summaryrefslogtreecommitdiff
AgeCommit message (Expand)AuthorFilesLines
80 min.i965/fs: Don't make a name for a vector splitting temporaryIan Romanick1-3/+8
4 hoursi965/fs: Properly calculate the number of instructions in calculate_register_...Jason Ekstrand1-1/+3
4 hoursi965/fs: Use the GRF for FB writes on gen >= 7Jason Ekstrand6-71/+142
4 hoursi965/fs: Handle COMPR4 in LOAD_PAYLOADJason Ekstrand2-1/+36
4 hoursi965/fs: Constant propagate into LOAD_PAYLOADJason Ekstrand1-0/+1
4 hoursi965/fs: Add split_virtual_grfs and compute_to_mrf after lower_load_payloadJason Ekstrand1-0/+2
4 hoursi965/fs: Add a an optional source to the FS_OPCODE_FB_WRITE instructionJason Ekstrand4-29/+28
4 hoursi965/fs: Use the GRF for UNTYPED_SURFACE_READ instructionsJason Ekstrand4-16/+24
4 hoursi965/fs: Use the GRF for UNTYPED_ATOMIC instructionsJason Ekstrand6-25/+36
4 hoursi965/fs: Add a function for getting a component of a 8 or 16-wide registerJason Ekstrand1-0/+10
4 hoursi965/fs: Use the instruction execution size directly for texture generationJason Ekstrand1-3/+10
4 hoursi965/fs: Use exec_size instead of force_uncompressed in dump_instructionJason Ekstrand1-6/+7
4 hoursi965/fs: Use instruction execution sizes instead of heuristicsJason Ekstrand3-23/+10
4 hoursi965/fs: Use instruction execution sizes to set compression stateJason Ekstrand1-6/+19
4 hoursi965/fs: Remove unneeded uses of force_uncompressedJason Ekstrand3-25/+9
4 hoursi965/fs: Derive force_uncompressed from instruction exec_sizeJason Ekstrand1-0/+3
4 hoursi965/fs: Make fs_reg::effective_width take fs_inst* instead of fs_visitor*Jason Ekstrand3-37/+43
4 hoursi965/fs: Better guess the width of LOAD_PAYLOADJason Ekstrand1-2/+9
4 hoursi965/fs: Add an exec_size field to fs_instJason Ekstrand5-32/+126
4 hoursi965/fs: Determine partial writes based on the destination widthJason Ekstrand2-5/+3
4 hoursi965/fs: Fix a bug in register coalesceJason Ekstrand1-0/+17
4 hoursi965/fs: Rework GEN5 texturing code to use fs_reg and offset()Jason Ekstrand1-39/+38
4 hoursi965/fs_reg: Allocate double the number of vgrfs in SIMD16 modeJason Ekstrand9-157/+371
4 hoursi965/fs: Handle printing of registers better.Jason Ekstrand1-2/+6
4 hoursi965/fs: Make half() divide the register width by 2 and use it moreJason Ekstrand2-5/+13
4 hoursi965/fs: Add a concept of a width to fs_regJason Ekstrand2-4/+78
4 hoursi965/fs: A little harmless refactoring of register_coalesceJason Ekstrand1-7/+7
4 hoursi965/fs: Copy propagate partial reads.Jason Ekstrand2-20/+64
4 hoursi965/fs: Refactor fs_inst::is_send_from_grf()Jason Ekstrand1-9/+16
4 hoursi965/fs: Clean up emit_fb_writesJason Ekstrand2-112/+85
4 hoursi965/fs: Print BAD_FILE registers in dump_instructionJason Ekstrand1-1/+1
4 hoursi965/fs: Make compact_virtual_grfs an optimization passJason Ekstrand2-8/+13
4 hoursi965/fs: Make null_reg_* const members of fs_visitor instead of globalsJason Ekstrand3-3/+12
4 hoursi965/fs: Use the var_from_vgrf helper function instead of doing it manuallyJason Ekstrand1-4/+4
4 hoursi965/fs: Fix a bug with dead_code_eliminate on large writesJason Ekstrand1-1/+1
4 hoursi965/fs: Use the UW type for the destination of VARYING_PULL_CONSTANT_LOAD in...Jason Ekstrand1-2/+2
4 hoursi965/fs: Use offset a lot more placesJason Ekstrand4-93/+78
4 hoursi965/fs: fix a comment in compact_virtual_grfsJason Ekstrand1-1/+1
4 hoursi965/fs: Rewrite fs_visitor::split_virtual_grfsJason Ekstrand1-47/+86
4 hoursi965/fs_live_variables: Use var_from_vgrf insead of repeating the calculationJason Ekstrand1-2/+2
4 hoursi965/fs: Manually generate the meta fast-clear shaderJason Ekstrand2-90/+35
3 daysi965/fs: Don't invalidate live intervals in saturate propagation.Matt Turner1-2/+1
3 daysi965/fs: Ignore mov.sat instructions in interference check in sat prop.Matt Turner1-1/+2
3 daysi965/fs: Walk instructions in reverse in saturate propagation.Matt Turner1-3/+3
6 daysi965/fs: Don't iterate between blocks with inst->next/prev.Matt Turner5-47/+34
6 daysi965/fs: Make count_to_loop_end() use basic blocks.Matt Turner1-15/+16
6 daysi965/fs: Don't use instruction list after calculating the cfg.Matt Turner3-22/+22
6 daysi965/fs: Preserve the CFG in a few more places.Matt Turner2-15/+21
8 daysi965/fs: Remove direct fs_visitor brw_wm_prog_key dependenceJordan Justen5-9/+36
8 daysi965/fs: Use brw_sampler_prog_key_data instead of brw_wm_prog_key::texJordan Justen1-12/+29