2041 lines · plain
1# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 52# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -start-before=greedy,2 -stop-after=virtregrewriter,2 -verify-regalloc -verify-machineinstrs -o - %s | FileCheck %s3 4# There aren't enough VGPRs for %0 to stay in a VGPR. %0 should be5# split and inflated to AV_512. The VGPR version of the instruction6# should be replaced with the AGPR version, and the copies eliminated.7 8---9name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e6410tracksRegLiveness: true11machineFunctionInfo:12 isEntryFunction: true13 stackPtrOffsetReg: '$sgpr32'14 occupancy: 1015 sgprForEXECCopy: '$sgpr100_sgpr101'16body: |17 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e6418 ; CHECK: bb.0:19 ; CHECK-NEXT: successors: %bb.1(0x80000000)20 ; CHECK-NEXT: {{ $}}21 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr022 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 023 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec24 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr025 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr126 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc27 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr828 ; CHECK-NEXT: {{ $}}29 ; CHECK-NEXT: bb.1:30 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)31 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr132 ; CHECK-NEXT: {{ $}}33 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)34 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec35 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc36 ; CHECK-NEXT: S_BRANCH %bb.237 ; CHECK-NEXT: {{ $}}38 ; CHECK-NEXT: bb.2:39 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF40 ; CHECK-NEXT: {{ $}}41 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr742 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr1543 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr2344 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr3145 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr3946 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr4747 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr5548 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr6349 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec50 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)51 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)52 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)53 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)54 ; CHECK-NEXT: S_ENDPGM 055 bb.0:56 S_NOP 0, implicit-def $agpr057 renamable $sgpr0 = S_MOV_B32 058 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec59 renamable $sgpr1 = COPY renamable $sgpr060 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr161 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc62 %0.sub9:vreg_512_align2 = COPY %0.sub863 64 bb.1:65 liveins: $vcc66 67 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)68 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec69 S_CBRANCH_VCCNZ %bb.1, implicit $vcc70 S_BRANCH %bb.271 72 bb.2:73 ; No VGPRs available for %074 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr775 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr1576 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr2377 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr3178 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr3979 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr4780 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr5581 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr6382 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec83 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)84 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)85 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)86 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)87 S_ENDPGM 088 89...90 91# Same, except spam some other uses of the register that we shouldn't92# break the liveness of93 94---95name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_multi_load96tracksRegLiveness: true97machineFunctionInfo:98 isEntryFunction: true99 stackPtrOffsetReg: '$sgpr32'100 occupancy: 10101 sgprForEXECCopy: '$sgpr100_sgpr101'102body: |103 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_multi_load104 ; CHECK: bb.0:105 ; CHECK-NEXT: successors: %bb.1(0x80000000)106 ; CHECK-NEXT: {{ $}}107 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0108 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0109 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec110 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0111 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1112 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc113 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8114 ; CHECK-NEXT: {{ $}}115 ; CHECK-NEXT: bb.1:116 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)117 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr1118 ; CHECK-NEXT: {{ $}}119 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)120 ; CHECK-NEXT: renamable $agpr2_agpr3 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 8, 0, implicit $exec :: (load (s64), addrspace 1)121 ; CHECK-NEXT: renamable $agpr4_agpr5_agpr6 = GLOBAL_LOAD_DWORDX3 undef renamable $vgpr0_vgpr1, 16, 0, implicit $exec :: (load (s96), align 16, addrspace 1)122 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 28, 0, 0, implicit $mode, implicit $exec123 ; CHECK-NEXT: S_NOP 0, implicit renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15124 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR undef renamable $vgpr0, renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)125 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc126 ; CHECK-NEXT: S_BRANCH %bb.2127 ; CHECK-NEXT: {{ $}}128 ; CHECK-NEXT: bb.2:129 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF130 ; CHECK-NEXT: {{ $}}131 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7132 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15133 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23134 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31135 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39136 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47137 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55138 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63139 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec140 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)141 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)142 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)143 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)144 ; CHECK-NEXT: S_ENDPGM 0145 bb.0:146 S_NOP 0, implicit-def $agpr0147 renamable $sgpr0 = S_MOV_B32 0148 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec149 renamable $sgpr1 = COPY renamable $sgpr0150 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1151 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc152 %0.sub9:vreg_512_align2 = COPY %0.sub8153 154 bb.1:155 liveins: $vcc156 157 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)158 %0.sub2_sub3:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 8, 0, implicit $exec :: (load (s64), addrspace 1)159 %0.sub4_sub5_sub6:vreg_512_align2 = GLOBAL_LOAD_DWORDX3 undef %3:vreg_64_align2, 16, 0, implicit $exec :: (load (s96), addrspace 1)160 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 28, 0, 0, implicit $mode, implicit $exec161 S_NOP 0, implicit %0162 GLOBAL_STORE_DWORDX4_SADDR undef %4:vgpr_32, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)163 S_CBRANCH_VCCNZ %bb.1, implicit $vcc164 S_BRANCH %bb.2165 166 bb.2:167 ; No VGPRs available for %0168 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7169 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15170 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23171 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31172 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39173 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47174 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55175 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63176 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec177 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)178 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)179 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)180 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)181 S_ENDPGM 0182 183...184 185# Non-mac variant, src2 is other VGPR. We cannot replace this case186# unless we also try to reassign src2 to an AGPR.187---188name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64189tracksRegLiveness: true190machineFunctionInfo:191 isEntryFunction: true192 stackPtrOffsetReg: '$sgpr32'193 occupancy: 10194 sgprForEXECCopy: '$sgpr100_sgpr101'195body: |196 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64197 ; CHECK: bb.0:198 ; CHECK-NEXT: successors: %bb.1(0x80000000)199 ; CHECK-NEXT: {{ $}}200 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0201 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0202 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec203 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0204 ; CHECK-NEXT: renamable $vgpr2_vgpr3 = COPY killed renamable $sgpr0_sgpr1205 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc206 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8207 ; CHECK-NEXT: {{ $}}208 ; CHECK-NEXT: bb.1:209 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)210 ; CHECK-NEXT: liveins: $vcc, $vgpr2_vgpr3211 ; CHECK-NEXT: {{ $}}212 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)213 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr2_vgpr3, $vgpr2_vgpr3, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec214 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc215 ; CHECK-NEXT: S_BRANCH %bb.2216 ; CHECK-NEXT: {{ $}}217 ; CHECK-NEXT: bb.2:218 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF219 ; CHECK-NEXT: {{ $}}220 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7221 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15222 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23223 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31224 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39225 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47226 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55227 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63228 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec229 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)230 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)231 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)232 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)233 ; CHECK-NEXT: S_ENDPGM 0234 bb.0:235 S_NOP 0, implicit-def $agpr0236 renamable $sgpr0 = S_MOV_B32 0237 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec238 renamable $sgpr1 = COPY renamable $sgpr0239 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1240 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc241 %0.sub9:vreg_512_align2 = COPY %0.sub8242 243 bb.1:244 liveins: $vcc245 246 undef %4.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)247 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %4, 0, 0, 0, implicit $mode, implicit $exec248 S_CBRANCH_VCCNZ %bb.1, implicit $vcc249 S_BRANCH %bb.2250 251 bb.2:252 ; No VGPRs available for %0253 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7254 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15255 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23256 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31257 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39258 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47259 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55260 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63261 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec262 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)263 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)264 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)265 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)266 S_ENDPGM 0267 268...269 270# TODO: The two MFMAs are both rewritable as a set, but if we look at271# one at a time the register class constraints of the other block the272# transform.273---274name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses275tracksRegLiveness: true276machineFunctionInfo:277 isEntryFunction: true278 stackPtrOffsetReg: '$sgpr32'279 occupancy: 10280 sgprForEXECCopy: '$sgpr100_sgpr101'281body: |282 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses283 ; CHECK: bb.0:284 ; CHECK-NEXT: successors: %bb.1(0x80000000)285 ; CHECK-NEXT: {{ $}}286 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0287 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0288 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec289 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0290 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1291 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc292 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8293 ; CHECK-NEXT: {{ $}}294 ; CHECK-NEXT: bb.1:295 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)296 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr1297 ; CHECK-NEXT: {{ $}}298 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)299 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec300 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, killed $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec301 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc302 ; CHECK-NEXT: S_BRANCH %bb.2303 ; CHECK-NEXT: {{ $}}304 ; CHECK-NEXT: bb.2:305 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF306 ; CHECK-NEXT: {{ $}}307 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7308 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15309 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23310 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31311 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39312 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47313 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55314 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63315 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec316 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)317 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)318 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)319 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)320 ; CHECK-NEXT: S_ENDPGM 0321 bb.0:322 S_NOP 0, implicit-def $agpr0323 renamable $sgpr0 = S_MOV_B32 0324 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec325 renamable $sgpr1 = COPY renamable $sgpr0326 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1327 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc328 %0.sub9:vreg_512_align2 = COPY %0.sub8329 330 bb.1:331 liveins: $vcc332 333 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)334 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec335 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec336 S_CBRANCH_VCCNZ %bb.1, implicit $vcc337 S_BRANCH %bb.2338 339 bb.2:340 ; No VGPRs available for %0341 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7342 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15343 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23344 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31345 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39346 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47347 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55348 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63349 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec350 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)351 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)352 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)353 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)354 S_ENDPGM 0355 356...357 358# The second MFMA has additional references to the rewritten register359# %0 in src0/src1, not just src2.360---361name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses_repeated_operand362tracksRegLiveness: true363machineFunctionInfo:364 isEntryFunction: true365 stackPtrOffsetReg: '$sgpr32'366 occupancy: 10367 sgprForEXECCopy: '$sgpr100_sgpr101'368body: |369 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses_repeated_operand370 ; CHECK: bb.0:371 ; CHECK-NEXT: successors: %bb.1(0x80000000)372 ; CHECK-NEXT: {{ $}}373 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0374 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0375 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec376 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0377 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1378 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc379 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8380 ; CHECK-NEXT: {{ $}}381 ; CHECK-NEXT: bb.1:382 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)383 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr1384 ; CHECK-NEXT: {{ $}}385 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)386 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec387 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 killed $agpr2_agpr3, $agpr0_agpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec388 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc389 ; CHECK-NEXT: S_BRANCH %bb.2390 ; CHECK-NEXT: {{ $}}391 ; CHECK-NEXT: bb.2:392 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF393 ; CHECK-NEXT: {{ $}}394 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7395 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15396 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23397 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31398 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39399 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47400 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55401 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63402 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec403 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)404 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)405 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)406 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)407 ; CHECK-NEXT: S_ENDPGM 0408 bb.0:409 S_NOP 0, implicit-def $agpr0410 renamable $sgpr0 = S_MOV_B32 0411 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec412 renamable $sgpr1 = COPY renamable $sgpr0413 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1414 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc415 %0.sub9:vreg_512_align2 = COPY %0.sub8416 417 bb.1:418 liveins: $vcc419 420 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)421 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec422 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %0.sub2_sub3, %0.sub0_sub1, %0, 0, 0, 0, implicit $mode, implicit $exec423 S_CBRANCH_VCCNZ %bb.1, implicit $vcc424 S_BRANCH %bb.2425 426 bb.2:427 ; No VGPRs available for %0428 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7429 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15430 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23431 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31432 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39433 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47434 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55435 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63436 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec437 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)438 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)439 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)440 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)441 S_ENDPGM 0442 443...444 445 446---447name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses_cannot_rewrite_final_use448tracksRegLiveness: true449machineFunctionInfo:450 isEntryFunction: true451 stackPtrOffsetReg: '$sgpr32'452 occupancy: 10453 sgprForEXECCopy: '$sgpr100_sgpr101'454body: |455 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_two_chained_uses_cannot_rewrite_final_use456 ; CHECK: bb.0:457 ; CHECK-NEXT: successors: %bb.1(0x80000000)458 ; CHECK-NEXT: {{ $}}459 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0460 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0461 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec462 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0463 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1464 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc465 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8466 ; CHECK-NEXT: {{ $}}467 ; CHECK-NEXT: bb.1:468 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)469 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr1470 ; CHECK-NEXT: {{ $}}471 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)472 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec473 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, killed $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec474 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc475 ; CHECK-NEXT: S_BRANCH %bb.2476 ; CHECK-NEXT: {{ $}}477 ; CHECK-NEXT: bb.2:478 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF479 ; CHECK-NEXT: {{ $}}480 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7481 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15482 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23483 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31484 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39485 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47486 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55487 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63488 ; CHECK-NEXT: renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15489 ; CHECK-NEXT: INLINEASM &"; use $0 ", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15490 ; CHECK-NEXT: S_ENDPGM 0491 bb.0:492 S_NOP 0, implicit-def $agpr0493 renamable $sgpr0 = S_MOV_B32 0494 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec495 renamable $sgpr1 = COPY renamable $sgpr0496 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1497 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc498 %0.sub9:vreg_512_align2 = COPY %0.sub8499 500 bb.1:501 liveins: $vcc502 503 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)504 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec505 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec506 S_CBRANCH_VCCNZ %bb.1, implicit $vcc507 S_BRANCH %bb.2508 509 bb.2:510 ; No VGPRs available for %0511 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7512 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15513 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23514 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31515 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39516 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47517 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55518 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63519 INLINEASM &"; use $0 ", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, %0:vreg_512_align2520 S_ENDPGM 0521 522...523 524# There is a rewrite candidate, but it is used by another MFMA which525# does not have a tied result.526---527name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_2nd_use_not_tied528tracksRegLiveness: true529machineFunctionInfo:530 isEntryFunction: true531 stackPtrOffsetReg: '$sgpr32'532 occupancy: 10533 sgprForEXECCopy: '$sgpr100_sgpr101'534body: |535 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_2nd_use_not_tied536 ; CHECK: bb.0:537 ; CHECK-NEXT: successors: %bb.1(0x80000000)538 ; CHECK-NEXT: {{ $}}539 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0540 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0541 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec542 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0543 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = COPY killed renamable $sgpr0_sgpr1544 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc545 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8546 ; CHECK-NEXT: {{ $}}547 ; CHECK-NEXT: bb.1:548 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)549 ; CHECK-NEXT: liveins: $vcc, $vgpr16_vgpr17550 ; CHECK-NEXT: {{ $}}551 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)552 ; CHECK-NEXT: renamable $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec553 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, killed $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec554 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc555 ; CHECK-NEXT: S_BRANCH %bb.2556 ; CHECK-NEXT: {{ $}}557 ; CHECK-NEXT: bb.2:558 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF559 ; CHECK-NEXT: {{ $}}560 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7561 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15562 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23563 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31564 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39565 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47566 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55567 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63568 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec569 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)570 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)571 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)572 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)573 ; CHECK-NEXT: S_ENDPGM 0574 bb.0:575 S_NOP 0, implicit-def $agpr0576 renamable $sgpr0 = S_MOV_B32 0577 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec578 renamable $sgpr1 = COPY renamable $sgpr0579 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1580 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc581 %0.sub9:vreg_512_align2 = COPY %0.sub8582 583 bb.1:584 liveins: $vcc585 586 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %2:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)587 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec588 %3:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec589 S_CBRANCH_VCCNZ %bb.1, implicit $vcc590 S_BRANCH %bb.2591 592 bb.2:593 ; No VGPRs available for %0 or %3594 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7595 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15596 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23597 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31598 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39599 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47600 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55601 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63602 %4:vgpr_32 = V_MOV_B32_e32 0, implicit $exec603 GLOBAL_STORE_DWORDX4_SADDR %4, %3.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)604 GLOBAL_STORE_DWORDX4_SADDR %4, %3.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)605 GLOBAL_STORE_DWORDX4_SADDR %4, %3.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)606 GLOBAL_STORE_DWORDX4_SADDR %4, %3.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)607 S_ENDPGM 0608 609...610 611# Chain of 2 untied cases612---613name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain614tracksRegLiveness: true615machineFunctionInfo:616 isEntryFunction: true617 stackPtrOffsetReg: '$sgpr32'618 occupancy: 10619 sgprForEXECCopy: '$sgpr100_sgpr101'620body: |621 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain622 ; CHECK: bb.0:623 ; CHECK-NEXT: successors: %bb.1(0x80000000)624 ; CHECK-NEXT: {{ $}}625 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0626 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0627 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec628 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0629 ; CHECK-NEXT: renamable $vgpr18_vgpr19 = COPY killed renamable $sgpr0_sgpr1630 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc631 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8632 ; CHECK-NEXT: {{ $}}633 ; CHECK-NEXT: bb.1:634 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)635 ; CHECK-NEXT: liveins: $vcc, $vgpr18_vgpr19636 ; CHECK-NEXT: {{ $}}637 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)638 ; CHECK-NEXT: early-clobber renamable $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31 = V_MFMA_F32_32X32X8F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec639 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, killed $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec640 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc641 ; CHECK-NEXT: S_BRANCH %bb.2642 ; CHECK-NEXT: {{ $}}643 ; CHECK-NEXT: bb.2:644 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF645 ; CHECK-NEXT: {{ $}}646 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7647 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15648 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23649 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31650 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39651 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47652 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55653 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63654 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec655 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)656 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)657 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)658 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)659 ; CHECK-NEXT: S_ENDPGM 0660 bb.0:661 S_NOP 0, implicit-def $agpr0662 renamable $sgpr0 = S_MOV_B32 0663 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec664 renamable $sgpr1 = COPY renamable $sgpr0665 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1666 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc667 %0.sub9:vreg_512_align2 = COPY %0.sub8668 669 bb.1:670 liveins: $vcc671 672 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %2:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)673 %3:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec674 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %3, 0, 0, 0, implicit $mode, implicit $exec675 S_CBRANCH_VCCNZ %bb.1, implicit $vcc676 S_BRANCH %bb.2677 678 bb.2:679 ; No VGPRs available for %0 or %4680 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7681 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15682 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23683 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31684 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39685 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47686 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55687 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63688 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec689 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)690 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)691 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)692 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)693 S_ENDPGM 0694 695...696# Chain of 2 untied cases, but the use isn't in src2.697---698name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain_non_src2699tracksRegLiveness: true700machineFunctionInfo:701 isEntryFunction: true702 stackPtrOffsetReg: '$sgpr32'703 occupancy: 10704 sgprForEXECCopy: '$sgpr100_sgpr101'705body: |706 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain_non_src2707 ; CHECK: bb.0:708 ; CHECK-NEXT: successors: %bb.1(0x80000000)709 ; CHECK-NEXT: {{ $}}710 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0711 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0712 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec713 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0714 ; CHECK-NEXT: renamable $vgpr18_vgpr19 = COPY killed renamable $sgpr0_sgpr1715 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc716 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8717 ; CHECK-NEXT: {{ $}}718 ; CHECK-NEXT: bb.1:719 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)720 ; CHECK-NEXT: liveins: $vcc, $vgpr18_vgpr19721 ; CHECK-NEXT: {{ $}}722 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)723 ; CHECK-NEXT: early-clobber renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31, 0, 0, 0, implicit $mode, implicit $exec724 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 killed $vgpr4_vgpr5, $vgpr8_vgpr9, undef $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec725 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc726 ; CHECK-NEXT: S_BRANCH %bb.2727 ; CHECK-NEXT: {{ $}}728 ; CHECK-NEXT: bb.2:729 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF730 ; CHECK-NEXT: {{ $}}731 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7732 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15733 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23734 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31735 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39736 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47737 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55738 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63739 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec740 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)741 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)742 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)743 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)744 ; CHECK-NEXT: S_ENDPGM 0745 bb.0:746 S_NOP 0, implicit-def $agpr0747 renamable $sgpr0 = S_MOV_B32 0748 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec749 renamable $sgpr1 = COPY renamable $sgpr0750 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1751 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc752 %0.sub9:vreg_512_align2 = COPY %0.sub8753 754 bb.1:755 liveins: $vcc756 757 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %2:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)758 %3:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec759 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %3.sub4_sub5, %3.sub8_sub9, undef %6:vreg_512_align2, 0, 0, 0, implicit $mode, implicit $exec760 S_CBRANCH_VCCNZ %bb.1, implicit $vcc761 S_BRANCH %bb.2762 763 bb.2:764 ; No VGPRs available for %0 or %4765 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7766 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15767 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23768 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31769 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39770 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47771 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55772 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63773 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec774 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)775 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)776 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)777 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)778 S_ENDPGM 0779 780...781 782# Chain of 2 untied cases, but the second mfma is a different size and783# uses a subregister.784---785name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain_subreg786tracksRegLiveness: true787machineFunctionInfo:788 isEntryFunction: true789 stackPtrOffsetReg: '$sgpr32'790 occupancy: 10791 sgprForEXECCopy: '$sgpr100_sgpr101'792body: |793 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_chain_subreg794 ; CHECK: bb.0:795 ; CHECK-NEXT: successors: %bb.1(0x80000000)796 ; CHECK-NEXT: {{ $}}797 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr0798 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0799 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec800 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0801 ; CHECK-NEXT: renamable $vgpr18_vgpr19 = COPY killed renamable $sgpr0_sgpr1802 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc803 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8804 ; CHECK-NEXT: {{ $}}805 ; CHECK-NEXT: bb.1:806 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)807 ; CHECK-NEXT: liveins: $vcc, $vgpr18_vgpr19808 ; CHECK-NEXT: {{ $}}809 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)810 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec811 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_16X16X16F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, killed $agpr2_agpr3_agpr4_agpr5, 0, 0, 0, implicit $mode, implicit $exec812 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc813 ; CHECK-NEXT: S_BRANCH %bb.2814 ; CHECK-NEXT: {{ $}}815 ; CHECK-NEXT: bb.2:816 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3817 ; CHECK-NEXT: {{ $}}818 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7819 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15820 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23821 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31822 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39823 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47824 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55825 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63826 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec827 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)828 ; CHECK-NEXT: S_ENDPGM 0829 bb.0:830 S_NOP 0, implicit-def $agpr0831 renamable $sgpr0 = S_MOV_B32 0832 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec833 renamable $sgpr1 = COPY renamable $sgpr0834 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1835 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc836 %0.sub9:vreg_512_align2 = COPY %0.sub8837 838 bb.1:839 liveins: $vcc840 841 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %2:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)842 %3:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec843 %4:vreg_128_align2 = V_MFMA_F32_16X16X16F16_vgprcd_e64 %1, %1, %3.sub2_sub3_sub4_sub5, 0, 0, 0, implicit $mode, implicit $exec844 S_CBRANCH_VCCNZ %bb.1, implicit $vcc845 S_BRANCH %bb.2846 847 bb.2:848 ; No VGPRs available for %0 or %4849 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7850 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15851 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23852 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31853 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39854 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47855 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55856 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63857 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec858 GLOBAL_STORE_DWORDX4_SADDR %5, %4, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)859 S_ENDPGM 0860 861...862 863---864name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_local_split865tracksRegLiveness: true866machineFunctionInfo:867 isEntryFunction: true868 stackPtrOffsetReg: '$sgpr32'869 occupancy: 10870 sgprForEXECCopy: '$sgpr100_sgpr101'871body: |872 bb.0:873 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_local_split874 ; CHECK: S_NOP 0, implicit-def $agpr0875 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0876 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec877 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0878 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1879 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc880 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8881 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)882 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 killed $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec883 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7884 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15885 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23886 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31887 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39888 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47889 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55890 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63891 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec892 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15893 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr10_vgpr11_vgpr12_vgpr13, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)894 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr14_vgpr15_vgpr16_vgpr17, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)895 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr2_vgpr3_vgpr4_vgpr5, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)896 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $vgpr6_vgpr7_vgpr8_vgpr9, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)897 ; CHECK-NEXT: S_ENDPGM 0898 S_NOP 0, implicit-def $agpr0899 renamable $sgpr0 = S_MOV_B32 0900 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec901 renamable $sgpr1 = COPY renamable $sgpr0902 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1903 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc904 %0.sub9:vreg_512_align2 = COPY %0.sub8905 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)906 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec907 ; No VGPRs available for %0908 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7909 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15910 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23911 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31912 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39913 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47914 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55915 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63916 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec917 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)918 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)919 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)920 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)921 S_ENDPGM 0922 923...924 925# Performs a split and inflate around the single instruction926---927name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_instruction_split928tracksRegLiveness: true929machineFunctionInfo:930 isEntryFunction: true931 stackPtrOffsetReg: '$sgpr32'932 occupancy: 10933 sgprForEXECCopy: '$sgpr100_sgpr101'934body: |935 bb.0:936 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_instruction_split937 ; CHECK: S_NOP 0, implicit-def $agpr0938 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 0939 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec940 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr0941 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr1942 ; CHECK-NEXT: SI_SPILL_AV64_SAVE killed $vgpr0_vgpr1, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)943 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc944 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr8945 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)946 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7947 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15948 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23949 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31950 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39951 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47952 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55953 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63954 ; CHECK-NEXT: renamable $vgpr2_vgpr3 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)955 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 killed $vgpr2_vgpr3, $vgpr2_vgpr3, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec956 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7957 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15958 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23959 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31960 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39961 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47962 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55963 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63964 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec965 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15966 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr10_vgpr11_vgpr12_vgpr13, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)967 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr14_vgpr15_vgpr16_vgpr17, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)968 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr2_vgpr3_vgpr4_vgpr5, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)969 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $vgpr6_vgpr7_vgpr8_vgpr9, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)970 ; CHECK-NEXT: S_ENDPGM 0971 S_NOP 0, implicit-def $agpr0972 renamable $sgpr0 = S_MOV_B32 0973 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec974 renamable $sgpr1 = COPY renamable $sgpr0975 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr1976 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc977 %0.sub9:vreg_512_align2 = COPY %0.sub8978 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)979 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7980 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15981 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23982 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31983 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39984 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47985 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55986 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63987 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec988 ; No VGPRs available for %0989 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7990 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15991 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23992 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31993 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39994 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47995 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55996 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63997 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec998 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)999 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1000 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1001 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1002 S_ENDPGM 01003 1004...1005 1006# Performs a split and inflate around the single instruction, non-tied case1007---1008name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_instruction_split1009tracksRegLiveness: true1010machineFunctionInfo:1011 isEntryFunction: true1012 stackPtrOffsetReg: '$sgpr32'1013 occupancy: 101014 sgprForEXECCopy: '$sgpr100_sgpr101'1015body: |1016 bb.0:1017 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_instruction_split1018 ; CHECK: S_NOP 0, implicit-def $agpr01019 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01020 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1021 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01022 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11023 ; CHECK-NEXT: SI_SPILL_AV64_SAVE killed $vgpr0_vgpr1, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)1024 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1025 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81026 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1027 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71028 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151029 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231030 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311031 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391032 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471033 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551034 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631035 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)1036 ; CHECK-NEXT: renamable $agpr16_agpr17 = COPY killed renamable $agpr0_agpr11037 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 killed $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1038 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71039 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151040 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231041 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311042 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391043 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471044 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551045 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631046 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1047 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr151048 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr10_vgpr11_vgpr12_vgpr13, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1049 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr14_vgpr15_vgpr16_vgpr17, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1050 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr2_vgpr3_vgpr4_vgpr5, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1051 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $vgpr6_vgpr7_vgpr8_vgpr9, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1052 ; CHECK-NEXT: S_ENDPGM 01053 S_NOP 0, implicit-def $agpr01054 renamable $sgpr0 = S_MOV_B32 01055 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1056 renamable $sgpr1 = COPY renamable $sgpr01057 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11058 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1059 %0.sub9:vreg_512_align2 = COPY %0.sub81060 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1061 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71062 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151063 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231064 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311065 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391066 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471067 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551068 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631069 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1070 ; No VGPRs available for %01071 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71072 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151073 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231074 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311075 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391076 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471077 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551078 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631079 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1080 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1081 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1082 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1083 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1084 S_ENDPGM 01085 1086...1087 1088# This case does not fully use %0 after the MFMA. As a result,1089# SplitKits insert a copy bundle for the subset of used lanes instead1090# of a simple copy.1091---1092name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_instruction_split_partial_uses_only1093tracksRegLiveness: true1094machineFunctionInfo:1095 isEntryFunction: true1096 stackPtrOffsetReg: '$sgpr32'1097 occupancy: 101098 sgprForEXECCopy: '$sgpr100_sgpr101'1099body: |1100 bb.0:1101 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_instruction_split_partial_uses_only1102 ; CHECK: S_NOP 0, implicit-def $agpr01103 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01104 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1105 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01106 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11107 ; CHECK-NEXT: SI_SPILL_AV64_SAVE killed $vgpr0_vgpr1, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)1108 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1109 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81110 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1111 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71112 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151113 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231114 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311115 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391116 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471117 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551118 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631119 ; CHECK-NEXT: renamable $vgpr2_vgpr3 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)1120 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 killed $vgpr2_vgpr3, $vgpr2_vgpr3, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec1121 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71122 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151123 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231124 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311125 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391126 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471127 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551128 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631129 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1130 ; CHECK-NEXT: renamable $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = COPY killed renamable $agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr151131 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5 = COPY renamable $agpr0_agpr1_agpr2_agpr31132 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr10_vgpr11_vgpr12_vgpr13, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1133 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr14_vgpr15_vgpr16_vgpr17, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1134 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $vgpr2_vgpr3_vgpr4_vgpr5, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1135 ; CHECK-NEXT: S_ENDPGM 01136 S_NOP 0, implicit-def $agpr01137 renamable $sgpr0 = S_MOV_B32 01138 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1139 renamable $sgpr1 = COPY renamable $sgpr01140 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11141 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1142 %0.sub9:vreg_512_align2 = COPY %0.sub81143 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1144 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71145 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151146 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231147 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311148 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391149 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471150 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551151 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631152 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1153 ; No VGPRs available for %01154 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71155 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151156 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231157 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311158 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391159 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471160 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551161 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631162 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1163 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1164 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1165 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1166 S_ENDPGM 01167 1168...1169 1170# Untied version of previous. This case does not fully use %4 after1171# the MFMA. As a result, SplitKits insert a copy bundle for the subset1172# of used lanes instead of a simple copy,1173---1174name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_instruction_split_partial_uses_only1175tracksRegLiveness: true1176machineFunctionInfo:1177 isEntryFunction: true1178 stackPtrOffsetReg: '$sgpr32'1179 occupancy: 101180 sgprForEXECCopy: '$sgpr100_sgpr101'1181body: |1182 bb.0:1183 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_instruction_split_partial_uses_only1184 ; CHECK: S_NOP 0, implicit-def $agpr01185 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01186 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1187 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01188 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11189 ; CHECK-NEXT: SI_SPILL_AV64_SAVE killed $vgpr0_vgpr1, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)1190 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1191 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81192 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1193 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71194 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151195 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231196 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311197 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391198 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471199 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551200 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631201 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)1202 ; CHECK-NEXT: renamable $agpr16_agpr17 = COPY killed renamable $agpr0_agpr11203 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 killed $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1204 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71205 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151206 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231207 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311208 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391209 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471210 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551211 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631212 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1213 ; CHECK-NEXT: renamable $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = COPY killed renamable $agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr151214 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5 = COPY renamable $agpr0_agpr1_agpr2_agpr31215 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr10_vgpr11_vgpr12_vgpr13, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1216 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $vgpr14_vgpr15_vgpr16_vgpr17, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1217 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $vgpr2_vgpr3_vgpr4_vgpr5, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1218 ; CHECK-NEXT: S_ENDPGM 01219 S_NOP 0, implicit-def $agpr01220 renamable $sgpr0 = S_MOV_B32 01221 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1222 renamable $sgpr1 = COPY renamable $sgpr01223 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11224 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1225 %0.sub9:vreg_512_align2 = COPY %0.sub81226 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1227 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71228 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151229 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231230 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311231 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391232 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471233 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551234 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631235 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1236 ; No VGPRs available for %41237 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71238 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151239 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231240 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311241 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391242 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471243 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551244 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631245 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1246 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1247 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1248 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1249 S_ENDPGM 01250 1251...1252 1253---1254name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_same_subreg1255tracksRegLiveness: true1256machineFunctionInfo:1257 isEntryFunction: true1258 stackPtrOffsetReg: '$sgpr32'1259 occupancy: 101260 sgprForEXECCopy: '$sgpr100_sgpr101'1261body: |1262 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_same_subreg1263 ; CHECK: bb.0:1264 ; CHECK-NEXT: successors: %bb.1(0x80000000)1265 ; CHECK-NEXT: {{ $}}1266 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01267 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01268 ; CHECK-NEXT: renamable $vgpr10 = V_MOV_B32_e32 0, implicit $exec1269 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01270 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11271 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1272 ; CHECK-NEXT: renamable $vgpr11 = COPY renamable $vgpr101273 ; CHECK-NEXT: {{ $}}1274 ; CHECK-NEXT: bb.1:1275 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1276 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr11277 ; CHECK-NEXT: {{ $}}1278 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5 = GLOBAL_LOAD_DWORDX4 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s128), addrspace 1)1279 ; CHECK-NEXT: renamable $vgpr6_vgpr7_vgpr8_vgpr9 = GLOBAL_LOAD_DWORDX4 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s128), addrspace 1)1280 ; CHECK-NEXT: renamable $vgpr10_vgpr11_vgpr12_vgpr13 = GLOBAL_LOAD_DWORDX4 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s128), addrspace 1)1281 ; CHECK-NEXT: renamable $vgpr14_vgpr15_vgpr16_vgpr17 = GLOBAL_LOAD_DWORDX4 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s128), addrspace 1)1282 ; CHECK-NEXT: renamable $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, 0, 0, 0, implicit $mode, implicit $exec1283 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1284 ; CHECK-NEXT: S_BRANCH %bb.21285 ; CHECK-NEXT: {{ $}}1286 ; CHECK-NEXT: bb.2:1287 ; CHECK-NEXT: liveins: $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33:0x00000000FFFFFFFF1288 ; CHECK-NEXT: {{ $}}1289 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = COPY killed renamable $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr171290 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71291 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151292 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231293 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311294 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391295 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471296 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551297 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631298 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1299 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1300 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1301 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1302 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1303 ; CHECK-NEXT: S_ENDPGM 01304 bb.0:1305 S_NOP 0, implicit-def $agpr01306 renamable $sgpr0 = S_MOV_B32 01307 undef %0.sub8:vreg_1024_align2 = V_MOV_B32_e32 0, implicit $exec1308 renamable $sgpr1 = COPY renamable $sgpr01309 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11310 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1311 %0.sub9:vreg_1024_align2 = COPY %0.sub81312 1313 bb.1:1314 liveins: $vcc1315 1316 %0.sub0_sub1_sub2_sub3:vreg_1024_align2 = GLOBAL_LOAD_DWORDX4 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s128), addrspace 1)1317 %0.sub4_sub5_sub6_sub7:vreg_1024_align2 = GLOBAL_LOAD_DWORDX4 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s128), addrspace 1)1318 %0.sub8_sub9_sub10_sub11:vreg_1024_align2 = GLOBAL_LOAD_DWORDX4 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s128), addrspace 1)1319 %0.sub12_sub13_sub14_sub15:vreg_1024_align2 = GLOBAL_LOAD_DWORDX4 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s128), addrspace 1)1320 %0.sub0_sub1_sub2_sub3_sub4_sub5_sub6_sub7_sub8_sub9_sub10_sub11_sub12_sub13_sub14_sub15:vreg_1024_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0.sub0_sub1_sub2_sub3_sub4_sub5_sub6_sub7_sub8_sub9_sub10_sub11_sub12_sub13_sub14_sub15, 0, 0, 0, implicit $mode, implicit $exec1321 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1322 S_BRANCH %bb.21323 1324 bb.2:1325 ; No VGPRs available for %01326 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71327 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151328 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231329 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311330 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391331 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471332 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551333 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631334 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1335 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1336 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1337 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1338 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1339 S_ENDPGM 01340 1341...1342 1343---1344name: chained_mfma_dst_user_is_vgpr1345tracksRegLiveness: true1346machineFunctionInfo:1347 isEntryFunction: true1348 stackPtrOffsetReg: '$sgpr32'1349 occupancy: 101350 sgprForEXECCopy: '$sgpr100_sgpr101'1351body: |1352 ; CHECK-LABEL: name: chained_mfma_dst_user_is_vgpr1353 ; CHECK: bb.0:1354 ; CHECK-NEXT: successors: %bb.1(0x80000000)1355 ; CHECK-NEXT: {{ $}}1356 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01357 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01358 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1359 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01360 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = COPY killed renamable $sgpr0_sgpr11361 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1362 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81363 ; CHECK-NEXT: {{ $}}1364 ; CHECK-NEXT: bb.1:1365 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1366 ; CHECK-NEXT: liveins: $vcc, $vgpr16_vgpr171367 ; CHECK-NEXT: {{ $}}1368 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1369 ; CHECK-NEXT: early-clobber renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 0, 0, 0, implicit $mode, implicit $exec1370 ; CHECK-NEXT: early-clobber renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33, 0, 0, 0, implicit $mode, implicit $exec1371 ; CHECK-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151372 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1373 ; CHECK-NEXT: S_BRANCH %bb.21374 ; CHECK-NEXT: {{ $}}1375 ; CHECK-NEXT: bb.2:1376 ; CHECK-NEXT: liveins: $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33:0x00000000FFFFFFFF1377 ; CHECK-NEXT: {{ $}}1378 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71379 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151380 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = COPY killed renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr331381 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231382 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311383 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391384 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471385 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551386 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631387 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1388 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1389 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1390 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1391 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1392 ; CHECK-NEXT: S_ENDPGM 01393 bb.0:1394 successors: %bb.1(0x80000000)1395 1396 S_NOP 0, implicit-def $agpr01397 renamable $sgpr0 = S_MOV_B32 01398 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1399 renamable $sgpr1 = COPY renamable $sgpr01400 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11401 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1402 %0.sub9:vreg_512_align2 = COPY %0.sub81403 1404 bb.1:1405 successors: %bb.1(0x40000000), %bb.2(0x40000000)1406 liveins: $vcc1407 1408 undef %2.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1409 early-clobber %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %2, 0, 0, 0, implicit $mode, implicit $exec1410 early-clobber %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1411 INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, %41412 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1413 S_BRANCH %bb.21414 1415 bb.2:1416 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71417 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151418 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231419 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311420 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391421 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471422 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551423 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631424 %6:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1425 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1426 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1427 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1428 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1429 S_ENDPGM 01430 1431...1432 1433# TODO: In this trivial case, the single copy required is cheaper than1434# the tuple copy.1435---1436name: chained_mfma_dst_user_is_vgpr_small_subreg1437tracksRegLiveness: true1438machineFunctionInfo:1439 isEntryFunction: true1440 stackPtrOffsetReg: '$sgpr32'1441 occupancy: 101442 sgprForEXECCopy: '$sgpr100_sgpr101'1443body: |1444 ; CHECK-LABEL: name: chained_mfma_dst_user_is_vgpr_small_subreg1445 ; CHECK: bb.0:1446 ; CHECK-NEXT: successors: %bb.1(0x80000000)1447 ; CHECK-NEXT: {{ $}}1448 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01449 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01450 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1451 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01452 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = COPY killed renamable $sgpr0_sgpr11453 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1454 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81455 ; CHECK-NEXT: {{ $}}1456 ; CHECK-NEXT: bb.1:1457 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1458 ; CHECK-NEXT: liveins: $vcc, $vgpr16_vgpr171459 ; CHECK-NEXT: {{ $}}1460 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1461 ; CHECK-NEXT: early-clobber renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 0, 0, 0, implicit $mode, implicit $exec1462 ; CHECK-NEXT: early-clobber renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33, 0, 0, 0, implicit $mode, implicit $exec1463 ; CHECK-NEXT: dead renamable $vgpr0 = nofpexcept V_CVT_F16_F32_e32 killed $vgpr0, implicit $mode, implicit $exec1464 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1465 ; CHECK-NEXT: S_BRANCH %bb.21466 ; CHECK-NEXT: {{ $}}1467 ; CHECK-NEXT: bb.2:1468 ; CHECK-NEXT: liveins: $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33:0x00000000FFFFFFFF1469 ; CHECK-NEXT: {{ $}}1470 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71471 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151472 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = COPY killed renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr331473 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231474 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311475 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391476 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471477 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551478 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631479 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1480 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1481 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1482 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1483 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1484 ; CHECK-NEXT: S_ENDPGM 01485 bb.0:1486 successors: %bb.1(0x80000000)1487 1488 S_NOP 0, implicit-def $agpr01489 renamable $sgpr0 = S_MOV_B32 01490 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1491 renamable $sgpr1 = COPY renamable $sgpr01492 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11493 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1494 %0.sub9:vreg_512_align2 = COPY %0.sub81495 1496 bb.1:1497 successors: %bb.1(0x40000000), %bb.2(0x40000000)1498 liveins: $vcc1499 1500 undef %2.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1501 early-clobber %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %2, 0, 0, 0, implicit $mode, implicit $exec1502 early-clobber %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1503 %5:vgpr_32 = nofpexcept V_CVT_F16_F32_e32 %4.sub0, implicit $mode, implicit $exec1504 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1505 S_BRANCH %bb.21506 1507 bb.2:1508 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71509 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151510 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231511 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311512 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391513 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471514 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551515 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631516 %6:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1517 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1518 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1519 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1520 GLOBAL_STORE_DWORDX4_SADDR %6, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1521 S_ENDPGM 01522 1523...1524 1525# Transitive user of the register is an MFMA with non-register src21526---1527name: chained_mfma_dst_user_has_imm_src21528tracksRegLiveness: true1529machineFunctionInfo:1530 isEntryFunction: true1531 stackPtrOffsetReg: '$sgpr32'1532 occupancy: 101533 sgprForEXECCopy: '$sgpr100_sgpr101'1534body: |1535 ; CHECK-LABEL: name: chained_mfma_dst_user_has_imm_src21536 ; CHECK: bb.0:1537 ; CHECK-NEXT: successors: %bb.1(0x80000000)1538 ; CHECK-NEXT: {{ $}}1539 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01540 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01541 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1542 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01543 ; CHECK-NEXT: renamable $vgpr18_vgpr19 = COPY killed renamable $sgpr0_sgpr11544 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1545 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81546 ; CHECK-NEXT: {{ $}}1547 ; CHECK-NEXT: bb.1:1548 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1549 ; CHECK-NEXT: liveins: $vcc, $vgpr18_vgpr191550 ; CHECK-NEXT: {{ $}}1551 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1552 ; CHECK-NEXT: early-clobber renamable $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31 = V_MFMA_F32_32X32X8F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec1553 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr18_vgpr19, $vgpr18_vgpr19, killed $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1554 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X4F16_e64 $agpr0_agpr1, $vgpr18_vgpr19, 0, 0, 0, 0, implicit $mode, implicit $exec1555 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1556 ; CHECK-NEXT: S_BRANCH %bb.21557 ; CHECK-NEXT: {{ $}}1558 ; CHECK-NEXT: bb.2:1559 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF1560 ; CHECK-NEXT: {{ $}}1561 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71562 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151563 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231564 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311565 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391566 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471567 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551568 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631569 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1570 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1571 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1572 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1573 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1574 ; CHECK-NEXT: S_ENDPGM 01575 bb.0:1576 successors: %bb.1(0x80000000)1577 1578 S_NOP 0, implicit-def $agpr01579 renamable $sgpr0 = S_MOV_B32 01580 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1581 renamable $sgpr1 = COPY renamable $sgpr01582 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11583 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1584 %0.sub9:vreg_512_align2 = COPY %0.sub81585 1586 bb.1:1587 successors: %bb.1(0x40000000), %bb.2(0x40000000)1588 liveins: $vcc1589 1590 undef %2.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1591 early-clobber %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %2, 0, 0, 0, implicit $mode, implicit $exec1592 early-clobber %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1593 %4.sub0_sub1_sub2_sub3:vreg_512_align2 = V_MFMA_F32_4X4X4F16_vgprcd_e64 %4.sub0_sub1, %1, 0, 0, 0, 0, implicit $mode, implicit $exec1594 1595 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1596 S_BRANCH %bb.21597 1598 bb.2:1599 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71600 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151601 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231602 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311603 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391604 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471605 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551606 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631607 %6:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1608 GLOBAL_STORE_DWORDX4_SADDR %6, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1609 GLOBAL_STORE_DWORDX4_SADDR %6, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1610 GLOBAL_STORE_DWORDX4_SADDR %6, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1611 GLOBAL_STORE_DWORDX4_SADDR %6, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1612 S_ENDPGM 01613 1614...1615 1616---1617name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_has_untied_user1618tracksRegLiveness: true1619machineFunctionInfo:1620 isEntryFunction: true1621 stackPtrOffsetReg: '$sgpr32'1622 occupancy: 101623 sgprForEXECCopy: '$sgpr100_sgpr101'1624body: |1625 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_has_untied_user1626 ; CHECK: bb.0:1627 ; CHECK-NEXT: successors: %bb.1(0x80000000)1628 ; CHECK-NEXT: {{ $}}1629 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01630 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01631 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1632 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01633 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = COPY killed renamable $sgpr0_sgpr11634 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1635 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81636 ; CHECK-NEXT: {{ $}}1637 ; CHECK-NEXT: bb.1:1638 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1639 ; CHECK-NEXT: liveins: $vcc, $vgpr16_vgpr171640 ; CHECK-NEXT: {{ $}}1641 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1642 ; CHECK-NEXT: renamable $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1643 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, killed $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1644 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1645 ; CHECK-NEXT: S_BRANCH %bb.21646 ; CHECK-NEXT: {{ $}}1647 ; CHECK-NEXT: bb.2:1648 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF1649 ; CHECK-NEXT: {{ $}}1650 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71651 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151652 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231653 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311654 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391655 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471656 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551657 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631658 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1659 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1660 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1661 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1662 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1663 ; CHECK-NEXT: S_ENDPGM 01664 bb.0:1665 S_NOP 0, implicit-def $agpr01666 renamable $sgpr0 = S_MOV_B32 01667 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1668 renamable $sgpr1 = COPY renamable $sgpr01669 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11670 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1671 %0.sub9:vreg_512_align2 = COPY %0.sub81672 1673 bb.1:1674 liveins: $vcc1675 1676 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1677 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1678 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1679 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1680 S_BRANCH %bb.21681 1682 bb.2:1683 ; No VGPRs available for %01684 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71685 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151686 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231687 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311688 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391689 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471690 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551691 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631692 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1693 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1694 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1695 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1696 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1697 S_ENDPGM 01698 1699...1700 1701---1702name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_has_untied_user_with_vgpr_use1703tracksRegLiveness: true1704machineFunctionInfo:1705 isEntryFunction: true1706 stackPtrOffsetReg: '$sgpr32'1707 occupancy: 101708 sgprForEXECCopy: '$sgpr100_sgpr101'1709body: |1710 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_mac_vgprcd_e64_has_untied_user_with_vgpr_use1711 ; CHECK: bb.0:1712 ; CHECK-NEXT: successors: %bb.1(0x80000000)1713 ; CHECK-NEXT: {{ $}}1714 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01715 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01716 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1717 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01718 ; CHECK-NEXT: renamable $vgpr16_vgpr17 = COPY killed renamable $sgpr0_sgpr11719 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1720 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81721 ; CHECK-NEXT: {{ $}}1722 ; CHECK-NEXT: bb.1:1723 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1724 ; CHECK-NEXT: liveins: $vcc, $vgpr16_vgpr171725 ; CHECK-NEXT: {{ $}}1726 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1727 ; CHECK-NEXT: renamable $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 0, 0, 0, implicit $mode, implicit $exec1728 ; CHECK-NEXT: early-clobber renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33 = V_MFMA_F32_32X32X8F16_vgprcd_e64 $vgpr16_vgpr17, $vgpr16_vgpr17, killed $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, 0, 0, 0, implicit $mode, implicit $exec1729 ; CHECK-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr331730 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1731 ; CHECK-NEXT: S_BRANCH %bb.21732 ; CHECK-NEXT: {{ $}}1733 ; CHECK-NEXT: bb.2:1734 ; CHECK-NEXT: liveins: $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33:0x00000000FFFFFFFF1735 ; CHECK-NEXT: {{ $}}1736 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71737 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151738 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = COPY killed renamable $vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr331739 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231740 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311741 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391742 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471743 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551744 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631745 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1746 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1747 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1748 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1749 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1750 ; CHECK-NEXT: S_ENDPGM 01751 bb.0:1752 S_NOP 0, implicit-def $agpr01753 renamable $sgpr0 = S_MOV_B32 01754 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1755 renamable $sgpr1 = COPY renamable $sgpr01756 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11757 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1758 %0.sub9:vreg_512_align2 = COPY %0.sub81759 1760 bb.1:1761 liveins: $vcc1762 1763 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1764 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1765 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1766 INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 39321609 /* reguse:VReg_512_Align2 */, %41767 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1768 S_BRANCH %bb.21769 1770 bb.2:1771 ; No VGPRs available for %01772 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71773 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151774 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231775 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311776 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391777 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471778 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551779 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631780 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1781 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1782 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1783 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1784 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1785 S_ENDPGM 01786 1787...1788 1789---1790name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_has_tied_user1791tracksRegLiveness: true1792machineFunctionInfo:1793 isEntryFunction: true1794 stackPtrOffsetReg: '$sgpr32'1795 occupancy: 101796 sgprForEXECCopy: '$sgpr100_sgpr101'1797body: |1798 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_has_tied_user1799 ; CHECK: bb.0:1800 ; CHECK-NEXT: successors: %bb.1(0x80000000)1801 ; CHECK-NEXT: {{ $}}1802 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01803 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01804 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1805 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01806 ; CHECK-NEXT: renamable $vgpr2_vgpr3 = COPY killed renamable $sgpr0_sgpr11807 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1808 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81809 ; CHECK-NEXT: {{ $}}1810 ; CHECK-NEXT: bb.1:1811 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1812 ; CHECK-NEXT: liveins: $vcc, $vgpr2_vgpr31813 ; CHECK-NEXT: {{ $}}1814 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1815 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr2_vgpr3, $vgpr2_vgpr3, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1816 ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_mac_e64 $vgpr2_vgpr3, $vgpr2_vgpr3, killed $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15, 0, 0, 0, implicit $mode, implicit $exec1817 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1818 ; CHECK-NEXT: S_BRANCH %bb.21819 ; CHECK-NEXT: {{ $}}1820 ; CHECK-NEXT: bb.2:1821 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF1822 ; CHECK-NEXT: {{ $}}1823 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71824 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151825 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231826 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311827 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391828 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471829 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551830 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631831 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1832 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1833 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1834 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1835 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1836 ; CHECK-NEXT: S_ENDPGM 01837 bb.0:1838 S_NOP 0, implicit-def $agpr01839 renamable $sgpr0 = S_MOV_B32 01840 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1841 renamable $sgpr1 = COPY renamable $sgpr01842 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11843 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1844 %0.sub9:vreg_512_align2 = COPY %0.sub81845 1846 bb.1:1847 liveins: $vcc1848 1849 undef %0.sub0_sub1:vreg_512_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)1850 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0, 0, 0, 0, implicit $mode, implicit $exec1851 %4:vreg_512_align2 = V_MFMA_F32_32X32X8F16_mac_vgprcd_e64 %1, %1, %4, 0, 0, 0, implicit $mode, implicit $exec1852 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1853 S_BRANCH %bb.21854 1855 bb.2:1856 ; No VGPRs available for %01857 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71858 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151859 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231860 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311861 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391862 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471863 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551864 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631865 %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1866 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1867 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1868 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1869 GLOBAL_STORE_DWORDX4_SADDR %5, %4.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1870 S_ENDPGM 01871 1872...1873 1874# Non-mac variant, src2 is an immediate.1875---1876name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_imm_src21877tracksRegLiveness: true1878machineFunctionInfo:1879 isEntryFunction: true1880 stackPtrOffsetReg: '$sgpr32'1881 occupancy: 101882 sgprForEXECCopy: '$sgpr100_sgpr101'1883body: |1884 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_imm_src21885 ; CHECK: bb.0:1886 ; CHECK-NEXT: successors: %bb.1(0x80000000)1887 ; CHECK-NEXT: {{ $}}1888 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01889 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01890 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1891 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01892 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11893 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1894 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81895 ; CHECK-NEXT: {{ $}}1896 ; CHECK-NEXT: bb.1:1897 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1898 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr11899 ; CHECK-NEXT: {{ $}}1900 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, 0, 0, 0, 0, implicit $mode, implicit $exec1901 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1902 ; CHECK-NEXT: S_BRANCH %bb.21903 ; CHECK-NEXT: {{ $}}1904 ; CHECK-NEXT: bb.2:1905 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15:0x00000000FFFFFFFF1906 ; CHECK-NEXT: {{ $}}1907 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71908 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151909 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231910 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311911 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391912 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471913 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551914 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631915 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec1916 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1917 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1918 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1919 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1920 ; CHECK-NEXT: S_ENDPGM 01921 bb.0:1922 S_NOP 0, implicit-def $agpr01923 renamable $sgpr0 = S_MOV_B32 01924 undef %0.sub8:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec1925 renamable $sgpr1 = COPY renamable $sgpr01926 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr11927 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1928 %0.sub9:vreg_512_align2 = COPY %0.sub81929 1930 bb.1:1931 liveins: $vcc1932 1933 %0:vreg_512_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, 0, 0, 0, 0, implicit $mode, implicit $exec1934 S_CBRANCH_VCCNZ %bb.1, implicit $vcc1935 S_BRANCH %bb.21936 1937 bb.2:1938 ; No VGPRs available for %01939 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71940 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151941 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231942 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311943 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391944 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471945 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551946 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631947 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec1948 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)1949 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)1950 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)1951 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)1952 S_ENDPGM 01953 1954...1955 1956# Non-mac variant, src2 is the same VGPR, but a different subregister.1957---1958name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_src2_different_subreg1959tracksRegLiveness: true1960machineFunctionInfo:1961 isEntryFunction: true1962 stackPtrOffsetReg: '$sgpr32'1963 occupancy: 101964 sgprForEXECCopy: '$sgpr100_sgpr101'1965body: |1966 ; CHECK-LABEL: name: inflate_result_to_agpr__V_MFMA_F32_32X32X8F16_vgprcd_e64_src2_different_subreg1967 ; CHECK: bb.0:1968 ; CHECK-NEXT: successors: %bb.1(0x80000000)1969 ; CHECK-NEXT: {{ $}}1970 ; CHECK-NEXT: S_NOP 0, implicit-def $agpr01971 ; CHECK-NEXT: renamable $sgpr0 = S_MOV_B32 01972 ; CHECK-NEXT: renamable $vgpr8 = V_MOV_B32_e32 0, implicit $exec1973 ; CHECK-NEXT: renamable $sgpr1 = COPY renamable $sgpr01974 ; CHECK-NEXT: renamable $vgpr0_vgpr1 = COPY killed renamable $sgpr0_sgpr11975 ; CHECK-NEXT: renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc1976 ; CHECK-NEXT: dead renamable $vgpr9 = COPY renamable $vgpr81977 ; CHECK-NEXT: {{ $}}1978 ; CHECK-NEXT: bb.1:1979 ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)1980 ; CHECK-NEXT: liveins: $vcc, $vgpr0_vgpr11981 ; CHECK-NEXT: {{ $}}1982 ; CHECK-NEXT: renamable $agpr0_agpr1 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1983 ; CHECK-NEXT: renamable $agpr16_agpr17 = GLOBAL_LOAD_DWORDX2 undef renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s64), addrspace 1)1984 ; CHECK-NEXT: early-clobber renamable $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15 = V_MFMA_F32_32X32X8F16_e64 $vgpr0_vgpr1, $vgpr0_vgpr1, $agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31, 0, 0, 0, implicit $mode, implicit $exec1985 ; CHECK-NEXT: S_CBRANCH_VCCNZ %bb.1, implicit $vcc1986 ; CHECK-NEXT: S_BRANCH %bb.21987 ; CHECK-NEXT: {{ $}}1988 ; CHECK-NEXT: bb.2:1989 ; CHECK-NEXT: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7_agpr8_agpr9_agpr10_agpr11_agpr12_agpr13_agpr14_agpr15_agpr16_agpr17_agpr18_agpr19_agpr20_agpr21_agpr22_agpr23_agpr24_agpr25_agpr26_agpr27_agpr28_agpr29_agpr30_agpr31:0x00000000FFFFFFFF1990 ; CHECK-NEXT: {{ $}}1991 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr71992 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr151993 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr231994 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr311995 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr391996 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr471997 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr551998 ; CHECK-NEXT: S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr631999 ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 0, implicit $exec2000 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr8_agpr9_agpr10_agpr11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)2001 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr12_agpr13_agpr14_agpr15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)2002 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR renamable $vgpr0, renamable $agpr0_agpr1_agpr2_agpr3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)2003 ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr0, killed renamable $agpr4_agpr5_agpr6_agpr7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)2004 ; CHECK-NEXT: S_ENDPGM 02005 bb.0:2006 S_NOP 0, implicit-def $agpr02007 renamable $sgpr0 = S_MOV_B32 02008 undef %0.sub8:vreg_1024_align2 = V_MOV_B32_e32 0, implicit $exec2009 renamable $sgpr1 = COPY renamable $sgpr02010 %1:vreg_64_align2 = COPY killed renamable $sgpr0_sgpr12011 renamable $vcc = S_AND_B64 $exec, -1, implicit-def dead $scc2012 %0.sub9:vreg_1024_align2 = COPY %0.sub82013 2014 bb.1:2015 liveins: $vcc2016 2017 undef %0.sub0_sub1:vreg_1024_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)2018 %0.sub16_sub17:vreg_1024_align2 = GLOBAL_LOAD_DWORDX2 undef %3:vreg_64_align2, 0, 0, implicit $exec :: (load (s64), addrspace 1)2019 %0.sub0_sub1_sub2_sub3_sub4_sub5_sub6_sub7_sub8_sub9_sub10_sub11_sub12_sub13_sub14_sub15:vreg_1024_align2 = V_MFMA_F32_32X32X8F16_vgprcd_e64 %1, %1, %0.sub16_sub17_sub18_sub19_sub20_sub21_sub22_sub23_sub24_sub25_sub26_sub27_sub28_sub29_sub30_sub31, 0, 0, 0, implicit $mode, implicit $exec2020 S_CBRANCH_VCCNZ %bb.1, implicit $vcc2021 S_BRANCH %bb.22022 2023 bb.2:2024 ; No VGPRs available for %02025 S_NOP 0, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr72026 S_NOP 0, implicit-def $vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr152027 S_NOP 0, implicit-def $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr232028 S_NOP 0, implicit-def $vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr312029 S_NOP 0, implicit-def $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr392030 S_NOP 0, implicit-def $vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr472031 S_NOP 0, implicit-def $vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr552032 S_NOP 0, implicit-def $vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr632033 %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec2034 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub8_sub9_sub10_sub11, undef $sgpr0_sgpr1, 32, 0, implicit $exec :: (store (s128), align 32, addrspace 1)2035 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub12_sub13_sub14_sub15, undef $sgpr0_sgpr1, 48, 0, implicit $exec :: (store (s128), addrspace 1)2036 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub0_sub1_sub2_sub3, undef $sgpr0_sgpr1, 0, 0, implicit $exec :: (store (s128), align 128, addrspace 1)2037 GLOBAL_STORE_DWORDX4_SADDR %2, %0.sub4_sub5_sub6_sub7, killed undef $sgpr0_sgpr1, 16, 0, implicit $exec :: (store (s128), addrspace 1)2038 S_ENDPGM 02039 2040...2041