[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes. #76500

kosarev · 2023-12-28T12:40:34Z

No description provided.

llvmbot · 2023-12-28T12:41:04Z

@llvm/pr-subscribers-backend-amdgpu

@llvm/pr-subscribers-llvm-globalisel

Author: Ivan Kosarev (kosarev)

Changes

Patch is 34.74 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/76500.diff

10 Files Affected:

(modified) llvm/lib/Target/AMDGPU/SIRegisterInfo.td (-14)
(modified) llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-inline-asm.ll (+16-16)
(modified) llvm/test/CodeGen/AMDGPU/branch-relax-indirect-branch.mir (+1-1)
(modified) llvm/test/CodeGen/AMDGPU/branch-relax-no-terminators.mir (+1-1)
(modified) llvm/test/CodeGen/AMDGPU/coalescer-early-clobber-subreg.mir (+4-4)
(modified) llvm/test/CodeGen/AMDGPU/inline-asm.i128.ll (+12-12)
(modified) llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll (+12-12)
(modified) llvm/test/CodeGen/AMDGPU/peephole-fold-imm.mir (+2-2)
(modified) llvm/test/CodeGen/AMDGPU/spill-vector-superclass.ll (+2-2)
(modified) llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/amdgpu_isel.ll.expected (+2-2)

diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
index 981da13fe08952..36480057b81949 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
@@ -588,20 +588,6 @@ def Reg16Types : RegisterTypes<[i16, f16]>;
 def Reg32Types : RegisterTypes<[i32, f32, v2i16, v2f16, p2, p3, p5, p6]>;
 
 let HasVGPR = 1 in {
-def VGPR_LO16 : SIRegisterClass<"AMDGPU", Reg16Types.types, 16,
-                              (add (sequence "VGPR%u_LO16", 0, 255))> {
-  let AllocationPriority = 0;
-  let Size = 16;
-  let GeneratePressureSet = 0;
-}
-
-def VGPR_HI16 : SIRegisterClass<"AMDGPU", Reg16Types.types, 16,
-                              (add (sequence "VGPR%u_HI16", 0, 255))> {
-  let AllocationPriority = 0;
-  let Size = 16;
-  let GeneratePressureSet = 0;
-}
-
 // VOP3 and VINTERP can access 256 lo and 256 hi registers.
 def VGPR_16 : SIRegisterClass<"AMDGPU",  Reg16Types.types, 16,
                             (add (interleave (sequence "VGPR%u_LO16", 0, 255),
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-inline-asm.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-inline-asm.ll
index 0f85c80954aca3..bb37e54e3b5665 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-inline-asm.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-inline-asm.ll
@@ -66,7 +66,7 @@ define amdgpu_kernel void @asm_simple_agpr_clobber() {
 define i32 @asm_vgpr_early_clobber() {
   ; CHECK-LABEL: name: asm_vgpr_early_clobber
   ; CHECK: bb.1 (%ir-block.0):
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 7; v_mov_b32 $1, 7", 1 /* sideeffect attdialect */, 2490379 /* regdef-ec:VGPR_32 */, def early-clobber %8, 2490379 /* regdef-ec:VGPR_32 */, def early-clobber %9, !0
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 7; v_mov_b32 $1, 7", 1 /* sideeffect attdialect */, 2228235 /* regdef-ec:VGPR_32 */, def early-clobber %8, 2228235 /* regdef-ec:VGPR_32 */, def early-clobber %9, !0
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY %9
   ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[COPY]], [[COPY1]]
@@ -94,7 +94,7 @@ entry:
 define i32 @test_single_vgpr_output() nounwind {
   ; CHECK-LABEL: name: test_single_vgpr_output
   ; CHECK: bb.1.entry:
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 7", 0 /* attdialect */, 2490378 /* regdef:VGPR_32 */, def %8
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 7", 0 /* attdialect */, 2228234 /* regdef:VGPR_32 */, def %8
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -106,7 +106,7 @@ entry:
 define i32 @test_single_sgpr_output_s32() nounwind {
   ; CHECK-LABEL: name: test_single_sgpr_output_s32
   ; CHECK: bb.1.entry:
-  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2621450 /* regdef:SReg_32 */, def %8
+  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2359306 /* regdef:SReg_32 */, def %8
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -119,7 +119,7 @@ entry:
 define float @test_multiple_register_outputs_same() #0 {
   ; CHECK-LABEL: name: test_multiple_register_outputs_same
   ; CHECK: bb.1 (%ir-block.0):
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 0; v_mov_b32 $1, 1", 0 /* attdialect */, 2490378 /* regdef:VGPR_32 */, def %8, 2490378 /* regdef:VGPR_32 */, def %9
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 0; v_mov_b32 $1, 1", 0 /* attdialect */, 2228234 /* regdef:VGPR_32 */, def %8, 2228234 /* regdef:VGPR_32 */, def %9
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY %9
   ; CHECK-NEXT:   [[FADD:%[0-9]+]]:_(s32) = G_FADD [[COPY]], [[COPY1]]
@@ -136,7 +136,7 @@ define float @test_multiple_register_outputs_same() #0 {
 define double @test_multiple_register_outputs_mixed() #0 {
   ; CHECK-LABEL: name: test_multiple_register_outputs_mixed
   ; CHECK: bb.1 (%ir-block.0):
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 0; v_add_f64 $1, 0, 0", 0 /* attdialect */, 2490378 /* regdef:VGPR_32 */, def %8, 3801098 /* regdef:VReg_64 */, def %9
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, 0; v_add_f64 $1, 0, 0", 0 /* attdialect */, 2228234 /* regdef:VGPR_32 */, def %8, 3538954 /* regdef:VReg_64 */, def %9
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s64) = COPY %9
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64)
@@ -171,7 +171,7 @@ define amdgpu_kernel void @test_input_vgpr_imm() {
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[C]](s32)
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 v0, $0", 1 /* sideeffect attdialect */, 2490377 /* reguse:VGPR_32 */, [[COPY1]]
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 v0, $0", 1 /* sideeffect attdialect */, 2228233 /* reguse:VGPR_32 */, [[COPY1]]
   ; CHECK-NEXT:   S_ENDPGM 0
   call void asm sideeffect "v_mov_b32 v0, $0", "v"(i32 42)
   ret void
@@ -185,7 +185,7 @@ define amdgpu_kernel void @test_input_sgpr_imm() {
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY [[C]](s32)
-  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 s0, $0", 1 /* sideeffect attdialect */, 2621449 /* reguse:SReg_32 */, [[COPY1]]
+  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 s0, $0", 1 /* sideeffect attdialect */, 2359305 /* reguse:SReg_32 */, [[COPY1]]
   ; CHECK-NEXT:   S_ENDPGM 0
   call void asm sideeffect "s_mov_b32 s0, $0", "s"(i32 42)
   ret void
@@ -212,7 +212,7 @@ define float @test_input_vgpr(i32 %src) nounwind {
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
-  ; CHECK-NEXT:   INLINEASM &"v_add_f32 $0, 1.0, $1", 0 /* attdialect */, 2490378 /* regdef:VGPR_32 */, def %9, 2490377 /* reguse:VGPR_32 */, [[COPY1]]
+  ; CHECK-NEXT:   INLINEASM &"v_add_f32 $0, 1.0, $1", 0 /* attdialect */, 2228234 /* regdef:VGPR_32 */, def %9, 2228233 /* reguse:VGPR_32 */, [[COPY1]]
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY %9
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY2]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -227,7 +227,7 @@ define i32 @test_memory_constraint(ptr addrspace(3) %a) nounwind {
   ; CHECK-NEXT:   liveins: $vgpr0
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
-  ; CHECK-NEXT:   INLINEASM &"ds_read_b32 $0, $1", 8 /* mayload attdialect */, 2490378 /* regdef:VGPR_32 */, def %9, 262158 /* mem:m */, [[COPY]](p3)
+  ; CHECK-NEXT:   INLINEASM &"ds_read_b32 $0, $1", 8 /* mayload attdialect */, 2228234 /* regdef:VGPR_32 */, def %9, 262158 /* mem:m */, [[COPY]](p3)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY %9
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -244,7 +244,7 @@ define i32 @test_vgpr_matching_constraint(i32 %a) nounwind {
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
   ; CHECK-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY]], [[C]]
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[AND]](s32)
-  ; CHECK-NEXT:   INLINEASM &";", 1 /* sideeffect attdialect */, 2490378 /* regdef:VGPR_32 */, def %11, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
+  ; CHECK-NEXT:   INLINEASM &";", 1 /* sideeffect attdialect */, 2228234 /* regdef:VGPR_32 */, def %11, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY %11
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY2]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -256,13 +256,13 @@ define i32 @test_vgpr_matching_constraint(i32 %a) nounwind {
 define i32 @test_sgpr_matching_constraint() nounwind {
   ; CHECK-LABEL: name: test_sgpr_matching_constraint
   ; CHECK: bb.1.entry:
-  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2621450 /* regdef:SReg_32 */, def %8
+  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2359306 /* regdef:SReg_32 */, def %8
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
-  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 8", 0 /* attdialect */, 2621450 /* regdef:SReg_32 */, def %10
+  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 8", 0 /* attdialect */, 2359306 /* regdef:SReg_32 */, def %10
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY %10
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY [[COPY1]](s32)
-  ; CHECK-NEXT:   INLINEASM &"s_add_u32 $0, $1, $2", 0 /* attdialect */, 2621450 /* regdef:SReg_32 */, def %12, 2621449 /* reguse:SReg_32 */, [[COPY2]], 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3)
+  ; CHECK-NEXT:   INLINEASM &"s_add_u32 $0, $1, $2", 0 /* attdialect */, 2359306 /* regdef:SReg_32 */, def %12, 2359305 /* reguse:SReg_32 */, [[COPY2]], 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3)
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY %12
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY4]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
@@ -285,7 +285,7 @@ define void @test_many_matching_constraints(i32 %a, i32 %b, i32 %c) nounwind {
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY2]](s32)
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[COPY1]](s32)
-  ; CHECK-NEXT:   INLINEASM &"; ", 1 /* sideeffect attdialect */, 2490378 /* regdef:VGPR_32 */, def %11, 2490378 /* regdef:VGPR_32 */, def %12, 2490378 /* regdef:VGPR_32 */, def %13, 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3), 2147614729 /* reguse tiedto:$2 */, [[COPY4]](tied-def 7), 2147549193 /* reguse tiedto:$1 */, [[COPY5]](tied-def 5)
+  ; CHECK-NEXT:   INLINEASM &"; ", 1 /* sideeffect attdialect */, 2228234 /* regdef:VGPR_32 */, def %11, 2228234 /* regdef:VGPR_32 */, def %12, 2228234 /* regdef:VGPR_32 */, def %13, 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3), 2147614729 /* reguse tiedto:$2 */, [[COPY4]](tied-def 7), 2147549193 /* reguse tiedto:$1 */, [[COPY5]](tied-def 5)
   ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY %11
   ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY %12
   ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY %13
@@ -306,10 +306,10 @@ define void @test_many_matching_constraints(i32 %a, i32 %b, i32 %c) nounwind {
 define i32 @test_sgpr_to_vgpr_move_matching_constraint() nounwind {
   ; CHECK-LABEL: name: test_sgpr_to_vgpr_move_matching_constraint
   ; CHECK: bb.1.entry:
-  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2621450 /* regdef:SReg_32 */, def %8
+  ; CHECK-NEXT:   INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 2359306 /* regdef:SReg_32 */, def %8
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY %8
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
-  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, $1", 0 /* attdialect */, 2490378 /* regdef:VGPR_32 */, def %10, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
+  ; CHECK-NEXT:   INLINEASM &"v_mov_b32 $0, $1", 0 /* attdialect */, 2228234 /* regdef:VGPR_32 */, def %10, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY %10
   ; CHECK-NEXT:   $vgpr0 = COPY [[COPY2]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0
diff --git a/llvm/test/CodeGen/AMDGPU/branch-relax-indirect-branch.mir b/llvm/test/CodeGen/AMDGPU/branch-relax-indirect-branch.mir
index 1aafcdcbdfa0cb..253e7e278aaff5 100644
--- a/llvm/test/CodeGen/AMDGPU/branch-relax-indirect-branch.mir
+++ b/llvm/test/CodeGen/AMDGPU/branch-relax-indirect-branch.mir
@@ -68,7 +68,7 @@ body:             |
   ; CHECK-NEXT:   successors: %bb.3(0x04000000), %bb.7(0x7c000000)
   ; CHECK-NEXT:   liveins: $vcc_hi, $vcc_lo, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr18, $sgpr19, $sgpr20, $sgpr21, $sgpr22, $sgpr23, $sgpr24, $sgpr25, $sgpr26, $sgpr27, $sgpr28, $sgpr29, $sgpr30, $sgpr31, $sgpr34, $sgpr35, $sgpr36, $sgpr37, $sgpr38, $sgpr39, $sgpr40, $sgpr41, $sgpr42, $sgpr43, $sgpr44, $sgpr45, $sgpr46, $sgpr47, $sgpr48, $sgpr49, $sgpr50, $sgpr51, $sgpr52, $sgpr53, $sgpr54, $sgpr55, $sgpr56, $sgpr57, $sgpr58, $sgpr59, $sgpr60, $sgpr61, $sgpr62, $sgpr63, $sgpr64, $sgpr65, $sgpr66, $sgpr67, $sgpr68, $sgpr69, $sgpr70, $sgpr71, $sgpr72, $sgpr73, $sgpr74, $sgpr75, $sgpr76, $sgpr77, $sgpr78, $sgpr79, $sgpr80, $sgpr81, $sgpr82, $sgpr83, $sgpr84, $sgpr85, $sgpr86, $sgpr87, $sgpr88, $sgpr89, $sgpr90, $sgpr91, $sgpr92, $sgpr93, $sgpr94, $sgpr95, $sgpr96, $sgpr97, $sgpr98, $sgpr99, $sgpr100, $sgpr101, $vgpr0, $vgpr1
   ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT:   INLINEASM &"v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64", 1 /* sideeffect attdialect */, 2097162 /* regdef:VS_32 */, def renamable $sgpr4
+  ; CHECK-NEXT:   INLINEASM &"v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64", 1 /* sideeffect attdialect */, 2097162 /* regdef:VRegOrLds_32 */, def renamable $sgpr4
   ; CHECK-NEXT:   S_CMP_LG_U32 killed renamable $sgpr4, 0, implicit-def $scc
   ; CHECK-NEXT:   S_CBRANCH_SCC0 %bb.3, implicit killed $scc
   ; CHECK-NEXT: {{  $}}
diff --git a/llvm/test/CodeGen/AMDGPU/branch-relax-no-terminators.mir b/llvm/test/CodeGen/AMDGPU/branch-relax-no-terminators.mir
index 15879eb1e2fbd0..e4d9fbfb170575 100644
--- a/llvm/test/CodeGen/AMDGPU/branch-relax-no-terminators.mir
+++ b/llvm/test/CodeGen/AMDGPU/branch-relax-no-terminators.mir
@@ -68,7 +68,7 @@ body:             |
   ; CHECK-NEXT:   successors: %bb.3(0x04000000), %bb.7(0x7c000000)
   ; CHECK-NEXT:   liveins: $vcc_hi, $vcc_lo, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr18, $sgpr19, $sgpr20, $sgpr21, $sgpr22, $sgpr23, $sgpr24, $sgpr25, $sgpr26, $sgpr27, $sgpr28, $sgpr29, $sgpr30, $sgpr31, $sgpr34, $sgpr35, $sgpr36, $sgpr37, $sgpr38, $sgpr39, $sgpr40, $sgpr41, $sgpr42, $sgpr43, $sgpr44, $sgpr45, $sgpr46, $sgpr47, $sgpr48, $sgpr49, $sgpr50, $sgpr51, $sgpr52, $sgpr53, $sgpr54, $sgpr55, $sgpr56, $sgpr57, $sgpr58, $sgpr59, $sgpr60, $sgpr61, $sgpr62, $sgpr63, $sgpr64, $sgpr65, $sgpr66, $sgpr67, $sgpr68, $sgpr69, $sgpr70, $sgpr71, $sgpr72, $sgpr73, $sgpr74, $sgpr75, $sgpr76, $sgpr77, $sgpr78, $sgpr79, $sgpr80, $sgpr81, $sgpr82, $sgpr83, $sgpr84, $sgpr85, $sgpr86, $sgpr87, $sgpr88, $sgpr89, $sgpr90, $sgpr91, $sgpr92, $sgpr93, $sgpr94, $sgpr95, $sgpr96, $sgpr97, $sgpr98, $sgpr99, $sgpr100, $sgpr101, $vgpr0, $vgpr1
   ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT:   INLINEASM &"v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64", 1 /* sideeffect attdialect */, 2097162 /* regdef:VS_32 */, def renamable $sgpr4
+  ; CHECK-NEXT:   INLINEASM &"v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64\0A    v_nop_e64", 1 /* sideeffect attdialect */, 2097162 /* regdef:VRegOrLds_32 */, def renamable $sgpr4
   ; CHECK-NEXT:   S_CMP_LG_U32 killed renamable $sgpr4, 0, implicit-def $scc
   ; CHECK-NEXT:   S_CBRANCH_SCC0 %bb.3, implicit killed $scc
   ; CHECK-NEXT: {{  $}}
diff --git a/llvm/test/CodeGen/AMDGPU/coalescer-early-clobber-subreg.mir b/llvm/test/CodeGen/AMDGPU/coalescer-early-clobber-subreg.mir
index 45c8a98f3dcbc3..968cf21bd6d567 100644
--- a/llvm/test/CodeGen/AMDGPU/coalescer-early-clobber-subreg.mir
+++ b/llvm/test/CodeGen/AMDGPU/coalescer-early-clobber-subreg.mir
@@ -20,7 +20,7 @@ body:             |
     ; CHECK-LABEL: name: foo1
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef %2.sub0, 1835019 /* regdef-ec:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef early-clobber %2.sub1
+    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VS_32 */, def undef %2.sub0, 1835019 /* regdef-ec:VS_32 */, def undef early-clobber %2.sub1
     ; CHECK-NEXT: FLAT_STORE_DWORDX2 $vgpr0_vgpr1, %2, 0, 0, implicit $exec, implicit $flat_scr :: (store (s64))
     ; CHECK-NEXT: S_ENDPGM 0
     INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VGPR_32 */, def %0:vgpr_32, 1835019 /* regdef-ec:VGPR_32 */, def early-clobber %1:vgpr_32
@@ -41,7 +41,7 @@ body:             |
     ; CHECK-LABEL: name: foo2
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef early-clobber %2.sub1, 1835018 /* regdef:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef %2.sub0
+    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VS_32 */, def undef early-clobber %2.sub1, 1835018 /* regdef:VS_32 */, def undef %2.sub0
     ; CHECK-NEXT: FLAT_STORE_DWORDX2 $vgpr0_vgpr1, %2, 0, 0, implicit $exec, implicit $flat_scr :: (store (s64))
     ; CHECK-NEXT: S_ENDPGM 0
     INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VGPR_32 */, def early-clobber %1:vgpr_32, 1835018 /* regdef:VGPR_32 */, def %0:vgpr_32
@@ -62,7 +62,7 @@ body:             |
     ; CHECK-LABEL: name: foo3
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef %2.sub0, 1835019 /* regdef-ec:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef early-clobber %2.sub1
+    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VS_32 */, def undef %2.sub0, 1835019 /* regdef-ec:VS_32 */, def undef early-clobber %2.sub1
     ; CHECK-NEXT: FLAT_STORE_DWORDX2 $vgpr0_vgpr1, %2, 0, 0, implicit $exec, implicit $flat_scr :: (store (s64))
     ; CHECK-NEXT: S_ENDPGM 0
     INLINEASM &"", 0 /* attdialect */, 1835018 /* regdef:VGPR_32 */, def %1:vgpr_32, 1835019 /* regdef-ec:VGPR_32 */, def early-clobber %0:vgpr_32
@@ -83,7 +83,7 @@ body:             |
     ; CHECK-LABEL: name: foo4
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef early-clobber %2.sub1, 1835018 /* regdef:VS_16_and_SReg_1_with_lo16_in_TTMP_LO16 */, def undef %2.sub0
+    ; CHECK-NEXT: INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VS_32 */, def undef early-clobber %2.sub1, 1835018 /* regdef:VS_32 */, def undef %2.sub0
     ; CHECK-NEXT: FLAT_STORE_DWORDX2 $vgpr0_vgpr1, %2, 0, 0, implicit $exec, implicit $flat_scr :: (store (s64))
     ; CHECK-NEXT: S_ENDPGM 0
     INLINEASM &"", 0 /* attdialect */, 1835019 /* regdef-ec:VGPR_32 */, def early-clobber %0:vgpr_32, 1835018 /* regdef:VGPR_32 */, def %1:vgpr_32
diff --git a/llvm/test/CodeGen/AMDGPU/inline-asm.i128.ll b/llvm/test/CodeGen/AMDGPU/inline-asm.i128.ll
index 2d1f032070f01f..76b007c22b699c 100644
--- a/llvm/test/CodeGen/AMDGPU/inline-asm.i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/inline-asm.i128.ll
@@ -8,16 +8,16 @@
 define amdgpu_kernel void @s_input_output_i128() {
   ; GFX908-LABEL: name: s_input_output_i128
   ; GFX908: bb.0 (%ir-block.0):
-  ; GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7602186 /* regdef:SGPR_128 */, def %4
+  ; GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7340042 /* regdef:SGPR_128 */, def %4
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_128 = COPY %4
-  ; GFX908-NEXT:   INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 7602185 /* reguse:SGPR_128 */, [[COPY]]
+  ; GFX908-NEXT:   INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 7340041 /* reguse:SGPR_128 */, [[COPY]]
   ; GFX908-NEXT:   S_ENDPGM 0
   ;
   ; GFX90A-LABEL: name: s_input_output_i128
   ; GFX90A: bb.0 (%ir-block.0):
-  ; GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7602186 /* regdef:SGPR_128 */, def %4
+  ; GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7340042 /* regdef:SGPR_128 */, def %4
   ; GFX90A-NEXT:   [[COPY:%[0-9]+]]:sgpr_128 = COPY %4
-  ; GFX90A-NEXT:   INL...
[truncated]

[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes.

4cccb01

kosarev requested review from jayfoad, arsenm, Sisyph and rampitec December 28, 2023 12:40

llvmbot added backend:AMDGPU llvm:globalisel labels Dec 28, 2023

jayfoad approved these changes Dec 29, 2023

View reviewed changes

kosarev merged commit b6daac0 into llvm:main Dec 29, 2023
5 of 6 checks passed

kosarev deleted the t16-remove_lo_hi_classes branch December 29, 2023 12:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes. #76500

[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes. #76500

kosarev commented Dec 28, 2023

llvmbot commented Dec 28, 2023 •

edited

[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes. #76500

[AMDGPU][True16] Remove the VGPR_LO/HI16 register classes. #76500

Conversation

kosarev commented Dec 28, 2023

llvmbot commented Dec 28, 2023 • edited

llvmbot commented Dec 28, 2023 •

edited