[SPIR-V] [NFC] Verify cl_intel_subgroup_local_block_io extension in SPIR-V BE #118796
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
|
||
| ; CHECK-SPIRV-LABEL: Return | ||
|
|
||
| target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024" |
There was a problem hiding this comment.
Let's remove datalayout and triple please
vmaksimo
changed the title
|
@llvm/pr-subscribers-backend-spir-v Author: Viktoria Maximova (vmaksimo) ChangesFull diff: https://github.com/llvm/llvm-project/pull/118796.diff 3 Files Affected:
diff --git a/llvm/lib/Target/SPIRV/SPIRVBuiltins.td b/llvm/lib/Target/SPIRV/SPIRVBuiltins.td
index e29013d28aafe4..af3901c0e621eb 100644
--- a/llvm/lib/Target/SPIRV/SPIRVBuiltins.td
+++ b/llvm/lib/Target/SPIRV/SPIRVBuiltins.td
@@ -1197,10 +1197,10 @@ defm : DemangledIntelSubgroupsBuiltin<"shuffle_down", 3, 3, OpSubgroupShuffleDow
defm : DemangledIntelSubgroupsBuiltin<"shuffle_up", 3, 3, OpSubgroupShuffleUpINTEL>;
defm : DemangledIntelSubgroupsBuiltin<"shuffle_xor", 2, 2, OpSubgroupShuffleXorINTEL>;
foreach i = ["", "2", "4", "8"] in {
- // cl_intel_subgroups
+ // cl_intel_subgroups, cl_intel_subgroup_local_block_io
defm : DemangledIntelSubgroupsBuiltin<!strconcat("block_read", i), 1, 2, OpSubgroupBlockReadINTEL>;
defm : DemangledIntelSubgroupsBuiltin<!strconcat("block_write", i), 2, 3, OpSubgroupBlockWriteINTEL>;
- // cl_intel_subgroups_short
+ // cl_intel_subgroups_short, cl_intel_subgroup_local_block_io
defm : DemangledIntelSubgroupsBuiltin<!strconcat("block_read_ui", i), 1, 2, OpSubgroupBlockReadINTEL>;
defm : DemangledIntelSubgroupsBuiltin<!strconcat("block_write_ui", i), 2, 3, OpSubgroupBlockWriteINTEL>;
// cl_intel_media_block_io
@@ -1209,7 +1209,7 @@ foreach i = ["", "2", "4", "8"] in {
defm : DemangledIntelSubgroupsBuiltin<!strconcat("media_block_write", i), 5, 5, OpSubgroupImageMediaBlockWriteINTEL>;
defm : DemangledIntelSubgroupsBuiltin<!strconcat("media_block_write_ui", i), 5, 5, OpSubgroupImageMediaBlockWriteINTEL>;
}
-// cl_intel_subgroups_char, cl_intel_subgroups_short, cl_intel_subgroups_long, cl_intel_media_block_io
+// cl_intel_subgroups_char, cl_intel_subgroups_short, cl_intel_subgroups_long, cl_intel_media_block_io, cl_intel_subgroup_local_block_io
foreach i = ["", "2", "4", "8", "16"] in {
foreach j = ["c", "s", "l"] in {
defm : DemangledIntelSubgroupsBuiltin<!strconcat("block_read_u", j, i), 1, 2, OpSubgroupBlockReadINTEL>;
diff --git a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_sub_groups.ll b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_sub_groups.ll
index 13667f44389e7b..457b3a14fd8fcf 100644
--- a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_sub_groups.ll
+++ b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_sub_groups.ll
@@ -77,9 +77,6 @@
; CHECK-SPIRV-LABEL: Return
-target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
-target triple = "spir64"
-
%opencl.image2d_ro_t = type opaque
%opencl.image2d_wo_t = type opaque
diff --git a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_subgroup_local_block_io.ll b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_subgroup_local_block_io.ll
new file mode 100644
index 00000000000000..8274c917bbe216
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_subgroup_local_block_io.ll
@@ -0,0 +1,156 @@
+; Modified from: https://github.com/KhronosGroup/SPIRV-LLVM-Translator/test/extensions/INTEL/SPV_INTEL_subgroups/cl_intel_sub_groups.ll
+
+;Source:
+;void __kernel test(float2 x, uint c,
+; read_only image2d_t image_in,
+; write_only image2d_t image_out,
+; int2 coord,
+; __local uint* p,
+; __local ushort* sp,
+; __local uchar* cp,
+; __local ulong* lp) {
+;
+; uint2 ui2 = intel_sub_group_block_read2(image_in, coord);
+; intel_sub_group_block_write2(image_out, coord, ui2);
+; ui2 = intel_sub_group_block_read2(p);
+; intel_sub_group_block_write2(p, ui2);
+;
+; ushort2 us2 = intel_sub_group_block_read_us2(image_in, coord);
+; intel_sub_group_block_write_us2(image_out, coord, us2);
+; us2 = intel_sub_group_block_read_us2(sp);
+; intel_sub_group_block_write_us2(sp, us2);
+;
+; uchar2 uc2 = intel_sub_group_block_read_uc2(image_in, coord);
+; intel_sub_group_block_write_uc2(image_out, coord, uc2);
+; uc2 = intel_sub_group_block_read_uc2(cp);
+; intel_sub_group_block_write_uc2(cp, uc2);
+;
+; ulong2 ul2 = intel_sub_group_block_read_ul2(image_in, coord);
+; intel_sub_group_block_write_ul2(image_out, coord, ul2);
+; ul2 = intel_sub_group_block_read_ul2(lp);
+; intel_sub_group_block_write_ul2(lp, ul2);
+;}
+
+; RUN: not llc -O0 -mtriple=spirv32-unknown-unknown %s -o %t.spvt 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR
+
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown --spirv-ext=+SPV_INTEL_subgroups %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown --spirv-ext=+SPV_INTEL_subgroups %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-ERROR: LLVM ERROR: intel_sub_group_block_read2: the builtin requires the following SPIR-V extension: SPV_INTEL_subgroups
+
+; CHECK-DAG: Capability SubgroupBufferBlockIOINTEL
+; CHECK-DAG: Capability SubgroupImageBlockIOINTEL
+; CHECK: Extension "SPV_INTEL_subgroups"
+
+; CHECK-SPIRV-LABEL: Function
+; CHECK-SPIRV-LABEL: Label
+
+; CHECK: SubgroupImageBlockReadINTEL
+; CHECK: SubgroupImageBlockWriteINTEL
+; CHECK: SubgroupBlockReadINTEL
+; CHECK: SubgroupBlockWriteINTEL
+
+; CHECK: SubgroupImageBlockReadINTEL
+; CHECK: SubgroupImageBlockWriteINTEL
+; CHECK: SubgroupBlockReadINTEL
+; CHECK: SubgroupBlockWriteINTEL
+
+; CHECK: SubgroupImageBlockReadINTEL
+; CHECK: SubgroupImageBlockWriteINTEL
+; CHECK: SubgroupBlockReadINTEL
+; CHECK: SubgroupBlockWriteINTEL
+
+; CHECK: SubgroupImageBlockReadINTEL
+; CHECK: SubgroupImageBlockWriteINTEL
+; CHECK: SubgroupBlockReadINTEL
+; CHECK: SubgroupBlockWriteINTEL
+
+; CHECK-SPIRV-LABEL: Return
+
+%opencl.image2d_ro_t = type opaque
+%opencl.image2d_wo_t = type opaque
+
+; Function Attrs: convergent nounwind
+define spir_kernel void @test(<2 x float> %x, i32 %c, ptr addrspace(3) %image_in, ptr addrspace(3) %image_out, <2 x i32> %coord, ptr addrspace(3) %p, ptr addrspace(3) %sp, ptr addrspace(3) %cp, ptr addrspace(3) %lp) !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3 !kernel_arg_base_type !4 !kernel_arg_type_qual !5 !kernel_arg_name !6 {
+entry:
+ %call4 = tail call spir_func <2 x i32> @_Z27intel_sub_group_block_read214ocl_image2d_roDv2_i(ptr addrspace(3) %image_in, <2 x i32> %coord)
+ tail call spir_func void @_Z28intel_sub_group_block_write214ocl_image2d_woDv2_iDv2_j(ptr addrspace(3) %image_out, <2 x i32> %coord, <2 x i32> %call4)
+ %call5 = tail call spir_func <2 x i32> @_Z27intel_sub_group_block_read2PU3AS1Kj(ptr addrspace(3) %p)
+ tail call spir_func void @_Z28intel_sub_group_block_write2PU3AS1jDv2_j(ptr addrspace(3) %p, <2 x i32> %call5)
+
+ %call6 = tail call spir_func <2 x i16> @_Z30intel_sub_group_block_read_us214ocl_image2d_roDv2_i(ptr addrspace(3) %image_in, <2 x i32> %coord)
+ tail call spir_func void @_Z31intel_sub_group_block_write_us214ocl_image2d_woDv2_iDv2_t(ptr addrspace(3) %image_out, <2 x i32> %coord, <2 x i16> %call6)
+ %call7 = tail call spir_func <2 x i16> @_Z30intel_sub_group_block_read_us2PU3AS1Kt(ptr addrspace(3) %sp)
+ tail call spir_func void @_Z31intel_sub_group_block_write_us2PU3AS1tDv2_t(ptr addrspace(3) %sp, <2 x i16> %call7)
+
+ %call8 = tail call spir_func <2 x i8> @_Z30intel_sub_group_block_read_uc214ocl_image2d_roDv2_i(ptr addrspace(3) %image_in, <2 x i32> %coord)
+ tail call spir_func void @_Z31intel_sub_group_block_write_uc214ocl_image2d_woDv2_iDv2_h(ptr addrspace(3) %image_out, <2 x i32> %coord, <2 x i8> %call8)
+ %call9 = tail call spir_func <2 x i8> @_Z30intel_sub_group_block_read_uc2PU3AS1Kh(ptr addrspace(3) %cp)
+ tail call spir_func void @_Z31intel_sub_group_block_write_uc2PU3AS1hDv2_h(ptr addrspace(3) %cp, <2 x i8> %call9)
+
+ %call10 = tail call spir_func <2 x i64> @_Z30intel_sub_group_block_read_ul214ocl_image2d_roDv2_i(ptr addrspace(3) %image_in, <2 x i32> %coord)
+ tail call spir_func void @_Z31intel_sub_group_block_write_ul214ocl_image2d_woDv2_iDv2_m(ptr addrspace(3) %image_out, <2 x i32> %coord, <2 x i64> %call10)
+ %call11 = tail call spir_func <2 x i64> @_Z30intel_sub_group_block_read_ul2PU3AS1Km(ptr addrspace(3) %lp)
+ tail call spir_func void @_Z31intel_sub_group_block_write_ul2PU3AS1mDv2_m(ptr addrspace(3) %lp, <2 x i64> %call11)
+
+ ret void
+}
+
+; Function Attrs: convergent
+declare spir_func <2 x i32> @_Z27intel_sub_group_block_read214ocl_image2d_roDv2_i(ptr addrspace(3), <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func void @_Z28intel_sub_group_block_write214ocl_image2d_woDv2_iDv2_j(ptr addrspace(3), <2 x i32>, <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i32> @_Z27intel_sub_group_block_read2PU3AS1Kj(ptr addrspace(3))
+
+; Function Attrs: convergent
+declare spir_func void @_Z28intel_sub_group_block_write2PU3AS1jDv2_j(ptr addrspace(3), <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i16> @_Z30intel_sub_group_block_read_us214ocl_image2d_roDv2_i(ptr addrspace(3), <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_us214ocl_image2d_woDv2_iDv2_t(ptr addrspace(3), <2 x i32>, <2 x i16>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i16> @_Z30intel_sub_group_block_read_us2PU3AS1Kt(ptr addrspace(3))
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_us2PU3AS1tDv2_t(ptr addrspace(3), <2 x i16>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i8> @_Z30intel_sub_group_block_read_uc214ocl_image2d_roDv2_i(ptr addrspace(3), <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_uc214ocl_image2d_woDv2_iDv2_h(ptr addrspace(3), <2 x i32>, <2 x i8>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i8> @_Z30intel_sub_group_block_read_uc2PU3AS1Kh(ptr addrspace(3))
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_uc2PU3AS1hDv2_h(ptr addrspace(3), <2 x i8>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i64> @_Z30intel_sub_group_block_read_ul214ocl_image2d_roDv2_i(ptr addrspace(3), <2 x i32>)
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_ul214ocl_image2d_woDv2_iDv2_m(ptr addrspace(3), <2 x i32>, <2 x i64>)
+
+; Function Attrs: convergent
+declare spir_func <2 x i64> @_Z30intel_sub_group_block_read_ul2PU3AS1Km(ptr addrspace(3))
+
+; Function Attrs: convergent
+declare spir_func void @_Z31intel_sub_group_block_write_ul2PU3AS1mDv2_m(ptr addrspace(3), <2 x i64>)
+
+!opencl.ocl.version = !{!0}
+!opencl.spir.version = !{!0}
+
+!0 = !{i32 1, i32 2}
+!1 = !{i32 0, i32 0, i32 1, i32 1, i32 0, i32 1, i32 1, i32 1, i32 1}
+!2 = !{!"none", !"none", !"read_only", !"write_only", !"none", !"none", !"none", !"none", !"none"}
+!3 = !{!"float2", !"uint", !"image2d_t", !"image2d_t", !"int2", !"uint*", !"ushort*", !"uchar*", !"ulong*"}
+!4 = !{!"float __attribute__((ext_vector_type(2)))", !"uint", !"image2d_t", !"image2d_t", !"int __attribute__((ext_vector_type(2)))", !"uint*", !"ushort*", !"uchar*", !"ulong*"}
+!5 = !{!"", !"", !"", !"", !"", !"", !"", !"", !""}
+!6 = !{!"x", !"c", !"image_in", !"image_out", !"coord", !"p", !"sp", !"cp", !"lp"}
|
VyacheslavLevytskyy
approved these changes
Dec 5, 2024
|
@MrSidims Can you please have a look |
MrSidims
reviewed
Dec 5, 2024
| @@ -0,0 +1,156 @@ | |||
| ; Modified from: https://github.com/KhronosGroup/SPIRV-LLVM-Translator/test/extensions/INTEL/SPV_INTEL_subgroups/cl_intel_sub_groups.ll | |||
There was a problem hiding this comment.
Apologies, but what is a principal difference between this test and lets say https://github.com/llvm/llvm-project/blob/main/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_subgroups/cl_intel_sub_groups.ll ? From the first glance they seem to test the same functionality, though the naming of cl_intel_sub_groups.ll is missleading.
There was a problem hiding this comment.
This test does not have shuffle instructions.
But the actual main difference is that input pointer to the builtins is in Local (addrspace (3)) instead of Global (addrspace (1)). cl_intel_subgroup_local_block_io just allows to use local ptr instread of global ptr, so no different functionality is expected to be tested except address space
MrSidims
approved these changes
Dec 6, 2024
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/81/builds/2949 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This OpenCL extension extends the subgroup block read and write functions defined by
cl_intel_subgroups(and itschar,short, andlongversions) to support reading from and writing to pointers to the__localmemory address space in addition to pointers to the__globalmemory address space.The builtins are translated to SPIR-V using
SPV_INTEL_subgroupsextension.