Implement shaderSubgroupBroadcastDynamicId
OpGroupNonUniformBroadcast can be supplied either a constant value or an
intermediate. I've split this Op into a fast path for constants that use
the original code, and a slow path that handles intermediates.
Tests: dEQP-VK.subgroups.ballot_broadcast.*
Bug: b/169608683
Change-Id: Idc74f3fe7e906315c59cf5b3bf3a450046e37375
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/50671
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Sean Risser <srisser@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
diff --git a/src/Pipeline/SpirvShaderGroup.cpp b/src/Pipeline/SpirvShaderGroup.cpp
index 51db930..19e997f 100644
--- a/src/Pipeline/SpirvShaderGroup.cpp
+++ b/src/Pipeline/SpirvShaderGroup.cpp
@@ -139,12 +139,39 @@
case spv::OpGroupNonUniformBroadcast:
{
auto valueId = Object::ID(insn.word(4));
- auto id = SIMD::Int(GetConstScalarInt(insn.word(5)));
+ auto idId = Object::ID(insn.word(5));
Operand value(this, state, valueId);
- auto mask = CmpEQ(id, SIMD::Int(0, 1, 2, 3));
- for(auto i = 0u; i < type.componentCount; i++)
+
+ // Decide between the fast path for constants and the slow path for
+ // intermediates.
+ if(getObject(idId).kind == SpirvShader::Object::Kind::Constant)
{
- dst.move(i, OrAll(value.Int(i) & mask));
+ auto id = SIMD::Int(GetConstScalarInt(insn.word(5)));
+ auto mask = CmpEQ(id, SIMD::Int(0, 1, 2, 3));
+ for(auto i = 0u; i < type.componentCount; i++)
+ {
+ dst.move(i, OrAll(value.Int(i) & mask));
+ }
+ }
+ else
+ {
+ Operand id(this, state, idId);
+
+ SIMD::UInt active = As<SIMD::UInt>(state->activeLaneMask()); // Considers helper invocations active. See b/151137030
+ SIMD::UInt inactive = ~active;
+ SIMD::UInt filled = id.UInt(0) & active;
+
+ for(int j = 0; j < SIMD::Width - 1; j++)
+ {
+ filled |= filled.yzwx & inactive; // Populate inactive 'holes' with a live value
+ }
+
+ auto mask = CmpEQ(filled, SIMD::UInt(0, 1, 2, 3));
+
+ for(uint32_t i = 0u; i < type.componentCount; i++)
+ {
+ dst.move(i, OrAll(value.UInt(i) & mask));
+ }
}
break;
}
diff --git a/src/Vulkan/VkPhysicalDevice.cpp b/src/Vulkan/VkPhysicalDevice.cpp
index 225adf7..9ddc243 100644
--- a/src/Vulkan/VkPhysicalDevice.cpp
+++ b/src/Vulkan/VkPhysicalDevice.cpp
@@ -271,7 +271,7 @@
features->vulkanMemoryModelAvailabilityVisibilityChains = VK_FALSE;
features->shaderOutputViewportIndex = VK_FALSE;
features->shaderOutputLayer = VK_FALSE;
- features->subgroupBroadcastDynamicId = VK_FALSE;
+ features->subgroupBroadcastDynamicId = VK_TRUE;
}
void PhysicalDevice::getFeatures2(VkPhysicalDeviceFeatures2 *features) const
