diff options
| author | Tim Foley <tfoley@nvidia.com> | 2017-06-28 10:20:16 -0700 |
|---|---|---|
| committer | Tim Foley <tfoley@nvidia.com> | 2017-06-28 11:08:03 -0700 |
| commit | d601921b71ed44835e8d4fa6f13ff7aefcf7649d (patch) | |
| tree | 79b9227ef038d173d780a440035e616dc31104bb | |
| parent | 4b3936e2983dcecd36a3437bd6c7eef8d5fbbffa (diff) | |
Actually respect suffixes on numeric literals.
- Add logic to extract the value and suffix from a numeric literal
- This duplicates some of the lexing logic, but this is hard to avoid without redundant runtime work
- Note that I'm not using and stdlib string-to-number code. This should be more robust once it is working, but it is obviously error prone in the near term. The main up-sides to this are:
- We can handle binary integer literals
- We can handle hexadecimal floating-point literals without stdlib support
- We can hypothetically support digit separators, if we ever wanted
- The parser looks at the suffix characters sliced off by the lexer, and tries to pick a type to use for a literal
- It uses `NULL` if there is no suffix, to avoid some nasty order dependencies where the stdlib might need to parse a number before it has seen the definition of `int`
- Right now I only handle a few cases, so there may be bugs lurking here
- The emit logic needs to handle the fact that a literal node in the AST might have a non-default type attached.
- Right now I just quickly check for the most likely types, and emit the literal with a matching suffix. This doesn't seem robust if any source language supports a suffix for a type where a target has no corresponding suffix. In the long term some amount of casting is probably required.
| -rw-r--r-- | source/slang/check.cpp | 4 | ||||
| -rw-r--r-- | source/slang/diagnostic-defs.h | 3 | ||||
| -rw-r--r-- | source/slang/emit.cpp | 41 | ||||
| -rw-r--r-- | source/slang/lexer.cpp | 210 | ||||
| -rw-r--r-- | source/slang/lexer.h | 6 | ||||
| -rw-r--r-- | source/slang/parser.cpp | 136 | ||||
| -rw-r--r-- | source/slang/syntax.cpp | 5 | ||||
| -rw-r--r-- | source/slang/syntax.h | 4 | ||||
| -rw-r--r-- | tests/hlsl/simple/literal-typing.hlsl | 25 |
9 files changed, 421 insertions, 13 deletions
diff --git a/source/slang/check.cpp b/source/slang/check.cpp index 11efb2f7d..dfa726150 100644 --- a/source/slang/check.cpp +++ b/source/slang/check.cpp @@ -1887,6 +1887,10 @@ namespace Slang } virtual RefPtr<ExpressionSyntaxNode> VisitConstantExpression(ConstantExpressionSyntaxNode *expr) override { + // The expression might already have a type, determined by its suffix + if(expr->Type.type) + return expr; + switch (expr->ConstType) { case ConstantExpressionSyntaxNode::ConstantType::Int: diff --git a/source/slang/diagnostic-defs.h b/source/slang/diagnostic-defs.h index f92f43ddb..c85270e08 100644 --- a/source/slang/diagnostic-defs.h +++ b/source/slang/diagnostic-defs.h @@ -284,6 +284,9 @@ DIAGNOSTIC(39999, Error, expectedPostfixOperator, "function called as postfix op DIAGNOSTIC(39999, Error, notEnoughArguments, "not enough arguments to call (got $0, expected $1)") DIAGNOSTIC(39999, Error, tooManyArguments, "too many arguments to call (got $0, expected $1)") +DIAGNOSTIC(39999, Error, invalidIntegerLiteralSuffix, "invalid suffix '$0' on integer literal") +DIAGNOSTIC(39999, Error, invalidFloatingPOintLiteralSuffix, "invalid suffix '$0' on floating-point literal") + // // 4xxxx - IL code generation. // diff --git a/source/slang/emit.cpp b/source/slang/emit.cpp index 5272204f6..9a1d44641 100644 --- a/source/slang/emit.cpp +++ b/source/slang/emit.cpp @@ -219,6 +219,8 @@ static void Emit(EmitContext* context, double value) Emit(context, buffer); } +static void emitTokenWithLocation(EmitContext* context, Token const& token); + // Expressions // Determine if an expression should not be emitted when it is the base of @@ -900,14 +902,53 @@ static void EmitExprWithPrecedence(EmitContext* context, RefPtr<ExpressionSyntax { needClose = MaybeEmitParens(context, outerPrec, kPrecedence_Atomic); + char const* suffix = ""; + auto type = litExpr->Type.type; switch (litExpr->ConstType) { case ConstantExpressionSyntaxNode::ConstantType::Int: + if(!type) + { + // Special case for "rewrite" mode + emitTokenWithLocation(context, litExpr->token); + break; + } + if(type->Equals(ExpressionType::GetInt())) + {} + else if(type->Equals(ExpressionType::GetUInt())) + { + suffix = "u"; + } + else + { + assert(!"unimplemented"); + } Emit(context, litExpr->IntValue); + Emit(context, suffix); break; + + case ConstantExpressionSyntaxNode::ConstantType::Float: + if(!type) + { + // Special case for "rewrite" mode + emitTokenWithLocation(context, litExpr->token); + break; + } + if(type->Equals(ExpressionType::GetFloat())) + {} + else if(type->Equals(ExpressionType::getDoubleType())) + { + suffix = "l"; + } + else + { + assert(!"unimplemented"); + } Emit(context, litExpr->FloatValue); + Emit(context, suffix); break; + case ConstantExpressionSyntaxNode::ConstantType::Bool: Emit(context, litExpr->IntValue ? "true" : "false"); break; diff --git a/source/slang/lexer.cpp b/source/slang/lexer.cpp index 73fbb9605..786376baf 100644 --- a/source/slang/lexer.cpp +++ b/source/slang/lexer.cpp @@ -415,25 +415,34 @@ namespace Slang return tokenType; } - static bool maybeLexNumberExponent(Lexer* lexer, int base) + static bool isNumberExponent(char c, int base) { - switch( peek(lexer) ) + switch( c ) { default: return false; case 'e': case 'E': if(base != 10) return false; - advance(lexer); break; case 'p': case 'P': if(base != 16) return false; - advance(lexer); break; } - // we saw an exponent marker, so we must + return true; + } + + static bool maybeLexNumberExponent(Lexer* lexer, int base) + { + if(!isNumberExponent(peek(lexer), base)) + return false; + + // we saw an exponent marker + advance(lexer); + + // Now start to read the exponent switch( peek(lexer) ) { case '+': case '-': @@ -482,6 +491,197 @@ namespace Slang return tokenType; } + static int maybeReadDigit(char const** ioCursor, int base) + { + auto& cursor = *ioCursor; + + for(;;) + { + int digitVal = 0; + int c = *cursor; + switch(c) + { + default: + return -1; + + // TODO: need to decide on digit separator characters + case '_': + cursor++; + continue; + + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + cursor++; + return c - '0'; + + case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': + if(base > 10) + { + cursor++; + return c - 'a'; + } + return -1; + + case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': + if(base > 10) + { + cursor++; + return c - 'A'; + } + return -1; + } + } + } + + static int readOptionalBase(char const** ioCursor) + { + auto& cursor = *ioCursor; + if( *cursor == '0' ) + { + cursor++; + switch(*cursor) + { + case 'x': case 'X': + cursor++; + return 16; + + case 'b': case 'B': + cursor++; + return 2; + + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + return 8; + + default: + return 10; + } + } + + return 10; + } + + + + IntegerLiteralValue getIntegerLiteralValue(Token const& token, String* outSuffix) + { + IntegerLiteralValue value = 0; + + char const* cursor = token.Content.begin(); + char const* end = token.Content.end(); + + int base = readOptionalBase(&cursor); + + for( ;;) + { + int digit = maybeReadDigit(&cursor, base); + if(digit < 0) + break; + + value = value*base + digit; + } + + if(outSuffix) + { + *outSuffix = String(cursor, end); + } + + return value; + } + + FloatingPointLiteralValue getFloatingPointLiteralValue(Token const& token, String* outSuffix) + { + FloatingPointLiteralValue value = 0; + + char const* cursor = token.Content.begin(); + char const* end = token.Content.end(); + + int radix = readOptionalBase(&cursor); + + bool seenDot = false; + FloatingPointLiteralValue divisor = 1; + for( ;;) + { + if(*cursor == '.') + { + cursor++; + seenDot = true; + continue; + } + + int digit = maybeReadDigit(&cursor, radix); + if(digit < 0) + break; + + value = value*radix + digit; + + if(seenDot) + { + divisor *= radix; + } + } + + // Now read optional exponent + if(isNumberExponent(*cursor, radix)) + { + cursor++; + + bool exponentIsNegative = false; + switch(*cursor) + { + default: + break; + + case '-': + exponentIsNegative = true; + cursor++; + break; + + case '+': + cursor++; + break; + } + + int exponentRadix = 10; + int exponent = 0; + + for(;;) + { + int digit = maybeReadDigit(&cursor, exponentRadix); + if(digit < 0) + break; + + exponent = exponent*exponentRadix + digit; + } + + FloatingPointLiteralValue exponentBase = 10; + if(radix == 16) + { + exponentBase = 2; + } + + FloatingPointLiteralValue exponentValue = pow(exponentBase, exponent); + + if( exponentIsNegative ) + { + divisor *= exponentValue; + } + else + { + value *= exponentValue; + } + } + + value /= divisor; + + if(outSuffix) + { + *outSuffix = String(cursor, end); + } + + return value; + } + static void lexStringLiteralBody(Lexer* lexer, char quote) { for(;;) diff --git a/source/slang/lexer.h b/source/slang/lexer.h index 5bcfc0f4a..53bfc0999 100644 --- a/source/slang/lexer.h +++ b/source/slang/lexer.h @@ -91,6 +91,12 @@ namespace Slang // Helper routines for extracting values from tokens String getStringLiteralTokenValue(Token const& token); String getFileNameTokenValue(Token const& token); + + typedef unsigned long long IntegerLiteralValue; + typedef double FloatingPointLiteralValue; + + IntegerLiteralValue getIntegerLiteralValue(Token const& token, String* outSuffix = 0); + FloatingPointLiteralValue getFloatingPointLiteralValue(Token const& token, String* outSuffix = 0); } #endif
\ No newline at end of file diff --git a/source/slang/parser.cpp b/source/slang/parser.cpp index 5e93d63cb..6437deca2 100644 --- a/source/slang/parser.cpp +++ b/source/slang/parser.cpp @@ -3214,22 +3214,140 @@ namespace Slang } case TokenType::IntegerLiteral: - case TokenType::FloatingPointLiteral: { RefPtr<ConstantExpressionSyntaxNode> constExpr = new ConstantExpressionSyntaxNode(); - auto token = parser->tokenReader.AdvanceToken(); parser->FillPosition(constExpr.Ptr()); - if (token.Type == TokenType::IntegerLiteral) + + auto token = parser->tokenReader.AdvanceToken(); + constExpr->token = token; + + String suffix; + IntegerLiteralValue value = getIntegerLiteralValue(token, &suffix); + + // Look at any suffix on the value + char const* suffixCursor = suffix.begin(); + + RefPtr<ExpressionType> suffixType = nullptr; + if( suffixCursor && *suffixCursor ) { - constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::Int; - constExpr->IntValue = StringToInt(token.Content); + int lCount = 0; + int uCount = 0; + int unknownCount = 0; + while(*suffixCursor) + { + switch( *suffixCursor++ ) + { + case 'l': case 'L': + lCount++; + break; + + case 'u': case 'U': + uCount++; + break; + + default: + unknownCount++; + break; + } + } + + if(unknownCount) + { + parser->sink->diagnose(token, Diagnostics::invalidIntegerLiteralSuffix, suffix); + suffixType = ExpressionType::GetError(); + } + // `u` or `ul` suffix -> `uint` + else if(uCount == 1 && (lCount <= 1)) + { + suffixType = ExpressionType::GetUInt(); + } + // `l` suffix on integer -> `int` (== `long`) + else if(lCount == 1 && !uCount) + { + suffixType = ExpressionType::GetInt(); + } + // TODO: probably need `ll` and `ull` + // TODO: are there other suffixes we need to handle? + else + { + parser->sink->diagnose(token, Diagnostics::invalidIntegerLiteralSuffix, suffix); + suffixType = ExpressionType::GetError(); + } } - else if (token.Type == TokenType::FloatingPointLiteral) + + constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::Int; + constExpr->IntValue = value; + constExpr->Type = suffixType; + + return constExpr; + } + + + case TokenType::FloatingPointLiteral: + { + RefPtr<ConstantExpressionSyntaxNode> constExpr = new ConstantExpressionSyntaxNode(); + parser->FillPosition(constExpr.Ptr()); + + auto token = parser->tokenReader.AdvanceToken(); + constExpr->token = token; + + String suffix; + FloatingPointLiteralValue value = getFloatingPointLiteralValue(token, &suffix); + + // Look at any suffix on the value + char const* suffixCursor = suffix.begin(); + + RefPtr<ExpressionType> suffixType = nullptr; + if( suffixCursor && *suffixCursor ) { - constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::Float; - constExpr->FloatValue = (FloatingPointLiteralValue) StringToDouble(token.Content); + int fCount = 0; + int lCount = 0; + int unknownCount = 0; + while(*suffixCursor) + { + switch( *suffixCursor++ ) + { + case 'f': case 'F': + fCount++; + break; + + case 'l': case 'L': + lCount++; + break; + + default: + unknownCount++; + break; + } + } + + if(unknownCount) + { + parser->sink->diagnose(token, Diagnostics::invalidFloatingPOintLiteralSuffix, suffix); + suffixType = ExpressionType::GetError(); + } + // `f` suffix -> `float` + if(fCount == 1 && !lCount) + { + suffixType = ExpressionType::GetFloat(); + } + // `l` or `lf` suffix on float -> `double` + else if(lCount == 1 && (fCount <= 1)) + { + suffixType = ExpressionType::getDoubleType(); + } + // TODO: are there other suffixes we need to handle? + else + { + parser->sink->diagnose(token, Diagnostics::invalidFloatingPOintLiteralSuffix, suffix); + suffixType = ExpressionType::GetError(); + } } + constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::Float; + constExpr->FloatValue = value; + constExpr->Type = suffixType; + return constExpr; } @@ -3237,6 +3355,7 @@ namespace Slang { RefPtr<ConstantExpressionSyntaxNode> constExpr = new ConstantExpressionSyntaxNode(); auto token = parser->tokenReader.AdvanceToken(); + constExpr->token = token; parser->FillPosition(constExpr.Ptr()); constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::String; @@ -3269,6 +3388,7 @@ namespace Slang { RefPtr<ConstantExpressionSyntaxNode> constExpr = new ConstantExpressionSyntaxNode(); auto token = parser->tokenReader.AdvanceToken(); + constExpr->token = token; parser->FillPosition(constExpr.Ptr()); constExpr->ConstType = ConstantExpressionSyntaxNode::ConstantType::Bool; constExpr->IntValue = token.Content == "true" ? 1 : 0; diff --git a/source/slang/syntax.cpp b/source/slang/syntax.cpp index f78d3c8ec..6b9a74f7d 100644 --- a/source/slang/syntax.cpp +++ b/source/slang/syntax.cpp @@ -1397,6 +1397,11 @@ namespace Slang return sBuiltinTypes[(int)BaseType::Float].GetValue().Ptr(); } + ExpressionType* ExpressionType::getDoubleType() + { + return sBuiltinTypes[(int)BaseType::Double].GetValue().Ptr(); + } + ExpressionType* ExpressionType::GetInt() { return sBuiltinTypes[(int)BaseType::Int].GetValue().Ptr(); diff --git a/source/slang/syntax.h b/source/slang/syntax.h index 7289719f4..81d6f8e01 100644 --- a/source/slang/syntax.h +++ b/source/slang/syntax.h @@ -375,6 +375,7 @@ namespace Slang UInt, UInt64, Float, + Double, #if 0 Texture2D = 48, TextureCube = 49, @@ -485,6 +486,7 @@ namespace Slang static ExpressionType* GetBool(); static ExpressionType* GetFloat(); + static ExpressionType* getDoubleType(); static ExpressionType* GetInt(); static ExpressionType* GetUInt(); static ExpressionType* GetVoid(); @@ -1916,6 +1918,8 @@ namespace Slang class ConstantExpressionSyntaxNode : public ExpressionSyntaxNode { public: + Token token; + enum class ConstantType { Int, diff --git a/tests/hlsl/simple/literal-typing.hlsl b/tests/hlsl/simple/literal-typing.hlsl new file mode 100644 index 000000000..71acb0d92 --- /dev/null +++ b/tests/hlsl/simple/literal-typing.hlsl @@ -0,0 +1,25 @@ +//TEST:COMPARE_HLSL: -target dxbc-assembly -profile cs_5_0 -entry main + +// Confirm that we get the typing of literal suffixes correct + +// A type created to cause type-checking failures downstream +struct Bad { int bad; }; + +// We define two overloads for `foo()`. The "right" one takes +// an unsigned integer, and returns it. The "wrong" one takes +// a signed integer and returns a `Bad`. + +uint foo(uint x) { return x; } +Bad foo(int x) { Bad b; b.bad = x; return b; } + +// The shader entry point will call `foo()` on a literal +// with a suffix that should make it unsigned, so that +// we either respect the suffix and call the right overload, +// or ignore it and call the wrong one. + +RWStructuredBuffer<uint> b; +[numthreads(32,1,1)] +void main(uint3 tid : SV_DispatchThreadID) +{ + b[tid.x] = foo(99u); +}
\ No newline at end of file |
