This is an archive of the discontinued LLVM Phabricator instance.

Differential D151558

[mlir][llvm] Add rounding intrinsics
ClosedPublic

Authored by LukasSommerTu on May 26 2023, 6:04 AM.

Details

Reviewers
ftynse
dcaballe
nicolasvasilache
gysit
zero9178
Dinistro
Commits
rG196d89740c5e: [mlir][llvm] Add rounding intrinsics
Summary

Add some of the missing libm rounding intrinsics to the LLVM dialect:

  • llvm.rint
  • llvm.nearbyint
  • llvm.lround
  • llvm.llround
  • llvm.lrint
  • llvm.llrint

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

Diff Detail

Event Timeline

LukasSommerTu created this revision.May 26 2023, 6:04 AM
Herald added a reviewer: ftynse. · View Herald TranscriptMay 26 2023, 6:04 AM
Herald added a reviewer: dcaballe. · View Herald Transcript
Herald added a project: Restricted Project. · View Herald Transcript
Herald added subscribers: bviyer, Moerafaat, zero9178 and 22 others. · View Herald Transcript
LukasSommerTu requested review of this revision.May 26 2023, 6:04 AM
Herald added a reviewer: nicolasvasilache. · View Herald TranscriptMay 26 2023, 6:04 AM
Herald added a project: Restricted Project. · View Herald Transcript
Herald added subscribers: stephenneuendorffer, nicolasvasilache. · View Herald Transcript
LukasSommerTu added reviewers: gysit, zero9178, Dinistro.May 26 2023, 6:06 AM
Harbormaster completed remote builds in B234841: Diff 526038.May 26 2023, 6:45 AM
zero9178 accepted this revision.May 26 2023, 7:07 AM

LGTM!

This revision is now accepted and ready to land.May 26 2023, 7:07 AM
gysit accepted this revision.May 27 2023, 11:48 AM

Thanks,

LGTM modulo some nit comments!

mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td
133

nit: I know I did not do it consistently myself but can you postfix the intrinsic names with Op, e.g.:

LLVM_RIntOp
LLVM_NearbyIntOp

etc?

mlir/test/Target/LLVMIR/Import/intrinsic.ll
133

nit: Can you put a CHECK-LABEL as well for consistency

mlir/test/Target/LLVMIR/llvmir-intrinsics.mlir
163

nit; Can you put a CHECK-LABEL for consistency here and below.

LukasSommerTu updated this revision to Diff 526415.May 29 2023, 1:54 AM

Add postfix to name and missing CHECK-LABEL

LukasSommerTu marked 3 inline comments as done.May 29 2023, 1:55 AM

@zero9178 @gysit Thanks for your reviews!

I addressed the comments in the latest update.

I do not have commit access, so if you are happy with the latest update, could someone please commit this for me? Thanks!

Harbormaster completed remote builds in B235134: Diff 526415.May 29 2023, 2:17 AM
Closed by commit rG196d89740c5e: [mlir][llvm] Add rounding intrinsics (authored by LukasSommerTu, committed by zero9178). · Explain WhyMay 29 2023, 9:12 AM
This revision was automatically updated to reflect the committed changes.
zero9178 added a commit: rG196d89740c5e: [mlir][llvm] Add rounding intrinsics.

Revision Contents

PathSize
mlir/
include/
mlir/
Dialect/
LLVMIR/
12 lines
test/
Target/
LLVMIR/
Import/
86 lines
90 lines

Diff 526450

mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td

Show First 20 Lines • Show All 124 Lines • ▼ Show 20 Linesdef LLVM_PowIOp : LLVM_OneResultIntrOp<"powi", [], [0,1],
[Pure], /*requiresFastmath=*/1> { [Pure], /*requiresFastmath=*/1> {
let arguments = let arguments =
(ins LLVM_ScalarOrVectorOf<LLVM_AnyFloat>:$val, (ins LLVM_ScalarOrVectorOf<LLVM_AnyFloat>:$val,
AnySignlessInteger:$power, AnySignlessInteger:$power,
DefaultValuedAttr<LLVM_FastmathFlagsAttr, "{}">:$fastmathFlags); DefaultValuedAttr<LLVM_FastmathFlagsAttr, "{}">:$fastmathFlags);
let assemblyFormat = "`(` operands `)` custom<LLVMOpAttrs>(attr-dict) `:` " let assemblyFormat = "`(` operands `)` custom<LLVMOpAttrs>(attr-dict) `:` "
"functional-type(operands, results)"; "functional-type(operands, results)";
} }
def LLVM_RintOp : LLVM_UnaryIntrOpF<"rint">;
gysitUnsubmitted
Done
ReplyInline Actions

nit: I know I did not do it consistently myself but can you postfix the intrinsic names with Op, e.g.:

LLVM_RIntOp
LLVM_NearbyIntOp

etc?

gysit: nit: I know I did not do it consistently myself but can you postfix the intrinsic names with Op…
def LLVM_NearbyintOp : LLVM_UnaryIntrOpF<"nearbyint">;
class LLVM_IntRoundIntrOpBase<string func> :
LLVM_OneResultIntrOp<func, [0], [0], [Pure]> {
let arguments = (ins LLVM_AnyFloat:$val);
let assemblyFormat = "`(` operands `)` custom<LLVMOpAttrs>(attr-dict) `:` "
"functional-type(operands, results)";
}
def LLVM_LroundOp : LLVM_IntRoundIntrOpBase<"lround">;
def LLVM_LlroundOp : LLVM_IntRoundIntrOpBase<"llround">;
def LLVM_LrintOp : LLVM_IntRoundIntrOpBase<"lrint">;
def LLVM_LlrintOp : LLVM_IntRoundIntrOpBase<"llrint">;
def LLVM_BitReverseOp : LLVM_UnaryIntrOpI<"bitreverse">; def LLVM_BitReverseOp : LLVM_UnaryIntrOpI<"bitreverse">;
def LLVM_ByteSwapOp : LLVM_UnaryIntrOpI<"bswap">; def LLVM_ByteSwapOp : LLVM_UnaryIntrOpI<"bswap">;
def LLVM_CountLeadingZerosOp : LLVM_CountZerosIntrOp<"ctlz">; def LLVM_CountLeadingZerosOp : LLVM_CountZerosIntrOp<"ctlz">;
def LLVM_CountTrailingZerosOp : LLVM_CountZerosIntrOp<"cttz">; def LLVM_CountTrailingZerosOp : LLVM_CountZerosIntrOp<"cttz">;
def LLVM_CtPopOp : LLVM_UnaryIntrOpI<"ctpop">; def LLVM_CtPopOp : LLVM_UnaryIntrOpI<"ctpop">;
def LLVM_FshlOp : LLVM_TernarySameArgsIntrOpI<"fshl">; def LLVM_FshlOp : LLVM_TernarySameArgsIntrOpI<"fshl">;
def LLVM_FshrOp : LLVM_TernarySameArgsIntrOpI<"fshr">; def LLVM_FshrOp : LLVM_TernarySameArgsIntrOpI<"fshr">;
def LLVM_MaxNumOp : LLVM_BinarySameArgsIntrOpF<"maxnum">; def LLVM_MaxNumOp : LLVM_BinarySameArgsIntrOpF<"maxnum">;
▲ Show 20 Lines • Show All 892 Lines • Show Last 20 Lines

mlir/test/Target/LLVMIR/Import/intrinsic.ll

Show First 20 Lines • Show All 111 Lines • ▼ Show 20 Lines
; CHECK-LABEL: llvm.func @pow_test ; CHECK-LABEL: llvm.func @pow_test
define void @pow_test(float %0, float %1, <8 x float> %2, <8 x float> %3) { define void @pow_test(float %0, float %1, <8 x float> %2, <8 x float> %3) {
; CHECK: llvm.intr.pow(%{{.*}}, %{{.*}}) : (f32, f32) -> f32 ; CHECK: llvm.intr.pow(%{{.*}}, %{{.*}}) : (f32, f32) -> f32
%5 = call float @llvm.pow.f32(float %0, float %1) %5 = call float @llvm.pow.f32(float %0, float %1)
; CHECK: llvm.intr.pow(%{{.*}}, %{{.*}}) : (vector<8xf32>, vector<8xf32>) -> vector<8xf32> ; CHECK: llvm.intr.pow(%{{.*}}, %{{.*}}) : (vector<8xf32>, vector<8xf32>) -> vector<8xf32>
%6 = call <8 x float> @llvm.pow.v8f32(<8 x float> %2, <8 x float> %3) %6 = call <8 x float> @llvm.pow.v8f32(<8 x float> %2, <8 x float> %3)
ret void ret void
} }
; CHECK-LABEL: llvm.func @rint_test
define void @rint_test(float %0, double %1, <8 x float> %2, <8 x double> %3) {
; CHECK: llvm.intr.rint(%{{.*}}) : (f32) -> f32
%5 = call float @llvm.rint.f32(float %0)
; CHECK: llvm.intr.rint(%{{.*}}) : (f64) -> f64
%6 = call double @llvm.rint.f64(double %1)
; CHECK: llvm.intr.rint(%{{.*}}) : (vector<8xf32>) -> vector<8xf32>
%7 = call <8 x float> @llvm.rint.v8f32(<8 x float> %2)
; CHECK: llvm.intr.rint(%{{.*}}) : (vector<8xf64>) -> vector<8xf64>
%8 = call <8 x double> @llvm.rint.v8f64(<8 x double> %3)
ret void
}
; CHECK-LABEL: llvm.func @nearbyint_test
gysitUnsubmitted
Done
ReplyInline Actions

nit: Can you put a CHECK-LABEL as well for consistency

gysit: nit: Can you put a CHECK-LABEL as well for consistency
define void @nearbyint_test(float %0, double %1, <8 x float> %2, <8 x double> %3) {
; CHECK: llvm.intr.nearbyint(%{{.*}}) : (f32) -> f32
%5 = call float @llvm.nearbyint.f32(float %0)
; CHECK: llvm.intr.nearbyint(%{{.*}}) : (f64) -> f64
%6 = call double @llvm.nearbyint.f64(double %1)
; CHECK: llvm.intr.nearbyint(%{{.*}}) : (vector<8xf32>) -> vector<8xf32>
%7 = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> %2)
; CHECK: llvm.intr.nearbyint(%{{.*}}) : (vector<8xf64>) -> vector<8xf64>
%8 = call <8 x double> @llvm.nearbyint.v8f64(<8 x double> %3)
ret void
}
; CHECK-LABEL: llvm.func @lround_test
define void @lround_test(float %0, double %1) {
; CHECK: llvm.intr.lround(%{{.*}}) : (f32) -> i32
%3 = call i32 @llvm.lround.i32.f32(float %0)
; CHECK: llvm.intr.lround(%{{.*}}) : (f32) -> i64
%4 = call i64 @llvm.lround.i64.f32(float %0)
; CHECK: llvm.intr.lround(%{{.*}}) : (f64) -> i32
%5 = call i32 @llvm.lround.i32.f64(double %1)
; CHECK: llvm.intr.lround(%{{.*}}) : (f64) -> i64
%6 = call i64 @llvm.lround.i64.f64(double %1)
ret void
}
; CHECK-LABEL: llvm.func @llround_test
define void @llround_test(float %0, double %1) {
; CHECK: llvm.intr.llround(%{{.*}}) : (f32) -> i64
%3 = call i64 @llvm.llround.i64.f32(float %0)
; CHECK: llvm.intr.llround(%{{.*}}) : (f64) -> i64
%4 = call i64 @llvm.llround.i64.f64(double %1)
ret void
}
; CHECK-LABEL: llvm.func @lrint_test
define void @lrint_test(float %0, double %1) {
; CHECK: llvm.intr.lrint(%{{.*}}) : (f32) -> i32
%3 = call i32 @llvm.lrint.i32.f32(float %0)
; CHECK: llvm.intr.lrint(%{{.*}}) : (f32) -> i64
%4 = call i64 @llvm.lrint.i64.f32(float %0)
; CHECK: llvm.intr.lrint(%{{.*}}) : (f64) -> i32
%5 = call i32 @llvm.lrint.i32.f64(double %1)
; CHECK: llvm.intr.lrint(%{{.*}}) : (f64) -> i64
%6 = call i64 @llvm.lrint.i64.f64(double %1)
ret void
}
; CHECK-LABEL: llvm.func @llrint_test
define void @llrint_test(float %0, double %1) {
; CHECK: llvm.intr.llrint(%{{.*}}) : (f32) -> i64
%3 = call i64 @llvm.llrint.i64.f32(float %0)
; CHECK: llvm.intr.llrint(%{{.*}}) : (f64) -> i64
%4 = call i64 @llvm.llrint.i64.f64(double %1)
ret void
}
; CHECK-LABEL: llvm.func @bitreverse_test ; CHECK-LABEL: llvm.func @bitreverse_test
define void @bitreverse_test(i32 %0, <8 x i32> %1) { define void @bitreverse_test(i32 %0, <8 x i32> %1) {
; CHECK: llvm.intr.bitreverse(%{{.*}}) : (i32) -> i32 ; CHECK: llvm.intr.bitreverse(%{{.*}}) : (i32) -> i32
%3 = call i32 @llvm.bitreverse.i32(i32 %0) %3 = call i32 @llvm.bitreverse.i32(i32 %0)
; CHECK: llvm.intr.bitreverse(%{{.*}}) : (vector<8xi32>) -> vector<8xi32> ; CHECK: llvm.intr.bitreverse(%{{.*}}) : (vector<8xi32>) -> vector<8xi32>
%4 = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %1) %4 = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %1)
ret void ret void
} }
▲ Show 20 Lines • Show All 648 Lines • ▼ Show 20 Lines
declare float @llvm.floor.f32(float) declare float @llvm.floor.f32(float)
declare <8 x float> @llvm.floor.v8f32(<8 x float>) declare <8 x float> @llvm.floor.v8f32(<8 x float>)
declare float @llvm.cos.f32(float) declare float @llvm.cos.f32(float)
declare <8 x float> @llvm.cos.v8f32(<8 x float>) declare <8 x float> @llvm.cos.v8f32(<8 x float>)
declare float @llvm.copysign.f32(float, float) declare float @llvm.copysign.f32(float, float)
declare <8 x float> @llvm.copysign.v8f32(<8 x float>, <8 x float>) declare <8 x float> @llvm.copysign.v8f32(<8 x float>, <8 x float>)
declare float @llvm.pow.f32(float, float) declare float @llvm.pow.f32(float, float)
declare <8 x float> @llvm.pow.v8f32(<8 x float>, <8 x float>) declare <8 x float> @llvm.pow.v8f32(<8 x float>, <8 x float>)
declare float @llvm.rint.f32(float)
declare double @llvm.rint.f64(double)
declare <8 x float> @llvm.rint.v8f32(<8 x float>)
declare <8 x double> @llvm.rint.v8f64(<8 x double>)
declare float @llvm.nearbyint.f32(float)
declare double @llvm.nearbyint.f64(double)
declare <8 x float> @llvm.nearbyint.v8f32(<8 x float>)
declare <8 x double> @llvm.nearbyint.v8f64(<8 x double>)
declare i32 @llvm.lround.i32.f32(float)
declare i64 @llvm.lround.i64.f32(float)
declare i32 @llvm.lround.i32.f64(double)
declare i64 @llvm.lround.i64.f64(double)
declare i64 @llvm.llround.i64.f32(float)
declare i64 @llvm.llround.i64.f64(double)
declare i32 @llvm.lrint.i32.f32(float)
declare i64 @llvm.lrint.i64.f32(float)
declare i32 @llvm.lrint.i32.f64(double)
declare i64 @llvm.lrint.i64.f64(double)
declare i64 @llvm.llrint.i64.f32(float)
declare i64 @llvm.llrint.i64.f64(double)
declare i32 @llvm.bitreverse.i32(i32) declare i32 @llvm.bitreverse.i32(i32)
declare <8 x i32> @llvm.bitreverse.v8i32(<8 x i32>) declare <8 x i32> @llvm.bitreverse.v8i32(<8 x i32>)
declare i32 @llvm.bswap.i32(i32) declare i32 @llvm.bswap.i32(i32)
declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>) declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>)
declare i32 @llvm.ctlz.i32(i32, i1 immarg) declare i32 @llvm.ctlz.i32(i32, i1 immarg)
declare <8 x i32> @llvm.ctlz.v8i32(<8 x i32>, i1 immarg) declare <8 x i32> @llvm.ctlz.v8i32(<8 x i32>, i1 immarg)
declare i32 @llvm.cttz.i32(i32, i1 immarg) declare i32 @llvm.cttz.i32(i32, i1 immarg)
declare <8 x i32> @llvm.cttz.v8i32(<8 x i32>, i1 immarg) declare <8 x i32> @llvm.cttz.v8i32(<8 x i32>, i1 immarg)
▲ Show 20 Lines • Show All 149 Lines • Show Last 20 Lines

mlir/test/Target/LLVMIR/llvmir-intrinsics.mlir

Show First 20 Lines • Show All 128 Lines • ▼ Show 20 Lines
llvm.func @pow_test(%arg0: f32, %arg1: f32, %arg2: vector<8xf32>, %arg3: vector<8xf32>) { llvm.func @pow_test(%arg0: f32, %arg1: f32, %arg2: vector<8xf32>, %arg3: vector<8xf32>) {
// CHECK: call float @llvm.pow.f32 // CHECK: call float @llvm.pow.f32
"llvm.intr.pow"(%arg0, %arg1) : (f32, f32) -> f32 "llvm.intr.pow"(%arg0, %arg1) : (f32, f32) -> f32
// CHECK: call <8 x float> @llvm.pow.v8f32 // CHECK: call <8 x float> @llvm.pow.v8f32
"llvm.intr.pow"(%arg2, %arg3) : (vector<8xf32>, vector<8xf32>) -> vector<8xf32> "llvm.intr.pow"(%arg2, %arg3) : (vector<8xf32>, vector<8xf32>) -> vector<8xf32>
llvm.return llvm.return
} }
// CHECK-LABEL: @rint_test
llvm.func @rint_test(%arg0 : f32, %arg1 : f64, %arg2 : vector<8xf32>, %arg3 : vector<8xf64>) {
// CHECK: call float @llvm.rint.f32
"llvm.intr.rint"(%arg0) : (f32) -> f32
// CHECK: call double @llvm.rint.f64
"llvm.intr.rint"(%arg1) : (f64) -> f64
// CHECK: call <8 x float> @llvm.rint.v8f32
"llvm.intr.rint"(%arg2) : (vector<8xf32>) -> vector<8xf32>
// CHECK: call <8 x double> @llvm.rint.v8f64
"llvm.intr.rint"(%arg3) : (vector<8xf64>) -> vector<8xf64>
llvm.return
}
// CHECK-LABEL: @nearbyint_test
llvm.func @nearbyint_test(%arg0 : f32, %arg1 : f64, %arg2 : vector<8xf32>, %arg3 : vector<8xf64>) {
// CHECK: call float @llvm.nearbyint.f32
"llvm.intr.nearbyint"(%arg0) : (f32) -> f32
// CHECK: call double @llvm.nearbyint.f64
"llvm.intr.nearbyint"(%arg1) : (f64) -> f64
// CHECK: call <8 x float> @llvm.nearbyint.v8f32
"llvm.intr.nearbyint"(%arg2) : (vector<8xf32>) -> vector<8xf32>
// CHECK: call <8 x double> @llvm.nearbyint.v8f64
"llvm.intr.nearbyint"(%arg3) : (vector<8xf64>) -> vector<8xf64>
llvm.return
}
// CHECK-LABEL: @lround_test
gysitUnsubmitted
Done
ReplyInline Actions

nit; Can you put a CHECK-LABEL for consistency here and below.

gysit: nit; Can you put a CHECK-LABEL for consistency here and below.
llvm.func @lround_test(%arg0 : f32, %arg1 : f64) {
// CHECK: call i32 @llvm.lround.i32.f32
"llvm.intr.lround"(%arg0) : (f32) -> i32
// CHECK: call i64 @llvm.lround.i64.f32
"llvm.intr.lround"(%arg0) : (f32) -> i64
// CHECK: call i32 @llvm.lround.i32.f64
"llvm.intr.lround"(%arg1) : (f64) -> i32
// CHECK: call i64 @llvm.lround.i64.f64
"llvm.intr.lround"(%arg1) : (f64) -> i64
llvm.return
}
// CHECK-LABEL: @llround_test
llvm.func @llround_test(%arg0 : f32, %arg1 : f64) {
// CHECK: call i64 @llvm.llround.i64.f32
"llvm.intr.llround"(%arg0) : (f32) -> i64
// CHECK: call i64 @llvm.llround.i64.f64
"llvm.intr.llround"(%arg1) : (f64) -> i64
llvm.return
}
// CHECK-LABEL: @lrint_test
llvm.func @lrint_test(%arg0 : f32, %arg1 : f64) {
// CHECK: call i32 @llvm.lrint.i32.f32
"llvm.intr.lrint"(%arg0) : (f32) -> i32
// CHECK: call i64 @llvm.lrint.i64.f32
"llvm.intr.lrint"(%arg0) : (f32) -> i64
// CHECK: call i32 @llvm.lrint.i32.f64
"llvm.intr.lrint"(%arg1) : (f64) -> i32
// CHECK: call i64 @llvm.lrint.i64.f64
"llvm.intr.lrint"(%arg1) : (f64) -> i64
llvm.return
}
// CHECK-LABEL: @llrint_test
llvm.func @llrint_test(%arg0 : f32, %arg1 : f64) {
// CHECK: call i64 @llvm.llrint.i64.f32
"llvm.intr.llrint"(%arg0) : (f32) -> i64
// CHECK: call i64 @llvm.llrint.i64.f64
"llvm.intr.llrint"(%arg1) : (f64) -> i64
llvm.return
}
// CHECK-LABEL: @bitreverse_test // CHECK-LABEL: @bitreverse_test
llvm.func @bitreverse_test(%arg0: i32, %arg1: vector<8xi32>) { llvm.func @bitreverse_test(%arg0: i32, %arg1: vector<8xi32>) {
// CHECK: call i32 @llvm.bitreverse.i32 // CHECK: call i32 @llvm.bitreverse.i32
"llvm.intr.bitreverse"(%arg0) : (i32) -> i32 "llvm.intr.bitreverse"(%arg0) : (i32) -> i32
// CHECK: call <8 x i32> @llvm.bitreverse.v8i32 // CHECK: call <8 x i32> @llvm.bitreverse.v8i32
"llvm.intr.bitreverse"(%arg1) : (vector<8xi32>) -> vector<8xi32> "llvm.intr.bitreverse"(%arg1) : (vector<8xi32>) -> vector<8xi32>
llvm.return llvm.return
} }
▲ Show 20 Lines • Show All 715 Lines • ▼ Show 20 Lines
// CHECK-DAG: declare <8 x float> @llvm.fabs.v8f32(<8 x float>) #0 // CHECK-DAG: declare <8 x float> @llvm.fabs.v8f32(<8 x float>) #0
// CHECK-DAG: declare float @llvm.sqrt.f32(float) // CHECK-DAG: declare float @llvm.sqrt.f32(float)
// CHECK-DAG: declare <8 x float> @llvm.sqrt.v8f32(<8 x float>) #0 // CHECK-DAG: declare <8 x float> @llvm.sqrt.v8f32(<8 x float>) #0
// CHECK-DAG: declare float @llvm.ceil.f32(float) // CHECK-DAG: declare float @llvm.ceil.f32(float)
// CHECK-DAG: declare <8 x float> @llvm.ceil.v8f32(<8 x float>) #0 // CHECK-DAG: declare <8 x float> @llvm.ceil.v8f32(<8 x float>) #0
// CHECK-DAG: declare float @llvm.cos.f32(float) // CHECK-DAG: declare float @llvm.cos.f32(float)
// CHECK-DAG: declare <8 x float> @llvm.cos.v8f32(<8 x float>) #0 // CHECK-DAG: declare <8 x float> @llvm.cos.v8f32(<8 x float>) #0
// CHECK-DAG: declare float @llvm.copysign.f32(float, float) // CHECK-DAG: declare float @llvm.copysign.f32(float, float)
// CHECK-DAG: declare float @llvm.rint.f32(float)
// CHECK-DAG: declare double @llvm.rint.f64(double)
// CHECK-DAG: declare <8 x float> @llvm.rint.v8f32(<8 x float>)
// CHECK-DAG: declare <8 x double> @llvm.rint.v8f64(<8 x double>)
// CHECK-DAG: declare float @llvm.nearbyint.f32(float)
// CHECK-DAG: declare double @llvm.nearbyint.f64(double)
// CHECK-DAG: declare <8 x float> @llvm.nearbyint.v8f32(<8 x float>)
// CHECK-DAG: declare <8 x double> @llvm.nearbyint.v8f64(<8 x double>)
// CHECK-DAG: declare i32 @llvm.lround.i32.f32(float)
// CHECK-DAG: declare i64 @llvm.lround.i64.f32(float)
// CHECK-DAG: declare i32 @llvm.lround.i32.f64(double)
// CHECK-DAG: declare i64 @llvm.lround.i64.f64(double)
// CHECK-DAG: declare i64 @llvm.llround.i64.f32(float)
// CHECK-DAG: declare i64 @llvm.llround.i64.f64(double)
// CHECK-DAG: declare i32 @llvm.lrint.i32.f32(float)
// CHECK-DAG: declare i64 @llvm.lrint.i64.f32(float)
// CHECK-DAG: declare i32 @llvm.lrint.i32.f64(double)
// CHECK-DAG: declare i64 @llvm.lrint.i64.f64(double)
// CHECK-DAG: declare i64 @llvm.llrint.i64.f32(float)
// CHECK-DAG: declare i64 @llvm.llrint.i64.f64(double)
// CHECK-DAG: declare <12 x float> @llvm.matrix.multiply.v12f32.v64f32.v48f32(<64 x float>, <48 x float>, i32 immarg, i32 immarg, i32 immarg) // CHECK-DAG: declare <12 x float> @llvm.matrix.multiply.v12f32.v64f32.v48f32(<64 x float>, <48 x float>, i32 immarg, i32 immarg, i32 immarg)
// CHECK-DAG: declare <48 x float> @llvm.matrix.transpose.v48f32(<48 x float>, i32 immarg, i32 immarg) // CHECK-DAG: declare <48 x float> @llvm.matrix.transpose.v48f32(<48 x float>, i32 immarg, i32 immarg)
// CHECK-DAG: declare <48 x float> @llvm.matrix.column.major.load.v48f32.i64(ptr nocapture, i64, i1 immarg, i32 immarg, i32 immarg) // CHECK-DAG: declare <48 x float> @llvm.matrix.column.major.load.v48f32.i64(ptr nocapture, i64, i1 immarg, i32 immarg, i32 immarg)
// CHECK-DAG: declare void @llvm.matrix.column.major.store.v48f32.i64(<48 x float>, ptr nocapture writeonly, i64, i1 immarg, i32 immarg, i32 immarg) // CHECK-DAG: declare void @llvm.matrix.column.major.store.v48f32.i64(<48 x float>, ptr nocapture writeonly, i64, i1 immarg, i32 immarg, i32 immarg)
// CHECK-DAG: declare <7 x i1> @llvm.get.active.lane.mask.v7i1.i64(i64, i64) // CHECK-DAG: declare <7 x i1> @llvm.get.active.lane.mask.v7i1.i64(i64, i64)
// CHECK-DAG: declare <7 x float> @llvm.masked.load.v7f32.p0(ptr nocapture, i32 immarg, <7 x i1>, <7 x float>) // CHECK-DAG: declare <7 x float> @llvm.masked.load.v7f32.p0(ptr nocapture, i32 immarg, <7 x i1>, <7 x float>)
// CHECK-DAG: declare void @llvm.masked.store.v7f32.p0(<7 x float>, ptr nocapture, i32 immarg, <7 x i1>) // CHECK-DAG: declare void @llvm.masked.store.v7f32.p0(<7 x float>, ptr nocapture, i32 immarg, <7 x i1>)
// CHECK-DAG: declare <7 x float> @llvm.masked.gather.v7f32.v7p0(<7 x ptr>, i32 immarg, <7 x i1>, <7 x float>) // CHECK-DAG: declare <7 x float> @llvm.masked.gather.v7f32.v7p0(<7 x ptr>, i32 immarg, <7 x i1>, <7 x float>)
▲ Show 20 Lines • Show All 96 Lines • Show Last 20 Lines