Lines Matching full:ir3
1340 - ir3/ra: Add specialized shared register RA/spilling
1341 - ir3: Set branchstack earlier
1342 - ir3: Rewrite (jp) and branchstack handling
1343 - ir3: Calculate physical edges correctly
1344 - ir3: Fix comment thinko
1345 - ir3/ra: Fix bug with collect source handling
1347 - ir3/a7xx: Fix load_global_ir3 with immediate offset
1348 - ir3: Initial support for pushing globals with ldg.k
1372 - nir/divergence_analysis: Add ir3-specific intrinsics
1375 - ir3: Allow single-predecessor phis
1376 - ir3: Run divergence analysis at the end
1377 - ir3: Remove loop shared copy check
1378 - ir3: Use divergence analysis for (jp) and physical CFG
1384 - ir3/legalize: Fix intra-block state propagation with loops
1385 - ir3: Rewrite nop insertion
1388 - ir3: Don't use non-contiguous component masks for FS
1389 - ir3: Don't pack FS inlocs
1391 - ir3: Add scan_clusters_macro to ir3_valid_flags()
1392 - ir3: Add scan_clusters.macro to is_subgroup_cond_mov_macro()
1393 - ir3/ra: Don't demote movmsk instructions to non-shared
1578 - ir3: Fix "print" meta instruction synchronization
1579 - ir3: Add fullsync and fullnop ir3 dbg options for over-syncing
1584 - ir3: Add ldg.k instruction
1588 - turnip,ir3/a750: Implement consts loading via preamble
1590 - turnip,ir3/a750: Implement inline uniforms via ldg.k
1611 - ir3: Do not set clip/cull mask if no one writes clip/cull
3058 - ir3: fix setting shared flag on parallel copy arguments
3059 - ir3: optimize read_first.macro to a mov
3060 - ir3: fix printing of brcst.active and quad_shuffle
3061 - ir3: optimize subgroup operations using brcst.active
3062 - ir3: set reconvergence for scan_clusters.macro
3063 - ir3: add disassembly for flat.b
3064 - ir3: update a0/a1 users when cloning instructions
3065 - ir3: fix alignment of spill slots
3066 - ir3: validate instruction block pointer
3067 - ir3: add terminators to blocks
3068 - ir3: fix instruction count before kill_sched
3069 - ir3: print branch sources
3070 - ir3: remove OPC_B and brtype from cat0
3071 - ir3: remove comp1/2 from cat0
3072 - ir3: allow liveness calculation for different register types
3073 - ir3: allow finding SSA uses for a subset of uses
3074 - ir3: implement RA for predicate registers
3075 - ir3: validate no registers are invalid after RA
3076 - ir3: integrate predicates into RA validation
3077 - ir3: optimize bitwise ops that can directly write predicates
3078 - ir3: insert predicate conversions after their source
3079 - ir3: fold negations into cmps.ne zero
3081 - ir3: fold and/or and negations into branches
3083 - ir3: fix freeing incorrect register in loops
3084 - ir3: fix returning false instead of NULL
3087 - ir3: calculate SSA uses at the start of predicates RA
3088 - ir3: fix finding uses of reloaded defs in predicates RA
3089 - ir3-disasm: run clang-format
3090 - ir3-disasm: remove unused #includes
3091 - ir3-disasm: add options to specify GPU by chip ID or name
3092 - ir3-disasm: add option to disassemble hex number
3094 - ir3: remove unnecessary tessellation epilogue
3095 - ir3: model predt/predf without sources
3096 - ir3: add support for precolored sources in predicate RA
3097 - ir3: add support for predication
4247 - fd/rddecompiler: Disable IR3 cache for replay context
5026 - freedreno/ir3: Fix ldg/stg offset