from:"Venkataramanan Kumar via Phabricator via cfe\-commits"

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

2020-11-25 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

In D88712#2414171 , @spatel wrote:

> In D88712#2413877 , 
> @venkataramanan.kumar.llvm wrote:
>
>> In D88712#2413688 , @spatel wrote:
>>
>>> In D88712#2412874 , 
>>> @venkataramanan.kumar.llvm wrote:
>>>
 In D88712#2412366 , @MaskRay 
 wrote:

> In D88712#2411841 , 
> @venkataramanan.kumar.llvm wrote:
>
>> 
>
> For your example:
>
>   extern double log (double) asm ("" "log_finite") __attribute__ 
> ((nothrow));
>   
>   double mylog (double d) { return log(d); }
>
> The intention is to emit `log_finite`, no matter the `-ffast-math`. Both 
> GCC and Clang (after this patch) emit `jmp log_finite`.
>
> The previous behavior (according to your comment, with an older glibc) 
> was undesired. So I think the right suggestion is to upgrade glibc.

 Then there optimizations like vectorization, inst combine which works on 
 the LLVM intrinsics.  But they will be not happening now  with  log_finite 
 calls.
>>>
>>> I'm not sure about the expected semantics/lowering for the finite calls, 
>>> but can you add something under 
>>> LibCallSimplifier::optimizeFloatingPointLibCall() that would turn it into 
>>> the LLVM log intrinsic with appropriate FMF? Codegen would need to be 
>>> responsible for converting it back to finite call(s) if those are available?
>>
>> Hi Sanjay I thought codegen no longer lowers them back to finite calls 
>> https://reviews.llvm.org/rGcd0926d087a85c5ee1222ca80980b4440214a822
>
> There was a comment in D74712  that we might 
> be moving too fast. If you would like to revert/adjust that patch, raise a 
> bug or post a patch to discuss? If the goal is to be able to vectorize the 
> finite calls, then we will need some way to represent those. Alternatively, 
> we could change something in the cost models to enable more unrolling, etc?

The goal is able to vectorize the finite call.  On a newer glibc 2.31 same test 
case would undergo all the optimizations.   As you said we can try to lower in 
to intrinsic in LibCallSimplifier and then generate the finite calls in code 
generator again.

I will raise a bug as of now.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88712/new/

https://reviews.llvm.org/D88712

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

2020-11-24 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

In D88712#2413688 , @spatel wrote:

> In D88712#2412874 , 
> @venkataramanan.kumar.llvm wrote:
>
>> In D88712#2412366 , @MaskRay wrote:
>>
>>> In D88712#2411841 , 
>>> @venkataramanan.kumar.llvm wrote:
>>>
 
>>>
>>> For your example:
>>>
>>>   extern double log (double) asm ("" "log_finite") __attribute__ 
>>> ((nothrow));
>>>   
>>>   double mylog (double d) { return log(d); }
>>>
>>> The intention is to emit `log_finite`, no matter the `-ffast-math`. Both 
>>> GCC and Clang (after this patch) emit `jmp log_finite`.
>>>
>>> The previous behavior (according to your comment, with an older glibc) was 
>>> undesired. So I think the right suggestion is to upgrade glibc.
>>
>> Then there optimizations like vectorization, inst combine which works on the 
>> LLVM intrinsics.  But they will be not happening now  with  log_finite calls.
>
> I'm not sure about the expected semantics/lowering for the finite calls, but 
> can you add something under LibCallSimplifier::optimizeFloatingPointLibCall() 
> that would turn it into the LLVM log intrinsic with appropriate FMF? Codegen 
> would need to be responsible for converting it back to finite call(s) if 
> those are available?

Hi Sanjay I thought codegen no longer lowers them back to finite calls 
https://reviews.llvm.org/rGcd0926d087a85c5ee1222ca80980b4440214a822


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88712/new/

https://reviews.llvm.org/D88712

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

2020-11-23 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

In D88712#2412366 , @MaskRay wrote:

> In D88712#2411841 , 
> @venkataramanan.kumar.llvm wrote:
>
>> 
>
> For your example:
>
>   extern double log (double) asm ("" "log_finite") __attribute__ ((nothrow));
>   
>   double mylog (double d) { return log(d); }
>
> The intention is to emit `log_finite`, no matter the `-ffast-math`. Both GCC 
> and Clang (after this patch) emit `jmp log_finite`.
>
> The previous behavior (according to your comment, with an older glibc) was 
> undesired. So I think the right suggestion is to upgrade glibc.

Then there optimizations like vectorization, inst combine which works on the 
LLVM intrinsics.  But they will be not happening now  with  log_finite calls.  
is it possible to add an option to disable this feature, in case anyone who 
wants to avail these optimization?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88712/new/

https://reviews.llvm.org/D88712

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

2020-11-23 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

Hi,

I am on Ubuntu 18 machine and it has finite math header .
This header is included by the glibc 2.27.  This header has this following 
definition.
extern double log (double) __asm__ ("" "__log_finite") __attribute__ 
((__nothrow__ ));

Consider the following test case

#include 
double mylog (double d) {

  return log(d);

}

Before this patch clang generates  -O2 -ffast-math
--Snip--
; Function Attrs: nounwind readnone uwtable
define dso_local double @mlog(double %d) local_unnamed_addr #0 {
entry:

  %0 = tail call fast double @llvm.log.f64(double %d)
  ret double %0

}
---Snip--

After this patch  on machines with lesser glibc versions,  for -O2 -ffast-math, 
clang generates

--Snip--
; Function Attrs: nounwind readnone uwtable
define dso_local double @mlog(double %d) local_unnamed_addr #0 {
entry:

  %call = tail call fast double @__log_finite(double %d) #2
  ret double %call

}
--Snip--

Note on latest Ubuntu 20.04.1 LTS with Glibc 2.31  header 
is not present
It is removed by glibc.  so there I see the @llvm.log.f64 intrinsic calls.

So on machines with lesser glibc versions where the header file is present this 
will be a problem.
Note the LLVM "opt" does optimizations like vectorization only on the 
intrinsic, but ignores the calls to  @__log_finite.

Example https://llvm.godbolt.org/z/765We8

I am not sure if this known issue.

It is causing some performance issues with some benchmarks  on my ubuntu 18 
machine.

Shall I file a PR for this ?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88712/new/

https://reviews.llvm.org/D88712

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-21 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

Thanks @spatel , @Florian  and @fpetrogalli  for the review comments and 
approval.   Can someone please commit it on my behalf.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-20 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 299280.
venkataramanan.kumar.llvm added a comment.

Remove an incorrect file that got attached with my earlier patch.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  clang/test/Driver/fveclib.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-VF2-VF8.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
+; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-X86-SAME: [2 x i8*] [
+; LIBMVEC-X86-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-X86-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC-X86:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC-X86:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-X86-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC-X86:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-X86-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-X86-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,373 @@
+; RUN: opt -vector-library=LIBMVEC-X86  -inject-tli-mappings -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK-LABEL:vector.body
+; CHECK:[[TMP5:%.*]] = call <4 x double> @_ZGVdN4v_sin(<4 x double> [[TMP4:%.*]])
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to double
+  %call = tail call double @sin(double %conv)
+  %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
+  store double %call, double* %arrayidx, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp eq i64 %iv.next, 1000
+  br i1 %exitcond,

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-20 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 299278.
venkataramanan.kumar.llvm added a comment.
Herald added a subscriber: dexonsmith.

Added a test case for testing  vector library calls for VF=2 and VF=8.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  clang/test/Driver/fveclib.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/CodeGen/X86/cpus-other.null
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-VF2-VF8.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
+; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-X86-SAME: [2 x i8*] [
+; LIBMVEC-X86-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-X86-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC-X86:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC-X86:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-X86-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC-X86:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-X86-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-X86-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,373 @@
+; RUN: opt -vector-library=LIBMVEC-X86  -inject-tli-mappings -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK-LABEL:vector.body
+; CHECK:[[TMP5:%.*]] = call <4 x double> @_ZGVdN4v_sin(<4 x double> [[TMP4:%.*]])
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv = sitofp i32 %tmp to double
+  %call = tail call double @sin(double %conv)
+  %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
+  store double %call, double* %arrayidx, align 4
+  %iv.next = add

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-13 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 297925.
venkataramanan.kumar.llvm added a comment.

Updated the patch as per review comments received.

The test cases are updated the checks based on the below comment from Francesco.

---Snip--
I think you are over-testing here. It is enough to check that inside the vector 
body there is a call to the vector function you have listed in the mapping.
---Snip--  
Florian also suggesting the same.

I still use metadata suggested by Sanjay . Currently testing only VF=4 . This 
is because we have float and double type lib calls in the test case and libmvec 
has vector call support for VF=4 .For VF say 8  there is vector call 
support for double types.

I can add few more test with meta data to cover floats with VF=8.   please let 
me know your suggestions.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  clang/test/Driver/fveclib.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/CodeGen/X86/cpus-other.null
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
+; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-X86-SAME: [2 x i8*] [
+; LIBMVEC-X86-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-X86-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC-X86:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC-X86:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-X86-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC-X86:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-X86-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-X86-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,373 @@
+; RUN: opt -vector-library=LIBMVEC-X86  -inject-tli-mappings -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-12 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 297654.
venkataramanan.kumar.llvm added a comment.

As per review comments from Sanjay,  updated the test case to use metadata.  
Also autogenerated the checks in the test cases using 
llvm/utils/update_test_checks.py.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
+; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-X86-SAME: [2 x i8*] [
+; LIBMVEC-X86-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-X86-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC-X86:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC-X86:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-X86-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC-X86:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-X86-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-X86-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,1010 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -vector-library=LIBMVEC-X86  -inject-tli-mappings -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+declare double @sin(double) #0
+declare float @sinf(float) #0
+declare double @llvm.sin.f64(double) #0
+declare float @llvm.sin.f32(float) #0
+
+declare double @cos(double) #0
+declare float @cosf(float) #0
+declare double @llvm.cos.f64(double) #0
+declare float @llvm.cos.f32(float) #0
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:   vector.ph:
+; CHECK-NEXT:br label [[VECTOR_BODY:%.*]]
+; CHECK:

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-11 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 297480.
venkataramanan.kumar.llvm added a comment.

Changed library naming to LIBMVEC-X86 as per comments and also selected based 
on Target Tripple in clang. 
I am still working on auto generating FileCheck for the test cases.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
+; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC-X86
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-X86-SAME: [2 x i8*] [
+; LIBMVEC-X86-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-X86-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC-X86:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC-X86:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-X86-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC-X86:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-X86-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-X86-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,325 @@
+; RUN: opt -vector-library=LIBMVEC-X86  -inject-tli-mappings -force-vector-width=4 -force-vector-interleave=1 -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+declare double @sin(double) #0
+declare float @sinf(float) #0
+declare double @llvm.sin.f64(double) #0
+declare float @llvm.sin.f32(float) #0
+
+declare double @cos(double) #0
+declare float @cosf(float) #0
+declare double @llvm.cos.f64(double) #0
+declare float @llvm.cos.f32(float) #0
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK:[[TMP5:%.*]] = call <4 x double> @_ZGVdN4v_sin(<4 x double> [[TMP4:%.*]])
+; CHECK:ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-11 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

In D88154#2314352 , @spatel wrote:

> In D88154#2310653 , 
> @venkataramanan.kumar.llvm wrote:
>
>> In D88154#2290205 , @abique wrote:
>>
>>> Looks good to me.
>>> Regarding the tests, it seems that you check if auto-vectorization takes 
>>> advantages of libmvec?
>>> Would it be interesting to have a test which declares a vector and call the 
>>> builtin sin on it?
>>>
>>> Thank you very much for the changes! :)
>>
>> do we we have built-in support for sin that takes vector types?
>>
>> I tried
>>
>> __m128d compute_sin(__m128d x)
>> {
>>
>>   return __builtin_sin(x);
>>
>> }
>>
 error: passing '__m128d' (vector of 2 'double' values) to parameter of 
 incompatible type 'double'
>
> We have LLVM intrinsics for sin/cos that may use vector types:
> http://llvm.org/docs/LangRef.html#llvm-sin-intrinsic
> ...but I don't know of a way to produce those directly from C source.

Ok I  see intrinsic for fp128 type in LangRef.

---Snip--
declare fp128 @llvm.cos.f128(fp128)

define fp128 @test_cos(float %float, double %double, fp128 %fp128) {

  %cosfp128 = call fp128 @llvm.cos.f128(fp128 %fp128)
  ret fp128 %cosfp128

}
--Snip--

f128  is treated a long double I see call to cosl.

  vmovaps %xmm2, %xmm0
   jmp cosl   

so I am not sure how to generate vector calls via built-ins.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-06 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

Pinging for review comments.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-04 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

In D88154#2290205 , @abique wrote:

> Looks good to me.
> Regarding the tests, it seems that you check if auto-vectorization takes 
> advantages of libmvec?
> Would it be interesting to have a test which declares a vector and call the 
> builtin sin on it?
>
> Thank you very much for the changes! :)

do we we have built-in support for sin that takes vector types?

I tried

__m128d compute_sin(__m128d x)
{

  return __builtin_sin(x);

}

>> error: passing '__m128d' (vector of 2 'double' values) to parameter of 
>> incompatible type 'double'


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-10-04 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm updated this revision to Diff 296035.
venkataramanan.kumar.llvm added a reviewer: spatel.
venkataramanan.kumar.llvm added a comment.
Herald added a subscriber: pengfei.

Selection of Glibc vector math library is enabled via the option 
-fvec-lib=libmvec .


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC
+; RUN: opt -vector-library=LIBMVEC -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-SAME: [2 x i8*] [
+; LIBMVEC-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-NOT:  _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,325 @@
+; RUN: opt -vector-library=LIBMVEC  -inject-tli-mappings -force-vector-width=4 -force-vector-interleave=1 -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+declare double @sin(double) #0
+declare float @sinf(float) #0
+declare double @llvm.sin.f64(double) #0
+declare float @llvm.sin.f32(float) #0
+
+declare double @cos(double) #0
+declare float @cosf(float) #0
+declare double @llvm.cos.f64(double) #0
+declare float @llvm.cos.f32(float) #0
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK:[[TMP5:%.*]] = call <4 x double> @_ZGVdN4v_sin(<4 x double> [[TMP4:%.*]])
+; CHECK:ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %tmp = trunc i64 %iv to i32
+  %conv

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-09-23 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm added a comment.

Initial version I supported the following vector functions (VF 2 and 4 ).

sin
cos
exp
pow
log

Also added test cases similar to SVML under X86. I am not sure about other 
targets.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88154/new/

https://reviews.llvm.org/D88154

___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

2020-09-23 Thread Venkataramanan Kumar via Phabricator via cfe-commits

venkataramanan.kumar.llvm created this revision.
venkataramanan.kumar.llvm added reviewers: Florian, abique.
Herald added subscribers: llvm-commits, cfe-commits, dang, dmgreen, hiraditya.
Herald added projects: clang, LLVM.
venkataramanan.kumar.llvm requested review of this revision.

Initial support for vectorization using Libmvec (GLIBC vector math library).


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D88154

Files:
  clang/include/clang/Basic/CodeGenOptions.def
  clang/include/clang/Basic/CodeGenOptions.h
  clang/include/clang/Driver/Options.td
  clang/lib/CodeGen/BackendUtil.cpp
  clang/lib/Frontend/CompilerInvocation.cpp
  clang/test/Driver/autocomplete.c
  llvm/include/llvm/Analysis/TargetLibraryInfo.h
  llvm/include/llvm/Analysis/VecFuncs.def
  llvm/lib/Analysis/TargetLibraryInfo.cpp
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
  llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
  llvm/test/Transforms/Util/add-TLI-mappings.ll

Index: llvm/test/Transforms/Util/add-TLI-mappings.ll
===
--- llvm/test/Transforms/Util/add-TLI-mappings.ll
+++ llvm/test/Transforms/Util/add-TLI-mappings.ll
@@ -3,6 +3,8 @@
 ; RUN: opt -vector-library=MASSV  -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=MASSV  -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,MASSV
 ; RUN: opt -vector-library=Accelerate -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
+; RUN: opt -vector-library=LIBMVEC -inject-tli-mappings-S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC
+; RUN: opt -vector-library=LIBMVEC -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,LIBMVEC
 ; RUN: opt -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s  --check-prefixes=COMMON,ACCELERATE
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -21,6 +23,9 @@
 ; MASSV-SAME: i8* bitcast (<4 x float> (<4 x float>)* @__log10f4_massv to i8*)
 ; ACCELERATE-SAME:  [1 x i8*] [
 ; ACCELERATE-SAME:i8* bitcast (<4 x float> (<4 x float>)* @vlog10f to i8*)
+; LIBMVEC-SAME: [2 x i8*] [
+; LIBMVEC-SAME:   i8* bitcast (<2 x double> (<2 x double>)* @_ZGVbN2v_sin to i8*),
+; LIBMVEC-SAME:   i8* bitcast (<4 x double> (<4 x double>)* @_ZGVdN4v_sin to i8*)
 ; COMMON-SAME:  ], section "llvm.metadata"
 
 define double @sin_f64(double %in) {
@@ -28,6 +33,7 @@
 ; SVML: call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; MASSV:call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; ACCELERATE:   call double @sin(double %{{.*}})
+; LIBMVEC:  call double @sin(double %{{.*}}) #[[SIN:[0-9]+]]
 ; No mapping of "sin" to a vector function for Accelerate.
 ; ACCELERATE-NOT: _ZGV_LLVM_{{.*}}_sin({{.*}})
   %call = tail call double @sin(double %in)
@@ -39,10 +45,12 @@
 define float @call_llvm.log10.f32(float %in) {
 ; COMMON-LABEL: @call_llvm.log10.f32(
 ; SVML: call float @llvm.log10.f32(float %{{.*}})
+; LIBMVEC:  call float @llvm.log10.f32(float %{{.*}})
 ; MASSV:call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; ACCELERATE:   call float @llvm.log10.f32(float %{{.*}}) #[[LOG10:[0-9]+]]
 ; No mapping of "llvm.log10.f32" to a vector function for SVML.
 ; SVML-NOT: _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
+; LIBMVEC-NOT:  _ZGV_LLVM_{{.*}}_llvm.log10.f32({{.*}})
   %call = tail call float @llvm.log10.f32(float %in)
   ret float %call
 }
@@ -62,3 +70,7 @@
 
 ; ACCELERATE:  attributes #[[LOG10]] = { "vector-function-abi-variant"=
 ; ACCELERATE-SAME:   "_ZGV_LLVM_N4v_llvm.log10.f32(vlog10f)" }
+
+; LIBMVEC:  attributes #[[SIN]] = { "vector-function-abi-variant"=
+; LIBMVEC-SAME:   "_ZGV_LLVM_N2v_sin(_ZGVbN2v_sin),
+; LIBMVEC-SAME:   _ZGV_LLVM_N4v_sin(_ZGVdN4v_sin)" }
Index: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
===
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
@@ -0,0 +1,325 @@
+; RUN: opt -vector-library=LIBMVEC  -inject-tli-mappings -force-vector-width=4 -force-vector-interleave=1 -loop-vectorize -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+declare double @sin(double) #0
+declare float @sinf(float) #0
+declare double @llvm.sin.f64(double) #0
+declare float @llvm.sin.f32(float) #0
+
+declare double @cos(double) #0
+declare float @cosf(float) #0
+declare double @llvm.cos.f64(double) #0
+declare float @llvm.cos.f32(float) #0
+
+define void @sin_f64(double* nocapture %varray) {
+; CHECK-LABEL: @sin_f64(
+; CHECK:[[TMP5:%.*]] = call <4 x double> @_ZGVdN4v_sin(<4 x double> [[TMP4:%.*]])
+; CHECK:ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

[PATCH] D88712: [CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

16 matches

Site Navigation

Mail list logo

Footer information