From a868378b996caf682c1b342b0ce25221b9632bac Mon Sep 17 00:00:00 2001
From: Orkuncakilkaya <orkuncakilkaya@gmail.com>
Date: Tue, 15 Sep 2026 20:22:12 +0300
Subject: [PATCH] feat(vk): call shapes for the acceleration-structure commands
 the interposer lacks

NVIDIA Streamline's interposer exports none of VK_KHR_acceleration_structure's
commands, so ray tracing looks each up per device with vkGetDeviceProcAddr and
calls it through a pointer: create (four pointers -> result), destroy (device,
handle, allocator), build sizes (device, type, info, counts, sizes), the
command-buffer build (buffer, count, infos, ranges) and the device address
(device, info -> u64). Both runtimes assemble; nothing uses them yet.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---
 runtime/native/vk.ludic  |  7 +++++++
 runtime/native/vk_mac.ll | 32 ++++++++++++++++++++++++++++++++
 runtime/native/vk_win.ll | 32 ++++++++++++++++++++++++++++++++
 3 files changed, 71 insertions(+)

diff --git a/runtime/native/vk.ludic b/runtime/native/vk.ludic
index 7ad3fdfd..2a6985ac 100644
--- a/runtime/native/vk.ludic
+++ b/runtime/native/vk.ludic
@@ -46,3 +46,10 @@ extern function vk_sl_call_ip(fn: pointer, a: int, b: pointer) -> int = "lsl_cal
 # a command-buffer command with three counts through a pointer (vkCmdDrawMeshTasksEXT, which the
 # Streamline interposer does not export; looked up with vkGetDeviceProcAddr)
 extern function vk_sl_call_piii(fn: pointer, a: pointer, b: int, c: int, d: int) = "lsl_call_piii"
+
+# acceleration-structure commands through a pointer from vkGetDeviceProcAddr (the interposer exports none)
+extern function vk_sl_call_pppp(fn: pointer, a: pointer, b: pointer, c: pointer, d: pointer) -> int = "lsl_call_pppp"
+extern function vk_sl_call_plp(fn: pointer, a: pointer, b: long, c: pointer) = "lsl_call_plp"
+extern function vk_sl_call_pippp(fn: pointer, a: pointer, b: int, c: pointer, d: pointer, e: pointer) = "lsl_call_pippp"
+extern function vk_sl_call_pipp(fn: pointer, a: pointer, b: int, c: pointer, d: pointer) = "lsl_call_pipp"
+extern function vk_sl_call_pp_l(fn: pointer, a: pointer, b: pointer) -> long = "lsl_call_pp_l"
diff --git a/runtime/native/vk_mac.ll b/runtime/native/vk_mac.ll
index 52c27be3..51c976db 100644
--- a/runtime/native/vk_mac.ll
+++ b/runtime/native/vk_mac.ll
@@ -190,3 +190,35 @@ entry:
   call void %fn(ptr %a0, i32 %a1, i32 %a2, i32 %a3)
   ret void
 }
+
+; ---- acceleration structures: commands the Streamline interposer does not export --------------------
+; vkCreateAccelerationStructureKHR(device, info, allocator, out) -> VkResult
+define i32 @lsl_call_pppp(ptr %fn, ptr %a0, ptr %a1, ptr %a2, ptr %a3) {
+entry:
+  %r = call i32 %fn(ptr %a0, ptr %a1, ptr %a2, ptr %a3)
+  ret i32 %r
+}
+; vkDestroyAccelerationStructureKHR(device, handle, allocator)
+define void @lsl_call_plp(ptr %fn, ptr %a0, i64 %a1, ptr %a2) {
+entry:
+  call void %fn(ptr %a0, i64 %a1, ptr %a2)
+  ret void
+}
+; vkGetAccelerationStructureBuildSizesKHR(device, buildType, info, maxPrimitiveCounts, sizes)
+define void @lsl_call_pippp(ptr %fn, ptr %a0, i32 %a1, ptr %a2, ptr %a3, ptr %a4) {
+entry:
+  call void %fn(ptr %a0, i32 %a1, ptr %a2, ptr %a3, ptr %a4)
+  ret void
+}
+; vkCmdBuildAccelerationStructuresKHR(commandBuffer, infoCount, infos, buildRangeInfos)
+define void @lsl_call_pipp(ptr %fn, ptr %a0, i32 %a1, ptr %a2, ptr %a3) {
+entry:
+  call void %fn(ptr %a0, i32 %a1, ptr %a2, ptr %a3)
+  ret void
+}
+; vkGetAccelerationStructureDeviceAddressKHR(device, info) -> VkDeviceAddress
+define i64 @lsl_call_pp_l(ptr %fn, ptr %a0, ptr %a1) {
+entry:
+  %r = call i64 %fn(ptr %a0, ptr %a1)
+  ret i64 %r
+}
diff --git a/runtime/native/vk_win.ll b/runtime/native/vk_win.ll
index 1d816b94..5bca3814 100644
--- a/runtime/native/vk_win.ll
+++ b/runtime/native/vk_win.ll
@@ -331,3 +331,35 @@ entry:
   call void %fn(ptr %a0, i32 %a1, i32 %a2, i32 %a3)
   ret void
 }
+
+; ---- acceleration structures: commands the Streamline interposer does not export --------------------
+; vkCreateAccelerationStructureKHR(device, info, allocator, out) -> VkResult
+define i32 @lsl_call_pppp(ptr %fn, ptr %a0, ptr %a1, ptr %a2, ptr %a3) {
+entry:
+  %r = call i32 %fn(ptr %a0, ptr %a1, ptr %a2, ptr %a3)
+  ret i32 %r
+}
+; vkDestroyAccelerationStructureKHR(device, handle, allocator)
+define void @lsl_call_plp(ptr %fn, ptr %a0, i64 %a1, ptr %a2) {
+entry:
+  call void %fn(ptr %a0, i64 %a1, ptr %a2)
+  ret void
+}
+; vkGetAccelerationStructureBuildSizesKHR(device, buildType, info, maxPrimitiveCounts, sizes)
+define void @lsl_call_pippp(ptr %fn, ptr %a0, i32 %a1, ptr %a2, ptr %a3, ptr %a4) {
+entry:
+  call void %fn(ptr %a0, i32 %a1, ptr %a2, ptr %a3, ptr %a4)
+  ret void
+}
+; vkCmdBuildAccelerationStructuresKHR(commandBuffer, infoCount, infos, buildRangeInfos)
+define void @lsl_call_pipp(ptr %fn, ptr %a0, i32 %a1, ptr %a2, ptr %a3) {
+entry:
+  call void %fn(ptr %a0, i32 %a1, ptr %a2, ptr %a3)
+  ret void
+}
+; vkGetAccelerationStructureDeviceAddressKHR(device, info) -> VkDeviceAddress
+define i64 @lsl_call_pp_l(ptr %fn, ptr %a0, ptr %a1) {
+entry:
+  %r = call i64 %fn(ptr %a0, ptr %a1)
+  ret i64 %r
+}