AMDGPU: Copy correct predicates for SDWA reals #116288
Open
+27
−23
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
There are a lot of messes in the special case predicate handling. Currently broad let blocks override specific predicates with more general cases. For instructions with SDWA, the HasSDWA predicate was overriding the SubtargetPredicate for the instruction. This fixes enough to properly disallow new instructions that support SDWA on older targets.
|
@llvm/pr-subscribers-backend-amdgpu Author: Matt Arsenault (arsenm) ChangesThere are a lot of messes in the special case This fixes enough to properly disallow new instructions Full diff: https://github.com/llvm/llvm-project/pull/116288.diff 5 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index c8ae010414dc40..d7feaef8c4a97d 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -2103,8 +2103,10 @@ def NotHasMinMaxDenormModes : Predicate<"!Subtarget->supportsMinMaxDenormModes()
def HasFminFmaxLegacy : Predicate<"Subtarget->hasFminFmaxLegacy()">;
-def HasSDWA : Predicate<"Subtarget->hasSDWA()">,
- AssemblerPredicate<(all_of FeatureSDWA, FeatureVolcanicIslands)>;
+def HasSDWA : Predicate<"Subtarget->hasSDWA()">;
+
+def HasSDWA8 : Predicate<"Subtarget->hasSDWA()">,
+ AssemblerPredicate<(all_of (not FeatureGFX9Insts), FeatureSDWA)>;
def HasSDWA9 :
Predicate<"Subtarget->hasSDWA()">,
diff --git a/llvm/lib/Target/AMDGPU/VOP1Instructions.td b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
index c743eb43e3465c..f7a66a08209397 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -1268,7 +1268,7 @@ multiclass VOP1_Real_vi <bits<10> op> {
if !cast<VOP1_Pseudo>(NAME#"_e32").Pfl.HasExtSDWA then
def _sdwa_vi :
- VOP_SDWA_Real <!cast<VOP1_SDWA_Pseudo>(NAME#"_sdwa")>,
+ VOP_SDWA8_Real <!cast<VOP1_SDWA_Pseudo>(NAME#"_sdwa")>,
VOP1_SDWAe <op{7-0}, !cast<VOP1_SDWA_Pseudo>(NAME#"_sdwa").Pfl>;
if !cast<VOP1_Pseudo>(NAME#"_e32").Pfl.HasExtSDWA9 then
@@ -1474,7 +1474,7 @@ def : GCNPat <
// GFX9
//===----------------------------------------------------------------------===//
-let AssemblerPredicate = isGFX9Only, DecoderNamespace = "GFX9" in {
+let DecoderNamespace = "GFX9" in {
multiclass VOP1_Real_gfx9 <bits<10> op> {
defm NAME : VOP1_Real_e32e64_vi <op>;
diff --git a/llvm/lib/Target/AMDGPU/VOP2Instructions.td b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
index 925b60561c9d68..c0d38fa52b3446 100644
--- a/llvm/lib/Target/AMDGPU/VOP2Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
@@ -766,16 +766,16 @@ defm V_SUBB_U32 : VOP2bInst <"v_subb_u32", VOP2b_I32_I1_I32_I32_I1, null_frag, "
defm V_SUBBREV_U32 : VOP2bInst <"v_subbrev_u32", VOP2b_I32_I1_I32_I32_I1, null_frag, "v_subb_u32">;
-let SubtargetPredicate = HasAddNoCarryInsts, isReMaterializable = 1 in {
+let SubtargetPredicate = HasAddNoCarryInsts, isReMaterializable = 1 in {
defm V_SUB_U32 : VOP2Inst <"v_sub_u32", VOP_I32_I32_I32_ARITH, null_frag, "v_sub_u32">;
defm V_SUBREV_U32 : VOP2Inst <"v_subrev_u32", VOP_I32_I32_I32_ARITH, null_frag, "v_sub_u32">;
}
-let SubtargetPredicate = HasAddNoCarryInsts, isReMaterializable = 1, isAdd = 1 in {
+let SubtargetPredicate = HasAddNoCarryInsts, isReMaterializable = 1, isAdd = 1 in {
defm V_ADD_U32 : VOP2Inst_VOPD <"v_add_u32", VOP_I32_I32_I32_ARITH, 0x10, "v_add_nc_u32", null_frag, "v_add_u32">;
}
-let isAdd = 1 in {
+let isAdd = 1 in {
defm V_ADD_CO_U32 : VOP2bInst <"v_add_co_u32", VOP2b_I32_I1_I32_I32, null_frag, "v_add_co_u32">;
defm V_ADDC_U32 : VOP2bInst <"v_addc_u32", VOP2b_I32_I1_I32_I32_I1, null_frag, "v_addc_u32">;
}
@@ -2290,10 +2290,10 @@ multiclass Base_VOP2_Real_e32e64_vi <bits<6> op> :
} // End AssemblerPredicate = isGFX8GFX9, DecoderNamespace = "GFX8"
-multiclass VOP2_SDWA_Real <bits<6> op> {
+multiclass VOP2_SDWA8_Real <bits<6> op> {
if !cast<VOP2_Pseudo>(NAME#"_e32").Pfl.HasExtSDWA then
def _sdwa_vi :
- VOP_SDWA_Real <!cast<VOP2_SDWA_Pseudo>(NAME#"_sdwa")>,
+ VOP_SDWA8_Real <!cast<VOP2_SDWA_Pseudo>(NAME#"_sdwa")>,
VOP2_SDWAe <op{5-0}, !cast<VOP2_SDWA_Pseudo>(NAME#"_sdwa").Pfl>;
}
@@ -2321,7 +2321,7 @@ multiclass VOP2be_Real_e32e64_vi_only <bits<6> op, string OpName, string AsmName
}
if !cast<VOP2_Pseudo>(OpName#"_e32").Pfl.HasExtSDWA then
def _sdwa_vi :
- VOP_SDWA_Real <!cast<VOP2_SDWA_Pseudo>(OpName#"_sdwa")>,
+ VOP_SDWA8_Real <!cast<VOP2_SDWA_Pseudo>(OpName#"_sdwa")>,
VOP2_SDWAe <op{5-0}, !cast<VOP2_SDWA_Pseudo>(OpName#"_sdwa").Pfl> {
VOP2_SDWA_Pseudo ps = !cast<VOP2_SDWA_Pseudo>(OpName#"_sdwa");
let AsmString = AsmName # ps.AsmOperands;
@@ -2337,7 +2337,7 @@ multiclass VOP2be_Real_e32e64_vi_only <bits<6> op, string OpName, string AsmName
} // End AssemblerPredicate = isGFX8Only, DecoderNamespace = "GFX8"
-let AssemblerPredicate = isGFX9Only, DecoderNamespace = "GFX9" in {
+let DecoderNamespace = "GFX9" in {
multiclass VOP2be_Real_e32e64_gfx9 <bits<6> op, string OpName, string AsmName> {
def _e32_gfx9 :
@@ -2386,10 +2386,10 @@ multiclass VOP2_Real_e32e64_gfx9 <bits<6> op> {
VOP2_DPPe<op, !cast<VOP2_DPP_Pseudo>(NAME#"_dpp")>;
}
-} // End AssemblerPredicate = isGFX9Only, DecoderNamespace = "GFX9"
+} // End DecoderNamespace = "GFX9"
multiclass VOP2_Real_e32e64_vi <bits<6> op> :
- Base_VOP2_Real_e32e64_vi<op>, VOP2_SDWA_Real<op>, VOP2_SDWA9_Real<op> {
+ Base_VOP2_Real_e32e64_vi<op>, VOP2_SDWA8_Real<op>, VOP2_SDWA9_Real<op> {
if !cast<VOP2_Pseudo>(NAME#"_e32").Pfl.HasExtDPP then
def _dpp_vi :
@@ -2401,7 +2401,7 @@ defm V_CNDMASK_B32 : VOP2_Real_e32e64_vi <0x0>;
defm V_ADD_F32 : VOP2_Real_e32e64_vi <0x1>;
defm V_SUB_F32 : VOP2_Real_e32e64_vi <0x2>;
defm V_SUBREV_F32 : VOP2_Real_e32e64_vi <0x3>;
-let AssemblerPredicate = isGCN3ExcludingGFX90A in
+let OtherPredicates = [isGCN3ExcludingGFX90A] in
defm V_MUL_LEGACY_F32 : VOP2_Real_e32e64_vi <0x4>;
defm V_MUL_F32 : VOP2_Real_e32e64_vi <0x5>;
defm V_MUL_I32_I24 : VOP2_Real_e32e64_vi <0x6>;
@@ -2431,6 +2431,7 @@ defm V_ADDC_U32 : VOP2be_Real_e32e64_vi_only <0x1c, "V_ADDC_U32", "
defm V_SUBB_U32 : VOP2be_Real_e32e64_vi_only <0x1d, "V_SUBB_U32", "v_subb_u32">;
defm V_SUBBREV_U32 : VOP2be_Real_e32e64_vi_only <0x1e, "V_SUBBREV_U32", "v_subbrev_u32">;
+let AssemblerPredicate = isGFX9Only in {
defm V_ADD_CO_U32 : VOP2be_Real_e32e64_gfx9 <0x19, "V_ADD_CO_U32", "v_add_co_u32">;
defm V_SUB_CO_U32 : VOP2be_Real_e32e64_gfx9 <0x1a, "V_SUB_CO_U32", "v_sub_co_u32">;
defm V_SUBREV_CO_U32 : VOP2be_Real_e32e64_gfx9 <0x1b, "V_SUBREV_CO_U32", "v_subrev_co_u32">;
@@ -2441,6 +2442,7 @@ defm V_SUBBREV_CO_U32 : VOP2be_Real_e32e64_gfx9 <0x1e, "V_SUBBREV_U32", "v_s
defm V_ADD_U32 : VOP2_Real_e32e64_gfx9 <0x34>;
defm V_SUB_U32 : VOP2_Real_e32e64_gfx9 <0x35>;
defm V_SUBREV_U32 : VOP2_Real_e32e64_gfx9 <0x36>;
+} // End AssemblerPredicate = isGFX9Only
defm V_BFM_B32 : VOP2_Real_e64only_vi <0x293>;
defm V_BCNT_U32_B32 : VOP2_Real_e64only_vi <0x28b>;
@@ -2518,7 +2520,7 @@ defm V_XNOR_B32 : VOP2_Real_e32e64_vi <0x3d>;
} // End SubtargetPredicate = HasDLInsts
-let AssemblerPredicate = isGFX90APlus, DecoderNamespace = "GFX90A" in {
+let DecoderNamespace = "GFX90A" in {
multiclass VOP2_Real_e32_gfx90a <bits<6> op> {
def _e32_gfx90a :
VOP2_Real<!cast<VOP2_Pseudo>(NAME#"_e32"), SIEncodingFamily.GFX90A>,
@@ -2551,7 +2553,7 @@ let SubtargetPredicate = HasFmacF64Inst in {
defm V_FMAC_F64 : VOP2_Real_e32e64_gfx90a <0x4>;
} // End SubtargetPredicate = HasFmacF64Inst
-let SubtargetPredicate = isGFX90APlus, IsSingle = 1 in {
+let IsSingle = 1 in {
defm V_MUL_LEGACY_F32 : VOP2_Real_e64_gfx90a <0x2a1>;
}
diff --git a/llvm/lib/Target/AMDGPU/VOPCInstructions.td b/llvm/lib/Target/AMDGPU/VOPCInstructions.td
index d6e08dce130ced..f4ccae1decb1df 100644
--- a/llvm/lib/Target/AMDGPU/VOPCInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOPCInstructions.td
@@ -2290,7 +2290,7 @@ multiclass VOPC_Real_vi <bits<10> op> {
if !cast<VOPC_Pseudo>(NAME#"_e32").Pfl.HasExtSDWA then
def _sdwa_vi :
- VOP_SDWA_Real <!cast<VOPC_SDWA_Pseudo>(NAME#"_sdwa")>,
+ VOP_SDWA8_Real <!cast<VOPC_SDWA_Pseudo>(NAME#"_sdwa")>,
VOPC_SDWAe <op{7-0}, !cast<VOPC_SDWA_Pseudo>(NAME#"_sdwa").Pfl>;
if !cast<VOPC_Pseudo>(NAME#"_e32").Pfl.HasExtSDWA9 then
diff --git a/llvm/lib/Target/AMDGPU/VOPInstructions.td b/llvm/lib/Target/AMDGPU/VOPInstructions.td
index aab5dc7465d938..24fe24b1a53141 100644
--- a/llvm/lib/Target/AMDGPU/VOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOPInstructions.td
@@ -650,7 +650,7 @@ class VOP_SDWA_Pseudo <string opName, VOPProfile P, list<dag> pattern=[]> :
let Uses = !if(ReadsModeReg, [MODE, EXEC], [EXEC]);
let SubtargetPredicate = HasSDWA;
- let AssemblerPredicate = HasSDWA;
+ //let AssemblerPredicate = HasSDWA;
let AsmVariantName = !if(P.HasExtSDWA, AMDGPUAsmVariants.SDWA,
AMDGPUAsmVariants.Disable);
let DecoderNamespace = "GFX8";
@@ -658,7 +658,7 @@ class VOP_SDWA_Pseudo <string opName, VOPProfile P, list<dag> pattern=[]> :
VOPProfile Pfl = P;
}
-class VOP_SDWA_Real <VOP_SDWA_Pseudo ps> :
+class VOP_SDWA8_Real <VOP_SDWA_Pseudo ps> :
InstSI <ps.OutOperandList, ps.InOperandList, ps.Mnemonic # ps.AsmOperands, []>,
SIMCInstr <ps.PseudoInstr, SIEncodingFamily.SDWA> {
@@ -676,7 +676,7 @@ class VOP_SDWA_Real <VOP_SDWA_Pseudo ps> :
// Copy relevant pseudo op flags
let SubtargetPredicate = ps.SubtargetPredicate;
- let AssemblerPredicate = ps.AssemblerPredicate;
+ let AssemblerPredicate = HasSDWA8;
let AsmMatchConverter = ps.AsmMatchConverter;
let AsmVariantName = ps.AsmVariantName;
let UseNamedOperandTable = ps.UseNamedOperandTable;
@@ -708,7 +708,7 @@ class Base_VOP_SDWA9_Real <VOP_SDWA_Pseudo ps> :
let Constraints = ps.Constraints;
let DisableEncoding = ps.DisableEncoding;
- let SubtargetPredicate = HasSDWA9;
+ let SubtargetPredicate = ps.SubtargetPredicate;
let AssemblerPredicate = HasSDWA9;
let OtherPredicates = ps.OtherPredicates;
let AsmVariantName = !if(ps.Pfl.HasExtSDWA9, AMDGPUAsmVariants.SDWA9,
@@ -735,7 +735,7 @@ class VOP_SDWA9_Real <VOP_SDWA_Pseudo ps> :
SIMCInstr <ps.PseudoInstr, SIEncodingFamily.SDWA9>;
class Base_VOP_SDWA10_Real<VOP_SDWA_Pseudo ps> : Base_VOP_SDWA9_Real<ps> {
- let SubtargetPredicate = HasSDWA10;
+ let SubtargetPredicate = ps.SubtargetPredicate;
let AssemblerPredicate = HasSDWA10;
let DecoderNamespace = "GFX10";
}
@@ -1508,7 +1508,7 @@ class VOP3_DPP16_t16_Helper<bits<10> op, VOP_DPP_Pseudo ps,
let SchedRW = ps.SchedRW;
let Uses = ps.Uses;
let AssemblerPredicate = HasDPP16;
- let SubtargetPredicate = HasDPP16;
+ let SubtargetPredicate = ps.SubtargetPredicate;
let OtherPredicates = ps.OtherPredicates;
}
|
arsenm
requested review from
broxigarchen,
jayfoad,
kosarev,
rampitec and
Sisyph
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There are a lot of messes in the special case
predicate handling. Currently broad let blocks
override specific predicates with more general
cases. For instructions with SDWA, the HasSDWA
predicate was overriding the SubtargetPredicate
for the instruction.
This fixes enough to properly disallow new instructions
that support SDWA on older targets.