[clang-tools-extra] [mlir] [llvm] [clang] [MLIR] Enable GPU Dialect to SYCL runtime integration (PR #71430)
@@ -0,0 +1,56 @@
+// RUN: mlir-opt %s
-pass-pipeline='builtin.module(spirv-attach-target{ver=v1.0
caps=Addresses,Int64,Kernel},convert-gpu-to-spirv{use-64bit-index=true},gpu.module(spirv.module(spirv-lower-abi-attrs,spirv-update-vce)),func.func(llvm-request-c-wrappers),convert-scf-to-cf,convert-cf-to-llvm,convert-arith-to-llvm,convert-math-to-llvm,convert-func-to-llvm,gpu-to-llvm{use-bare-pointers-for-kernels=true},gpu-module-to-binary,expand-strided-metadata,lower-affine,finalize-memref-to-llvm,reconcile-unrealized-casts)'
\
+// RUN: | mlir-cpu-runner \
+// RUN: --shared-libs=%mlir_sycl_runtime \
+// RUN: --shared-libs=%mlir_runner_utils \
+// RUN: --entry-point-result=void \
+// RUN: | FileCheck %s
+
+module @add attributes {gpu.container_module} {
+ memref.global "private" constant @__constant_2x2x2xf32_0 : memref<2x2x2xf32>
= dense<[[[1.1, 2.2], [3.3, 4.4]], [[5.5, 6.6], [7.7, 8.8 ]]]>
+ memref.global "private" constant @__constant_2x2x2xf32 : memref<2x2x2xf32> =
dense<[[[1.2, 2.3], [4.5, 5.8]], [[7.2, 8.3], [10.5, 11.8]]]>
+ func.func @main() {
+%0 = memref.get_global @__constant_2x2x2xf32 : memref<2x2x2xf32>
+%1 = memref.get_global @__constant_2x2x2xf32_0 : memref<2x2x2xf32>
+%2 = call @test(%0, %1) : (memref<2x2x2xf32>, memref<2x2x2xf32>) ->
memref<2x2x2xf32>
+%cast = memref.cast %2 : memref<2x2x2xf32> to memref<*xf32>
+call @printMemrefF32(%cast) : (memref<*xf32>) -> ()
+return
+ }
+ func.func private @printMemrefF32(memref<*xf32>)
+ func.func @test(%arg0: memref<2x2x2xf32>, %arg1: memref<2x2x2xf32>) ->
memref<2x2x2xf32> {
Dinistro wrote:
Ultra nit drive by comment: Could you indent the body of this function as you
did for the other function bodies?
https://github.com/llvm/llvm-project/pull/71430
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
https://github.com/Dinistro commented: This is missing tests for the LLVM import and export. I suspect that this is currently still part of the function's passthrough dictionary, which should be changed. https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
https://github.com/Dinistro edited https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
@@ -411,6 +412,13 @@ void Flang::addTargetOptions(const ArgList &Args,
}
// TODO: Add target specific flags, ABI, mtune option etc.
+ if (const Arg *A = Args.getLastArg(options::OPT_mtune_EQ)) {
+CmdArgs.push_back("-tune-cpu");
+if (strcmp(A->getValue(), "native") == 0)
Dinistro wrote:
Nit: `strcmp` is a C function. I assume that `A->getValue()` returns a
StringRef, or something else that supports normal `==` comparison.
https://github.com/llvm/llvm-project/pull/95043
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
@@ -58,6 +58,15 @@ def FramePointerKindAttr : LLVM_Attr<"FramePointerKind",
"framePointerKind"> {
let assemblyFormat = "`<` $framePointerKind `>`";
}
+//===--===//
+// TuneCPUAttr
+//===--===//
+
+//def TuneCPUAttr : LLVM_Attr<"TuneCPU", "tuneCPU"> {
+ //let parameters = (ins "tuneCPU::tuneCPU":$tuneCPU);
+ //let assemblyFormat = "`<` $tuneCPU `>`";
+//}
Dinistro wrote:
Dead code?
https://github.com/llvm/llvm-project/pull/95043
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
@@ -0,0 +1,9 @@ +; RUN: mlir-translate -import-llvm -split-input-file %s | FileCheck %s + +; CHECK-LABEL: llvm.func @tune_cpu() +; CHECK-SAME: tune_cpu = "pentium4" Dinistro wrote: Can you add a check that ensures that this is not also placed in the passthrough dictionary? I don't see any part in the code that would avoid this. Adding it to the `ExplicitAttributes` set in `ModuleImport.cpp` should be enough. https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
@@ -411,6 +412,13 @@ void Flang::addTargetOptions(const ArgList &Args,
}
// TODO: Add target specific flags, ABI, mtune option etc.
+ if (const Arg *A = Args.getLastArg(options::OPT_mtune_EQ)) {
+CmdArgs.push_back("-tune-cpu");
+if (strcmp(A->getValue(), "native") == 0)
Dinistro wrote:
A few lines further up, the string literal get's wrapped into a `StringRef`,
which seems to work.
`if (A->getValue() == StringRef{"Accelerate"})`
https://github.com/llvm/llvm-project/pull/95043
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
https://github.com/Dinistro approved this pull request. The MLIR side looks good to me % one remaining nit comment. Thanks for addressing the comments. https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
https://github.com/Dinistro edited https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [mlir] [flang] Add basic -mtune support (PR #95043)
@@ -0,0 +1,9 @@ +; RUN: mlir-translate -import-llvm -split-input-file %s | FileCheck %s + +; CHECK-LABEL: llvm.func @tune_cpu() +; CHECK-SAME: tune_cpu = "pentium4" Dinistro wrote: Given that we are both not entirely sure what happens, it might be sensible to check for that, just to be sure 🙂 https://github.com/llvm/llvm-project/pull/95043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [mlir] Remove the `x86_mmx` IR type. (PR #98505)
Dinistro wrote: The MLIR side is looking good, but I cannot say something about the rest. https://github.com/llvm/llvm-project/pull/98505 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] fix(clang/**.py): fix comparison to None (PR #94014)
https://github.com/Dinistro commented: That seems a reasonable change. LGTM! https://github.com/llvm/llvm-project/pull/94014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] fix(clang/**.py): fix comparison to None (PR #94014)
https://github.com/Dinistro approved this pull request. https://github.com/llvm/llvm-project/pull/94014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] fix(clang/**.py): fix comparison to None (PR #94014)
Dinistro wrote: Do you have commit rights to land this or should I do it? https://github.com/llvm/llvm-project/pull/94014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [lldb] [llvm] [mlir] Fix typo "instrinsic" (PR #112899)
https://github.com/Dinistro approved this pull request. LGTM, thanks for the fixes 😄 https://github.com/llvm/llvm-project/pull/112899 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] fix(clang/**.py): fix comparison to None (PR #94014)
Dinistro wrote: Then I'll merge it for you. https://github.com/llvm/llvm-project/pull/94014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] fix(clang/**.py): fix comparison to None (PR #94014)
https://github.com/Dinistro closed https://github.com/llvm/llvm-project/pull/94014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [LLVM][IR] Use splat syntax when printing Constant[Data]Vector. (PR #112548)
Dinistro wrote: We have a, temporary, odd use-case where we need to feed the textual IR back to a tool that is built with LLVM 17. For now, we were able to bypass all incompatibilities by simple workarounds on either the LLVM head, or on the LLVM 17 side. Unfortunately, this change seems the first change that leads to a non-trivial breakage that we cannot temporary circumvent. I'm aware that our use case is not officially supported, I just wanted to check what is supposed to happen. We will find a way to deal with this. Either way, thanks for the quick answer and thanks for working on improvements like these 🙂 https://github.com/llvm/llvm-project/pull/112548 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [LLVM][IR] Use splat syntax when printing Constant[Data]Vector. (PR #112548)
Dinistro wrote:
I stumbled over this change while trying to integrate LLVM head into our
downstream project. Is there any way to enforce that this still prints textual
IR in the old format, i.e., without "splat"? I found the following flags, but
they do not seem to provide the expected behavior.
```
--use-constant-int-for-fixed-length-splat
--use-constant-fp-for-fixed-length-splat
--use-constant-int-for-scalable-splat
--use-constant-fp-for-scalable-splat
```
What I would require is some way of forcing `opt` to roundtrip on the following
IR file **without** transforming this to splat syntax:
```
define <2 x double> @test(<2 x double> %in) {
%res = fmul <2 x double> %in,
ret <2 x double> %res
}
```
Ofc., we can find other kinds of workarounds, but I wanted to first to confirm
that this is indeed expected behavior or not.
https://github.com/llvm/llvm-project/pull/112548
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
