caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
ec056f54589d937ae3f74d795b94de886cbb7ec7
clang-p2996/llvm/test/CodeGen/AMDGPU/llvm.mulo.ll
Matt Arsenault d246cc618a PeepholeOpt: Do not add subregister indexes to reg_sequence operands (#124111) 
Given the rest of the pass just gives up when it needs to compose
subregisters, folding a subregister extract directly into a reg_sequence
is counterproductive. Later fold attempts in the function will give up
on the subregister operand, preventing looking up through the reg_sequence.

It may still be profitable to do these folds if we start handling
the composes. There are some test regressions, but this mostly
looks better.
2025-01-30 20:42:02 +07:00

859 lines
36 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefix=SI %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX11 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX12 %s
define { i64, i1 } @umulo_i64_v_v(i64 %x, i64 %y) {
; SI-LABEL: umulo_i64_v_v:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_mul_hi_u32 v4, v1, v2
; SI-NEXT: v_mul_lo_u32 v5, v1, v2
; SI-NEXT: v_mul_hi_u32 v6, v0, v3
; SI-NEXT: v_mul_lo_u32 v7, v0, v3
; SI-NEXT: v_mul_hi_u32 v8, v0, v2
; SI-NEXT: v_mul_hi_u32 v9, v1, v3
; SI-NEXT: v_mul_lo_u32 v3, v1, v3
; SI-NEXT: v_mul_lo_u32 v0, v0, v2
; SI-NEXT: v_add_i32_e32 v1, vcc, v8, v7
; SI-NEXT: v_addc_u32_e32 v2, vcc, 0, v6, vcc
; SI-NEXT: v_add_i32_e32 v6, vcc, v1, v5
; SI-NEXT: v_add_i32_e64 v1, s[4:5], v1, v5
; SI-NEXT: v_addc_u32_e32 v2, vcc, v2, v4, vcc
; SI-NEXT: v_addc_u32_e32 v4, vcc, 0, v9, vcc
; SI-NEXT: v_add_i32_e32 v2, vcc, v2, v3
; SI-NEXT: v_addc_u32_e32 v3, vcc, 0, v4, vcc
; SI-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[2:3]
; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; SI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: umulo_i64_v_v:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v4, v1
; GFX9-NEXT: v_mad_u64_u32 v[5:6], s[4:5], v0, v3, 0
; GFX9-NEXT: v_mad_u64_u32 v[0:1], s[4:5], v0, v2, 0
; GFX9-NEXT: v_mad_u64_u32 v[7:8], s[4:5], v4, v2, 0
; GFX9-NEXT: v_add_co_u32_e32 v9, vcc, v1, v5
; GFX9-NEXT: v_mad_u64_u32 v[2:3], s[4:5], v4, v3, 0
; GFX9-NEXT: v_addc_co_u32_e32 v6, vcc, 0, v6, vcc
; GFX9-NEXT: v_add_co_u32_e32 v4, vcc, v9, v7
; GFX9-NEXT: v_addc_co_u32_e32 v4, vcc, v6, v8, vcc
; GFX9-NEXT: v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT: v_add_co_u32_e32 v2, vcc, v4, v2
; GFX9-NEXT: v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[2:3]
; GFX9-NEXT: v_add3_u32 v1, v1, v5, v7
; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: umulo_i64_v_v:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v4, v1
; GFX10-NEXT: v_mad_u64_u32 v[5:6], s4, v0, v3, 0
; GFX10-NEXT: v_mad_u64_u32 v[0:1], s4, v0, v2, 0
; GFX10-NEXT: v_mad_u64_u32 v[7:8], s4, v4, v2, 0
; GFX10-NEXT: v_mad_u64_u32 v[2:3], s4, v4, v3, 0
; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v1, v5
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, 0, v6, vcc_lo
; GFX10-NEXT: v_add3_u32 v1, v1, v5, v7
; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v4, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v4, vcc_lo, v6, v8, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v4, v2
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: umulo_i64_v_v:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_dual_mov_b32 v4, v0 :: v_dual_mov_b32 v5, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_mad_u64_u32 v[6:7], null, v4, v3, 0
; GFX11-NEXT: v_mad_u64_u32 v[0:1], null, v4, v2, 0
; GFX11-NEXT: v_mad_u64_u32 v[8:9], null, v5, v2, 0
; GFX11-NEXT: v_mad_u64_u32 v[10:11], null, v5, v3, 0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, v1, v6
; GFX11-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v7, vcc_lo
; GFX11-NEXT: v_add3_u32 v1, v1, v6, v8
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, v2, v8
; GFX11-NEXT: v_add_co_ci_u32_e32 v2, vcc_lo, v3, v9, vcc_lo
; GFX11-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v11, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, v2, v10
; GFX11-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_cmp_ne_u64_e32 vcc_lo, 0, v[2:3]
; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-LABEL: umulo_i64_v_v:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT: s_wait_expcnt 0x0
; GFX12-NEXT: s_wait_samplecnt 0x0
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v4, v1
; GFX12-NEXT: v_mad_co_u64_u32 v[5:6], null, v0, v3, 0
; GFX12-NEXT: v_mad_co_u64_u32 v[0:1], null, v0, v2, 0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_mad_co_u64_u32 v[7:8], null, v4, v2, 0
; GFX12-NEXT: v_mad_co_u64_u32 v[2:3], null, v4, v3, 0
; GFX12-NEXT: v_add_co_u32 v4, vcc_lo, v1, v5
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, 0, v6, vcc_lo
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_add3_u32 v1, v1, v5, v7
; GFX12-NEXT: v_add_co_u32 v4, vcc_lo, v4, v7
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v4, vcc_lo, v6, v8, vcc_lo
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX12-NEXT: v_add_co_u32 v2, vcc_lo, v4, v2
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX12-NEXT: v_cmp_ne_u64_e32 vcc_lo, 0, v[2:3]
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX12-NEXT: s_setpc_b64 s[30:31]
bb:
%umulo = tail call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %x, i64 %y)
ret { i64, i1 } %umulo
}
define { i64, i1 } @smulo_i64_v_v(i64 %x, i64 %y) {
; SI-LABEL: smulo_i64_v_v:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_mul_hi_u32 v6, v1, v2
; SI-NEXT: v_mul_lo_u32 v5, v1, v2
; SI-NEXT: v_mul_hi_u32 v7, v0, v3
; SI-NEXT: v_mul_lo_u32 v8, v0, v3
; SI-NEXT: v_mul_hi_u32 v9, v0, v2
; SI-NEXT: v_mul_hi_i32 v10, v1, v3
; SI-NEXT: v_mul_lo_u32 v11, v1, v3
; SI-NEXT: v_mul_lo_u32 v4, v0, v2
; SI-NEXT: v_add_i32_e32 v8, vcc, v9, v8
; SI-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
; SI-NEXT: v_add_i32_e32 v9, vcc, v8, v5
; SI-NEXT: v_add_i32_e64 v5, s[4:5], v8, v5
; SI-NEXT: v_addc_u32_e32 v8, vcc, v7, v6, vcc
; SI-NEXT: v_ashrrev_i32_e32 v6, 31, v5
; SI-NEXT: v_addc_u32_e32 v9, vcc, 0, v10, vcc
; SI-NEXT: v_mov_b32_e32 v7, v6
; SI-NEXT: v_add_i32_e32 v8, vcc, v8, v11
; SI-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc
; SI-NEXT: v_sub_i32_e32 v2, vcc, v8, v2
; SI-NEXT: v_subbrev_u32_e32 v10, vcc, 0, v9, vcc
; SI-NEXT: v_cmp_gt_i32_e32 vcc, 0, v1
; SI-NEXT: v_cndmask_b32_e32 v1, v9, v10, vcc
; SI-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc
; SI-NEXT: v_sub_i32_e32 v0, vcc, v2, v0
; SI-NEXT: v_subbrev_u32_e32 v8, vcc, 0, v1, vcc
; SI-NEXT: v_cmp_gt_i32_e32 vcc, 0, v3
; SI-NEXT: v_cndmask_b32_e32 v1, v1, v8, vcc
; SI-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc
; SI-NEXT: v_cmp_ne_u64_e32 vcc, v[0:1], v[6:7]
; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; SI-NEXT: v_mov_b32_e32 v0, v4
; SI-NEXT: v_mov_b32_e32 v1, v5
; SI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: smulo_i64_v_v:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v5, v0
; GFX9-NEXT: v_mov_b32_e32 v4, v1
; GFX9-NEXT: v_mad_u64_u32 v[6:7], s[4:5], v5, v3, 0
; GFX9-NEXT: v_mad_u64_u32 v[0:1], s[4:5], v5, v2, 0
; GFX9-NEXT: v_mad_u64_u32 v[8:9], s[4:5], v4, v2, 0
; GFX9-NEXT: v_add_co_u32_e32 v12, vcc, v1, v6
; GFX9-NEXT: v_mad_i64_i32 v[10:11], s[4:5], v4, v3, 0
; GFX9-NEXT: v_addc_co_u32_e32 v7, vcc, 0, v7, vcc
; GFX9-NEXT: v_add_co_u32_e32 v12, vcc, v12, v8
; GFX9-NEXT: v_addc_co_u32_e32 v7, vcc, v7, v9, vcc
; GFX9-NEXT: v_addc_co_u32_e32 v9, vcc, 0, v11, vcc
; GFX9-NEXT: v_add_co_u32_e32 v7, vcc, v7, v10
; GFX9-NEXT: v_addc_co_u32_e32 v9, vcc, 0, v9, vcc
; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v7, v2
; GFX9-NEXT: v_subbrev_co_u32_e32 v10, vcc, 0, v9, vcc
; GFX9-NEXT: v_cmp_gt_i32_e32 vcc, 0, v4
; GFX9-NEXT: v_cndmask_b32_e32 v2, v7, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v4, v9, v10, vcc
; GFX9-NEXT: v_sub_co_u32_e32 v5, vcc, v2, v5
; GFX9-NEXT: v_subbrev_co_u32_e32 v7, vcc, 0, v4, vcc
; GFX9-NEXT: v_cmp_gt_i32_e32 vcc, 0, v3
; GFX9-NEXT: v_add3_u32 v1, v1, v6, v8
; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v7, vcc
; GFX9-NEXT: v_ashrrev_i32_e32 v4, 31, v1
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v5, vcc
; GFX9-NEXT: v_mov_b32_e32 v5, v4
; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, v[2:3], v[4:5]
; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: smulo_i64_v_v:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v4, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: v_mad_u64_u32 v[6:7], s4, v4, v3, 0
; GFX10-NEXT: v_mad_u64_u32 v[0:1], s4, v4, v2, 0
; GFX10-NEXT: v_mad_u64_u32 v[8:9], s4, v5, v2, 0
; GFX10-NEXT: v_mad_i64_i32 v[10:11], s4, v5, v3, 0
; GFX10-NEXT: v_add_co_u32 v12, vcc_lo, v1, v6
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo
; GFX10-NEXT: v_add3_u32 v1, v1, v6, v8
; GFX10-NEXT: v_add_co_u32 v12, vcc_lo, v12, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, v7, v9, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v11, vcc_lo
; GFX10-NEXT: v_add_co_u32 v7, vcc_lo, v7, v10
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v9, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v2, vcc_lo, v7, v2
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v10, vcc_lo, 0, v9, vcc_lo
; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v5
; GFX10-NEXT: v_cndmask_b32_e32 v6, v7, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, v9, v10, vcc_lo
; GFX10-NEXT: v_ashrrev_i32_e32 v2, 31, v1
; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v6, v4
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v7, vcc_lo, 0, v5, vcc_lo
; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v3
; GFX10-NEXT: v_mov_b32_e32 v3, v2
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v4, v6, v4, vcc_lo
; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[4:5], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: smulo_i64_v_v:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_dual_mov_b32 v4, v0 :: v_dual_mov_b32 v5, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_mad_u64_u32 v[6:7], null, v4, v3, 0
; GFX11-NEXT: v_mad_u64_u32 v[0:1], null, v4, v2, 0
; GFX11-NEXT: v_mad_u64_u32 v[8:9], null, v5, v2, 0
; GFX11-NEXT: v_mad_i64_i32 v[10:11], null, v5, v3, 0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-NEXT: v_add_co_u32 v12, vcc_lo, v1, v6
; GFX11-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo
; GFX11-NEXT: v_add3_u32 v1, v1, v6, v8
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_add_co_u32 v12, vcc_lo, v12, v8
; GFX11-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, v7, v9, vcc_lo
; GFX11-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v11, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_add_co_u32 v7, vcc_lo, v7, v10
; GFX11-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v9, vcc_lo
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v7, v2
; GFX11-NEXT: v_subrev_co_ci_u32_e32 v10, vcc_lo, 0, v9, vcc_lo
; GFX11-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v5
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_cndmask_b32_e32 v6, v7, v2, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e32 v5, v9, v10, vcc_lo
; GFX11-NEXT: v_ashrrev_i32_e32 v2, 31, v1
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_sub_co_u32 v4, vcc_lo, v6, v4
; GFX11-NEXT: v_subrev_co_ci_u32_e32 v7, vcc_lo, 0, v5, vcc_lo
; GFX11-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v3
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-NEXT: v_mov_b32_e32 v3, v2
; GFX11-NEXT: v_dual_cndmask_b32 v4, v6, v4 :: v_dual_cndmask_b32 v5, v5, v7
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[4:5], v[2:3]
; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-LABEL: smulo_i64_v_v:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT: s_wait_expcnt 0x0
; GFX12-NEXT: s_wait_samplecnt 0x0
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_dual_mov_b32 v4, v0 :: v_dual_mov_b32 v5, v1
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_mad_co_u64_u32 v[6:7], null, v4, v3, 0
; GFX12-NEXT: v_mad_co_u64_u32 v[0:1], null, v4, v2, 0
; GFX12-NEXT: v_mad_co_u64_u32 v[8:9], null, v5, v2, 0
; GFX12-NEXT: v_mad_co_i64_i32 v[10:11], null, v5, v3, 0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
; GFX12-NEXT: v_add_co_u32 v12, vcc_lo, v1, v6
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo
; GFX12-NEXT: v_add3_u32 v1, v1, v6, v8
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_add_co_u32 v12, vcc_lo, v12, v8
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, v7, v9, vcc_lo
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v11, vcc_lo
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v7, vcc_lo, v7, v10
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0, v9, vcc_lo
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_sub_co_u32 v2, vcc_lo, v7, v2
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_subrev_co_ci_u32_e32 v10, vcc_lo, 0, v9, vcc_lo
; GFX12-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v5
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_cndmask_b32_e32 v6, v7, v2, vcc_lo
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_cndmask_b32_e32 v5, v9, v10, vcc_lo
; GFX12-NEXT: v_ashrrev_i32_e32 v2, 31, v1
; GFX12-NEXT: v_sub_co_u32 v4, vcc_lo, v6, v4
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_3)
; GFX12-NEXT: v_subrev_co_ci_u32_e32 v7, vcc_lo, 0, v5, vcc_lo
; GFX12-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v3
; GFX12-NEXT: v_mov_b32_e32 v3, v2
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_dual_cndmask_b32 v4, v6, v4 :: v_dual_cndmask_b32 v5, v5, v7
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[4:5], v[2:3]
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX12-NEXT: s_setpc_b64 s[30:31]
bb:
%smulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %x, i64 %y)
ret { i64, i1 } %smulo
}
define amdgpu_kernel void @umulo_i64_s(i64 %x, i64 %y) {
; SI-LABEL: umulo_i64_s:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: v_mov_b32_e32 v0, s2
; SI-NEXT: v_mul_hi_u32 v1, s1, v0
; SI-NEXT: s_mul_i32 s4, s1, s2
; SI-NEXT: v_mov_b32_e32 v2, s3
; SI-NEXT: v_mul_hi_u32 v3, s0, v2
; SI-NEXT: s_mul_i32 s5, s0, s3
; SI-NEXT: v_mul_hi_u32 v0, s0, v0
; SI-NEXT: v_mul_hi_u32 v2, s1, v2
; SI-NEXT: s_mul_i32 s1, s1, s3
; SI-NEXT: s_mul_i32 s2, s0, s2
; SI-NEXT: v_add_i32_e32 v4, vcc, s5, v0
; SI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
; SI-NEXT: v_add_i32_e32 v4, vcc, s4, v4
; SI-NEXT: v_addc_u32_e32 v1, vcc, v3, v1, vcc
; SI-NEXT: v_addc_u32_e32 v2, vcc, 0, v2, vcc
; SI-NEXT: v_add_i32_e32 v3, vcc, s5, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, s1, v1
; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v2, vcc
; SI-NEXT: v_add_i32_e32 v2, vcc, s4, v3
; SI-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; SI-NEXT: v_cndmask_b32_e64 v1, v2, 0, vcc
; SI-NEXT: s_and_b64 s[0:1], vcc, exec
; SI-NEXT: s_cselect_b32 s0, 0, s2
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: v_mov_b32_e32 v0, s0
; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; GFX9-LABEL: umulo_i64_s:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_mul_i32 s7, s0, s3
; GFX9-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX9-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX9-NEXT: s_add_u32 s9, s8, s7
; GFX9-NEXT: s_mul_i32 s6, s1, s2
; GFX9-NEXT: s_addc_u32 s5, 0, s5
; GFX9-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX9-NEXT: s_add_u32 s9, s9, s6
; GFX9-NEXT: s_mul_hi_u32 s10, s1, s3
; GFX9-NEXT: s_addc_u32 s4, s5, s4
; GFX9-NEXT: s_addc_u32 s5, s10, 0
; GFX9-NEXT: s_mul_i32 s1, s1, s3
; GFX9-NEXT: s_add_u32 s4, s4, s1
; GFX9-NEXT: s_addc_u32 s5, 0, s5
; GFX9-NEXT: s_add_i32 s1, s8, s7
; GFX9-NEXT: s_add_i32 s1, s1, s6
; GFX9-NEXT: s_mul_i32 s0, s0, s2
; GFX9-NEXT: s_cmp_lg_u64 s[4:5], 0
; GFX9-NEXT: s_cselect_b32 s1, 0, s1
; GFX9-NEXT: s_cselect_b32 s0, 0, s0
; GFX9-NEXT: v_mov_b32_e32 v0, s0
; GFX9-NEXT: v_mov_b32_e32 v1, s1
; GFX9-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_endpgm
;
; GFX10-LABEL: umulo_i64_s:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_mul_i32 s7, s0, s3
; GFX10-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX10-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX10-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX10-NEXT: s_mul_i32 s6, s1, s2
; GFX10-NEXT: s_mul_hi_u32 s9, s1, s3
; GFX10-NEXT: s_mul_i32 s1, s1, s3
; GFX10-NEXT: s_add_u32 s3, s8, s7
; GFX10-NEXT: s_addc_u32 s5, 0, s5
; GFX10-NEXT: s_add_u32 s3, s3, s6
; GFX10-NEXT: s_addc_u32 s3, s5, s4
; GFX10-NEXT: s_addc_u32 s5, s9, 0
; GFX10-NEXT: s_add_u32 s4, s3, s1
; GFX10-NEXT: s_addc_u32 s5, 0, s5
; GFX10-NEXT: s_add_i32 s1, s8, s7
; GFX10-NEXT: s_mul_i32 s0, s0, s2
; GFX10-NEXT: s_add_i32 s1, s1, s6
; GFX10-NEXT: s_cmp_lg_u64 s[4:5], 0
; GFX10-NEXT: s_cselect_b32 s0, 0, s0
; GFX10-NEXT: s_cselect_b32 s1, 0, s1
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: umulo_i64_s:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_mul_i32 s7, s0, s3
; GFX11-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX11-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX11-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX11-NEXT: s_mul_i32 s6, s1, s2
; GFX11-NEXT: s_mul_hi_u32 s9, s1, s3
; GFX11-NEXT: s_mul_i32 s1, s1, s3
; GFX11-NEXT: s_add_u32 s3, s8, s7
; GFX11-NEXT: s_addc_u32 s5, 0, s5
; GFX11-NEXT: s_add_u32 s3, s3, s6
; GFX11-NEXT: s_addc_u32 s3, s5, s4
; GFX11-NEXT: s_addc_u32 s5, s9, 0
; GFX11-NEXT: s_add_u32 s4, s3, s1
; GFX11-NEXT: s_addc_u32 s5, 0, s5
; GFX11-NEXT: s_add_i32 s1, s8, s7
; GFX11-NEXT: s_mul_i32 s0, s0, s2
; GFX11-NEXT: s_add_i32 s1, s1, s6
; GFX11-NEXT: s_cmp_lg_u64 s[4:5], 0
; GFX11-NEXT: s_cselect_b32 s0, 0, s0
; GFX11-NEXT: s_cselect_b32 s1, 0, s1
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_endpgm
;
; GFX12-LABEL: umulo_i64_s:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
; GFX12-NEXT: s_mov_b32 s5, 0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: s_mul_hi_u32 s7, s0, s3
; GFX12-NEXT: s_mul_i32 s6, s0, s3
; GFX12-NEXT: s_mul_hi_u32 s4, s0, s2
; GFX12-NEXT: s_mul_i32 s10, s1, s2
; GFX12-NEXT: s_add_nc_u64 s[6:7], s[4:5], s[6:7]
; GFX12-NEXT: s_mul_hi_u32 s9, s1, s2
; GFX12-NEXT: s_mul_hi_u32 s11, s1, s3
; GFX12-NEXT: s_add_co_u32 s4, s6, s10
; GFX12-NEXT: s_add_co_ci_u32 s4, s7, s9
; GFX12-NEXT: s_mul_i32 s8, s1, s3
; GFX12-NEXT: s_add_co_ci_u32 s9, s11, 0
; GFX12-NEXT: s_mul_u64 s[0:1], s[0:1], s[2:3]
; GFX12-NEXT: s_add_nc_u64 s[4:5], s[4:5], s[8:9]
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_cmp_lg_u64 s[4:5], 0
; GFX12-NEXT: s_cselect_b32 s0, 0, s0
; GFX12-NEXT: s_cselect_b32 s1, 0, s1
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX12-NEXT: s_endpgm
bb:
%umulo = tail call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %x, i64 %y)
%mul = extractvalue { i64, i1 } %umulo, 0
%overflow = extractvalue { i64, i1 } %umulo, 1
%res = select i1 %overflow, i64 0, i64 %mul
store i64 %res, ptr addrspace(1) undef
ret void
}
define amdgpu_kernel void @smulo_i64_s(i64 %x, i64 %y) {
; SI-LABEL: smulo_i64_s:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: v_mov_b32_e32 v0, s2
; SI-NEXT: v_mul_hi_u32 v1, s1, v0
; SI-NEXT: s_mul_i32 s4, s1, s2
; SI-NEXT: v_mov_b32_e32 v2, s3
; SI-NEXT: v_mul_hi_u32 v3, s0, v2
; SI-NEXT: s_mul_i32 s5, s0, s3
; SI-NEXT: v_mul_hi_u32 v0, s0, v0
; SI-NEXT: v_mul_hi_i32 v2, s1, v2
; SI-NEXT: s_mul_i32 s6, s1, s3
; SI-NEXT: s_mul_i32 s8, s0, s2
; SI-NEXT: v_readfirstlane_b32 s9, v1
; SI-NEXT: v_readfirstlane_b32 s10, v3
; SI-NEXT: v_readfirstlane_b32 s11, v0
; SI-NEXT: v_readfirstlane_b32 s12, v2
; SI-NEXT: v_add_i32_e32 v0, vcc, s5, v0
; SI-NEXT: s_add_u32 s5, s11, s5
; SI-NEXT: v_add_i32_e32 v2, vcc, s4, v0
; SI-NEXT: s_addc_u32 s10, 0, s10
; SI-NEXT: v_ashrrev_i32_e32 v0, 31, v2
; SI-NEXT: s_add_u32 s4, s5, s4
; SI-NEXT: v_mov_b32_e32 v1, v0
; SI-NEXT: s_addc_u32 s4, s10, s9
; SI-NEXT: s_addc_u32 s5, s12, 0
; SI-NEXT: s_add_u32 s4, s4, s6
; SI-NEXT: s_addc_u32 s5, 0, s5
; SI-NEXT: s_sub_u32 s2, s4, s2
; SI-NEXT: s_subb_u32 s6, s5, 0
; SI-NEXT: s_cmp_lt_i32 s1, 0
; SI-NEXT: s_cselect_b32 s1, s6, s5
; SI-NEXT: s_cselect_b32 s2, s2, s4
; SI-NEXT: s_sub_u32 s0, s2, s0
; SI-NEXT: s_subb_u32 s4, s1, 0
; SI-NEXT: s_cmp_lt_i32 s3, 0
; SI-NEXT: s_cselect_b32 s1, s4, s1
; SI-NEXT: s_cselect_b32 s0, s0, s2
; SI-NEXT: v_cmp_ne_u64_e32 vcc, s[0:1], v[0:1]
; SI-NEXT: v_cndmask_b32_e64 v1, v2, 0, vcc
; SI-NEXT: s_and_b64 s[0:1], vcc, exec
; SI-NEXT: s_cselect_b32 s0, 0, s8
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: v_mov_b32_e32 v0, s0
; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; GFX9-LABEL: smulo_i64_s:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_mul_i32 s7, s0, s3
; GFX9-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX9-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX9-NEXT: s_add_u32 s9, s8, s7
; GFX9-NEXT: s_mul_i32 s6, s1, s2
; GFX9-NEXT: s_addc_u32 s5, 0, s5
; GFX9-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX9-NEXT: s_add_u32 s9, s9, s6
; GFX9-NEXT: s_mul_hi_i32 s10, s1, s3
; GFX9-NEXT: s_addc_u32 s4, s5, s4
; GFX9-NEXT: s_addc_u32 s5, s10, 0
; GFX9-NEXT: s_mul_i32 s9, s1, s3
; GFX9-NEXT: s_add_u32 s4, s4, s9
; GFX9-NEXT: s_addc_u32 s5, 0, s5
; GFX9-NEXT: s_sub_u32 s9, s4, s2
; GFX9-NEXT: s_subb_u32 s10, s5, 0
; GFX9-NEXT: s_cmp_lt_i32 s1, 0
; GFX9-NEXT: s_cselect_b32 s4, s9, s4
; GFX9-NEXT: s_cselect_b32 s1, s10, s5
; GFX9-NEXT: s_sub_u32 s9, s4, s0
; GFX9-NEXT: s_subb_u32 s5, s1, 0
; GFX9-NEXT: s_cmp_lt_i32 s3, 0
; GFX9-NEXT: s_cselect_b32 s5, s5, s1
; GFX9-NEXT: s_cselect_b32 s4, s9, s4
; GFX9-NEXT: s_add_i32 s1, s8, s7
; GFX9-NEXT: s_add_i32 s1, s1, s6
; GFX9-NEXT: s_ashr_i32 s6, s1, 31
; GFX9-NEXT: s_mov_b32 s7, s6
; GFX9-NEXT: s_mul_i32 s0, s0, s2
; GFX9-NEXT: s_cmp_lg_u64 s[4:5], s[6:7]
; GFX9-NEXT: s_cselect_b32 s1, 0, s1
; GFX9-NEXT: s_cselect_b32 s0, 0, s0
; GFX9-NEXT: v_mov_b32_e32 v0, s0
; GFX9-NEXT: v_mov_b32_e32 v1, s1
; GFX9-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT: s_endpgm
;
; GFX10-LABEL: smulo_i64_s:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_mul_i32 s7, s0, s3
; GFX10-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX10-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX10-NEXT: s_mul_i32 s6, s1, s2
; GFX10-NEXT: s_add_u32 s11, s8, s7
; GFX10-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX10-NEXT: s_addc_u32 s5, 0, s5
; GFX10-NEXT: s_mul_hi_i32 s9, s1, s3
; GFX10-NEXT: s_add_u32 s11, s11, s6
; GFX10-NEXT: s_mul_i32 s10, s1, s3
; GFX10-NEXT: s_addc_u32 s4, s5, s4
; GFX10-NEXT: s_addc_u32 s5, s9, 0
; GFX10-NEXT: s_add_u32 s4, s4, s10
; GFX10-NEXT: s_addc_u32 s5, 0, s5
; GFX10-NEXT: s_sub_u32 s9, s4, s2
; GFX10-NEXT: s_subb_u32 s10, s5, 0
; GFX10-NEXT: s_cmp_lt_i32 s1, 0
; GFX10-NEXT: s_cselect_b32 s1, s9, s4
; GFX10-NEXT: s_cselect_b32 s4, s10, s5
; GFX10-NEXT: s_sub_u32 s9, s1, s0
; GFX10-NEXT: s_subb_u32 s5, s4, 0
; GFX10-NEXT: s_cmp_lt_i32 s3, 0
; GFX10-NEXT: s_mul_i32 s0, s0, s2
; GFX10-NEXT: s_cselect_b32 s5, s5, s4
; GFX10-NEXT: s_cselect_b32 s4, s9, s1
; GFX10-NEXT: s_add_i32 s1, s8, s7
; GFX10-NEXT: s_add_i32 s1, s1, s6
; GFX10-NEXT: s_ashr_i32 s6, s1, 31
; GFX10-NEXT: s_mov_b32 s7, s6
; GFX10-NEXT: s_cmp_lg_u64 s[4:5], s[6:7]
; GFX10-NEXT: s_cselect_b32 s0, 0, s0
; GFX10-NEXT: s_cselect_b32 s1, 0, s1
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: smulo_i64_s:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_mul_i32 s7, s0, s3
; GFX11-NEXT: s_mul_hi_u32 s8, s0, s2
; GFX11-NEXT: s_mul_hi_u32 s5, s0, s3
; GFX11-NEXT: s_mul_i32 s6, s1, s2
; GFX11-NEXT: s_add_u32 s11, s8, s7
; GFX11-NEXT: s_mul_hi_u32 s4, s1, s2
; GFX11-NEXT: s_addc_u32 s5, 0, s5
; GFX11-NEXT: s_mul_hi_i32 s9, s1, s3
; GFX11-NEXT: s_add_u32 s11, s11, s6
; GFX11-NEXT: s_mul_i32 s10, s1, s3
; GFX11-NEXT: s_addc_u32 s4, s5, s4
; GFX11-NEXT: s_addc_u32 s5, s9, 0
; GFX11-NEXT: s_add_u32 s4, s4, s10
; GFX11-NEXT: s_addc_u32 s5, 0, s5
; GFX11-NEXT: s_sub_u32 s9, s4, s2
; GFX11-NEXT: s_subb_u32 s10, s5, 0
; GFX11-NEXT: s_cmp_lt_i32 s1, 0
; GFX11-NEXT: s_cselect_b32 s1, s9, s4
; GFX11-NEXT: s_cselect_b32 s4, s10, s5
; GFX11-NEXT: s_sub_u32 s9, s1, s0
; GFX11-NEXT: s_subb_u32 s5, s4, 0
; GFX11-NEXT: s_cmp_lt_i32 s3, 0
; GFX11-NEXT: s_mul_i32 s0, s0, s2
; GFX11-NEXT: s_cselect_b32 s5, s5, s4
; GFX11-NEXT: s_cselect_b32 s4, s9, s1
; GFX11-NEXT: s_add_i32 s1, s8, s7
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-NEXT: s_add_i32 s1, s1, s6
; GFX11-NEXT: s_ashr_i32 s6, s1, 31
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-NEXT: s_mov_b32 s7, s6
; GFX11-NEXT: s_cmp_lg_u64 s[4:5], s[6:7]
; GFX11-NEXT: s_cselect_b32 s0, 0, s0
; GFX11-NEXT: s_cselect_b32 s1, 0, s1
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_endpgm
;
; GFX12-LABEL: smulo_i64_s:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
; GFX12-NEXT: s_mov_b32 s5, 0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: s_mul_hi_u32 s7, s0, s3
; GFX12-NEXT: s_mul_i32 s6, s0, s3
; GFX12-NEXT: s_mul_hi_u32 s4, s0, s2
; GFX12-NEXT: s_mul_i32 s10, s1, s2
; GFX12-NEXT: s_add_nc_u64 s[6:7], s[4:5], s[6:7]
; GFX12-NEXT: s_mul_hi_u32 s9, s1, s2
; GFX12-NEXT: s_mul_hi_i32 s11, s1, s3
; GFX12-NEXT: s_add_co_u32 s4, s6, s10
; GFX12-NEXT: s_add_co_ci_u32 s4, s7, s9
; GFX12-NEXT: s_mul_i32 s8, s1, s3
; GFX12-NEXT: s_add_co_ci_u32 s9, s11, 0
; GFX12-NEXT: s_cmp_lt_i32 s1, 0
; GFX12-NEXT: s_add_nc_u64 s[6:7], s[4:5], s[8:9]
; GFX12-NEXT: s_mov_b32 s4, s2
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: s_sub_nc_u64 s[8:9], s[6:7], s[4:5]
; GFX12-NEXT: s_mov_b32 s4, s0
; GFX12-NEXT: s_cselect_b32 s7, s9, s7
; GFX12-NEXT: s_cselect_b32 s6, s8, s6
; GFX12-NEXT: s_cmp_lt_i32 s3, 0
; GFX12-NEXT: s_sub_nc_u64 s[4:5], s[6:7], s[4:5]
; GFX12-NEXT: s_mul_u64 s[0:1], s[0:1], s[2:3]
; GFX12-NEXT: s_cselect_b32 s3, s5, s7
; GFX12-NEXT: s_cselect_b32 s2, s4, s6
; GFX12-NEXT: s_ashr_i32 s4, s1, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_mov_b32 s5, s4
; GFX12-NEXT: s_cmp_lg_u64 s[2:3], s[4:5]
; GFX12-NEXT: s_cselect_b32 s0, 0, s0
; GFX12-NEXT: s_cselect_b32 s1, 0, s1
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX12-NEXT: s_endpgm
bb:
%umulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %x, i64 %y)
%mul = extractvalue { i64, i1 } %umulo, 0
%overflow = extractvalue { i64, i1 } %umulo, 1
%res = select i1 %overflow, i64 0, i64 %mul
store i64 %res, ptr addrspace(1) undef
ret void
}
define { i64, i1 } @smulo_i64_v_4(i64 %i) {
; SI-LABEL: smulo_i64_v_4:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_lshl_b64 v[5:6], v[0:1], 2
; SI-NEXT: v_alignbit_b32 v4, v1, v0, 30
; SI-NEXT: v_ashr_i64 v[2:3], v[5:6], 2
; SI-NEXT: v_cmp_ne_u64_e32 vcc, v[2:3], v[0:1]
; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; SI-NEXT: v_mov_b32_e32 v0, v5
; SI-NEXT: v_mov_b32_e32 v1, v4
; SI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: smulo_i64_v_4:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX9-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX9-NEXT: v_ashrrev_i64 v[5:6], 2, v[4:5]
; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, v[5:6], v[0:1]
; GFX9-NEXT: v_mov_b32_e32 v0, v4
; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; GFX9-NEXT: v_mov_b32_e32 v1, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: smulo_i64_v_4:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX10-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX10-NEXT: v_ashrrev_i64 v[5:6], 2, v[4:5]
; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[5:6], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v0, v4
; GFX10-NEXT: v_mov_b32_e32 v1, v3
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: smulo_i64_v_4:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX11-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT: v_ashrrev_i64 v[5:6], 2, v[4:5]
; GFX11-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[5:6], v[0:1]
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
; GFX11-NEXT: v_dual_mov_b32 v0, v4 :: v_dual_mov_b32 v1, v3
; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-LABEL: smulo_i64_v_4:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT: s_wait_expcnt 0x0
; GFX12-NEXT: s_wait_samplecnt 0x0
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_lshlrev_b64_e32 v[4:5], 2, v[0:1]
; GFX12-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_ashrrev_i64 v[5:6], 2, v[4:5]
; GFX12-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[5:6], v[0:1]
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3)
; GFX12-NEXT: v_dual_mov_b32 v0, v4 :: v_dual_mov_b32 v1, v3
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX12-NEXT: s_setpc_b64 s[30:31]
bb:
%umulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %i, i64 4)
ret { i64, i1 } %umulo
}
define { i64, i1 } @umulo_i64_v_4(i64 %i) {
; SI-LABEL: umulo_i64_v_4:
; SI: ; %bb.0: ; %bb
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_and_b32_e32 v7, 0x3fffffff, v1
; SI-NEXT: v_mov_b32_e32 v6, v0
; SI-NEXT: v_lshl_b64 v[4:5], v[0:1], 2
; SI-NEXT: v_alignbit_b32 v3, v1, v0, 30
; SI-NEXT: v_cmp_ne_u64_e32 vcc, v[6:7], v[0:1]
; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; SI-NEXT: v_mov_b32_e32 v0, v4
; SI-NEXT: v_mov_b32_e32 v1, v3
; SI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: umulo_i64_v_4:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v7, 0x3fffffff, v1
; GFX9-NEXT: v_mov_b32_e32 v6, v0
; GFX9-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, v[6:7], v[0:1]
; GFX9-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; GFX9-NEXT: v_mov_b32_e32 v0, v4
; GFX9-NEXT: v_mov_b32_e32 v1, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: umulo_i64_v_4:
; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_and_b32_e32 v7, 0x3fffffff, v1
; GFX10-NEXT: v_mov_b32_e32 v6, v0
; GFX10-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX10-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[6:7], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v0, v4
; GFX10-NEXT: v_mov_b32_e32 v1, v3
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: umulo_i64_v_4:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_dual_mov_b32 v6, v0 :: v_dual_and_b32 v7, 0x3fffffff, v1
; GFX11-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX11-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[6:7], v[0:1]
; GFX11-NEXT: v_dual_mov_b32 v0, v4 :: v_dual_mov_b32 v1, v3
; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-LABEL: umulo_i64_v_4:
; GFX12: ; %bb.0: ; %bb
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT: s_wait_expcnt 0x0
; GFX12-NEXT: s_wait_samplecnt 0x0
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_dual_mov_b32 v6, v0 :: v_dual_and_b32 v7, 0x3fffffff, v1
; GFX12-NEXT: v_lshlrev_b64_e32 v[4:5], 2, v[0:1]
; GFX12-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[6:7], v[0:1]
; GFX12-NEXT: v_dual_mov_b32 v0, v4 :: v_dual_mov_b32 v1, v3
; GFX12-NEXT: s_wait_alu 0xfffd
; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX12-NEXT: s_setpc_b64 s[30:31]
bb:
%umulo = tail call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %i, i64 4)
ret { i64, i1 } %umulo
}
declare { i64, i1 } @llvm.umul.with.overflow.i64(i64, i64)
declare { i64, i1 } @llvm.smul.with.overflow.i64(i64, i64)
Reference in New Issue View Git Blame Copy Permalink