Lines Matching refs:index
1024 LLVMValueRef index)
1037 stride = LLVMBuildMul(builder, vb_stride, index, "");
1044 lp_build_print_value(gallivm, " instance index = ", index);
1066 LLVMValueRef index = lp_build_const_int32(gallivm, i);
1069 vs_type, aos, index);
1150 LLVMValueRef index,
1159 indices[1] = index;
1166 lp_build_printf(gallivm, " ---- %p storing attribute %d (io = %p)\n", data_ptr, index, io_ptr);
1282 lp_build_printf(gallivm, "io = %p, index %d, clipmask = %x\n",
1463 LLVMValueRef index;
1465 index = lp_build_const_int32(gallivm, i + scale_index_offset);
1466 scale_i = LLVMBuildGEP2(builder, elem_type, vp_ptr, &index, 1, "");
1468 index = lp_build_const_int32(gallivm, i + trans_index_offset);
1469 trans_i = LLVMBuildGEP2(builder, elem_type, vp_ptr, &index, 1, "");
1936 * The code can't handle vp transform when vs writes vp index neither
1990 * get index buffer overflows due to vsplit (which provides its own
2060 LLVMValueRef index = lp_build_const_int32(gallivm, i);
2061 ind_vec = LLVMBuildInsertElement(builder, ind_vec, index, index, "");
2127 * index = start_instance + (instance_id / divisor).
2160 lp_build_printf(gallivm, "velem %d, vbuf index = %u, vb_stride = %u\n",
2195 * beyond index buffer (or rather vsplit elt buffer) size.
2212 * detected if index > fetch_elt_max - but the correct condition
2213 * would be index >= fetch_elt_max (since this is just size of elts
2219 * not being zero will get a different fetch index than the valid
2220 * index 0. So, just rely on vsplit code preventing out-of-bounds
2222 * was no index buffer bound - the real buffer is never seen here, at
2223 * least not if there are index buffer overflows...
2276 * index bias and because indices inside our elements array have
2277 * already had index bias applied we need to subtract it here to
2278 * get back to the original index.
2280 * original start index and because we abuse the 'start' variable
2281 * to either represent the actual start index or the index at which
2284 * index out of our vertex id here.