[RISCV] Add RVVConstraint to SiFive custom matrix multiply instructions. #124055
Conversation
This file contains hidden or 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
…ply instructions. The instructions don't allow the vs1 encoded register to overlap vd. Confusingly these instructions order their operands vd, vs1, vs2 while every other vector instruction is vd, vs2, vs1. VS2Constraint really means check the first operand after vd which is vs1 in this case.
topperc
requested review from
4vtomat,
kito-cheng,
lukel97,
preames,
rofirrim and
wangpc-pp
Member
|
@llvm/pr-subscribers-backend-risc-v Author: Craig Topper (topperc) ChangesThe instructions don't allow the vs1 encoded register to overlap vd. Confusingly these instructions order their operands vd, vs1, vs2 while every other vector instruction is vd, vs2, vs1. VS2Constraint really means check the first operand after vd which is vs1 in this case. Full diff: https://github.com/llvm/llvm-project/pull/124055.diff 3 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
index 20adda91f6bde1..117789c0fb278a 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
@@ -202,7 +202,7 @@ let Predicates = [HasVendorXSfvcp], mayLoad = 0, mayStore = 0,
}
let Predicates = [HasVendorXSfvqmaccdod], DecoderNamespace = "XSfvqmaccdod",
- DestEEW = EEWSEWx4 in {
+ DestEEW = EEWSEWx4, RVVConstraint=VS2Constraint in {
def VQMACCU_2x8x2 : CustomSiFiveVMACC<0b101100, OPMVV, "sf.vqmaccu.2x8x2">;
def VQMACC_2x8x2 : CustomSiFiveVMACC<0b101101, OPMVV, "sf.vqmacc.2x8x2">;
def VQMACCUS_2x8x2 : CustomSiFiveVMACC<0b101110, OPMVV, "sf.vqmaccus.2x8x2">;
@@ -210,7 +210,7 @@ let Predicates = [HasVendorXSfvqmaccdod], DecoderNamespace = "XSfvqmaccdod",
}
let Predicates = [HasVendorXSfvqmaccqoq], DecoderNamespace = "XSfvqmaccqoq",
- DestEEW = EEWSEWx4 in {
+ DestEEW = EEWSEWx4, RVVConstraint=VS2Constraint in {
def VQMACCU_4x8x4 : CustomSiFiveVMACC<0b111100, OPMVV, "sf.vqmaccu.4x8x4">;
def VQMACC_4x8x4 : CustomSiFiveVMACC<0b111101, OPMVV, "sf.vqmacc.4x8x4">;
def VQMACCUS_4x8x4 : CustomSiFiveVMACC<0b111110, OPMVV, "sf.vqmaccus.4x8x4">;
@@ -218,7 +218,7 @@ let Predicates = [HasVendorXSfvqmaccqoq], DecoderNamespace = "XSfvqmaccqoq",
}
let Predicates = [HasVendorXSfvfwmaccqqq], DecoderNamespace = "XSfvfwmaccqqq",
- DestEEW = EEWSEWx2 in {
+ DestEEW = EEWSEWx2, RVVConstraint=VS2Constraint in {
def VFWMACC_4x4x4 : CustomSiFiveVMACC<0b111100, OPFVV, "sf.vfwmacc.4x4x4">;
}
diff --git a/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s b/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s
new file mode 100644
index 00000000000000..357afb03af1826
--- /dev/null
+++ b/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s
@@ -0,0 +1,6 @@
+# RUN: not llvm-mc -triple=riscv64 -show-encoding -mattr=+v,+xsfvfwmaccqqq %s 2>&1 \
+# RUN: | FileCheck %s --check-prefixes=CHECK-ERROR
+
+sf.vfwmacc.4x4x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vfwmacc.4x4x4 v8, v8, v20{{$}}
diff --git a/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s b/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s
new file mode 100644
index 00000000000000..aaf35322eede6f
--- /dev/null
+++ b/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s
@@ -0,0 +1,34 @@
+# RUN: not llvm-mc -triple=riscv64 -show-encoding -mattr=+v,+xsfvqmaccqoq,+xsfvqmaccdod %s 2>&1 \
+# RUN: | FileCheck %s --check-prefix=CHECK-ERROR
+
+sf.vqmaccu.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccu.2x8x2 v8, v8, v20
+
+sf.vqmacc.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmacc.2x8x2 v8, v8, v20
+
+sf.vqmaccus.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccus.2x8x2 v8, v8, v20
+
+sf.vqmaccsu.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccsu.2x8x2 v8, v8, v20
+
+sf.vqmaccu.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccu.4x8x4 v8, v8, v20
+
+sf.vqmacc.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmacc.4x8x4 v8, v8, v20
+
+sf.vqmaccus.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccus.4x8x4 v8, v8, v20
+
+sf.vqmaccsu.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccsu.4x8x4 v8, v8, v20
|
Member
|
@llvm/pr-subscribers-mc Author: Craig Topper (topperc) ChangesThe instructions don't allow the vs1 encoded register to overlap vd. Confusingly these instructions order their operands vd, vs1, vs2 while every other vector instruction is vd, vs2, vs1. VS2Constraint really means check the first operand after vd which is vs1 in this case. Full diff: https://github.com/llvm/llvm-project/pull/124055.diff 3 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
index 20adda91f6bde1..117789c0fb278a 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
@@ -202,7 +202,7 @@ let Predicates = [HasVendorXSfvcp], mayLoad = 0, mayStore = 0,
}
let Predicates = [HasVendorXSfvqmaccdod], DecoderNamespace = "XSfvqmaccdod",
- DestEEW = EEWSEWx4 in {
+ DestEEW = EEWSEWx4, RVVConstraint=VS2Constraint in {
def VQMACCU_2x8x2 : CustomSiFiveVMACC<0b101100, OPMVV, "sf.vqmaccu.2x8x2">;
def VQMACC_2x8x2 : CustomSiFiveVMACC<0b101101, OPMVV, "sf.vqmacc.2x8x2">;
def VQMACCUS_2x8x2 : CustomSiFiveVMACC<0b101110, OPMVV, "sf.vqmaccus.2x8x2">;
@@ -210,7 +210,7 @@ let Predicates = [HasVendorXSfvqmaccdod], DecoderNamespace = "XSfvqmaccdod",
}
let Predicates = [HasVendorXSfvqmaccqoq], DecoderNamespace = "XSfvqmaccqoq",
- DestEEW = EEWSEWx4 in {
+ DestEEW = EEWSEWx4, RVVConstraint=VS2Constraint in {
def VQMACCU_4x8x4 : CustomSiFiveVMACC<0b111100, OPMVV, "sf.vqmaccu.4x8x4">;
def VQMACC_4x8x4 : CustomSiFiveVMACC<0b111101, OPMVV, "sf.vqmacc.4x8x4">;
def VQMACCUS_4x8x4 : CustomSiFiveVMACC<0b111110, OPMVV, "sf.vqmaccus.4x8x4">;
@@ -218,7 +218,7 @@ let Predicates = [HasVendorXSfvqmaccqoq], DecoderNamespace = "XSfvqmaccqoq",
}
let Predicates = [HasVendorXSfvfwmaccqqq], DecoderNamespace = "XSfvfwmaccqqq",
- DestEEW = EEWSEWx2 in {
+ DestEEW = EEWSEWx2, RVVConstraint=VS2Constraint in {
def VFWMACC_4x4x4 : CustomSiFiveVMACC<0b111100, OPFVV, "sf.vfwmacc.4x4x4">;
}
diff --git a/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s b/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s
new file mode 100644
index 00000000000000..357afb03af1826
--- /dev/null
+++ b/llvm/test/MC/RISCV/rvv/xsfvfwmacc-invalid.s
@@ -0,0 +1,6 @@
+# RUN: not llvm-mc -triple=riscv64 -show-encoding -mattr=+v,+xsfvfwmaccqqq %s 2>&1 \
+# RUN: | FileCheck %s --check-prefixes=CHECK-ERROR
+
+sf.vfwmacc.4x4x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vfwmacc.4x4x4 v8, v8, v20{{$}}
diff --git a/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s b/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s
new file mode 100644
index 00000000000000..aaf35322eede6f
--- /dev/null
+++ b/llvm/test/MC/RISCV/rvv/xsfvqmacc-invalid.s
@@ -0,0 +1,34 @@
+# RUN: not llvm-mc -triple=riscv64 -show-encoding -mattr=+v,+xsfvqmaccqoq,+xsfvqmaccdod %s 2>&1 \
+# RUN: | FileCheck %s --check-prefix=CHECK-ERROR
+
+sf.vqmaccu.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccu.2x8x2 v8, v8, v20
+
+sf.vqmacc.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmacc.2x8x2 v8, v8, v20
+
+sf.vqmaccus.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccus.2x8x2 v8, v8, v20
+
+sf.vqmaccsu.2x8x2 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccsu.2x8x2 v8, v8, v20
+
+sf.vqmaccu.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccu.4x8x4 v8, v8, v20
+
+sf.vqmacc.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmacc.4x8x4 v8, v8, v20
+
+sf.vqmaccus.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccus.4x8x4 v8, v8, v20
+
+sf.vqmaccsu.4x8x4 v8, v8, v20
+# CHECK-ERROR: the destination vector register group cannot overlap the source vector register group{{$}}
+# CHECK-ERROR-LABEL: sf.vqmaccsu.4x8x4 v8, v8, v20
|
topperc
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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 instructions don't allow the vs1 encoded register to overlap vd. Confusingly these instructions order their operands vd, vs1, vs2 while every other vector instruction is vd, vs2, vs1. So we need to use VS2Constraint for this since it checks the first operand after vd.
2 of the 3 extensions have instruction that produce result with EMUL=2*LMUL. This makes them subject to the widening constraints for vs2. So for these extensions we use WidenV which includes VS2Constraint.