| # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py |
| # RUN: llc -mtriple=amdgcn -mcpu=gfx1250 -run-pass post-RA-hazard-rec %s -o - | FileCheck -check-prefix=GFX1250 %s |
| |
| # WMMA writes: D0, WMMA reads: A0/B0/Index0 |
| # VALU writes: D1, VALU reads: Use1 |
| # Hards could be: |
| # RAW: D0 overlaps Use1 |
| # WAW: D0 overlaps D1 |
| # WAR: A0/B0/Index0 overlaps D1 |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1_with_4_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1_with_4_valus_in_between |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr26 = V_MOV_B32_e32 26, implicit $exec |
| ; GFX1250-NEXT: $vgpr27 = V_MOV_B32_e32 27, implicit $exec |
| ; GFX1250-NEXT: $vgpr28 = V_MOV_B32_e32 28, implicit $exec |
| ; GFX1250-NEXT: $vgpr29 = V_MOV_B32_e32 29, implicit $exec |
| ; GFX1250-NEXT: $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $vgpr26 = V_MOV_B32_e32 26, implicit $exec |
| $vgpr27 = V_MOV_B32_e32 27, implicit $exec |
| $vgpr28 = V_MOV_B32_e32 28, implicit $exec |
| $vgpr29 = V_MOV_B32_e32 29, implicit $exec |
| $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1_with_4_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_D0_overlaps_Use1_with_4_salus_in_between |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 3 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 3 |
| $vgpr25 = V_ADD_F32_e32 $vgpr24, $vgpr16, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr16 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $vgpr16 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f32_16x16x32_bf16_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f32_16x16x32_bf16_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr24, $vgpr25, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr27 = V_ADD_F32_e32 $vgpr22, $vgpr26, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $vgpr27 = V_ADD_F32_e32 $vgpr22, $vgpr26, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1_with_4_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1_with_4_valus_in_between |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr26 = V_MOV_B32_e32 26, implicit $exec |
| ; GFX1250-NEXT: $vgpr27 = V_MOV_B32_e32 27, implicit $exec |
| ; GFX1250-NEXT: $vgpr28 = V_MOV_B32_e32 28, implicit $exec |
| ; GFX1250-NEXT: $vgpr29 = V_MOV_B32_e32 29, implicit $exec |
| ; GFX1250-NEXT: $vgpr31 = V_ADD_F32_e32 $vgpr22, $vgpr30, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $vgpr26 = V_MOV_B32_e32 26, implicit $exec |
| $vgpr27 = V_MOV_B32_e32 27, implicit $exec |
| $vgpr28 = V_MOV_B32_e32 28, implicit $exec |
| $vgpr29 = V_MOV_B32_e32 29, implicit $exec |
| $vgpr31 = V_ADD_F32_e32 $vgpr22, $vgpr30, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1_with_4_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_Use1_with_4_salus_in_between |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 3 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr27 = V_ADD_F32_e32 $vgpr22, $vgpr26, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 3 |
| $vgpr27 = V_ADD_F32_e32 $vgpr22, $vgpr26, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr22 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $vgpr22 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_f16_16x16x64_fp8_fp8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_f16_16x16x64_fp8_fp8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| $vgpr22_vgpr23_vgpr24_vgpr25 = V_WMMA_F16_16X16X64_FP8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 8, killed $vgpr22_vgpr23_vgpr24_vgpr25, 0, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr26, $vgpr27, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1_with_4_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1_with_4_valus_in_between |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| ; GFX1250-NEXT: $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| ; GFX1250-NEXT: $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1_with_4_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_Use1_with_4_salus_in_between |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 3 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 3 |
| $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_NoF8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_NoF8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr16 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 2, 2, 0, 0, implicit $exec |
| $vgpr16 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1_with_8_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1_with_8_valus_in_between |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| ; GFX1250-NEXT: $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| ; GFX1250-NEXT: $vgpr44 = V_MOV_B32_e32 44, implicit $exec |
| ; GFX1250-NEXT: $vgpr45 = V_MOV_B32_e32 45, implicit $exec |
| ; GFX1250-NEXT: $vgpr46 = V_MOV_B32_e32 46, implicit $exec |
| ; GFX1250-NEXT: $vgpr47 = V_MOV_B32_e32 47, implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| $vgpr44 = V_MOV_B32_e32 44, implicit $exec |
| $vgpr45 = V_MOV_B32_e32 45, implicit $exec |
| $vgpr46 = V_MOV_B32_e32 46, implicit $exec |
| $vgpr47 = V_MOV_B32_e32 47, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1_with_8_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_D0_overlaps_Use1_with_8_salus_in_between |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 3 |
| ; GFX1250-NEXT: $sgpr4 = S_MOV_B32 4 |
| ; GFX1250-NEXT: $sgpr5 = S_MOV_B32 5 |
| ; GFX1250-NEXT: $sgpr6 = S_MOV_B32 6 |
| ; GFX1250-NEXT: $sgpr7 = S_MOV_B32 7 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 3 |
| $sgpr4 = S_MOV_B32 4 |
| $sgpr5 = S_MOV_B32 5 |
| $sgpr6 = S_MOV_B32 6 |
| $sgpr7 = S_MOV_B32 7 |
| $vgpr32 = V_ADD_F32_e32 $vgpr16, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr16 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $vgpr16 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_I32_16x16x64_IU8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_I32_16x16x64_IU8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23 = V_WMMA_I32_16X16X64_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr32, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 2, 0, 0, implicit $exec |
| $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1_with_8_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1_with_8_valus_in_between |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 2, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| ; GFX1250-NEXT: $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| ; GFX1250-NEXT: $vgpr44 = V_MOV_B32_e32 44, implicit $exec |
| ; GFX1250-NEXT: $vgpr45 = V_MOV_B32_e32 45, implicit $exec |
| ; GFX1250-NEXT: $vgpr46 = V_MOV_B32_e32 46, implicit $exec |
| ; GFX1250-NEXT: $vgpr47 = V_MOV_B32_e32 47, implicit $exec |
| ; GFX1250-NEXT: $vgpr49 = V_ADD_F32_e32 $vgpr32, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 2, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| $vgpr44 = V_MOV_B32_e32 44, implicit $exec |
| $vgpr45 = V_MOV_B32_e32 45, implicit $exec |
| $vgpr46 = V_MOV_B32_e32 46, implicit $exec |
| $vgpr47 = V_MOV_B32_e32 47, implicit $exec |
| $vgpr49 = V_ADD_F32_e32 $vgpr32, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1_with_8_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_Use1_with_8_salus_in_between |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 3 |
| ; GFX1250-NEXT: $sgpr4 = S_MOV_B32 4 |
| ; GFX1250-NEXT: $sgpr5 = S_MOV_B32 5 |
| ; GFX1250-NEXT: $sgpr6 = S_MOV_B32 6 |
| ; GFX1250-NEXT: $sgpr7 = S_MOV_B32 7 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 3 |
| $sgpr4 = S_MOV_B32 4 |
| $sgpr5 = S_MOV_B32 5 |
| $sgpr6 = S_MOV_B32 6 |
| $sgpr7 = S_MOV_B32 7 |
| $vgpr48 = V_ADD_F32_e32 $vgpr32, $vgpr47, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_wmma_F32_16x16x128_F8F6F4_F8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_wmma_F32_16x16x128_F8F6F4_F8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr16 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39 = V_WMMA_F32_16X16X128_F8F6F4_f8_f8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, killed $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 8, killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39, 0, 0, 0, 0, implicit $exec |
| $vgpr16 = V_ADD_F32_e32 $vgpr47, $vgpr48, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1_with_2_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1_with_2_valus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1_with_2_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_Use1_with_2_salus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $vgpr34 = V_ADD_F32_e32 $vgpr24, $vgpr33, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr24 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr24 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f32_16x16x64_bf16_Index0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f32_16x16x64_bf16_Index0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_F32_16X16X64_BF16_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32, 0, 0, 0, 0, 0, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr33, $vgpr34, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1_with_2_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1_with_2_valus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1_with_2_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_Use1_with_2_salus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $vgpr31 = V_ADD_F32_e32 $vgpr24, $vgpr30, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr24 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr24 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_f16_16x16x128_bf8_fp8_Index0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_f16_16x16x128_bf8_fp8_Index0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr28 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27 = V_SWMMAC_F16_16X16X128_BF8_FP8_w32_twoaddr killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27, killed $vgpr28_vgpr29, 0, 0, 0, implicit $exec |
| $vgpr28 = V_ADD_F32_e32 $vgpr30, $vgpr31, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr33, $vgpr24, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr34 = V_ADD_F32_e32 $vgpr33, $vgpr24, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1_with_4_valus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1_with_4_valus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| ; GFX1250-NEXT: $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| ; GFX1250-NEXT: $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| ; GFX1250-NEXT: $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr34, $vgpr24, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr40 = V_MOV_B32_e32 40, implicit $exec |
| $vgpr41 = V_MOV_B32_e32 41, implicit $exec |
| $vgpr42 = V_MOV_B32_e32 42, implicit $exec |
| $vgpr43 = V_MOV_B32_e32 43, implicit $exec |
| $vgpr34 = V_ADD_F32_e32 $vgpr34, $vgpr24, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1_with_4_salus_in_between |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_Use1_with_4_salus_in_between |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: $sgpr0 = S_MOV_B32 0 |
| ; GFX1250-NEXT: $sgpr1 = S_MOV_B32 1 |
| ; GFX1250-NEXT: $sgpr2 = S_MOV_B32 2 |
| ; GFX1250-NEXT: $sgpr3 = S_MOV_B32 4 |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr34 = V_ADD_F32_e32 $vgpr34, $vgpr24, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $sgpr0 = S_MOV_B32 0 |
| $sgpr1 = S_MOV_B32 1 |
| $sgpr2 = S_MOV_B32 2 |
| $sgpr3 = S_MOV_B32 4 |
| $vgpr34 = V_ADD_F32_e32 $vgpr34, $vgpr24, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_D0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr24 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr24 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_A0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_A0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr0 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr0 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_B0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_B0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr8 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr8 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| ... |
| |
| --- |
| name: test_swmmac_i32_16x16x128_iu8_Index0_overlaps_D1 |
| body: | |
| bb.0: |
| ; GFX1250-LABEL: name: test_swmmac_i32_16x16x128_iu8_Index0_overlaps_D1 |
| ; GFX1250: early-clobber $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: V_NOP_e32 implicit $exec |
| ; GFX1250-NEXT: $vgpr32 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = V_SWMMAC_I32_16X16X128_IU8_w32_twoaddr 8, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, 8, killed $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23, killed $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, killed $vgpr32_vgpr33, 0, 0, 0, 0, implicit $exec |
| $vgpr32 = V_ADD_F32_e32 $vgpr34, $vgpr35, implicit $mode, implicit $exec |
| ... |