matrox own presentation of displacement mapping (translated)

[ZokesPro]

-- where you understand the new technology " D-MAP " of DirectX 9 attention that 1 Juan Guardado person session of the Matrox which did announcement concerning D-MAP technology: March 19th - March 23rd (local time) Opening area: San Jose MC$enery Convention Center and the like The Matrox technical explanation the displacement mapping which being supported the Direct3d of DirectX 9 (the alias D3D9) in is decisive (the Displacement Mapping) with concerning the new technology which is called was done in San Jose in the Game Developers Conference which is held from 19 days to 24 days. As with this manuscript you understand the basis of this technology, it expects to the form of the D3D9 corresponding hardware which is visible from here. Dividing into this day and 2 times of tomorrow, it reports. -------------------------------------------------------------------------------- * What displacement mapping? - The conception, apex texture Displacement * mapping (Displacement Mapping: Below D-MAP) With, the Mapping= it sticks Displacement= displacement and substitution...... with it becomes the mind which is said. The expression, texture mapping probably hearing it has been. This the picture which is called texture with the technology which sticks to the 3d object which was formed with polygon, is the expression expedient which becomes most the basis of present 3d graphics. With the D-MAP it is not to stick image, it probably will be easy to represent intuition you catch in the manner that " apex information is stuck ". The expression, " apex texture mapping " was proposed even with session of the Matrox. " Apex is stuck ", with what thing? For example when we assume, there was a sphere, it becomes deformed in the form for example like the fish by sticking another apex information vis-a-vis this, or it is possible to become deformed in the head of the human where eye nose shape is attached. Furthermore if it compares is, if it applies the D-MAP to the 3d model which becomes the basis, like clay workmanship deformation is possible...... with it means to say. The person who is the technology which is called bump mapping as the expression which acquires unevenness vis-a-vis the 3d model which becomes the basis they probably are many things. Bump mapping is called the height map, it is something of the image that it sticks unevenness to polygon, unevenness was expressed (or housed the normal vector) making use of texture, but the D-MAP with the unevenness information the 3d model processing (conversion) is the technology which is done at geometry level. At the highest only the dynamic range of 8 - 16bit integer there was unevenness information of bump mapping, but because with the D-MAP it can utilize largest 32bit floating decimal, it means to be able to express rather complicated form.

* The merit of D-MAP? - With bump mapping useless reason For example when we assume with bump mapping unevenness information, could be expressed with 32bit floating decimal, how probably will be? Because very rich unevenness expression is possible vis-a-vis the object, when that 3d model was made to deform, it probably is to be able to create equal visual. To tell the truth this does not go well. Assuming, with general bump mapping at the time in rendering lighting was done at the pixel unit, shading in that pixel (only shadow processing) it is done. Because it cannot reproduce the shadow which the unevenness causes around, it has not become really correct image. The fact that this is displayed conceptually is the figure 1a. With general bump mapping as for the shadow of the convex part it is reflected on only the convex part. Originally if is, the shadow of the convex part is the expectation which reaches to around that base, but it cannot process this. So, the generally known self shadow is not reflected. If it is the delicate unevenness impression of the skin and roughly the impression et cetera of the rock wall, not doing strict shadow processing to there, that ‚Á ‚Û ‚ it is visible. But, when the dynamic range of the unevenness, it is very large, like the figure 1a this lie exposes. Because unevenness can do with the D-MAP at geometry level, also the self shadow is considered, if (you try to do...... is but). The figure 1b is the inner product bump mapping whose unevenness is large, but there is an unevenness truly, if is, even in the edge in sphere the convex the expectation where concave appears. Bump mapping being to be image based technology is this kind of weak point.

[ZokesPro]

Part 2

* Apex conservation - the character where appearance differs completely from the identical basic geometric model is formed with the D-MAP With graphics of photo realistic expression multi apex conversions of the 3d model are considerable. This flow is seen with high resolution conversion of indicatory resolution it continues for a while. As for apex information of such character model being placed in memory, it is the case that it is managed, but if the character model which had the complicated form where the number of apexes is many becomes many, that much also the memory consumption becomes many. As for the D-MAP you say it can become also the solution which solves such problem. If only the 3d model which had a certain basic form it is, the form expression where the or more of that is detailed says processing with the D-MAP. The picture A is the human body model of very simple form of 100 polygon front and back. As for the picture B it is the example which applies the D-MAP to the simple human body model of the same these 100 polygons. The simple human body model of identical shape it changed to the character which has the form classified by Mack due to the D-MAP...... with it is the case that it is said. The picture C is the completion shape which sticks texture to this. The extent which it created from the identical basic model and with cannot think, you are surprised to becoming the character of the lux which differs. < Picture A > This basic model < Picture B > When the D-MAP is applied to this. Being able to achieve the extent dynamic form change which it became deformed from the identical basic model and cannot think, it is < Picture C > Doing texture and bump mapping, completion shape. This the expression power of D-MAP Vague, it did, it becomes even if, but as for the human body ??? of the present 3d game there are about 1,000 polygons. Like this example if we assume you thought 1,000 polygon suitable characters could be expressed with the human body model of 100 polygons, this is plugged and means to be able to compress the number of apexes in 1/10. When we assume the appearance character in the game was expressed with D-MAP application on the basis of the human body model of this simple form of everything altogether game it becomes conservation considerable of the apex data quantity. If the apex data can be saved, therefore the reason which means also the number of apex the between the CPU-AGP transfers to decrease, pulling, it is the case that it becomes also bus conservation the between the CPU-AGP. But, there is also restriction of that appearance in the D-MAP. First, with the D-MAP doing, the fact that making the cube like the die the space ship of the acute angle surface being lazy deform means it is not possible. If there is no certain basic form, useless it puts out, the connection of the D-MAP does not become beautiful. Concerning this it mentions later. In addition, it seems like the human body model, case of the multi joint characters, the degree of freedom of the movement at the bone it does not have to set and must control. At that time, even excessively model form of crust side being too simple, when operating that character, there is a possibility the movement failing. You must examine making use of the D-MAP utility and the like which this will repeat cutting & trying, in the future probably will appear tries, it probably is the part. The basic geometric model however much being simple, it is not the case that after all finally as for the number of apexes being the case that it increases, in that sense, the number of processing of GPU core side apexes decreases with the D-MAP. In addition, form of the basic model which is handled being simple, saying, apex point total number may be small, becomes bus conservation the between the CPU-AGP local bus side of the GPU with the apex transfer which was increased by D-MAP processing system, that much becomes busy. Amount and the reverse side which in the chart have become easy become busy...... with it is saying. -------------------------------------------------------------------------------- * As for the D-MAP being similar to the N-PATCH, those which become fault To " as for the D-MAP following to the geometry data which is appointed, being able to express the thing and one mouth which make 3d object deform ", the actualization style very is complicated. To " making 3d object deform ", the fact that it becomes necessary as a processing system is, " automatic operation formation of be adapted and dynamic apex ". With bump mapping it just keeps doing the lighting calculation which corresponds to unevenness information the case of shading it is the case, but with the D-MAP unevenness is really formed. In other words being the case that it means that the polygon which forms unevenness is formed, this namely is not anything less than apex being formed. When you mention " automatic operation formation of apex ", it is special function of the ATI RADEON 8500 GPU, the person who is " TRUFORM " function it probably will be many. < Figure 2 > The TRUFORM the plane surface division which is called the N-PATCH (finally apex formation) is the function which the mechanism the ????????????? is done. Because normal vector (the Normal) it means N, the N-PATCH, it becomes the kind of meaning " normal vector (N) patching (the PATCH) of doing ". As for figure 2 the TRUFORM (N-PATCH) it is something which diagrams basic algorithm. With the N-PATCH you pay attention to the direction of normal vector of three apexes of triangle, between three apexes, it keeps deciding the new kind of apex where change of normal vector becomes natural. The fact that it keeps dividing polygon this way, call by result automatically forming the smooth curved surface it keeps is the N-PATCH. If the N-PATCH was applied, the ???? and the ???? it did, being the 3d model of the form where the numbers of polygons are few, being finally living thing, it can indicate the object which had the judo kana aspect...... with it is the case that it is said. Also the apex formation in the D-MAP, in the sense that has been similar to the concept of the N-PATCH apex is formed arithmetic,, but the place where it keeps forming apex on 3d object side with the D-MAP with the unevenness information which is prepared as an origin is different. (It continues to the next time).