update rspirv 0.11 to 0.12 #1118
Conversation
Firestar99
force-pushed
the
mesh-ext-shaders
branch
5 times, most recently
from
5dfcd5c to
2a3dbcb
Compare
|
While porting, I've noticed that |
Firestar99
force-pushed
the
mesh-ext-shaders
branch
from
19b955b to
1d67ab5
Compare
Firestar99
changed the title
Firestar99
force-pushed
the
mesh-ext-shaders
branch
from
9f49ec0 to
ff0d9ae
Compare
|
Basically ready to merge, apart from |
There was a problem hiding this comment.
Ahh, I see you got around to fixing up the BaryCoord stuff before I did, nice!
Leaving this review more as a note that I will check the subtler aspects in more depth (so we don't accidentally merge too soon), but what I've seen looks great, thank you so much for working on this!
| (OperandKind::FPDenormMode, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::FPDenormMode(x)), | ||
| Err(()) => self.err(format!("unknown FPDenormMode {word}")), | ||
| }, | ||
| (OperandKind::QuantizationModes, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::QuantizationModes(x)), | ||
| Err(()) => self.err(format!("unknown QuantizationModes {word}")), | ||
| }, | ||
| (OperandKind::FPOperationMode, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::FPOperationMode(x)), | ||
| Err(()) => self.err(format!("unknown FPOperationMode {word}")), | ||
| }, | ||
| (OperandKind::OverflowModes, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::OverflowModes(x)), | ||
| Err(()) => self.err(format!("unknown OverflowModes {word}")), | ||
| }, | ||
| (OperandKind::PackedVectorFormat, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::PackedVectorFormat(x)), | ||
| Err(()) => self.err(format!("unknown PackedVectorFormat {word}")), | ||
| }, | ||
| (OperandKind::HostAccessQualifier, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::HostAccessQualifier(x)), | ||
| Err(()) => self.err(format!("unknown HostAccessQualifier {word}")), | ||
| }, | ||
| (OperandKind::CooperativeMatrixOperands, Some(word)) => { | ||
| match parse_bitflags_operand(COOPERATIVE_MATRIX_OPERANDS, word) { | ||
| Some(x) => inst | ||
| .operands | ||
| .push(dr::Operand::CooperativeMatrixOperands(x)), | ||
| None => self.err(format!("Unknown CooperativeMatrixOperands {word}")), | ||
| } | ||
| } | ||
| (OperandKind::CooperativeMatrixLayout, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::CooperativeMatrixLayout(x)), | ||
| Err(()) => self.err(format!("unknown CooperativeMatrixLayout {word}")), | ||
| }, | ||
| (OperandKind::CooperativeMatrixUse, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::CooperativeMatrixUse(x)), | ||
| Err(()) => self.err(format!("unknown CooperativeMatrixUse {word}")), | ||
| }, | ||
| (OperandKind::InitializationModeQualifier, Some(word)) => match word.parse() { | ||
| Ok(x) => inst | ||
| .operands | ||
| .push(dr::Operand::InitializationModeQualifier(x)), | ||
| Err(()) => self.err(format!("unknown InitializationModeQualifier {word}")), | ||
| }, | ||
| (OperandKind::LoadCacheControl, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::LoadCacheControl(x)), | ||
| Err(()) => self.err(format!("unknown LoadCacheControl {word}")), | ||
| }, | ||
| (OperandKind::StoreCacheControl, Some(word)) => match word.parse() { | ||
| Ok(x) => inst.operands.push(dr::Operand::StoreCacheControl(x)), | ||
| Err(()) => self.err(format!("unknown StoreCacheControl {word}")), | ||
| }, |
There was a problem hiding this comment.
Wow I didn't realize this was so manual. Longer-term might be worth switching to SPIR-T's view of OperandKind, which handles immediate operands in a much more dynamic way, and contains a lot of the same information that .parse() uses in these match arms (but then I'm not sure you could create the typed rspirv Operand value out of that).
Firestar99
force-pushed
the
mesh-ext-shaders
branch
from
0ab6ec1 to
edfc23c
Compare
This comment was marked as resolved.
This comment was marked as resolved.
|
@Firestar99 #1127 just landed, you'll need to rebase (but I should be able to review concurrently). |
Firestar99
force-pushed
the
mesh-ext-shaders
branch
3 times, most recently
from
cd0cbf6 to
9bea24f
Compare
Firestar99
force-pushed
the
mesh-ext-shaders
branch
from
9bea24f to
0acafd5
Compare
The rspirv update is required to get the spirv symbols required for mesh shaders, which will be added in a separate PR.
Original mesh shader PR: #1019