v0, s6, m0
-; GFX942-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX942-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX942-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX942-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX942-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX942-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX942-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX942-NEXT: ; implicit-def: $vgpr2
-; GFX942-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX942-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX942-NEXT: ; %bb.7:
-; GFX942-NEXT: v_mov_b32_e32 v2, s2
-; GFX942-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX942-NEXT: ; %bb.8:
-; GFX942-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX942-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX942-NEXT: v_mov_b32_e32 v1, s2
; GFX942-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX942-NEXT: s_waitcnt lgkmcnt(0)
-; GFX942-NEXT: v_readfirstlane_b32 s2, v2
+; GFX942-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX942-NEXT: v_mov_b32_e32 v1, 0
-; GFX942-NEXT: s_nop 0
-; GFX942-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX942-NEXT: v_mov_b32_e32 v2, s2
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX942-NEXT: s_waitcnt lgkmcnt(0)
; GFX942-NEXT: global_store_dword v1, v0, s[0:1]
; GFX942-NEXT: s_endpgm
;
@@ -8809,44 +8538,15 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX11-NEXT: ; %bb.4:
; GFX11-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX11-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX11-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX11-NEXT: s_mov_b32 s0, exec_lo
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
+; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mul_f32 v0, 0x42280000, v0
; GFX11-NEXT: v_add_f32_e32 v0, s3, v0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_cndmask_b32_e64 v2, v0, s3, vcc_lo
-; GFX11-NEXT: ; implicit-def: $vgpr0
-; GFX11-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX11-NEXT: s_ctz_i32_b32 s1, s0
-; GFX11-NEXT: v_readfirstlane_b32 s3, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_readlane_b32 s6, v2, s1
-; GFX11-NEXT: v_writelane_b32 v0, s3, s1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: v_add_f32_e32 v1, s6, v1
-; GFX11-NEXT: s_lshl_b32 s1, 1, s1
-; GFX11-NEXT: s_and_not1_b32 s0, s0, s1
-; GFX11-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX11-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX11-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX11-NEXT: ; implicit-def: $vgpr2
-; GFX11-NEXT: s_and_saveexec_b32 s0, vcc_lo
-; GFX11-NEXT: s_xor_b32 s0, exec_lo, s0
-; GFX11-NEXT: ; %bb.7:
-; GFX11-NEXT: v_mov_b32_e32 v2, s2
-; GFX11-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX11-NEXT: ; %bb.8:
-; GFX11-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x0
+; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
+; GFX11-NEXT: ds_add_rtn_f32 v0, v1, v0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: v_readfirstlane_b32 s2, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_add_f32 v0, s2, v0
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
; GFX11-NEXT: global_store_b32 v1, v0, s[0:1]
; GFX11-NEXT: s_endpgm
;
@@ -8891,40 +8591,14 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX10-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s6
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
-; GFX10-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX10-NEXT: s_mov_b32 s0, exec_lo
+; GFX10-NEXT: v_mov_b32_e32 v1, s2
+; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX10-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX10-NEXT: v_add_f32_e32 v0, s3, v0
-; GFX10-NEXT: v_cndmask_b32_e64 v2, v0, s3, vcc_lo
-; GFX10-NEXT: ; implicit-def: $vgpr0
-; GFX10-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX10-NEXT: s_ff1_i32_b32 s1, s0
-; GFX10-NEXT: v_readfirstlane_b32 s3, v1
-; GFX10-NEXT: v_readlane_b32 s6, v2, s1
-; GFX10-NEXT: v_writelane_b32 v0, s3, s1
-; GFX10-NEXT: v_add_f32_e32 v1, s6, v1
-; GFX10-NEXT: s_lshl_b32 s1, 1, s1
-; GFX10-NEXT: s_andn2_b32 s0, s0, s1
-; GFX10-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX10-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX10-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX10-NEXT: ; implicit-def: $vgpr2
-; GFX10-NEXT: s_and_saveexec_b32 s0, vcc_lo
-; GFX10-NEXT: s_xor_b32 s0, exec_lo, s0
-; GFX10-NEXT: ; %bb.7:
-; GFX10-NEXT: v_mov_b32_e32 v2, s2
-; GFX10-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX10-NEXT: ; %bb.8:
-; GFX10-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
-; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX10-NEXT: s_waitcnt lgkmcnt(0)
-; GFX10-NEXT: v_readfirstlane_b32 s2, v2
+; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s3, vcc_lo
+; GFX10-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX10-NEXT: v_mov_b32_e32 v1, 0
-; GFX10-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: global_store_dword v1, v0, s[0:1]
; GFX10-NEXT: s_endpgm
;
@@ -8972,40 +8646,12 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX90A-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX90A-NEXT: v_add_f32_e32 v0, s8, v0
; GFX90A-NEXT: v_mov_b32_e32 v1, s8
-; GFX90A-NEXT: s_mov_b64 s[0:1], exec
-; GFX90A-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX90A-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX90A-NEXT: ; implicit-def: $vgpr0
-; GFX90A-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX90A-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX90A-NEXT: v_readfirstlane_b32 s6, v1
-; GFX90A-NEXT: s_mov_b32 m0, s3
-; GFX90A-NEXT: v_readlane_b32 s8, v2, s3
-; GFX90A-NEXT: v_writelane_b32 v0, s6, m0
-; GFX90A-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX90A-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX90A-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX90A-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX90A-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX90A-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX90A-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX90A-NEXT: ; implicit-def: $vgpr2
-; GFX90A-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX90A-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX90A-NEXT: ; %bb.7:
-; GFX90A-NEXT: v_mov_b32_e32 v2, s2
-; GFX90A-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX90A-NEXT: ; %bb.8:
-; GFX90A-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX90A-NEXT: v_mov_b32_e32 v1, s2
; GFX90A-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX90A-NEXT: s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT: v_readfirstlane_b32 s2, v2
-; GFX90A-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX90A-NEXT: v_mov_b32_e32 v2, s2
+; GFX90A-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX90A-NEXT: s_waitcnt lgkmcnt(0)
; GFX90A-NEXT: global_store_dword v1, v0, s[0:1]
; GFX90A-NEXT: s_endpgm
;
@@ -9053,40 +8699,12 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX908-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX908-NEXT: v_add_f32_e32 v0, s8, v0
; GFX908-NEXT: v_mov_b32_e32 v1, s8
-; GFX908-NEXT: s_mov_b64 s[0:1], exec
-; GFX908-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX908-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX908-NEXT: ; implicit-def: $vgpr0
-; GFX908-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX908-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX908-NEXT: v_readfirstlane_b32 s6, v1
-; GFX908-NEXT: s_mov_b32 m0, s3
-; GFX908-NEXT: v_readlane_b32 s8, v2, s3
-; GFX908-NEXT: v_writelane_b32 v0, s6, m0
-; GFX908-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX908-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX908-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX908-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX908-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX908-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX908-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX908-NEXT: ; implicit-def: $vgpr2
-; GFX908-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX908-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX908-NEXT: ; %bb.7:
-; GFX908-NEXT: v_mov_b32_e32 v2, s2
-; GFX908-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX908-NEXT: ; %bb.8:
-; GFX908-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX908-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
+; GFX908-NEXT: v_mov_b32_e32 v1, s2
; GFX908-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX908-NEXT: s_waitcnt lgkmcnt(0)
-; GFX908-NEXT: v_readfirstlane_b32 s2, v2
-; GFX908-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX908-NEXT: v_mov_b32_e32 v2, s2
+; GFX908-NEXT: ds_add_rtn_f32 v0, v1, v0
; GFX908-NEXT: v_mov_b32_e32 v1, 0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX908-NEXT: s_waitcnt lgkmcnt(0)
; GFX908-NEXT: global_store_dword v1, v0, s[0:1]
; GFX908-NEXT: s_endpgm
;
@@ -9103,8 +8721,8 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: s_add_i32 s3, s3, 4
; GFX8-NEXT: v_readfirstlane_b32 s6, v1
-; GFX8-NEXT: ; implicit-def: $vgpr1
; GFX8-NEXT: s_mov_b32 m0, -1
+; GFX8-NEXT: ; implicit-def: $vgpr1
; GFX8-NEXT: s_and_saveexec_b64 s[0:1], vcc
; GFX8-NEXT: ; %bb.1:
; GFX8-NEXT: s_lshl_b32 s7, s3, 3
@@ -9135,40 +8753,11 @@ define amdgpu_kernel void @local_ds_fadd_one_as(ptr addrspace(1) %out, ptr addrs
; GFX8-NEXT: v_mul_f32_e32 v0, 0x42280000, v0
; GFX8-NEXT: v_add_f32_e32 v0, s8, v0
; GFX8-NEXT: v_mov_b32_e32 v1, s8
-; GFX8-NEXT: s_mov_b64 s[0:1], exec
-; GFX8-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
-; GFX8-NEXT: v_bfrev_b32_e32 v1, 1
-; GFX8-NEXT: ; implicit-def: $vgpr0
-; GFX8-NEXT: .LBB29_5: ; %ComputeLoop
-; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: s_ff1_i32_b64 s3, s[0:1]
-; GFX8-NEXT: v_readfirstlane_b32 s6, v1
-; GFX8-NEXT: s_mov_b32 m0, s3
-; GFX8-NEXT: v_readlane_b32 s8, v2, s3
-; GFX8-NEXT: v_writelane_b32 v0, s6, m0
-; GFX8-NEXT: s_lshl_b64 s[6:7], 1, s3
-; GFX8-NEXT: s_andn2_b64 s[0:1], s[0:1], s[6:7]
-; GFX8-NEXT: v_add_f32_e32 v1, s8, v1
-; GFX8-NEXT: s_cbranch_scc1 .LBB29_5
-; GFX8-NEXT: ; %bb.6: ; %ComputeEnd
-; GFX8-NEXT: v_mbcnt_lo_u32_b32 v2, exec_lo, 0
-; GFX8-NEXT: v_mbcnt_hi_u32_b32 v2, exec_hi, v2
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX8-NEXT: ; implicit-def: $vgpr2
-; GFX8-NEXT: s_and_saveexec_b64 s[0:1], vcc
-; GFX8-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
-; GFX8-NEXT: ; %bb.7:
-; GFX8-NEXT: v_mov_b32_e32 v2, s2
-; GFX8-NEXT: s_mov_b32 m0, -1
-; GFX8-NEXT: ds_add_rtn_f32 v2, v2, v1
-; GFX8-NEXT: ; %bb.8:
-; GFX8-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_readfirstlane_b32 s2, v2
-; GFX8-NEXT: v_add_f32_e32 v0, s2, v0
; GFX8-NEXT: v_mov_b32_e32 v1, s2
-; GFX8-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc
+; GFX8-NEXT: ds_add_rtn_f32 v2, v1, v0
+; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v0, s0
; GFX8-NEXT: v_mov_b32_e32 v1, s1
; GFX8-NEXT: flat_store_dword v[0:1], v2
More information about the llvm-commits
mailing list