[AArch64] FP/SIMD is not mandatory for v8-R #79004
Merged
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
The FP/SIMD instructions are optional for v8-R, so they should not be marked as a dependency of HasV8_0rOps. This had the effect of disabling some v8R-specific system registers when any of these features was disabled. I've moved these features to be enabled by default for Cortex-R82 (currently the only v8-R AArch64 core), matching the previous behavior, and clang's default. Based on a patch by Simi Pallipurath <simi.pallipurath@arm.com>
|
@llvm/pr-subscribers-backend-aarch64 @llvm/pr-subscribers-mc Author: None (ostannard) ChangesThe FP/SIMD instructions are optional for v8-R, so they should not be marked as a dependency of HasV8_0rOps. This had the effect of disabling some v8R-specific system registers when any of these features was disabled. I've moved these features to be enabled by default for Cortex-R82 (currently the only v8-R AArch64 core), matching the previous behavior, and clang's default. Based on a patch by Simi Pallipurath <simi.pallipurath@arm.com> Full diff: https://github.com/llvm/llvm-project/pull/79004.diff 7 Files Affected:
diff --git a/llvm/lib/Target/AArch64/AArch64.td b/llvm/lib/Target/AArch64/AArch64.td
index 36700f73df4b20a..02fb01caf7e801f 100644
--- a/llvm/lib/Target/AArch64/AArch64.td
+++ b/llvm/lib/Target/AArch64/AArch64.td
@@ -81,7 +81,8 @@ def FeatureFMV : SubtargetFeature<"fmv", "HasFMV", "true",
"Enable Function Multi Versioning support.">;
def FeatureRDM : SubtargetFeature<"rdm", "HasRDM", "true",
- "Enable ARMv8.1 Rounding Double Multiply Add/Subtract instructions (FEAT_RDM)">;
+ "Enable ARMv8.1 Rounding Double Multiply Add/Subtract instructions (FEAT_RDM)",
+ [FeatureNEON]>;
def FeaturePAN : SubtargetFeature<
"pan", "HasPAN", "true",
@@ -321,7 +322,7 @@ def FeatureUseRSqrt : SubtargetFeature<
def FeatureDotProd : SubtargetFeature<
"dotprod", "HasDotProd", "true",
- "Enable dot product support (FEAT_DotProd)">;
+ "Enable dot product support (FEAT_DotProd)", [FeatureNEON]>;
def FeaturePAuth : SubtargetFeature<
"pauth", "HasPAuth", "true",
@@ -708,15 +709,14 @@ def HasV9_5aOps : SubtargetFeature<
def HasV8_0rOps : SubtargetFeature<
"v8r", "HasV8_0rOps", "true", "Support ARM v8r instructions",
[//v8.1
- FeatureCRC, FeaturePAN, FeatureRDM, FeatureLSE, FeatureCONTEXTIDREL2,
+ FeatureCRC, FeaturePAN, FeatureLSE, FeatureCONTEXTIDREL2,
//v8.2
FeatureRAS, FeaturePsUAO, FeatureCCPP, FeaturePAN_RWV,
//v8.3
- FeatureComplxNum, FeatureCCIDX, FeatureJS,
- FeaturePAuth, FeatureRCPC,
+ FeatureCCIDX, FeaturePAuth, FeatureRCPC,
//v8.4
- FeatureDotProd, FeatureTRACEV8_4, FeatureTLB_RMI,
- FeatureFlagM, FeatureDIT, FeatureSEL2, FeatureRCPC_IMMO,
+ FeatureTRACEV8_4, FeatureTLB_RMI, FeatureFlagM, FeatureDIT, FeatureSEL2,
+ FeatureRCPC_IMMO,
// Not mandatory in v8.0-R, but included here on the grounds that it
// only enables names of system registers
FeatureSpecRestrict
@@ -1420,7 +1420,8 @@ def ProcessorFeatures {
FeaturePerfMon, FeatureSPE, FeatureSPE_EEF];
list<SubtargetFeature> R82 = [HasV8_0rOps, FeaturePerfMon, FeatureFullFP16,
FeatureFP16FML, FeatureSSBS, FeaturePredRes,
- FeatureSB];
+ FeatureSB, FeatureRDM, FeatureDotProd,
+ FeatureComplxNum, FeatureJS];
list<SubtargetFeature> X1 = [HasV8_2aOps, FeatureCrypto, FeatureFPARMv8,
FeatureNEON, FeatureRCPC, FeaturePerfMon,
FeatureSPE, FeatureFullFP16, FeatureDotProd,
diff --git a/llvm/test/MC/AArch64/armv8.1a-rdma.s b/llvm/test/MC/AArch64/armv8.1a-rdma.s
index 907e4149454ae44..6e9e97a528fb99a 100644
--- a/llvm/test/MC/AArch64/armv8.1a-rdma.s
+++ b/llvm/test/MC/AArch64/armv8.1a-rdma.s
@@ -1,6 +1,6 @@
// RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.1a -show-encoding < %s 2> %t | FileCheck %s
// RUN: FileCheck --check-prefix=CHECK-ERROR < %t %s
-// RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r -show-encoding < %s 2> %t | FileCheck %s
+// RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r,+rdm -show-encoding < %s 2> %t | FileCheck %s
// RUN: FileCheck --check-prefix=CHECK-ERROR < %t %s
.text
diff --git a/llvm/test/MC/AArch64/armv8.2a-dotprod-errors.s b/llvm/test/MC/AArch64/armv8.2a-dotprod-errors.s
index 3c2fca56dd6b8b7..b3736a281122a91 100644
--- a/llvm/test/MC/AArch64/armv8.2a-dotprod-errors.s
+++ b/llvm/test/MC/AArch64/armv8.2a-dotprod-errors.s
@@ -1,6 +1,6 @@
// RUN: not llvm-mc -triple aarch64 -mattr=+dotprod -show-encoding < %s 2> %t
// RUN: FileCheck --check-prefix=CHECK-ERROR < %t %s
-// RUN: not llvm-mc -triple aarch64 -mattr=+v8r -show-encoding < %s 2> %t
+// RUN: not llvm-mc -triple aarch64 -mattr=+v8r,+dotprod -show-encoding < %s 2> %t
// RUN: FileCheck --check-prefix=CHECK-ERROR < %t %s
udot v0.2s, v1.8b, v2.4b[4]
diff --git a/llvm/test/MC/AArch64/armv8.2a-dotprod.s b/llvm/test/MC/AArch64/armv8.2a-dotprod.s
index 9c4a6cad7e07a65..a49ed14d8677aa1 100644
--- a/llvm/test/MC/AArch64/armv8.2a-dotprod.s
+++ b/llvm/test/MC/AArch64/armv8.2a-dotprod.s
@@ -12,13 +12,13 @@
// RUN: llvm-mc -triple aarch64 -mcpu=neoverse-n2 -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
// RUN: llvm-mc -triple aarch64 -mcpu=tsv110 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
// RUN: llvm-mc -triple aarch64 -mcpu=cortex-r82 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
-// RUN: llvm-mc -triple aarch64 -mattr=+v8r -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
+// RUN: llvm-mc -triple aarch64 -mattr=+v8r,+dotprod -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
// RUN: llvm-mc -triple aarch64 -mcpu=ampere1 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
// RUN: llvm-mc -triple aarch64 -mcpu=ampere1a -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
// RUN: not llvm-mc -triple aarch64 -mattr=+v8.2a -show-encoding < %s 2> %t
// RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
-// RUN: not llvm-mc -triple aarch64 -mattr=+v8r,-dotprod -show-encoding < %s 2> %t
+// RUN: not llvm-mc -triple aarch64 -mattr=+v8r -show-encoding < %s 2> %t
// RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
// RUN: not llvm-mc -triple aarch64 -mcpu=cortex-r82 -mattr=-dotprod -show-encoding < %s 2> %t
// RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
diff --git a/llvm/test/MC/AArch64/armv8r-sysreg.s b/llvm/test/MC/AArch64/armv8r-sysreg.s
index 0f9cb05622d2ad2..46a23d5827b2956 100644
--- a/llvm/test/MC/AArch64/armv8r-sysreg.s
+++ b/llvm/test/MC/AArch64/armv8r-sysreg.s
@@ -1,4 +1,7 @@
// RUN: llvm-mc -triple aarch64 -show-encoding -mattr=+v8r -o - %s | FileCheck %s
+// RUN: llvm-mc -triple aarch64 -show-encoding -mattr=+v8r,-fp-armv8,-rdm,-dotprod,-complxnum,-jsconv -o - %s | FileCheck %s
+// RUN: llvm-mc -triple aarch64 -show-encoding -mcpu=cortex-r82 -o - %s | FileCheck %s
+// RUN: llvm-mc -triple aarch64 -show-encoding -mcpu=cortex-r82 -mattr=-fp-armv8,-rdm,-dotprod,-complxnum,-jsconv -o - %s | FileCheck %s
.text
mrs x0, VSCTLR_EL2
mrs x0, MPUIR_EL1
diff --git a/llvm/test/MC/Disassembler/AArch64/armv8.3a-complex.txt b/llvm/test/MC/Disassembler/AArch64/armv8.3a-complex.txt
index f920639173ad285..ea8b498fdc6e735 100644
--- a/llvm/test/MC/Disassembler/AArch64/armv8.3a-complex.txt
+++ b/llvm/test/MC/Disassembler/AArch64/armv8.3a-complex.txt
@@ -1,7 +1,7 @@
# RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.3a,-fullfp16 --disassemble < %s 2>%t | FileCheck %s --check-prefix=CHECK
# RUN: FileCheck %s < %t --check-prefix=NO-FP16
# RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.3a,+fullfp16 --disassemble < %s 2>%t | FileCheck %s --check-prefix=CHECK --check-prefix=FP16
-# RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r --disassemble < %s 2>%t | FileCheck %s --check-prefix=CHECK
+# RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r,+complxnum --disassemble < %s 2>%t | FileCheck %s --check-prefix=CHECK
# RUN: FileCheck %s < %t --check-prefix=NO-FP16
# RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=-v8.3a,+fullfp16 --disassemble < %s 2>&1 | FileCheck %s --check-prefix=NO-V83A
diff --git a/llvm/test/MC/Disassembler/AArch64/armv8.3a-js.txt b/llvm/test/MC/Disassembler/AArch64/armv8.3a-js.txt
index 66c8c3001e98242..22144cf64d965da 100644
--- a/llvm/test/MC/Disassembler/AArch64/armv8.3a-js.txt
+++ b/llvm/test/MC/Disassembler/AArch64/armv8.3a-js.txt
@@ -1,4 +1,4 @@
# RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.3a --disassemble < %s | FileCheck %s
-# RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r --disassemble < %s | FileCheck %s
+# RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r,+jsconv --disassemble < %s | FileCheck %s
# CHECK: fjcvtzs w0, d0
[0x00,0x00,0x7e,0x1e]
|
statham-arm
reviewed
Jan 22, 2024
|
The windows buildkite job failed with |
statham-arm
approved these changes
Jan 24, 2024
This was referenced Apr 5, 2024
ostannard
pushed a commit
that referenced
this pull request
May 7, 2024
This was [addressed for AArch64 here](#79004), but the same applies to ARM. Move the enablement of neon+fp64 to `-mcpu=cortex-r52`, which optionally supports these features.
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.
The FP/SIMD instructions are optional for v8-R, so they should not be marked as a dependency of HasV8_0rOps. This had the effect of disabling some v8R-specific system registers when any of these features was disabled.
I've moved these features to be enabled by default for Cortex-R82 (currently the only v8-R AArch64 core), matching the previous behavior, and clang's default.
Based on a patch by Simi Pallipurath simi.pallipurath@arm.com