Skip to content

[X86] Rename AVX512 VEXTRACT/INSERT??x? to VEXTRACT/INSERT??X? #116826

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 2 commits into from
Nov 20, 2024
Merged

[X86] Rename AVX512 VEXTRACT/INSERT??x? to VEXTRACT/INSERT??X? #116826

merged 2 commits into from
Nov 20, 2024

Conversation

RKSimon
Copy link
Collaborator

@RKSimon RKSimon commented Nov 19, 2024

Use uppercase in the subvector description ("32x2" -> "32X4" etc.) - matches what we already do in VBROADCAST??X?, and we try to use uppercase for all x86 instruction mnemonics anyway (and lowercase just for the arg description suffix).

@RKSimon
[X86] Rename AVX512 VEXTRACT/INSERT??x? to VEXTRACT/INSERT??X?
99109d5 
Use uppercase in the subvector description ("32x2" -> "32X4" etc.) - matches what we already do in VBROADCAST??X?, and we try to use uppercase for all x86 instruction mnemonics anyway (and lowercase just for the arg description suffix).
@llvmbot
Copy link
Member

llvmbot commented Nov 19, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-globalisel
@llvm/pr-subscribers-tablegen

@llvm/pr-subscribers-backend-x86

Author: Simon Pilgrim (RKSimon)

Changes

Use uppercase in the subvector description ("32x2" -> "32X4" etc.) - matches what we already do in VBROADCAST??X?, and we try to use uppercase for all x86 instruction mnemonics anyway (and lowercase just for the arg description suffix).

Patch is 55.67 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/116826.diff

11 Files Affected:

  • (modified) llvm/lib/Target/X86/GISel/X86InstructionSelector.cpp (+6-6)
  • (modified) llvm/lib/Target/X86/X86InstrAVX512.td (+103-103)
  • (modified) llvm/lib/Target/X86/X86InstrInfo.cpp (+4-4)
  • (modified) llvm/lib/Target/X86/X86ReplaceableInstrs.def (+24-24)
  • (modified) llvm/lib/Target/X86/X86SchedIceLake.td (+8-8)
  • (modified) llvm/lib/Target/X86/X86SchedSapphireRapids.td (+3-3)
  • (modified) llvm/lib/Target/X86/X86SchedSkylakeServer.td (+8-8)
  • (modified) llvm/test/TableGen/x86-fold-tables.inc (+48-48)
  • (modified) llvm/test/TableGen/x86-instr-mapping.inc (+16-16)
  • (modified) llvm/utils/TableGen/X86ManualFoldTables.def (+12-12)
  • (modified) llvm/utils/TableGen/X86ManualInstrMapping.def (+8-8)
diff --git a/llvm/lib/Target/X86/GISel/X86InstructionSelector.cpp b/llvm/lib/Target/X86/GISel/X86InstructionSelector.cpp
index d425a0d507524a..ee456a11d58441 100644
--- a/llvm/lib/Target/X86/GISel/X86InstructionSelector.cpp
+++ b/llvm/lib/Target/X86/GISel/X86InstructionSelector.cpp
@@ -1254,16 +1254,16 @@ bool X86InstructionSelector::selectExtract(MachineInstr &I,
 
   if (SrcTy.getSizeInBits() == 256 && DstTy.getSizeInBits() == 128) {
     if (HasVLX)
-      I.setDesc(TII.get(X86::VEXTRACTF32x4Z256rri));
+      I.setDesc(TII.get(X86::VEXTRACTF32X4Z256rri));
     else if (HasAVX)
       I.setDesc(TII.get(X86::VEXTRACTF128rri));
     else
       return false;
   } else if (SrcTy.getSizeInBits() == 512 && HasAVX512) {
     if (DstTy.getSizeInBits() == 128)
-      I.setDesc(TII.get(X86::VEXTRACTF32x4Zrri));
+      I.setDesc(TII.get(X86::VEXTRACTF32X4Zrri));
     else if (DstTy.getSizeInBits() == 256)
-      I.setDesc(TII.get(X86::VEXTRACTF64x4Zrri));
+      I.setDesc(TII.get(X86::VEXTRACTF64X4Zrri));
     else
       return false;
   } else
@@ -1387,16 +1387,16 @@ bool X86InstructionSelector::selectInsert(MachineInstr &I,
 
   if (DstTy.getSizeInBits() == 256 && InsertRegTy.getSizeInBits() == 128) {
     if (HasVLX)
-      I.setDesc(TII.get(X86::VINSERTF32x4Z256rri));
+      I.setDesc(TII.get(X86::VINSERTF32X4Z256rri));
     else if (HasAVX)
       I.setDesc(TII.get(X86::VINSERTF128rri));
     else
       return false;
   } else if (DstTy.getSizeInBits() == 512 && HasAVX512) {
     if (InsertRegTy.getSizeInBits() == 128)
-      I.setDesc(TII.get(X86::VINSERTF32x4Zrri));
+      I.setDesc(TII.get(X86::VINSERTF32X4Zrri));
     else if (InsertRegTy.getSizeInBits() == 256)
-      I.setDesc(TII.get(X86::VINSERTF64x4Zrri));
+      I.setDesc(TII.get(X86::VINSERTF64X4Zrri));
     else
       return false;
   } else
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index 7adc72f8bc7d06..4094a1ba9b7903 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -427,24 +427,24 @@ multiclass vinsert_for_type<ValueType EltVT32, int Opcode128,
                             X86FoldableSchedWrite sched> {
 
   let Predicates = [HasVLX] in
-    defm NAME # "32x4Z256" : vinsert_for_size<Opcode128,
+    defm NAME # "32X4Z256" : vinsert_for_size<Opcode128,
                                  X86VectorVTInfo< 4, EltVT32, VR128X>,
                                  X86VectorVTInfo< 8, EltVT32, VR256X>,
                                  vinsert128_insert, sched>, EVEX_V256;
 
-  defm NAME # "32x4Z" : vinsert_for_size<Opcode128,
+  defm NAME # "32X4Z" : vinsert_for_size<Opcode128,
                                  X86VectorVTInfo< 4, EltVT32, VR128X>,
                                  X86VectorVTInfo<16, EltVT32, VR512>,
                                  vinsert128_insert, sched>, EVEX_V512;
 
-  defm NAME # "64x4Z" : vinsert_for_size<Opcode256,
+  defm NAME # "64X4Z" : vinsert_for_size<Opcode256,
                                  X86VectorVTInfo< 4, EltVT64, VR256X>,
                                  X86VectorVTInfo< 8, EltVT64, VR512>,
                                  vinsert256_insert, sched>, REX_W, EVEX_V512;
 
   // Even with DQI we'd like to only use these instructions for masking.
   let Predicates = [HasVLX, HasDQI] in
-    defm NAME # "64x2Z256" : vinsert_for_size_split<Opcode128,
+    defm NAME # "64X2Z256" : vinsert_for_size_split<Opcode128,
                                    X86VectorVTInfo< 2, EltVT64, VR128X>,
                                    X86VectorVTInfo< 4, EltVT64, VR256X>,
                                    null_frag, vinsert128_insert, sched>,
@@ -452,13 +452,13 @@ multiclass vinsert_for_type<ValueType EltVT32, int Opcode128,
 
   // Even with DQI we'd like to only use these instructions for masking.
   let Predicates = [HasDQI] in {
-    defm NAME # "64x2Z" : vinsert_for_size_split<Opcode128,
+    defm NAME # "64X2Z" : vinsert_for_size_split<Opcode128,
                                  X86VectorVTInfo< 2, EltVT64, VR128X>,
                                  X86VectorVTInfo< 8, EltVT64, VR512>,
                                  null_frag, vinsert128_insert, sched>,
                                  REX_W, EVEX_V512;
 
-    defm NAME # "32x8Z" : vinsert_for_size_split<Opcode256,
+    defm NAME # "32X8Z" : vinsert_for_size_split<Opcode256,
                                    X86VectorVTInfo< 8, EltVT32, VR256X>,
                                    X86VectorVTInfo<16, EltVT32, VR512>,
                                    null_frag, vinsert256_insert, sched>,
@@ -472,47 +472,47 @@ defm VINSERTI : vinsert_for_type<i32, 0x38, i64, 0x3a, WriteShuffle256>;
 
 // Codegen pattern with the alternative types,
 // Even with AVX512DQ we'll still use these for unmasked operations.
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z256", v2f64x_info, v4f64x_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z256", v2f64x_info, v4f64x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z256", v2i64x_info, v4i64x_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z256", v2i64x_info, v4i64x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
 
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z", v2f64x_info, v8f64_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z", v2f64x_info, v8f64_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z", v2i64x_info, v8i64_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z", v2i64x_info, v8i64_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
 
-defm : vinsert_for_size_lowering<"VINSERTF64x4Z", v8f32x_info, v16f32_info,
+defm : vinsert_for_size_lowering<"VINSERTF64X4Z", v8f32x_info, v16f32_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTI64x4Z", v8i32x_info, v16i32_info,
+defm : vinsert_for_size_lowering<"VINSERTI64X4Z", v8i32x_info, v16i32_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
 
 // Codegen pattern with the alternative types insert VEC128 into VEC256
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z256", v8i16x_info, v16i16x_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z256", v8i16x_info, v16i16x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z256", v16i8x_info, v32i8x_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z256", v16i8x_info, v32i8x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z256", v8f16x_info, v16f16x_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z256", v8f16x_info, v16f16x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z256", v8bf16x_info, v16bf16x_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z256", v8bf16x_info, v16bf16x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
 // Codegen pattern with the alternative types insert VEC128 into VEC512
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z", v8i16x_info, v32i16_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z", v8i16x_info, v32i16_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTI32x4Z", v16i8x_info, v64i8_info,
+defm : vinsert_for_size_lowering<"VINSERTI32X4Z", v16i8x_info, v64i8_info,
                vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z", v8f16x_info, v32f16_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z", v8f16x_info, v32f16_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTF32x4Z", v8bf16x_info, v32bf16_info,
+defm : vinsert_for_size_lowering<"VINSERTF32X4Z", v8bf16x_info, v32bf16_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
 // Codegen pattern with the alternative types insert VEC256 into VEC512
-defm : vinsert_for_size_lowering<"VINSERTI64x4Z", v16i16x_info, v32i16_info,
+defm : vinsert_for_size_lowering<"VINSERTI64X4Z", v16i16x_info, v32i16_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTI64x4Z", v32i8x_info, v64i8_info,
+defm : vinsert_for_size_lowering<"VINSERTI64X4Z", v32i8x_info, v64i8_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTF64x4Z", v16f16x_info, v32f16_info,
+defm : vinsert_for_size_lowering<"VINSERTF64X4Z", v16f16x_info, v32f16_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_size_lowering<"VINSERTF64x4Z", v16bf16x_info, v32bf16_info,
+defm : vinsert_for_size_lowering<"VINSERTF64X4Z", v16bf16x_info, v32bf16_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
 
 
@@ -568,81 +568,81 @@ let Predicates = p in {
 }
 }
 
-defm : vinsert_for_mask_cast<"VINSERTF32x4Z256", v2f64x_info, v4f64x_info,
+defm : vinsert_for_mask_cast<"VINSERTF32X4Z256", v2f64x_info, v4f64x_info,
                              v8f32x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x2Z256", v4f32x_info, v8f32x_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X2Z256", v4f32x_info, v8f32x_info,
                              v4f64x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI, HasVLX]>;
 
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z256", v2i64x_info, v4i64x_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z256", v2i64x_info, v4i64x_info,
                              v8i32x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z256", v8i16x_info, v16i16x_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z256", v8i16x_info, v16i16x_info,
                              v8i32x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z256", v16i8x_info, v32i8x_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z256", v16i8x_info, v32i8x_info,
                              v8i32x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x2Z256", v4i32x_info, v8i32x_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X2Z256", v4i32x_info, v8i32x_info,
                              v4i64x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI, HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x2Z256", v8i16x_info, v16i16x_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X2Z256", v8i16x_info, v16i16x_info,
                              v4i64x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI, HasVLX]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x2Z256", v16i8x_info, v32i8x_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X2Z256", v16i8x_info, v32i8x_info,
                              v4i64x_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI, HasVLX]>;
 
-defm : vinsert_for_mask_cast<"VINSERTF32x4Z", v2f64x_info, v8f64_info,
+defm : vinsert_for_mask_cast<"VINSERTF32X4Z", v2f64x_info, v8f64_info,
                              v16f32_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x2Z", v4f32x_info, v16f32_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X2Z", v4f32x_info, v16f32_info,
                              v8f64_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI]>;
 
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z", v2i64x_info, v8i64_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z", v2i64x_info, v8i64_info,
                              v16i32_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z", v8i16x_info, v32i16_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z", v8i16x_info, v32i16_info,
                              v16i32_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x4Z", v16i8x_info, v64i8_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X4Z", v16i8x_info, v64i8_info,
                              v16i32_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x2Z", v4i32x_info, v16i32_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X2Z", v4i32x_info, v16i32_info,
                              v8i64_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x2Z", v8i16x_info, v32i16_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X2Z", v8i16x_info, v32i16_info,
                              v8i64_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x2Z", v16i8x_info, v64i8_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X2Z", v16i8x_info, v64i8_info,
                              v8i64_info, vinsert128_insert,
                              INSERT_get_vinsert128_imm, [HasDQI]>;
 
-defm : vinsert_for_mask_cast<"VINSERTF32x8Z", v4f64x_info, v8f64_info,
+defm : vinsert_for_mask_cast<"VINSERTF32X8Z", v4f64x_info, v8f64_info,
                              v16f32_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTF64x4Z", v8f32x_info, v16f32_info,
+defm : vinsert_for_mask_cast<"VINSERTF64X4Z", v8f32x_info, v16f32_info,
                              v8f64_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasAVX512]>;
 
-defm : vinsert_for_mask_cast<"VINSERTI32x8Z", v4i64x_info, v8i64_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X8Z", v4i64x_info, v8i64_info,
                              v16i32_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x8Z", v16i16x_info, v32i16_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X8Z", v16i16x_info, v32i16_info,
                              v16i32_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTI32x8Z", v32i8x_info, v64i8_info,
+defm : vinsert_for_mask_cast<"VINSERTI32X8Z", v32i8x_info, v64i8_info,
                              v16i32_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasDQI]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x4Z", v8i32x_info, v16i32_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X4Z", v8i32x_info, v16i32_info,
                              v8i64_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x4Z", v16i16x_info, v32i16_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X4Z", v16i16x_info, v32i16_info,
                              v8i64_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasAVX512]>;
-defm : vinsert_for_mask_cast<"VINSERTI64x4Z", v32i8x_info, v64i8_info,
+defm : vinsert_for_mask_cast<"VINSERTI64X4Z", v32i8x_info, v64i8_info,
                              v8i64_info, vinsert256_insert,
                              INSERT_get_vinsert256_imm, [HasAVX512]>;
 
@@ -732,19 +732,19 @@ multiclass vextract_for_type<ValueType EltVT32, int Opcode128,
                              ValueType EltVT64, int Opcode256,
                              SchedWrite SchedRR, SchedWrite SchedMR> {
   let Predicates = [HasAVX512] in {
-    defm NAME # "32x4Z" : vextract_for_size<Opcode128,
+    defm NAME # "32X4Z" : vextract_for_size<Opcode128,
                                    X86VectorVTInfo<16, EltVT32, VR512>,
                                    X86VectorVTInfo< 4, EltVT32, VR128X>,
                                    vextract128_extract, SchedRR, SchedMR>,
                                        EVEX_V512, EVEX_CD8<32, CD8VT4>;
-    defm NAME # "64x4Z" : vextract_for_size<Opcode256,
+    defm NAME # "64X4Z" : vextract_for_size<Opcode256,
                                    X86VectorVTInfo< 8, EltVT64, VR512>,
                                    X86VectorVTInfo< 4, EltVT64, VR256X>,
                                    vextract256_extract, SchedRR, SchedMR>,
                                        REX_W, EVEX_V512, EVEX_CD8<64, CD8VT4>;
   }
   let Predicates = [HasVLX] in
-    defm NAME # "32x4Z256" : vextract_for_size<Opcode128,
+    defm NAME # "32X4Z256" : vextract_for_size<Opcode128,
                                  X86VectorVTInfo< 8, EltVT32, VR256X>,
                                  X86VectorVTInfo< 4, EltVT32, VR128X>,
                                  vextract128_extract, SchedRR, SchedMR>,
@@ -752,7 +752,7 @@ multiclass vextract_for_type<ValueType EltVT32, int Opcode128,
 
   // Even with DQI we'd like to only use these instructions for masking.
   let Predicates = [HasVLX, HasDQI] in
-    defm NAME # "64x2Z256" : vextract_for_size_split<Opcode128,
+    defm NAME # "64X2Z256" : vextract_for_size_split<Opcode128,
                                  X86VectorVTInfo< 4, EltVT64, VR256X>,
                                  X86VectorVTInfo< 2, EltVT64, VR128X>,
                                  null_frag, vextract128_extract, SchedRR, SchedMR>,
@@ -760,12 +760,12 @@ multiclass vextract_for_type<ValueType EltVT32, int Opcode128,
 
   // Even with DQI we'd like to only use these instructions for masking.
   let Predicates = [HasDQI] in {
-    defm NAME # "64x2Z" : vextract_for_size_split<Opcode128,
+    defm NAME # "64X2Z" : vextract_for_size_split<Opcode128,
                                  X86VectorVTInfo< 8, EltVT64, VR512>,
                                  X86VectorVTInfo< 2, EltVT64, VR128X>,
                                  null_frag, vextract128_extract, SchedRR, SchedMR>,
                                      REX_W, EVEX_V512, EVEX_CD8<64, CD8VT2>;
-    defm NAME # "32x8Z" : vextract_for_size_split<Opcode256,
+    defm NAME # "32X8Z" : vextract_for_size_split<Opcode256,
                                  X86VectorVTInfo<16, EltVT32, VR512>,
                                  X86VectorVTInfo< 8, EltVT32, VR256X>,
                                  null_frag, vextract256_extract, SchedRR, SchedMR>,
@@ -779,48 +779,48 @@ defm VEXTRACTI : vextract_for_type<i32, 0x39, i64, 0x3b, WriteShuffle256, WriteV
 
 // extract_subvector codegen patterns with the alternative types.
 // Even with AVX512DQ we'll still use these for unmasked operations.
-defm : vextract_for_size_lowering<"VEXTRACTF32x4Z", v8f64_info, v2f64x_info,
+defm : vextract_for_size_lowering<"VEXTRACTF32X4Z", v8f64_info, v2f64x_info,
           vextract128_extract, EXTRACT_get_vextract128_imm, [HasAVX512]>;
-defm : vextract_for_size_lowering<"VEXTRACTI32x4Z", v8i64_info, v2i64x_info,
+defm : vextract_for_size_lowering<"VEXTRACTI32X4Z", v8i64_info, v2i64x_info,
           vextract128_extract, EXTRACT_get_vextract128_imm, [HasAVX512]>;
 
-defm : vextract_for_size_lowering<"VEXTRACTF64x4Z", v16f32_info, v8f32x_info,
+defm : vextract_for_size_lowering<"VEXTRACTF64X4Z", v16f32_info, v8f32x_info,
           vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
-defm : vextract_for_size_lowering<"VEXTRACTI64x4Z", v16i32_info, v8i32x_info,
+defm : vextract_for_size_lowering<"VEXTRACTI64X4Z", v16i32_info, v8i32x_info,
           vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
 
-defm : vextract_for_size_lowering<"VEXTRACTF32x4Z256", v4f64x_info, v2f64x_info,
+defm : vextract_for_size_lowering<"VEXTRACTF32X4Z256", v4f64x_info, v2f64x_info,
           vextract128_extract, EXTRACT_get_vextract128_imm, [HasVLX]>;
-defm : vextract_for_size_lowering<"VEXTRACTI32x4Z256", v4i64x_info, v2i64x_info,
+defm...
[truncated]

phoebewang
phoebewang approved these changes Nov 20, 2024
View reviewed changes
Copy link
Contributor

@phoebewang phoebewang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

KanRobert
KanRobert approved these changes Nov 20, 2024
View reviewed changes
Copy link
Contributor

@KanRobert KanRobert left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@RKSimon RKSimon merged commit 3a5cf6d into llvm:main Nov 20, 2024
7 of 9 checks passed
@RKSimon RKSimon deleted the x86-avx512-subvector-uppercase branch November 20, 2024 08:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@phoebewang phoebewang phoebewang approved these changes

@KanRobert KanRobert KanRobert approved these changes

Assignees
No one assigned
Labels
backend:X86 llvm:globalisel tablegen
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@RKSimon @llvmbot @phoebewang @KanRobert