ed/TimothyLottes.github.io
1
0
Fork 0
mirror of https://github.com/gomson/TimothyLottes.github.io.git synced 2026-08-04 14:48:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add files via upload

Browse Source
This commit is contained in:
TimothyLottes
2016-11-10 19:24:50 -05:00
committed by GitHub
parent 249d7859e3
commit 3375c9ac01
3 changed files with 68 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
20150309.html
+18
View File
@@ -0,0 +1,18 @@
<html><head><link rel="stylesheet" href="style.css"></head><body><div class="page">
<h1>20150309 - Quick Thoughts on Strobbed LCD Displays for VR Perf Proxy</h1>
<br>
Using numbers from <a href="http://www.roadtovr.com/valve-talks-advanced-vr-rendering-live-blog-5pm-pst/">Alex Vlachos's Vive VR Talk at GDC</a>, Vive with stencil optimization and super-sampling required for regular rendering: 378 Mpix/s (rectangular projection), or with just stencil optimization for ray tracing: 193 Mpix/s (derived from the Valve numbers, no 1.4x multiplier, shoot rays in warped view directly). Interesting that tracing or non-standard rendering tech might be able to leverage a 50% cut in the number of pixels to process.<br>
<br>
Looking for a non-VR perf proxy? Something non-VR to target to get in the ballpark for frame cost for VR? For the regular rendering target: 2560x1440 at 100 Hz is still under the Mpix/s target. The 100 Hz figure is the highest ULMB strobbed refresh supported by the first strobbed IPS LCD panel: <a href="http://www.tftcentral.co.uk/reviews/acer_xb270hu.htm">Acer XB270HU</a>. Other than the viewing angle advantage of that IPS panel, it has roughly the same {gamut, brightness, etc} as the 1080p TN <a href="http://www.tftcentral.co.uk/reviews/benq_xl2720z.htm">BenQ XL2720Z</a> (roughly $450), and the TN has the advantage of supporting ULMB strobbed refresh all the way up to 144 Hz. For 1440p personally I think the larger <a href="http://orwww2.benq.com/product/monitor/xl2730z/">BenQ XL2730Z</a> is a better option than the Acer: the BenQ supports the VESA Standard Adaptive-Sync instead of being vendor locked for adaptive frame rate. Getting back to the numbers: 1920x1080 at 120 Hz is roughly 1.3x the "ray tracing" target for Vive, and 1920x1080 at 144 Hz is roughly 0.79x the "regular rendering" target (so would need some amount of super-sampling to be a perf proxy). It would be relatively easy to use the 1080p BenQ as a low persistence non-VR perf proxy for either case.<br>
<br>
</div></body></html>
20150420.html
+48
View File
@@ -0,0 +1,48 @@
<html><head><link rel="stylesheet" href="style.css"></head><body><div class="page">
<h1>20150420 - From Scratch Bug 2 : Source-Less Programming</h1>
<br>
<i>This is a disruptive idea which comes back periodically: source-less programming.
Is it possible to efficiently program at a level even lower than an assembler?</i><br>
<br>
The general idea is that the editor is now something similar to an enhanced hex editor which edits a binary image directly.
Lowest memory is split into three parts {{running image, annotations for edit image}, edit image}.
The {running image, annotations for edit image} is the boot image.
The {edit image} is working space which gets snapshot replacement for {running image} on a safe transition zone.
The "annotation" section is what enables human understanding of the binary image.<br>
<br>
<b>Words</b>
<br>
One way to keep the system very simple is to always work in 32-bit words.
A 32-bit word in memory is one of four things {data, opcode, absolute address, rip relative address}.
Data is easily handled by the hex editor part.
The annotation could provide a name for the data or a comment.
Opcodes in x86 can be complex but it is easy to simplify.
Start with something similar to forth zero-operand and one-operand operations (calls, etc).
Make all operations padded to 32-bit alignment (x86-64 can use the 2e ignored prefix to pad).
A call for instance becomes {32-bit opcode for call, 32-bit rip relative branch address}.
Or a global load becomes {32-bit opcode for load, 32-bit rip relative branch address}.
Annotation memory can provide a name for the opcode.
Annotation can provide a tag for each word in memory which marks
if the memory is a relative or absolute address (word gets a different color based on tag similar to color forth).
Addresses can be auto annotated by displaying the annotation associated with the destination.
Editor works on the {edit image},
with insert and delete of words automatically adjusting all the words tagged as address
(effectively relinking at edit time).
The {edit image} can also keep a mapping of original {running image} address
so that it is possible to view the live values of any data.
Editor provides something to easily be able to write an annotation
and have the associated opcode or address automatically get updated.
For example type the opcode name and the 32-bit value is automatically updated.
Very simple and yet powerful minimal tool.
</div></body></html>
index.html
+2
View File
@@ -97,12 +97,14 @@ Below this is active random migration (400 prior posts still to filter through)
<a href="20150423.html">20150423 - Source-Less Programming : 3</a><br>
<a href="20150422.html">20150422 - Source-Less Programming : 2</a><br>
<a href="20150421.html">20150421 - Look No Triangles : Scatter vs Gather</a><br>
<a href="20150420.html">20150420 - From Scratch Bug 2 : Source-Less Programming</a><br>
<a href="20150416.html">20150416 - Pixel Art and Slot Mask Pitch</a><br>
<a href="20150415.html">20150415 - Indie vs Real Slug Fest</a><br>
<a href="20150414.html">20150414 - From Scratch Bug</a><br>
<a href="20150406.html">20150406 - End of an Era</a><br>
<a href="20150403.html">20150403 - Why I'm Using Fedex From Now On</a><br>
<a href="20150328.html">20150328 - Stills From My Talk</a><br>
<a href="20150309.html">20150309 - Quick Thoughts on Strobbed LCD Displays for VR Perf Proxy</a><br>
<a href="20150327.html">20150327 - Other CRT Options</a><br>
<a href="20150308.html">20150308 - CRTs</a><br>
<br>