[spirv] Don't emit Float16 capability. (#1109)

Capability Float16 is not allowed by Vulkan 1.0 specification.
SPV_AMD_gpu_shader_half_float should be used if 16bit floats are used.

docs/SPIR-V.rst

@@ -359,7 +359,7 @@ type instructions:
 ``uint``/``dword``/``uin32_t``                         ``OpTypeInt 32 0``
 ``uint16_t``                   ``-enable-16bit-types`` ``OpTypeInt 16 0`` ``Int16``
 ``half``                                               ``OpTypeFloat 32``
-``half``/``float16_t``         ``-enable-16bit-types`` ``OpTypeFloat 16`` ``Float16`` ``SPV_AMD_gpu_shader_half_float``
+``half``/``float16_t``         ``-enable-16bit-types`` ``OpTypeFloat 16``             ``SPV_AMD_gpu_shader_half_float``
 ``float``/``float32_t``                                ``OpTypeFloat 32``
 ``snorm float``                                        ``OpTypeFloat 32``
 ``unorm float``                                        ``OpTypeFloat 32``
@@ -392,8 +392,8 @@ https://github.com/Microsoft/DirectXShaderCompiler/wiki/16-Bit-Scalar-Types
 ``min16int``                           ``OpTypeInt 32 1`` ``RelaxedPrecision``
 ``min12int``                           ``OpTypeInt 32 1`` ``RelaxedPrecision``
 ``min16uint``                          ``OpTypeInt 32 0`` ``RelaxedPrecision``
-``min16float`` ``-enable-16bit-types`` ``OpTypeFloat 16``                      ``Float16``  ``SPV_AMD_gpu_shader_half_float``
-``min10float`` ``-enable-16bit-types`` ``OpTypeFloat 16``                      ``Float16``  ``SPV_AMD_gpu_shader_half_float``
+``min16float`` ``-enable-16bit-types`` ``OpTypeFloat 16``                                   ``SPV_AMD_gpu_shader_half_float``
+``min10float`` ``-enable-16bit-types`` ``OpTypeFloat 16``                                   ``SPV_AMD_gpu_shader_half_float``
 ``min16int``   ``-enable-16bit-types`` ``OpTypeInt 16 1``                      ``Int16``
 ``min12int``   ``-enable-16bit-types`` ``OpTypeInt 16 1``                      ``Int16``
 ``min16uint``  ``-enable-16bit-types`` ``OpTypeInt 16 0``                      ``Int16``

tools/clang/lib/SPIRV/ModuleBuilder.cpp

@@ -847,12 +847,17 @@ IMPL_GET_PRIMITIVE_TYPE(Float32)
 
 #undef IMPL_GET_PRIMITIVE_TYPE
 
+// Note: At the moment, Float16 capability should not be added for Vulkan 1.0.
+// It is not a required capability, and adding the SPV_AMD_gpu_half_float does
+// not enable this capability. Any driver that supports float16 in Vulkan 1.0
+// should accept this extension.
 #define IMPL_GET_PRIMITIVE_TYPE_WITH_CAPABILITY(ty, cap)                       \
                                                                                \
   uint32_t ModuleBuilder::get##ty##Type() {                                    \
-    requireCapability(spv::Capability::cap);                                   \
     if (spv::Capability::cap == spv::Capability::Float16)                      \
       theModule.addExtension("SPV_AMD_gpu_shader_half_float");                 \
+    else                                                                       \
+      requireCapability(spv::Capability::cap);                                 \
     const Type *type = Type::get##ty(theContext);                              \
     const uint32_t typeId = theContext.getResultIdForType(type);               \
     theModule.addType(type, typeId);                                           \

tools/clang/test/CodeGenSPIRV/constant.scalar.16bit.enabled.hlsl

@@ -15,8 +15,8 @@
 // min16uint:  uint16_t(warning)
 // uint16_t:   uint16_t
 
-// CHECK: OpCapability Float16
 // CHECK: OpCapability Int16
+// CHECK: OpExtension "SPV_AMD_gpu_shader_half_float"
 
 // CHECK-NOT: OpDecorate %c_half RelaxedPrecision
 // CHECK-NOT: OpDecorate %c_min10float RelaxedPrecision