Add parameters and possible values in documentation

src/README.md

@@ -10,6 +10,12 @@ The algorithms that has constrainsts for parameters, includes it in the related
 | `PRECISION` | Whether to use single-precision floating-point or double-precision floating-point for the computations. | {`float` (32), `double` (64)} |
 
 ### BFS
+| Parameter | Description | Search Space |
+| --- | --- | --- |
+| `BLOCK_SIZE` | The block size used for launching the kernel. | {1, ..., `Max GPU threads per block` or 1000 if size=1} |
+| `CHUNK_FACTOR` | Factor multiplied with a chunk size which is the work per thread. | {1, 2, 4, 8} |
+| `TEXTURE_MEMORY_EA1` | Not use texture memory, use texture reference (older version) or use texture object (newer version). | {0, 1, 2} |
+| `TEXTURE_MEMORY_EAA` | Not use texture memory, use texture reference (older version) or use texture object (newer version). | {0, 1, 2} |
 
 ### Molecular Dynamics  (MD)
 | Parameter | Description | Search Space |
@@ -20,6 +26,20 @@ The algorithms that has constrainsts for parameters, includes it in the related
 | `WORK_PER_THREAD` | The amount of work performed by each GPU thread. | {1, ..., 5} |
 
 ### SPMV
+| Parameter | Description | Search Space |
+| --- | --- | --- |
+| `BLOCK_SIZE` | The block size used for the kernel. | {1, ..., `Max GPU threads per block`} |
+| `PRECISION` | Use single or double precision on variables. | {32, 64} |
+| `FORMAT` | Which format to use for SpMV. <br> 0: ELLPACK-R <br>1: CSR Normal Scalar <br>2: CSR Padded Scalar <br>3: CSR Normal Vector <br>4: CSR Padded Vector | {0, 1, 2, 3, 4} |
+| `UNROLL_LOOP_2` | Not unroll or unroll loop. | {0, 1} |
+| `TEXTURE_MEMORY` | Not use texture memory or use the newer version of texture memory (texture objects). | {0, 1} |
+
+#### Constraints
+| Constraint | Description |
+| --- | --- |
+| `FORMAT < 3 or BLOCK_SIZE % 32 == 0` | Format 3 or 4, which is the CSR Vector formats, needs to have block size that is divisible by 32. |
+| `FORMAT > 2 or UNROLL_LOOP_2 < 1` | The loop exists only in the CSR vector kernel, which means that the loop unroll parameter should be 0 when the format is something else than CSR vector. |
+
 
 ### Sort
 | Parameter | Description | Search Space |
@@ -41,6 +61,16 @@ The algorithms that has constrainsts for parameters, includes it in the related
 | `(8 * SCAN_DATA_SIZE * SCAN_BLOCK_SIZE + 128) ≤ available_shared_memory` | Available shared memory for the selected GPU can not be less than needed shared memory in the `reorderData` kernel. |
 
 ### MD5 Hash
+| Parameter | Description | Search Space |
+| --- | --- | --- |
+| `BLOCK_SIZE` | The block size used for the kernel. | {1, ..., `Max GPU threads per block`} |
+| `ROUND_STYLE` | Which MD5 round style to use. | {0, 1} |
+| `UNROLL_LOOP_1` | Not unroll or unroll loop 1. | {0, 1} |
+| `UNROLL_LOOP_2` | Not unroll or unroll loop 2. | {0, 1} |
+| `UNROLL_LOOP_3` | Not unroll or unroll loop 3. | {0, 1} |
+| `INLINE_1` | Not inline or inline kernel 1. | {0, 1} |
+| `INLINE_2` | Not inline or inline kernel 2. | {0, 1} |
+| `WORK_PER_THREAD_FACTOR` | Factor for setting how much work a thread should do. | {1, 2, 3, 4, 5} |
 
 ### Reduction
 | Parameter | Description | Search Space |
@@ -58,6 +88,24 @@ The algorithms that has constrainsts for parameters, includes it in the related
 | `TEXTURE_MEMORY` | Whether to use texture memory or not for the input data to the `reduce` kernel. | {False (0), True (1)} |
 
 ### Scan
+| Parameter | Description | Search Space |
+| --- | --- | --- |
+| `BLOCK_SIZE` | The block size used for the kernels. | {16, 64, 128, 256, 512} |
+| `GRID_SIZE` | The grid size used for the kernels. | {1, 2, 4, 8, 16, 32, 64, 128, 256, 512} |
+| `PRECISION` | Use single or double precision. | {32, 64} |
+| `UNROLL_LOOP_1` | Not unroll or unroll loop 1. | {0, 1} |
+| `UNROLL_LOOP_2` | Not unroll or unroll loop 2. | {0, 1} |
+| `USE_FAST_MATH` | If fast math will be used when compiling. | {0, 1} |
+| `OPTIMIZATION_LEVEL_HOST` | Compiler optimization level for host code. | {0, 1, 2, 3} |
+| `OPTIMIZATION_LEVEL_DEVICE` | Compiler optimization level for GPU code. | {0, 1, 2, 3} |
+| `MAX_REGISTERS` | The amount of max registers to be used for the GPU kernels. -1 means to not set this compiler option. | {-1, 20, 40, 60, 80, 100, 120} |
+| `GPUS` | The number of GPUs used in the computations.  | {1, ..., `Connected GPUs`} |
+
+#### Constraints
+| Constraint | Description |
+| --- | --- |
+| `GRID_SIZE <= BLOCK_SIZE` | Grid size needs to be less than or equal to block size. |
+
 
 ### Stencil 2D
 | Parameter | Description | Search Space |