[MLIR][LLVM] Add ftz and fuse FP ops related function attribute support #97812

runseny · 2024-07-05T10:31:11Z

Adds denormal-fp-math-f32, denormal-fp-math, fp-contract to llvmFuncOp attributes.

denormal-fp-math-f32 and denormal-fp-math can enable the ftz, that is , flushing denormal to zero.

fp-contract can enable the fma fusion such as mul + add -> fma

github-actions · 2024-07-05T10:31:28Z

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be
notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write
permissions for the repository. In which case you can instead tag reviewers by
name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review
by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate
is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

llvmbot · 2024-07-05T10:31:57Z

@llvm/pr-subscribers-mlir-llvm

@llvm/pr-subscribers-mlir

Author: None (runseny)

Changes

Adds denormal-fp-math-f32, denormal-fp-math, fp-contract to llvmFuncOp attributes.

denormal-fp-math-f32 and denormal-fp-math can enable the ftz, that is , flushing denormal to zero.

fp-contract can enable the fma fusion such as mul + add -> fma

Full diff: https://github.com/llvm/llvm-project/pull/97812.diff

6 Files Affected:

(modified) mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td (+4-1)
(modified) mlir/lib/Target/LLVMIR/ModuleImport.cpp (+15)
(modified) mlir/lib/Target/LLVMIR/ModuleTranslation.cpp (+12)
(modified) mlir/test/Dialect/LLVMIR/func.mlir (+19)
(modified) mlir/test/Target/LLVMIR/Import/function-attributes.ll (+66)
(modified) mlir/test/Target/LLVMIR/fp-math-function-attributes.mlir (+99)

diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
index 3774bda05eb2be..373847265a5d34 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
@@ -1456,7 +1456,10 @@ def LLVM_LLVMFuncOp : LLVM_Op<"func", [
     OptionalAttr<BoolAttr>:$no_signed_zeros_fp_math,
     OptionalAttr<UnitAttr>:$no_inline,
     OptionalAttr<UnitAttr>:$always_inline,
-    OptionalAttr<UnitAttr>:$optimize_none
+    OptionalAttr<UnitAttr>:$optimize_none,
+    OptionalAttr<StrAttr>:$denormal_fp_math,
+    OptionalAttr<StrAttr>:$denormal_fp_math_f32,
+    OptionalAttr<StrAttr>:$fp_contract
   );
 
   let regions = (region AnyRegion:$body);
diff --git a/mlir/lib/Target/LLVMIR/ModuleImport.cpp b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
index 9b917db5e7dfe3..8c99452cf43b78 100644
--- a/mlir/lib/Target/LLVMIR/ModuleImport.cpp
+++ b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
@@ -1685,6 +1685,9 @@ static constexpr std::array kExplicitAttributes{
     StringLiteral("target-features"),
     StringLiteral("unsafe-fp-math"),
     StringLiteral("vscale_range"),
+    StringLiteral("denormal-fp-math"),
+    StringLiteral("denormal-fp-math-f32"),
+    StringLiteral("fp-contract"),
 };
 
 static void processPassthroughAttrs(llvm::Function *func, LLVMFuncOp funcOp) {
@@ -1823,6 +1826,18 @@ void ModuleImport::processFunctionAttributes(llvm::Function *func,
   if (llvm::Attribute attr = func->getFnAttribute("no-signed-zeros-fp-math");
       attr.isStringAttribute())
     funcOp.setNoSignedZerosFpMath(attr.getValueAsBool());
+
+  if (llvm::Attribute attr = func->getFnAttribute("denormal-fp-math");
+      attr.isStringAttribute())
+    funcOp.setDenormalFpMathAttr(StringAttr::get(context, attr.getValueAsString()));
+
+  if (llvm::Attribute attr = func->getFnAttribute("denormal-fp-math-f32");
+      attr.isStringAttribute())
+    funcOp.setDenormalFpMathF32Attr(StringAttr::get(context, attr.getValueAsString()));
+
+  if (llvm::Attribute attr = func->getFnAttribute("fp-contract");
+      attr.isStringAttribute())
+    funcOp.setFpContractAttr(StringAttr::get(context, attr.getValueAsString()));
 }
 
 DictionaryAttr
diff --git a/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp b/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
index 40196a5c760f96..457e03383a0c62 100644
--- a/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
@@ -1351,6 +1351,18 @@ LogicalResult ModuleTranslation::convertOneFunction(LLVMFuncOp func) {
     llvmFunc->addFnAttr("no-signed-zeros-fp-math",
                         llvm::toStringRef(*noSignedZerosFpMath));
 
+  if (auto DenormalFPMath = func.getDenormalFpMath())
+    llvmFunc->addFnAttr("denormal-fp-math",
+                        *DenormalFPMath);
+
+  if (auto DenormalFPMathF32 = func.getDenormalFpMathF32())
+    llvmFunc->addFnAttr("denormal-fp-math-f32",
+                        *DenormalFPMathF32);
+
+  if (auto FpContract = func.getFpContract())
+    llvmFunc->addFnAttr("fp-contract",
+                        *FpContract);
+
   // Add function attribute frame-pointer, if found.
   if (FramePointerKindAttr attr = func.getFramePointerAttr())
     llvmFunc->addFnAttr("frame-pointer",
diff --git a/mlir/test/Dialect/LLVMIR/func.mlir b/mlir/test/Dialect/LLVMIR/func.mlir
index d417942861940b..e0810a23697f8a 100644
--- a/mlir/test/Dialect/LLVMIR/func.mlir
+++ b/mlir/test/Dialect/LLVMIR/func.mlir
@@ -293,6 +293,25 @@ module {
     // CHECK-SAME: attributes {convergent}
     llvm.return
   }
+
+  llvm.func @denormal_fp_math_roundtrip() attributes {denormal_fp_math = "preserve-sign"} {
+    // CHECK: @denormal_fp_math_roundtrip
+    // CHECK-SAME: attributes {denormal_fp_math = "preserve-sign"}
+    llvm.return
+  }
+
+  llvm.func @denormal_fp_math_f32_roundtrip() attributes {denormal_fp_math_f32 = "preserve-sign"} {
+    // CHECK: @denormal_fp_math_f32_roundtrip
+    // CHECK-SAME: attributes {denormal_fp_math_f32 = "preserve-sign"}
+    llvm.return
+  }
+
+  llvm.func @fp_contract_roundtrip() attributes {fp_contract = "fast"} {
+    // CHECK: @fp_contract_roundtrip
+    // CHECK-SAME: attributes {fp_contract = "fast"}
+    llvm.return
+  }
+
 }
 
 // -----
diff --git a/mlir/test/Target/LLVMIR/Import/function-attributes.ll b/mlir/test/Target/LLVMIR/Import/function-attributes.ll
index 322ce6eadab4e4..d87048950b3e4e 100644
--- a/mlir/test/Target/LLVMIR/Import/function-attributes.ll
+++ b/mlir/test/Target/LLVMIR/Import/function-attributes.ll
@@ -344,6 +344,72 @@ declare void @func_attr_no_signed_zeros_fp_math_true() "no-signed-zeros-fp-math"
 ; CHECK-SAME: attributes {no_signed_zeros_fp_math = false}
 declare void @func_attr_no_signed_zeros_fp_math_false() "no-signed-zeros-fp-math"="false"
 
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_ieee
+; CHECK-SAME: attributes {denormal_fp_math = "ieee"}
+declare void @func_attr_denormal_fp_math_ieee() "denormal-fp-math"="ieee"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_preserve_sign
+; CHECK-SAME: attributes {denormal_fp_math = "preserve-sign"}
+declare void @func_attr_denormal_fp_math_preserve_sign() "denormal-fp-math"="preserve-sign"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_positive_zero
+; CHECK-SAME: attributes {denormal_fp_math = "positive-zero"}
+declare void @func_attr_denormal_fp_math_positive_zero() "denormal-fp-math"="positive-zero"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_dynamic
+; CHECK-SAME: attributes {denormal_fp_math = "dynamic"}
+declare void @func_attr_denormal_fp_math_dynamic() "denormal-fp-math"="dynamic"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_f32_ieee
+; CHECK-SAME: attributes {denormal_fp_math_f32 = "ieee"}
+declare void @func_attr_denormal_fp_math_f32_ieee() "denormal-fp-math-f32"="ieee"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_f32_preserve_sign
+; CHECK-SAME: attributes {denormal_fp_math_f32 = "preserve-sign"}
+declare void @func_attr_denormal_fp_math_f32_preserve_sign() "denormal-fp-math-f32"="preserve-sign"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_f32_positive_zero
+; CHECK-SAME: attributes {denormal_fp_math_f32 = "positive-zero"}
+declare void @func_attr_denormal_fp_math_f32_positive_zero() "denormal-fp-math-f32"="positive-zero"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_denormal_fp_math_f32_dynamic
+; CHECK-SAME: attributes {denormal_fp_math_f32 = "dynamic"}
+declare void @func_attr_denormal_fp_math_f32_dynamic() "denormal-fp-math-f32"="dynamic"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_fp_contract_fast
+; CHECK-SAME: attributes {fp_contract = "fast"}
+declare void @func_attr_fp_contract_fast() "fp-contract"="fast"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_fp_contract_on
+; CHECK-SAME: attributes {fp_contract = "on"}
+declare void @func_attr_fp_contract_on() "fp-contract"="on"
+
+; // -----
+
+; CHECK-LABEL: @func_attr_fp_contract_off
+; CHECK-SAME: attributes {fp_contract = "off"}
+declare void @func_attr_fp_contract_off() "fp-contract"="off"
+
 // -----
 
 ; CHECK-LABEL: @noinline_attribute
diff --git a/mlir/test/Target/LLVMIR/fp-math-function-attributes.mlir b/mlir/test/Target/LLVMIR/fp-math-function-attributes.mlir
index 4877c1137e3cd7..cd14fdf86b0f53 100644
--- a/mlir/test/Target/LLVMIR/fp-math-function-attributes.mlir
+++ b/mlir/test/Target/LLVMIR/fp-math-function-attributes.mlir
@@ -87,3 +87,102 @@ llvm.func @no_signed_zeros_fp_math_func_false() attributes {no_signed_zeros_fp_m
   llvm.return
 }
 // CHECK: attributes #[[ATTRS]] = { "no-signed-zeros-fp-math"="false" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_func_ieee()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_func_ieee() attributes {denormal_fp_math = "ieee"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math"="ieee" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_func_preserve_sign()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_func_preserve_sign() attributes {denormal_fp_math = "preserve-sign"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math"="preserve-sign" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_func_positive_zero()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_func_positive_zero() attributes {denormal_fp_math = "positive-zero"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math"="positive-zero" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_func_dynamic()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_func_dynamic() attributes {denormal_fp_math = "dynamic"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math"="dynamic" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_f32_func_ieee()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_f32_func_ieee() attributes {denormal_fp_math_f32 = "ieee"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math-f32"="ieee" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_f32_func_preserve_sign()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_f32_func_preserve_sign() attributes {denormal_fp_math_f32 = "preserve-sign"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math-f32"="preserve-sign" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_f32_func_positive_zero()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_f32_func_positive_zero() attributes {denormal_fp_math_f32 = "positive-zero"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math-f32"="positive-zero" }
+
+// -----
+
+// CHECK-LABEL: define void @denormal_fp_math_f32_func_dynamic()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @denormal_fp_math_f32_func_dynamic() attributes {denormal_fp_math_f32 = "dynamic"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "denormal-fp-math-f32"="dynamic" }
+
+// -----
+
+// CHECK-LABEL: define void @fp_contract_func_fast()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @fp_contract_func_fast() attributes {fp_contract = "fast"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "fp-contract"="fast" }
+
+// -----
+
+// CHECK-LABEL: define void @fp_contract_func_on()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @fp_contract_func_on() attributes {fp_contract = "on"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "fp-contract"="on" }
+
+// -----
+
+// CHECK-LABEL: define void @fp_contract_func_off()
+// CHECK-SAME: #[[ATTRS:[0-9]+]]
+llvm.func @fp_contract_func_off() attributes {fp_contract = "off"}  {
+  llvm.return
+}
+// CHECK: attributes #[[ATTRS]] = { "fp-contract"="off" }

runseny · 2024-07-05T10:35:26Z

@asb @zero9178 Could you help review this mr? This mr is similar with the previous mr: #79812, Thanks in advance!

gysit

Thanks!

I added some mostly nit comments.

gysit · 2024-07-05T15:34:37Z

mlir/test/Target/LLVMIR/Import/function-attributes.ll

+
+; CHECK-LABEL: @func_attr_denormal_fp_math_ieee
+; CHECK-SAME: attributes {denormal_fp_math = "ieee"}
+declare void @func_attr_denormal_fp_math_ieee() "denormal-fp-math"="ieee"


Since these are string attributes there is probably not much value in testing different string values for every attribute (i.e. ieee, preserve-sign)? Or is there specific reason for testing the different strings? Instead I would have a test for every attribute with possible different strings? So three tests here and three tests in the file below which exercises the lowering to LLVM IR.

Or is there specific reason for testing the different strings?

These different strings are the general cmd options in codegen such as denormalFpMath related and fp-contract. So i tested all these general cmd options for functional completeness.

Ok! I saw there is already prior art above.

I think the point I tried to make is that testing different values would make a lot of sense if the stored value would be an enum. Then we want to test that all enum values exist and are handled properly. In this case, we store a string so if one works all other strings should work as well (there are no separate code paths for different enum values). For maintenance reasons, I would thus keep the test set minimal if there is no extra functionality (in the sense of code coverage) that is tested.

gysit · 2024-07-05T15:37:06Z

mlir/lib/Target/LLVMIR/ModuleImport.cpp

@@ -1685,6 +1685,9 @@ static constexpr std::array kExplicitAttributes{
    StringLiteral("target-features"),
    StringLiteral("unsafe-fp-math"),
    StringLiteral("vscale_range"),
+    StringLiteral("denormal-fp-math"),
+    StringLiteral("denormal-fp-math-f32"),
+    StringLiteral("fp-contract"),


nit: Can you sort the array alphabetically. It looks like it was sorted before.

Yeah, i have sorted it.

gysit · 2024-07-05T15:39:45Z

mlir/lib/Target/LLVMIR/ModuleTranslation.cpp

@@ -1351,6 +1351,18 @@ LogicalResult ModuleTranslation::convertOneFunction(LLVMFuncOp func) {
    llvmFunc->addFnAttr("no-signed-zeros-fp-math",
                        llvm::toStringRef(*noSignedZerosFpMath));

+  if (auto DenormalFPMath = func.getDenormalFpMath())


Suggested change

if (auto DenormalFPMath = func.getDenormalFpMath())

if (auto denormalFPMath = func.getDenormalFpMath())

nit: the variable names should start lowercase (below as well).

Thanks, modified. The diff please see: diff, please take a look again :)

gysit · 2024-07-05T15:42:23Z

mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td

+    OptionalAttr<UnitAttr>:$optimize_none,
+    OptionalAttr<StrAttr>:$denormal_fp_math,
+    OptionalAttr<StrAttr>:$denormal_fp_math_f32,
+    OptionalAttr<StrAttr>:$fp_contract


nit: Are these flags related to the other fp_math flags above? Would it make sense to group them together?

Yeah, these flags are related to fp_math, so i grouped these flags with other fastMath flags.

gysit

LGTM

Thanks for addressing the comments! I still think there is some redundancy in terms of test cases. Feel free to trim them down if you think my argument makes sense, or otherwise keep them :).

gysit · 2024-07-08T06:41:01Z

mlir/test/Target/LLVMIR/Import/function-attributes.ll

+
+; CHECK-LABEL: @func_attr_denormal_fp_math_ieee
+; CHECK-SAME: attributes {denormal_fp_math = "ieee"}
+declare void @func_attr_denormal_fp_math_ieee() "denormal-fp-math"="ieee"


Ok! I saw there is already prior art above.

I think the point I tried to make is that testing different values would make a lot of sense if the stored value would be an enum. Then we want to test that all enum values exist and are handled properly. In this case, we store a string so if one works all other strings should work as well (there are no separate code paths for different enum values). For maintenance reasons, I would thus keep the test set minimal if there is no extra functionality (in the sense of code coverage) that is tested.

runseny · 2024-07-08T08:34:09Z

LGTM

Thanks for addressing the comments! I still think there is some redundancy in terms of test cases. Feel free to trim them down if you think my argument makes sense, or otherwise keep them :).

Yeah, you are right, thanks for this suggestion. Now i only test one string for each attribute. @gysit

runseny · 2024-07-08T09:11:15Z

@zero9178 Hi, i still need an extra approval, Could you help take a look?

github-actions · 2024-07-08T09:35:42Z

✅ With the latest revision this PR passed the C/C++ code formatter.

zero9178

Not super familiar with these fp functions, but given it mirrors LLVM, its their mess to figure and first and foremost 🙂
LGTM with clang-format addressed. Feel free to ping if no one merged this in a day once CI is green.

Given that some of these only have a finite set of limited values afaik, we could turn these into an enum attribute in the future. But this can be a future iteration.

@zero9178 Hi, i still need an extra approval, Could you help take a look?

FYI, there is no such general rule in LLVM that you need to have more than one approval, although you're of course free to get more eyes on the review 🙂! Sometimes the first reviewer will ask you for either more time before landing or another approval if they do not feel comfortable being the only approval, but that is not applicable here.

runseny · 2024-07-09T05:25:54Z

Not super familiar with these fp functions, but given it mirrors LLVM, its their mess to figure and first and foremost 🙂 LGTM with clang-format addressed. Feel free to ping if no one merged this in a day once CI is green.

Given that some of these only have a finite set of limited values afaik, we could turn these into an enum attribute in the future. But this can be a future iteration.

@zero9178 Hi, i still need an extra approval, Could you help take a look?

FYI, there is no such general rule in LLVM that you need to have more than one approval, although you're of course free to get more eyes on the review 🙂! Sometimes the first reviewer will ask you for either more time before landing or another approval if they do not feel comfortable being the only approval, but that is not applicable here.

Thanks for your detail explanation. Now, the CI of this mr passed, could you help merge it? thanks. @zero9178

github-actions · 2024-07-09T08:35:00Z

@runseny Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested
by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as
the builds can include changes from many authors. It is not uncommon for your
change to be included in a build that fails due to someone else's changes, or
infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself.
This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!

…rt (llvm#97812) Adds `denormal-fp-math-f32`, `denormal-fp-math`, `fp-contract` to llvmFuncOp attributes. `denormal-fp-math-f32` and `denormal-fp-math` can enable the ftz, that is , flushing denormal to zero. `fp-contract` can enable the fma fusion such as `mul + add -> fma`

llvmbot added mlir:llvm mlir labels Jul 5, 2024

gysit reviewed Jul 5, 2024

View reviewed changes

runseny force-pushed the ftz branch 2 times, most recently from c296c22 to 6938bf8 Compare July 6, 2024 14:15

runseny requested a review from gysit July 8, 2024 00:40

gysit approved these changes Jul 8, 2024

View reviewed changes

runseny force-pushed the ftz branch from 6938bf8 to 8af308f Compare July 8, 2024 08:27

zero9178 approved these changes Jul 8, 2024

View reviewed changes

runseny force-pushed the ftz branch from 8af308f to 94331e0 Compare July 9, 2024 01:42

[MLIR][LLVM] Add ftz and fuse FP ops related function attribute support

4ec769b

runseny force-pushed the ftz branch from 94331e0 to 4ec769b Compare July 9, 2024 01:44

zero9178 merged commit e5865ec into llvm:main Jul 9, 2024
7 checks passed

	if (auto DenormalFPMath = func.getDenormalFpMath())
	if (auto denormalFPMath = func.getDenormalFpMath())

[MLIR][LLVM] Add ftz and fuse FP ops related function attribute support #97812

[MLIR][LLVM] Add ftz and fuse FP ops related function attribute support #97812

Uh oh!

Conversation

runseny commented Jul 5, 2024

Uh oh!

github-actions bot commented Jul 5, 2024

Uh oh!

llvmbot commented Jul 5, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

runseny commented Jul 5, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

gysit left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

runseny Jul 6, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

runseny Jul 6, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gysit left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

runseny commented Jul 8, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

runseny commented Jul 8, 2024

Uh oh!

github-actions bot commented Jul 8, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zero9178 left a comment

Choose a reason for hiding this comment

Uh oh!

runseny commented Jul 9, 2024

Uh oh!

Uh oh!

github-actions bot commented Jul 9, 2024

Uh oh!

Uh oh!

llvmbot commented Jul 5, 2024 •

edited

Loading

runseny commented Jul 5, 2024 •

edited

Loading

runseny Jul 6, 2024 •

edited

Loading

runseny Jul 6, 2024 •

edited

Loading

runseny commented Jul 8, 2024 •

edited

Loading

github-actions bot commented Jul 8, 2024 •

edited

Loading