class LlvmProfdata

This is the minimum alignment required for LiveData::counters::data().

The object can be default-constructed and copied into, but cannot be used

in its default-constructed state.

If not compiled with instrumentation at all, then all the link-time

references in the real implementation below won't work. So provide stubs.

If not compiled with instrumentation at all, then all the link-time

references in the real implementation below won't work. So provide stubs.

If not compiled with instrumentation at all, then all the link-time

references in the real implementation below won't work. So provide stubs.

If not compiled with instrumentation at all, then all the link-time

references in the real implementation below won't work. So provide stubs.

If not compiled with instrumentation at all, then all the link-time

references in the real implementation below won't work. So provide stubs.

This returns true if single byte counters are used.

It is only public to be used by tests.

If the data appears to be valid llvm-profdata format with a build ID, then

return the subspan that is just the build ID bytes themselves. Otherwise

return an empty span. This does only minimal format validation that is

sufficient to find the build ID safely, and does not guarantee that the

other sizes in the header are valid.

If the data appears to be valid llvm-profdata format with a build ID, then

return the subspan that is just the build ID bytes themselves. Otherwise

return an empty span. This does only minimal format validation that is

sufficient to find the build ID safely, and does not guarantee that the

other sizes in the header are valid.

If the data appears to be valid llvm-profdata format with a build ID, then

return the subspan that is just the build ID bytes themselves. Otherwise

return an empty span. This does only minimal format validation that is

sufficient to find the build ID safely, and does not guarantee that the

other sizes in the header are valid.

If the data appears to be valid llvm-profdata format with a build ID, then

return the subspan that is just the build ID bytes themselves. Otherwise

return an empty span. This does only minimal format validation that is

sufficient to find the build ID safely, and does not guarantee that the

other sizes in the header are valid.

If the data appears to be valid llvm-profdata format with a build ID, then

return the subspan that is just the build ID bytes themselves. Otherwise

return an empty span. This does only minimal format validation that is

sufficient to find the build ID safely, and does not guarantee that the

other sizes in the header are valid.

Return true if data appears to be a valid llvm-profdata dump whose build

ID matches the one passed to Init.

Return true if data appears to be a valid llvm-profdata dump whose build

ID matches the one passed to Init.

Return true if data appears to be a valid llvm-profdata dump whose build

ID matches the one passed to Init.

Return true if data appears to be a valid llvm-profdata dump whose build

ID matches the one passed to Init.

Return true if data appears to be a valid llvm-profdata dump whose build

ID matches the one passed to Init.

Copy out the current live data values from their link-time locations where

they have accumulated since startup. The data.counters buffer must be at

least counters_size_bytes() and must be aligned as uint64, and the

data.bitmap must be at least bitmap_size_bytes(), usually the return value

of WriteFixedData or VerifyMatch.

This is like CopyLiveData, but instead of overwriting the buffer, it

merges the data with the existing live data values in the buffer.

This returns the size of the data blob to be published.

The return value is zero if there is no data to publish.

After CopyLiveData or MergeLiveData has prepared the buffers, start using

them for live data updates. This can be called again later to switch to

different buffers.

These return the offset and size within the blob of the aligned uint64_t[]

counters array.

These return the offset and size within the blob of the char[] bitmap

bytes array.

This must be passed a span of at least size_bytes() whose pointer must be

aligned to kAlign bytes. Write the fixed metadata into the buffer, but

leave the live data areas in the buffer untouched. Returns the subspans

covering the live data.

This must be passed a span of at least size_bytes() whose pointer must be

aligned to kAlign bytes. Write the fixed metadata into the buffer, but

leave the live data areas in the buffer untouched. Returns the subspans

covering the live data.

This must be passed a span of at least size_bytes() whose pointer must be

aligned to kAlign bytes. Write the fixed metadata into the buffer, but

leave the live data areas in the buffer untouched. Returns the subspans

covering the live data.

This must be passed a span of at least size_bytes() whose pointer must be

aligned to kAlign bytes. Write the fixed metadata into the buffer, but

leave the live data areas in the buffer untouched. Returns the subspans

covering the live data.

This must be passed a span of at least size_bytes() whose pointer must be

aligned to kAlign bytes. Write the fixed metadata into the buffer, but

leave the live data areas in the buffer untouched. Returns the subspans

covering the live data.

Verify the contents after Match(data) is true, causing assertion failures

if the data was corrupted. After this, the data is verified to match what

WriteFixedData would have written. Returns the subspans covering the

live data, just as WriteFixedData would have done.

Verify the contents after Match(data) is true, causing assertion failures

if the data was corrupted. After this, the data is verified to match what

WriteFixedData would have written. Returns the subspans covering the

live data, just as WriteFixedData would have done.

Verify the contents after Match(data) is true, causing assertion failures

if the data was corrupted. After this, the data is verified to match what

WriteFixedData would have written. Returns the subspans covering the

live data, just as WriteFixedData would have done.

Verify the contents after Match(data) is true, causing assertion failures

if the data was corrupted. After this, the data is verified to match what

WriteFixedData would have written. Returns the subspans covering the

live data, just as WriteFixedData would have done.

Verify the contents after Match(data) is true, causing assertion failures

if the data was corrupted. After this, the data is verified to match what

WriteFixedData would have written. Returns the subspans covering the

live data, just as WriteFixedData would have done.

This merges the from values into the to values.

This resets the runtime after UseLiveData so that the original link-time

counter locations will be updated hereafter. It's only used in tests.