[llvm] d46f8c6 - ValueTracking: Handle exp/exp2 in computeKnownFPClass
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Mon Apr 24 11:25:19 PDT 2023
Author: Matt Arsenault
Date: 2023-04-24T14:25:06-04:00
New Revision: d46f8c6ec98e3965139d055cc27c34e946f74a1d
URL: https://github.com/llvm/llvm-project/commit/d46f8c6ec98e3965139d055cc27c34e946f74a1d
DIFF: https://github.com/llvm/llvm-project/commit/d46f8c6ec98e3965139d055cc27c34e946f74a1d.diff
LOG: ValueTracking: Handle exp/exp2 in computeKnownFPClass
Added:
Modified:
llvm/lib/Analysis/ValueTracking.cpp
llvm/test/Transforms/Attributor/nofpclass-exp.ll
Removed:
################################################################################
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index d361bf436517..b2d1406e98f3 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -4596,6 +4596,22 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,
Known.knownNot(fcNan);
break;
}
+ case Intrinsic::exp:
+ case Intrinsic::exp2: {
+ Known.knownNot(fcNegative);
+ if ((InterestedClasses & fcNan) == fcNone)
+ break;
+
+ KnownFPClass KnownSrc;
+ computeKnownFPClass(II->getArgOperand(0), DemandedElts,
+ InterestedClasses, KnownSrc, Depth + 1, Q, TLI);
+ if (KnownSrc.isKnownNeverNaN()) {
+ Known.knownNot(fcNan);
+ Known.SignBit = false;
+ }
+
+ break;
+ }
case Intrinsic::arithmetic_fence: {
computeKnownFPClass(II->getArgOperand(0), DemandedElts,
InterestedClasses, Known, Depth + 1, Q, TLI);
diff --git a/llvm/test/Transforms/Attributor/nofpclass-exp.ll b/llvm/test/Transforms/Attributor/nofpclass-exp.ll
index afc1c846eb43..9c82c60f14ed 100644
--- a/llvm/test/Transforms/Attributor/nofpclass-exp.ll
+++ b/llvm/test/Transforms/Attributor/nofpclass-exp.ll
@@ -5,9 +5,9 @@ declare float @llvm.exp.f32(float)
declare float @llvm.exp2.f32(float)
define float @ret_exp(float %arg0) {
-; CHECK-LABEL: define float @ret_exp
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp
; CHECK-SAME: (float [[ARG0:%.*]]) #[[ATTR1:[0-9]+]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2:[0-9]+]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2:[0-9]+]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -15,9 +15,9 @@ define float @ret_exp(float %arg0) {
}
define float @ret_exp_noinf(float nofpclass(inf) %arg0) {
-; CHECK-LABEL: define float @ret_exp_noinf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp_noinf
; CHECK-SAME: (float nofpclass(inf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -25,9 +25,9 @@ define float @ret_exp_noinf(float nofpclass(inf) %arg0) {
}
define float @ret_exp_nopinf(float nofpclass(pinf) %arg0) {
-; CHECK-LABEL: define float @ret_exp_nopinf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp_nopinf
; CHECK-SAME: (float nofpclass(pinf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -35,9 +35,9 @@ define float @ret_exp_nopinf(float nofpclass(pinf) %arg0) {
}
define float @ret_exp_noninf(float nofpclass(ninf) %arg0) {
-; CHECK-LABEL: define float @ret_exp_noninf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp_noninf
; CHECK-SAME: (float nofpclass(ninf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -45,9 +45,9 @@ define float @ret_exp_noninf(float nofpclass(ninf) %arg0) {
}
define float @ret_exp_nonan(float nofpclass(nan) %arg0) {
-; CHECK-LABEL: define float @ret_exp_nonan
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp_nonan
; CHECK-SAME: (float nofpclass(nan) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -55,9 +55,9 @@ define float @ret_exp_nonan(float nofpclass(nan) %arg0) {
}
define float @ret_exp_nonan_noinf(float nofpclass(nan inf) %arg0) {
-; CHECK-LABEL: define float @ret_exp_nonan_noinf
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp_nonan_noinf
; CHECK-SAME: (float nofpclass(nan inf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -65,9 +65,9 @@ define float @ret_exp_nonan_noinf(float nofpclass(nan inf) %arg0) {
}
define float @ret_exp_nonan_noinf_nozero(float nofpclass(nan inf zero) %arg0) {
-; CHECK-LABEL: define float @ret_exp_nonan_noinf_nozero
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp_nonan_noinf_nozero
; CHECK-SAME: (float nofpclass(nan inf zero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -75,9 +75,9 @@ define float @ret_exp_nonan_noinf_nozero(float nofpclass(nan inf zero) %arg0) {
}
define float @ret_exp_noinf_nozero(float nofpclass(inf zero) %arg0) {
-; CHECK-LABEL: define float @ret_exp_noinf_nozero
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp_noinf_nozero
; CHECK-SAME: (float nofpclass(inf zero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -85,9 +85,9 @@ define float @ret_exp_noinf_nozero(float nofpclass(inf zero) %arg0) {
}
define float @ret_exp_noinf_nonegzero(float nofpclass(inf nzero) %arg0) {
-; CHECK-LABEL: define float @ret_exp_noinf_nonegzero
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp_noinf_nonegzero
; CHECK-SAME: (float nofpclass(inf nzero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp.f32(float %arg0)
@@ -95,10 +95,10 @@ define float @ret_exp_noinf_nonegzero(float nofpclass(inf nzero) %arg0) {
}
define float @ret_exp_positive_source(i32 %arg) {
-; CHECK-LABEL: define float @ret_exp_positive_source
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp_positive_source
; CHECK-SAME: (i32 [[ARG:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[UITOFP:%.*]] = uitofp i32 [[ARG]] to float
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[UITOFP]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp.f32(float [[UITOFP]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%uitofp = uitofp i32 %arg to float
@@ -108,10 +108,10 @@ define float @ret_exp_positive_source(i32 %arg) {
; Could produce a nan because we don't know if the multiply is negative.
define float @ret_exp_unknown_sign(float nofpclass(nan) %arg0, float nofpclass(nan) %arg1) {
-; CHECK-LABEL: define float @ret_exp_unknown_sign
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp_unknown_sign
; CHECK-SAME: (float nofpclass(nan) [[ARG0:%.*]], float nofpclass(nan) [[ARG1:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[UNKNOWN_SIGN_NOT_NAN:%.*]] = fmul nnan float [[ARG0]], [[ARG1]]
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp.f32(float [[UNKNOWN_SIGN_NOT_NAN]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp.f32(float [[UNKNOWN_SIGN_NOT_NAN]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%unknown.sign.not.nan = fmul nnan float %arg0, %arg1
@@ -120,9 +120,9 @@ define float @ret_exp_unknown_sign(float nofpclass(nan) %arg0, float nofpclass(n
}
define float @ret_exp2(float %arg0) {
-; CHECK-LABEL: define float @ret_exp2
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2
; CHECK-SAME: (float [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -130,9 +130,9 @@ define float @ret_exp2(float %arg0) {
}
define float @ret_exp2_noinf(float nofpclass(inf) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_noinf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2_noinf
; CHECK-SAME: (float nofpclass(inf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -140,9 +140,9 @@ define float @ret_exp2_noinf(float nofpclass(inf) %arg0) {
}
define float @ret_exp2_nopinf(float nofpclass(pinf) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_nopinf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2_nopinf
; CHECK-SAME: (float nofpclass(pinf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -150,9 +150,9 @@ define float @ret_exp2_nopinf(float nofpclass(pinf) %arg0) {
}
define float @ret_exp2_noninf(float nofpclass(ninf) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_noninf
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2_noninf
; CHECK-SAME: (float nofpclass(ninf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -160,9 +160,9 @@ define float @ret_exp2_noninf(float nofpclass(ninf) %arg0) {
}
define float @ret_exp2_nonan(float nofpclass(nan) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_nonan
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp2_nonan
; CHECK-SAME: (float nofpclass(nan) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -170,9 +170,9 @@ define float @ret_exp2_nonan(float nofpclass(nan) %arg0) {
}
define float @ret_exp2_nonan_noinf(float nofpclass(nan inf) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_nonan_noinf
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp2_nonan_noinf
; CHECK-SAME: (float nofpclass(nan inf) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -180,9 +180,9 @@ define float @ret_exp2_nonan_noinf(float nofpclass(nan inf) %arg0) {
}
define float @ret_exp2_nonan_noinf_nozero(float nofpclass(nan inf zero) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_nonan_noinf_nozero
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp2_nonan_noinf_nozero
; CHECK-SAME: (float nofpclass(nan inf zero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -190,9 +190,9 @@ define float @ret_exp2_nonan_noinf_nozero(float nofpclass(nan inf zero) %arg0) {
}
define float @ret_exp2_noinf_nozero(float nofpclass(inf zero) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_noinf_nozero
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2_noinf_nozero
; CHECK-SAME: (float nofpclass(inf zero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -200,9 +200,9 @@ define float @ret_exp2_noinf_nozero(float nofpclass(inf zero) %arg0) {
}
define float @ret_exp2_noinf_nonegzero(float nofpclass(inf nzero) %arg0) {
-; CHECK-LABEL: define float @ret_exp2_noinf_nonegzero
+; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @ret_exp2_noinf_nonegzero
; CHECK-SAME: (float nofpclass(inf nzero) [[ARG0:%.*]]) #[[ATTR1]] {
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[ARG0]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%call = call float @llvm.exp2.f32(float %arg0)
@@ -210,10 +210,10 @@ define float @ret_exp2_noinf_nonegzero(float nofpclass(inf nzero) %arg0) {
}
define float @ret_exp2_positive_source(i32 %arg) {
-; CHECK-LABEL: define float @ret_exp2_positive_source
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp2_positive_source
; CHECK-SAME: (i32 [[ARG:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[UITOFP:%.*]] = uitofp i32 [[ARG]] to float
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[UITOFP]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[UITOFP]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%uitofp = uitofp i32 %arg to float
@@ -223,10 +223,10 @@ define float @ret_exp2_positive_source(i32 %arg) {
; Could produce a nan because we don't know if the multiply is negative.
define float @ret_exp2_unknown_sign(float nofpclass(nan) %arg0, float nofpclass(nan) %arg1) {
-; CHECK-LABEL: define float @ret_exp2_unknown_sign
+; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @ret_exp2_unknown_sign
; CHECK-SAME: (float nofpclass(nan) [[ARG0:%.*]], float nofpclass(nan) [[ARG1:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[UNKNOWN_SIGN_NOT_NAN:%.*]] = fmul nnan float [[ARG0]], [[ARG1]]
-; CHECK-NEXT: [[CALL:%.*]] = call float @llvm.exp2.f32(float [[UNKNOWN_SIGN_NOT_NAN]]) #[[ATTR2]]
+; CHECK-NEXT: [[CALL:%.*]] = call nofpclass(nan ninf nzero nsub nnorm) float @llvm.exp2.f32(float [[UNKNOWN_SIGN_NOT_NAN]]) #[[ATTR2]]
; CHECK-NEXT: ret float [[CALL]]
;
%unknown.sign.not.nan = fmul nnan float %arg0, %arg1
More information about the llvm-commits
mailing list