The case reached the compared path once host imports kept a guest-byte
snapshot, so it passed without the fix. It now runs over watched memory
without an import on every host, and over a watched host import on Linux
(the font atlas's path, with no valid snapshot), checks the image is on
that path, and runs as its own ctest test.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A write-back keeps the CPU's bytes for a block stamped since the layer's
generation. Stamps can fall on bytes the CPU did not change (PPSA12544's
font atlas: the stamped pages still hold the allocator fill the image was
uploaded from), and the GPU results there were lost. Images up to 1 MiB
keep the guest bytes of their generation point; a stamped range that still
equals them stores the GPU results.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The CFG/GOT filter's entry point collector checked header and symbol
positions as `offset + size > file size`, which wraps for offsets
near 2^64. A section header or symbol table offset there passed the
check, read bytes before the input buffer, and could reach an
in-bounds table through the wrapped address, adding entry points that
keep relocations alive. Check every position by subtraction instead.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A CMASK that was neither all fast-cleared nor all expanded threw, so a
title that fast clears a target and then draws over part of it before the
eliminate could not run past it.
The CMASK is now decoded tile by tile: one nibble per 8x8 tile in
1024x512 metablocks of 4 KiB, addressed with the pipe-aligned
SW_64KB_Z_X CMASK equation for 16 pipes (addrlib
GFX10_CMASK_SW_PATTERN, Gfx10Lib::HwlComputeCmaskAddrFromCoord). Code 0
stores CB_COLOR_CLEAR_WORD into the tile's texels, 0xF keeps them; any
other code, a cleared tile of a DCC target and any other CMASK swizzle
mode throw. Edge tiles are included and the CMASK is then stored as
expanded.
For a resident target whose CMASK is host-imported the clear is a
compute pass recorded in stream order: it reads the nibbles from the
import, stores the clear into the cleared tiles through a UINT view of
the image and fills the CMASK expanded, so nothing waits for pending GPU
writes of the CMASK. Invalid codes it reads set a bit that the next
materialization throws on. Otherwise the CPU reads the CMASK and stores
the cleared tiles into guest memory, or fills the resident image when
every tile is cleared. A CMASK nothing has stored over since it was
expanded is skipped.
The decode requires CB_COLOR_ATTRIB3 FMASK_SW_MODE SW_64KB_Z_X with
CMASK_PIPE_ALIGNED, a 4 KiB aligned writable CMASK and texels of at most
64 bits, and CMASK_BASE_EXT is part of the draw key.
Tests: CMASK sizes and nibble addresses against addrlib; decode
rejections and draw key coverage; CPU eliminate of full, partial,
out-of-range-nibble and DCC decompress cases and invalid codes. The new
agc_driver_cmask_fast_clear GPU test draws over fast-cleared and partly
fast-cleared linear and SW_64KB_R_X targets, eliminates partial, edge
and checkerboard CMASKs through the compute pass and checks that an
invalid code it reads throws.
A wave64 vertex program reduces a float key across the wave with the same shape as the integer half-wave reductions: keys selected under the all-lanes EXEC with -inf (v_max_f32) or +inf (v_min_f32) outside the entry EXEC, four DPP row_shr steps, v_permlanex16_b32, then v_readlane_b32 of lanes 31 and 63, combined by one more v_max_f32. On a 32-lane host subgroup the read of lane 63 threw that it is outside the subgroup.
The matcher now also accepts FPMax32 and FPMin32 steps, with their BitCasts and the f32 denormal flush the float mode may add around operands and results; the flush leaves the -inf and +inf keys unchanged. When the entry EXEC proves that lanes 32-63 hold the constant key, every step there combines that constant with itself. The existing v_max_f32/v_min_f32 lowering returns the operand unchanged for equal operands (NaN included), so lane 63 is exactly that constant, and the read returns it. No subgroup float reduction is emitted. Lane 31, the lanes on a 64-lane subgroup, and the combining v_max_f32 keep their existing lowering. So NaN and signed-zero results are the ones the shader computes today.
On Linux, sceKernelSync called sync(2), which blocks until all
dirty pages are written back. Unrelated host I/O can stall guest's
sceKernelSync, causing the guest_kernel_sync test to stall for minutes.
Share the Windows implementation instead.
Fsync each recorded path and its changed parent directories.
Use st_mtim and coarse realtime mark on Linux in place of FILETIME.
Every draw copied its index range and its vertex ranges out of guest
memory into new host buffers (CopyDrawInput) before the GPU could read
them. A range that lies in a host import is now bound from the import
itself, at its offset in it, the way the console's input assembler reads
guest memory directly.
A range is bound in place only when the batch has no write over it that
could still land (a recorded or in-flight GPU write after the pending
results are flushed, a completion label, a queued label store, a copied
buffer a dispatch or draw writes back) and its offset in the import is a
multiple of what the reader needs: the index size, or the strictest
format alignment of the vertex attributes that share the range
(VertexCopyPlan::alignments; a copy starts its range at offset 0, so
only an in-place range can be misaligned). Otherwise it is copied as
before. The read is noted on the batch (Recorder::ReadKind::DrawInput),
so a CPU store into the range waits for the draw like the other in-place
readers, and the barrier validation sees it. Imports gain vertex and
index buffer usage.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
FreeBSD sys_flock takes LOCK_UN over LOCK_EX over LOCK_SH, ignores other
bits and fails with EBADF when none of the three is set. The operation
reached the host unchanged, so malformed operations differed between
hosts: Linux rejects them with EINVAL and the Windows lock tracking threw.
Decode the operation before either host sees it.
flock mapped each call straight onto LockFileEx/UnlockFileEx, whose
byte-range locks stack per handle. LOCK_UN without a lock failed with
ERROR_NOT_LOCKED and aborted the process, a repeated LOCK_EX blocked
forever, converting between shared and exclusive failed, and a repeated
LOCK_SH needed as many unlocks.
Track the one lock each descriptor holds: a repeated mode and LOCK_UN
without a lock succeed, a conversion releases the old lock before taking
the new one, and close forgets the descriptor's lock.
Measured on RDNA2 (gfx1036) with tools/hw-oracle in wave32 and wave64:
s_call_b64 writes PC + 4 to the link pair and branches to
PC + 4 + signext(SIMM16) * 4, and s_setpc_b64 of the link returns;
s_swappc_b64 with SDST 125 branches and leaves a marker pair untouched, so
its decode as s_setpc_b64 matches, and with an SGPR destination the link is
PC + 4.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Claude-Session: https://claude.ai/code/session_012GLfKH7Sqe51wX254k1XsV
jweidenh