feat(examples): simplify vadd spv gen (#4)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
2026-06-05 00:10:24 +08:00 · 2026-03-26 22:55:36 +08:00
parent 4a1d72447b
commit 60203bf529
7 changed files with 30 additions and 68 deletions
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -39,9 +39,6 @@ jobs:
      - name: Build cmd/gen
        run: go build ./cmd/gen

-      - name: Build cmd/func2kernel
-        run: go build ./cmd/func2kernel
-
      - name: Run tests
        run: go test -v -count=1 ./...

--- a/README.md
+++ b/README.md
@@ -215,7 +215,6 @@ gozel/
 │   └── command.go            #   Command queues, lists, barriers
 ├── internal/zecall/          # purego FFI layer (loads ze_loader at runtime)
 ├── cmd/gen/                  # Code generator: parses L0 headers → Go source
-├── cmd/func2kernel/          # LLVM IR transformer for SPIR-V kernel preparation
 ├── spec/                     # Optional L0 SDK headers for dev purpose (input to cmd/gen)
 └── examples/
    ├── quick_start/          # The quick start shown in this README
@@ -267,18 +266,14 @@ This invokes `cmd/gen` with the local `spec/` directory as configured in `doc.go
 GPU kernels in the [examples](examples) folder are written in SYCL C++ and compiled to SPIR-V for embedding into Go programs, which is a little bit hacky. You can also use `ocloc`, which is a common practice and you can search for the build doc elsewhere. The build pipeline uses `go generate` directives:

 ```
-main.cpp ──clang++ -fsycl──▶ device_func.bc
+main.cpp ──clang++ -fsycl──▶ device_kern.bc
                              │
                        sycl-post-link
                              │
-                        ▼ device_func_0.bc
+                        ▼ device_kern_0.bc
                              │
                   clang++ -emit-llvm -S
                              │
-                        ▼ device_func.ll
-                              │
-                     cmd/func2kernel        ← transforms spir_func → spir_kernel
-                              │
                        ▼ device_kern.ll
                              │
                     llvm-spirv
@@ -297,5 +292,3 @@ cd examples/vadd
 go generate    # compiles main.cpp → main.spv
 go run main.go # runs vector addition on GPU
 ```
-
-The `cmd/func2kernel` tool handles the LLVM IR transformation (`spir_func` → `spir_kernel`, address space 4 → 1) required by the SPIR-V backend.
--- a/cmd/func2kernel/main.go
+++ b/cmd/func2kernel/main.go
@@ -1,28 +0,0 @@
-package main
-
-import (
-	"bufio"
-	"os"
-	"strings"
-)
-
-func main() {
-	f, err := os.Open(os.Args[1])
-	if err != nil {
-		panic(err)
-	}
-	defer f.Close()
-	fo, err := os.Create(os.Args[2])
-	if err != nil {
-		panic(err)
-	}
-	defer fo.Close()
-	scan := bufio.NewScanner(f)
-	for scan.Scan() {
-		t := scan.Text()
-		t = strings.ReplaceAll(t, " spir_func ", " spir_kernel ")
-		t = strings.ReplaceAll(t, "ptr addrspace(4)", "ptr addrspace(1)")
-		fo.WriteString(t)
-		fo.WriteString("\n")
-	}
-}
--- a/examples/vadd/main.cpp
+++ b/examples/vadd/main.cpp
@@ -1,6 +1,6 @@
 #include <sycl/sycl.hpp>

-extern "C" SYCL_EXTERNAL
+extern "C" SYCL_EXT_ONEAPI_FUNCTION_PROPERTY((sycl::ext::oneapi::experimental::nd_range_kernel<1>))
 void vector_add(float* a, float* b) {
    auto item = sycl::ext::oneapi::this_work_item::get_nd_item<1>();
    int idx = item.get_global_id(0);
--- a/examples/vadd/main.go
+++ b/examples/vadd/main.go
@@ -15,11 +15,9 @@ import (
 	"github.com/fumiama/gozel/ze"
 )

-//go:generate clang++ -fsycl -fsycl-device-only -fsycl-targets=spirv64 -faddrsig -Xclang -emit-llvm-bc main.cpp -o device_func.bc
-//go:generate sycl-post-link -symbols -split=auto -o device_func.table device_func.bc
-//go:generate clang++ -target spirv64-unknown-unknown -S -emit-llvm -x ir device_func_0.bc -o device_func.ll
-//go:generate go run ../../cmd/func2kernel device_func.ll device_kern.ll
-//go:generate clang++ -target spirv64-unknown-unknown -c -emit-llvm -x ir device_kern.ll -o device_kern.bc
+//go:generate clang++ -fsycl -fsycl-device-only -fsycl-targets=spirv64 -Xclang -emit-llvm-bc main.cpp -o device_kern.bc
+//go:generate sycl-post-link -symbols -split=auto -o device_kern.table device_kern.bc
+//go:generate clang++ -target spirv64-unknown-unknown -S -emit-llvm -x ir device_kern_0.bc -o device_kern.ll
 //go:generate llvm-spirv -o main.spv device_kern.bc
 //go:generate clang++ -target spirv64-unknown-unknown -S -emit-llvm -x ir device_kern.bc -o main.ll

@@ -131,7 +129,7 @@ func main() {
 	}
 	defer mod.Destroy()

-	krn, err := mod.KernelCreate("vector_add")
+	krn, err := mod.KernelCreate("__sycl_kernel_vector_add")
 	if err != nil {
 		panic(err)
 	}
--- a/examples/vadd/main.ll
+++ b/examples/vadd/main.ll
@@ -6,30 +6,31 @@ target triple = "spirv64-unknown-unknown"
@__spirv_BuiltInGlobalInvocationId = external local_unnamed_addr addrspace(1) constant <3 x i64>, align 32

 ; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite, inaccessiblemem: write)
-define spir_kernel void @vector_add(ptr addrspace(1) noundef captures(none) %0, ptr addrspace(1) noundef readonly captures(none) %1) local_unnamed_addr #0 !sycl_fixed_targets !7 {
-  %3 = load i64, ptr addrspace(1) @__spirv_BuiltInGlobalInvocationId, align 32, !noalias !8
+define spir_kernel void @__sycl_kernel_vector_add(ptr addrspace(1) noundef align 4 captures(none) %0, ptr addrspace(1) noundef readonly align 4 captures(none) %1) local_unnamed_addr #0 !kernel_arg_buffer_location !6 !sycl_fixed_targets !7 !sycl_kernel_omit_args !8 {
+  %3 = load i64, ptr addrspace(1) @__spirv_BuiltInGlobalInvocationId, align 32, !noalias !9
  %4 = icmp ult i64 %3, 2147483648
  tail call void @llvm.assume(i1 %4)
-  %5 = getelementptr inbounds nuw float, ptr addrspace(1) %1, i64 %3
-  %6 = load float, ptr addrspace(1) %5, align 4, !tbaa !15
-  %7 = getelementptr inbounds nuw float, ptr addrspace(1) %0, i64 %3
-  %8 = load float, ptr addrspace(1) %7, align 4, !tbaa !15
+  %5 = getelementptr inbounds float, ptr addrspace(1) %1, i64 %3
+  %6 = load float, ptr addrspace(1) %5, align 4, !tbaa !16
+  %7 = getelementptr inbounds float, ptr addrspace(1) %0, i64 %3
+  %8 = load float, ptr addrspace(1) %7, align 4, !tbaa !16
  %9 = fadd float %8, %6
-  store float %9, ptr addrspace(1) %7, align 4, !tbaa !15
+  store float %9, ptr addrspace(1) %7, align 4, !tbaa !16
  ret void
 }

 ; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)
 declare void @llvm.assume(i1 noundef) #1

-attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite, inaccessiblemem: write) "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "sycl-entry-point" "sycl-module-id"="main.cpp" "sycl-optlevel"="2" }
+declare dso_local spir_func i32 @_Z18__spirv_ocl_printfPU3AS2Kcz(ptr addrspace(2), ...)
+
+attributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite, inaccessiblemem: write) "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "sycl-module-id"="main.cpp" "sycl-nd-range-kernel"="1" "sycl-optlevel"="2" "uniform-work-group-size"="true" }
 attributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }

 !llvm.module.flags = !{!0, !1, !2}
 !opencl.spir.version = !{!3}
 !spirv.Source = !{!4}
 !llvm.ident = !{!5}
-!sycl-esimd-split-status = !{!6}

 !0 = !{i32 1, !"wchar_size", i32 4}
 !1 = !{i32 1, !"sycl-device", i32 1}
@@ -37,16 +38,17 @@ attributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessib
 !3 = !{i32 1, i32 2}
 !4 = !{i32 4, i32 100000}
 !5 = !{!"clang version 21.0.0git (https://github.com/intel/llvm d5f649b706f63b5c74e1929bc95db8de91085560)"}
-!6 = !{i8 0}
+!6 = !{i32 -1, i32 -1}
 !7 = !{}
-!8 = !{!9, !11, !13}
-!9 = distinct !{!9, !10, !"_ZN7__spirv29InitSizesSTGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEE8initSizeEv: argument 0"}
-!10 = distinct !{!10, !"_ZN7__spirv29InitSizesSTGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEE8initSizeEv"}
-!11 = distinct !{!11, !12, !"_ZN7__spirv22initGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEEET0_v: argument 0"}
-!12 = distinct !{!12, !"_ZN7__spirv22initGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEEET0_v"}
-!13 = distinct !{!13, !14, !"_ZNK4sycl3_V17nd_itemILi1EE13get_global_idEv: argument 0"}
-!14 = distinct !{!14, !"_ZNK4sycl3_V17nd_itemILi1EE13get_global_idEv"}
-!15 = !{!16, !16, i64 0}
-!16 = !{!"float", !17, i64 0}
-!17 = !{!"omnipotent char", !18, i64 0}
-!18 = !{!"Simple C++ TBAA"}
+!8 = !{i1 false, i1 false}
+!9 = !{!10, !12, !14}
+!10 = distinct !{!10, !11, !"_ZN7__spirv29InitSizesSTGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEE8initSizeEv: argument 0"}
+!11 = distinct !{!11, !"_ZN7__spirv29InitSizesSTGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEE8initSizeEv"}
+!12 = distinct !{!12, !13, !"_ZN7__spirv22initGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEEET0_v: argument 0"}
+!13 = distinct !{!13, !"_ZN7__spirv22initGlobalInvocationIdILi1EN4sycl3_V12idILi1EEEEET0_v"}
+!14 = distinct !{!14, !15, !"_ZNK4sycl3_V17nd_itemILi1EE13get_global_idEv: argument 0"}
+!15 = distinct !{!15, !"_ZNK4sycl3_V17nd_itemILi1EE13get_global_idEv"}
+!16 = !{!17, !17, i64 0}
+!17 = !{!"float", !18, i64 0}
+!18 = !{!"omnipotent char", !19, i64 0}
+!19 = !{!"Simple C++ TBAA"}
--- a/examples/vadd/main.spv
+++ b/examples/vadd/main.spv