diff options
author | Juan A. Suarez Romero <[email protected]> | 2016-07-18 07:27:56 +0000 |
---|---|---|
committer | Francisco Jerez <[email protected]> | 2017-04-14 14:56:07 -0700 |
commit | 0f1316d4dbc19f46acd5a738df25e632d95f4105 (patch) | |
tree | 2c2f69185eca82e06297758bfa22aa271e644f63 /src/intel/compiler | |
parent | 79af2563889098550de5d4a0955efbeb87a24565 (diff) |
i965/fs: double regioning parameters and execsize for DF in IVB/BYT
In IVB and BYT, both regioning parameters and execution sizes are measured as
32-bits element size.
So when we have something like:
mov(8) g2<1>DF g3<4,4,1>DF
We are not actually moving 8 doubles (our intention), but 4 doubles.
We need to double the parameters to cope with this issue. However,
horizontal strides don't behave as they're supposed to on IVB
for DF regions, they will cause each 32-bit half of DF sources to be
strided individually, and doubling the value won't make any difference.
v2:
- Use devinfo directly (Matt).
- Use Baytrail instead of Valleview (Matt).
- Use IvyBridge instead of Ivy (Matt)
- Double the exec_size in code emission (Curro)
v3:
- Change hstride doubling by an assert and fix commit log (Curro).
- Substitute remaining compiler->devinfo by devinfo (Curro).
v4:
- Fix comment (Curro).
Signed-off-by: Samuel Iglesias Gonsálvez <[email protected]>
Reviewed-by: Francisco Jerez <[email protected]>
Diffstat (limited to 'src/intel/compiler')
-rw-r--r-- | src/intel/compiler/brw_fs_generator.cpp | 50 |
1 files changed, 43 insertions, 7 deletions
diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp index aeed6a11977..bcb6608d3ae 100644 --- a/src/intel/compiler/brw_fs_generator.cpp +++ b/src/intel/compiler/brw_fs_generator.cpp @@ -53,13 +53,14 @@ brw_file_from_reg(fs_reg *reg) } static struct brw_reg -brw_reg_from_fs_reg(fs_inst *inst, fs_reg *reg, unsigned gen, bool compressed) +brw_reg_from_fs_reg(const struct gen_device_info *devinfo, fs_inst *inst, + fs_reg *reg, bool compressed) { struct brw_reg brw_reg; switch (reg->file) { case MRF: - assert((reg->nr & ~BRW_MRF_COMPR4) < BRW_MAX_MRF(gen)); + assert((reg->nr & ~BRW_MRF_COMPR4) < BRW_MAX_MRF(devinfo->gen)); /* Fallthrough */ case VGRF: if (reg->stride == 0) { @@ -92,6 +93,34 @@ brw_reg_from_fs_reg(fs_inst *inst, fs_reg *reg, unsigned gen, bool compressed) const unsigned width = MIN2(reg_width, phys_width); brw_reg = brw_vecn_reg(width, brw_file_from_reg(reg), reg->nr, 0); brw_reg = stride(brw_reg, width * reg->stride, width, reg->stride); + /* From the IvyBridge PRM (EU Changes by Processor Generation, page 13): + * "Each DF (Double Float) operand uses an element size of 4 rather + * than 8 and all regioning parameters are twice what the values + * would be based on the true element size: ExecSize, Width, + * HorzStride, and VertStride. Each DF operand uses a pair of + * channels and all masking and swizzing should be adjusted + * appropriately." + * + * From the IvyBridge PRM (Special Requirements for Handling Double + * Precision Data Types, page 71): + * "In Align1 mode, all regioning parameters like stride, execution + * size, and width must use the syntax of a pair of packed + * floats. The offsets for these data types must be 64-bit + * aligned. The execution size and regioning parameters are in terms + * of floats." + * + * Summarized: when handling DF-typed arguments, ExecSize, + * VertStride, and Width must be doubled. + * + * It applies to BayTrail too. + */ + if (devinfo->gen == 7 && !devinfo->is_haswell && + type_sz(reg->type) == 8) { + brw_reg.width++; + if (brw_reg.vstride > 0) + brw_reg.vstride++; + assert(brw_reg.hstride == BRW_HORIZONTAL_STRIDE_1); + } } brw_reg = retype(brw_reg, reg->type); @@ -1585,9 +1614,8 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width) brw_set_default_group(p, inst->group); for (unsigned int i = 0; i < inst->sources; i++) { - src[i] = brw_reg_from_fs_reg(inst, &inst->src[i], devinfo->gen, - compressed); - + src[i] = brw_reg_from_fs_reg(devinfo, inst, + &inst->src[i], compressed); /* The accumulator result appears to get used for the * conditional modifier generation. When negating a UD * value, there is a 33rd bit generated for the sign in the @@ -1598,7 +1626,8 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width) inst->src[i].type != BRW_REGISTER_TYPE_UD || !inst->src[i].negate); } - dst = brw_reg_from_fs_reg(inst, &inst->dst, devinfo->gen, compressed); + dst = brw_reg_from_fs_reg(devinfo, inst, + &inst->dst, compressed); brw_set_default_access_mode(p, BRW_ALIGN_1); brw_set_default_predicate_control(p, inst->predicate); @@ -1607,7 +1636,14 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width) brw_set_default_saturate(p, inst->saturate); brw_set_default_mask_control(p, inst->force_writemask_all); brw_set_default_acc_write_control(p, inst->writes_accumulator); - brw_set_default_exec_size(p, cvt(inst->exec_size) - 1); + + unsigned exec_size = inst->exec_size; + if (devinfo->gen == 7 && !devinfo->is_haswell && + (get_exec_type_size(inst) == 8 || type_sz(inst->dst.type) == 8)) { + exec_size *= 2; + } + + brw_set_default_exec_size(p, cvt(exec_size) - 1); assert(inst->force_writemask_all || inst->exec_size >= 4); assert(inst->force_writemask_all || inst->group % inst->exec_size == 0); |