Click on an instruction to jump to that page

 

abs

m3x2

ps

tex

tex3x2depth

add

m3x3

rcp

texbem

tex3x2pad

bem

m3x4

rsq

texbeml

tex3x3tex

cmp

m4x3

sincos

texcoord

texm3x3

cnd

m4x4

sub

texcrd

texm3x3pad

crs

mad

 

texdepth

texm3x3tex

dcl

max

 

texdp3

texm3x3vspec

def

min

 

texdp3tex

texreg2ar

dp2add

mov

 

texkill

texreg2gb

dp3

mul

 

texld

texreg2rgb

dp4

nop

 

texldb

 

exp

nrm

 

texldp

 

frc

phase

 

 

 

log

pow

 

 

 

lrp

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

abs (macro)

 ps 2.0

 

This macro computes the absolute value of the input register on an element by element basis.

One arithmetic instruction slot.

abs  Dest0, Source0

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Computes the absolute value of each element of Source0 and places the result in Dest0.

 

abs  r4, r4

 

this macro is equivalent to the following code.

 

cmp Dest0, Source0, Source0, -Source0

 

add

 ps 1.0 – 2.0

 

Adds two source colors and places the sum into a third register.

One arithmetic instruction slot.

add  Dest0, Source0, Source1

 

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

x

x

x

x

x

1.1

 

 

x

x

x

x

x

x

1.2

 

 

x

x

x

x

x

x

1.3

 

 

x

x

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

 

Adds the Source0 and Source1 registers and places the result in the Dest0 register.

 

add  r0,     r0,  c2

add r0,     r0,  1-r1

add  r0,     r0,  t0_bias

 

 

bem

 ps 1.4

 

Bump-Environment-Map. Apply a fake bump environment map transform.

Two arithmetic instruction slots.

bem  Dest0.rg, Source0, Source1

 

Legal Register Arguments

 

Dest

Source0

Source1

PS Version

v

c

t

r

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

x

 

x

 

x

 

 

 

x

1.4 Phase 2

 

 

 

 

 

 

 

 

 

 

 

 

 

This instruction takes two source registers, applies the 2x2 bump-map matrix to Source1, adds in Source0 and places the result in the destination register. This instruction only operates on the .rg components. The matrix values are set prior to the pixel shader through a SetTextureStage call setting the four D3DTSS_BUMPENVMATnn matrix elements. The destination register number determines the texture stage number. There are no restrictions on modifiers on the source registers. There are restrictions on using the bem instruction;

 

 

Note: This instruction works only in PS 1.4.

 

bem  r3.rg,  r0,  c2 // bump texture 3 * c2, add in r0

bem  r4.xy   c2,  r0 // bump texture 4 * r0, add in c2

 

 

cmp

 ps 1.2 - 2.0

 

Compare: Performs a conditional assignment based upon a comparison of the values in the source registers.

Two arithmetic slots for PS 1.2 & 1.3, One for PS 1.4 & 2.0

cmp  Dest0, Source0, Source1, Source2

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

x

x

x

x

x

x

1.3

 

 

x

x

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

2.0

 

 

x

x

x

x

x

x

 

If the value in Source 0 is greater-than or equal-to zero, then Source1 is placed in the destination; else Source2 is placed in the destination. This comparison is performed on an element-by-element basis. Dest0 must be a different register than any of the source registers.

 

Note: For PS 1.2-1.3 the source register can be any register, while the destination can be either a texture or a temporary registers. There can be a maximum of three cmp instruction is a pixel shader.

 

Bug: For PS 1.2 & 1.3, the destination register cannot be the same as a source register, but validation fails to catch this. The cmp instruction takes up two slots, but validation counts this as only one slot.

For PS 1.4 the source register can be temporary registers, constant registers, or vertex color registers for phase 2, while the source register can be temporary registers or constant registers for phase 1. The destination register must be a temporary register.

 

 

 

cmp  r0, r1, c1, c2

 

 

cnd

 ps 1.1 – 1.4

 

Conditional: Performs a conditional assignment based upon a comparison of the value in r0.a (PS 1.1-1.3) or another register (PS1.4) with 0.5.

One arithmetic instruction slot.

cnd  Dest0, Source0, Source1, Source2

 

 

Legal Register Arguments

 

Dest

Source0

Source1/2

PS Version

v

c

t

r

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

 

 

 

 

1.1

 

 

x

x

 

 

 

r0.a

x

x

x

x

1.2

 

 

x

x

 

 

 

r0.a

x

x

x

x

1.3

 

 

x

x

 

 

 

r0.a

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

x

x

 

x

 

Compares the value in r0.a against 0.5.  Source0 must be r0.a for PS 1.1 – 1.3. If Source0 is greater than 0.5, then Dest0 is set to the value in Source1. If not, then Dest0 is set to the value in Source3.

For PS 1.4, the cnd instruction operates on an element-by-element basis.

 

 

 

 

You can also use this command to compare between two values with an added step as shown below

 

cnd  r0,   r0.a, c1,  c2 // PS 1.1-PS 1.3

 

// to compare two values v0 and v1

sub  r0,  v0, v1_bias ; r0 = v0-(v1-0.5) = v0-v1+0.5

cnd  r0,  r0.a, c0, c1 ; if r0.a > 0.5 use c0, else c1

; thus v0-v1+0.5 > 0.5 = v0-v1 > 0 = v0 > v1

 

 

crs (macro)

 ps 2.0

 

The three component cross product.

Two arithmetic instruction slots.

crs  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Computes the three component cross product using the right-hand rule.  There cannot be ands swizzles on the source registers. Dest0 must have a write mask that is one of the following: .x,  .y, .z, .xy, .xz, .yz, or .xyz. The destination register should be different from the source registers.

 

crs  r1, v0, v3

 

this macro is equivalent to the following code sans write mask on Dest0.

 

mul Dest0, Source0.zxyw, Source1.yzxw

mad Dest0, Source0.yzxw, Source1.zxyw, -Dest0

 

 

 

 

 

 

 

dcl

 ps 2.0

 

Declare how a texture sampler register is to be used.

Takes no slots

dcl_<type> Dest0

 

In order to make it easier to optimize and verify shaders PS 2.0 now requires a declaration statement on all sampler, color input, and texture coordinate registers. All sampler registers must be declared before use in the shader. Dest0 will be a specific input register. The <type> tag is an indicator of the type of texture this register is to be used with. The allowable types are; 2d, cube, and volume.

 

Color input and texture coordinate register must specify a mask of which components are used. Texture coordinate register may specify the partial precision (_pp) modifier.

 

 Must appear before any arithmetic or texture instructions.

 

// sampler registers

dcl_2d     s1 // s1 will be using a 2D texture

dcl_cube   s2 // s2 will be using a cubemap

dcl_volume s3 // s3 will be using a 3D texture

 

// color and texture coordinate registers

dcl t1.rg

dcl_pp t2.rg

dcl v0.rgba

 

 

 

def

 ps 1.0 – 2.0

 

Defines the constants to be used in the shader.

Takes no slot or time.

def  Const1, float1, float2, float3 float4

 

The first argument is the name of the constant – e.g. c0, c1, etc. The remaining arguments are the four floating-point values to be placed in the constant register in a, r, b, b order. The constants must be in the [-1,1] range. The constant are available one the shader is loaded, but a subsequent call to SetPixelShaderConstant() will overwrite any constant values previously set.

 

def  c0, 0.0f, 0.0f, 0.0f, 0.0f

def  c1, 1.0, 0.5, 0.25, 0.125   

 

 

dp2add

 ps  2.0

 

Two-component dot product plus a scalar add. Can be used to emulate the bem instruction.

One arithmetic instruction slot.

dp2add Dest0, Source0, Source1, Source2

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

Computes the two component dot product of the Source0 and Source1 (elements .r and .g) and add in a scalar value from Source2 (which must have a replicate swizzle). The results are replicated to all elements of the destination

 

p2add  r0, v0, v1, c4.z

 

 

dp3

 ps 1.0 – 2.0

 

Three-component dot product (a.k.a. Dot-product three) of the color arguments, replicated to all channels of the output register.

One arithmetic instruction slot.

dp3  Dest0, Source0, Source1

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

x

x

x

x

x

1.1

 

 

x

x

x

x

x

x

1.2

 

 

x

x

x

x

x

x

1.3

 

 

x

x

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

2.0

 

 

x

x

x

x

x

x

 

Computes the dot product of the Source0 and Source1 color registers and places the result in all channels of the Dest0 register. Only the r, g and b values are used to compute the dot product, the alpha component is ignored. The dp3 instruction does not clamp the results to the range [0,1]. If you want to clamp the results, use the _sat modifier.  If a write-mask is used only the selected channels are written.

 

Note: Since the dp3 instruction is a vector operation it is always scheduled for the vector pipeline. Thus when used with instruction pairing it’s always got to be the vector operation. It can be co-issued if the dp3 is writing color channels while the other instruction is writing the alpha channel.

 

dp3      r0,     t0,  v2 ; rgba is set

dp3      r0.rgb, t0,  v2 ; only rgb is set

dp3      r0.a,   t0,  v2 ; only a is set

 

dp4

 ps 1.2 – 2.0

 

Four-component dot product (a.k.a. Dot-product four) of the color arguments, replicated to all channels of the output register.

Two arithmetic instruction  slots in PS 1.2 & 1.3, one slot in PS 1.4 & 2.0

dp4  Dest0, Source0, Source1

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

 

x

x

x

x

x

1.3

 

 

 

x

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

2.0

 

 

 

x

x

x

x

x

 

Computes the dot product of the Source0 and Source1 color registers and places the result in all channels of the Dest0 register. The r, g, b, and a values are used to compute the dot product. The dp4 instruction does not clamp the results to the range [0,1].  If you want to clamp the results, use the _sat modifier.  If a write-mask is used only the selected channels are written.

 

Note: This instruction cannot be co-issued with another instruction.

 

For PS 1.2 & 1.3: A maximum of four dp4’s are allowed in a single pixel shader.

 

Bugs PS 1.2 & 1.3: The destination register should not be the same as the source registers, and validation does not catch this. While this instruction takes two slots, it’s actually counted as taking one, so that it’s possible to overflow the instruction buffer.

 

dp4      r0,     t0,  v2 ; rgba is set

dp4      r0.rgb, t0,  v2 ; only rgb is set

dp4      r0.a,   t0,  v2 ; only a is set

 

 

exp

 ps  2.0

 

Computes the base two exponent of a scalar value.

One arithmetic instruction slot.

exp  Dest0, Source0

 

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Computes 2Source0, where Source0 must have a replicate swizzle.

 

Setup:

Store the value you want the exponent of in an element of Source0. Use the replicate swizzle to select this element.

 

Results:

All elements of Dest0 will contain the exponential value.

 

exp  r1,  r2.y // replicate exp(r2.y) in r1

 

frc

 ps  2.0

 

This instruction removes the integer part of the input register on an element-by-element basis.

One arithmetic instruction slot.

frc  Dest0, Source0

 

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

Takes the fractional parts of Source0’s elements and places them in Dest0’s elements on a per-element basis. The truncation used is Source0-floor(Source0), so the results are always positive, that is, any negative values in Source0 will result in the fraction necessary to subtract to reach the next integer < Source0. e.g. If Source0.x = -5.4, Dest0.x will equal 0.6.

 

Setup:

Store the value you want the fractional part of in Source0. Use may use a replicate swizzle.

 

Results:

Dest0 will contain the fractional values of Source0.

 

frc  r6, r2

 

 

log

 ps  2.0

 

Compute the base two logarithm of a scalar.

One arithmetic instruction slot.

log  Dest0, Source0

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

Computes the base two log of a Source0 element. You must us a replicate swizzle to select the element of Soruce0. Dest0 will be filled with the result. If the source value is zero, then the destination result will be MINUS_INFINITY. (Or at least a really big negative number)

 

log  r0,  r5.w // place log2(r5.w) in r0

 

lrp

 ps 1.0 – 2.0

 

Linear interpolation between two registers (a.k.a. “lerp”) using a fraction specified in a third register.

One arithmetic instruction slot.

add  Dest0, Source0, Source1, Source2

 

Legal Register Arguments

 

Dest

Source

PS Version

v

c

t

r

v

c

t

r

1.0

 

 

 

x

x

x

x

x

1.1

 

 

x

x

x

x

x

x

1.2

 

 

x

x

x

x

x

x

1.3

 

 

x

x

x

x

x

x

1.4 Phase 1

 

 

 

x

 

x

 

x

1.4 Phase 2

 

 

 

x

x

x

 

x

2.0

 

 

x

x

x

x

x

x

 

Source0 contains the fractional interpolant value. When Source0 is zero, Source2 is placed in the destination. When Source0 is one, Source1 is placed in the destination. Values in the [0-1] range interpolate between Source1 and Source2. If the value is outside the range [0,1] the result is indeterminate.

 

Note: You need to be careful that the value of the interpolant doesn’t exceed the [0,1] range. When computing the interpolant use the _sat modifier to clamp the value.

 

lrp  r0,  s0,  s1,  s2

 

m3x2 (macro)

 ps 2.0

 

Perform a 3 × 2 matrix multiply on a 3 element vector.

Two arithmetic instruction slots.

m3x2  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Multiplies a 3 element vector by a 2 × 3 matrix. The result is stored in Dest0.xy. The next register from Source1 is implicitly used. Dest0 must not be the same as any source registers, both explicit and implicit.

 

You cannot use swizzle or negate modifiers on Source1.

 

m3x2  r1, r3, c4 // c5 use implicit

 

this macro is equivalent to the following code.

 

dp3 Dest0.x, Source0, Source1

dp3 Dest0.y, Source0, Source1+1

 

 

m3x3 (macro)

 ps 2.0

 

Perform a 3 × 3 matrix multiply on a 3 element vector.

Three arithmetic instruction slots.

m3x3  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Multiplies a 3 element vector by a 3 × 3 matrix. The result is stored in Dest0.xyz. The next two registers from Source1 are implicitly used. Dest0 must not be the same as any source registers, both explicit and implicit.

 

You cannot use swizzle or negate modifiers on Source1.

 

m3x3  r1, r3, c4 // c5 and c6 use implicit

 

this macro is equivalent to the following code.

 

dp3 Dest0.x, Source0, Source1

dp3 Dest0.y, Source0, Source1+1

dp3 Dest0.z, Source0, Source1+2

 

 

m3x4 (macro)

 ps 2.0

 

Perform a 3 × 4 matrix multiply on a 3 element vector.

Four arithmetic instruction slots.

m3x4  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Multiplies a 3 element vector by a 4 × 3 matrix. The result is stored in Dest0.xyzw. The next three registers from Source1 are implicitly used. Dest0 must not be the same as any source registers, both explicit and implicit.

 

You cannot use swizzle or negate modifiers on Source1.

 

m3x3  r1, r3, c4 // c5, c6, and c7 use implicit

 

this macro is equivalent to the following code.

 

dp3 Dest0.x, Source0, Source1

dp3 Dest0.y, Source0, Source1+1

dp3 Dest0.z, Source0, Source1+2

dp3 Dest0.w, Source0, Source1+3

 

 

m4x3 (macro)

 ps 2.0

 

Perform a 3 × 4 matrix multiply on a 4 element vector.

Three arithmetic instruction slots.

m4x3  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Multiplies a 4 element vector by a 3 × 4 matrix. The result is stored in Dest0.xyz. The next two registers from Source1 are implicitly used. Dest0 must not be the same as any source registers, both explicit and implicit.

 

You cannot use swizzle or negate modifiers on Source1.

 

m4x3  r1, r3, c4 // c5 and c6 use implicit

 

this macro is equivalent to the following code.

 

dp4 Dest0.x, Source0, Source1

dp4 Dest0.y, Source0, Source1+1

dp4 Dest0.z, Source0, Source1+2

 

 

m4x4 (macro)

 ps 2.0

 

Perform a 4 × 4 matrix multiply on a 4 element vector.

Four arithmetic instruction slots.

m4x4  Dest0, Source0, Source1

 

Legal Register Arguments

 

Destination

Source

PS Version

v

c

t

r

v

c

t

r

s

1.0

 

 

 

 

 

 

 

 

 

1.1

 

 

 

 

 

 

 

 

 

1.2

 

 

 

 

 

 

 

 

 

1.3

 

 

 

 

 

 

 

 

 

1.4 Phase 1

 

 

 

 

 

 

 

 

 

1.4 Phase 2

 

 

 

 

 

 

 

 

 

2.0

 

 

 

x

x

x

x

x

 

 

Multiplies a 4 element vector by a 4 × 4 matrix. The result is stored in Dest0.xyzw. The next three registers from Source1 are implicitly used. Dest0 must not be the same as any source registers, both explicit and implicit.

 

You cannot use swizzle or negate modifiers on Source1.

 

m4x3  r1, r3, c4 // c5, c6 and c7 use implicit

 

this macro is equivalent to the following code.

 

dp4 Dest0.x, Source0, Source1

dp4 Dest0.y, Source0, Source1+1

dp4 Dest0.z, Source0, Source1+2

dp4 Dest0.w, Source0, Source1+3

 

mad

 ps 1.0 – 2.0

 

Multiply and Add. Multiplies two registers and then adds a third to the result then stores the result in the destination.

One arithmetic instruction slot.

mad  Dest0, Source0, Source1, Source2

 

Legal Register Arguments

<