diff --git a/colossalai/kernel/cuda_native/csrc/scaled_masked_softmax.h b/colossalai/kernel/cuda_native/csrc/scaled_masked_softmax.h index 1583030b8..d3e6f04e6 100644 --- a/colossalai/kernel/cuda_native/csrc/scaled_masked_softmax.h +++ b/colossalai/kernel/cuda_native/csrc/scaled_masked_softmax.h @@ -4,12 +4,12 @@ #pragma once #include +#include #include +#include + #include #include -#include -#include -#include namespace { @@ -17,37 +17,53 @@ template __device__ __inline__ void copy_vector(Datatype *dst, const Datatype *src); template <> -__device__ __inline__ void copy_vector(c10::BFloat16 *dst, const c10::BFloat16 *src) { *dst = *src; } - -template <> -__device__ __inline__ void copy_vector(c10::BFloat16 *dst, const c10::BFloat16 *src) { *((float2*) dst) = *((float2*) src); } - -template <> -__device__ __inline__ void copy_vector(c10::Half *dst, const c10::Half *src) { *dst = *src; } - -template <> -__device__ __inline__ void copy_vector(c10::Half *dst, const c10::Half *src) { *((float2*) dst) = *((float2*) src); } - -template <> -__device__ __inline__ void copy_vector(uint8_t *dst, const uint8_t *src) { *dst = *src; } - -template <> -__device__ __inline__ void copy_vector(uint8_t *dst, const uint8_t *src) {*((half2*) dst) = *((half2*) src); } - -int log2_ceil(int value) { - int log2_value = 0; - while ((1 << log2_value) < value) ++log2_value; - return log2_value; +__device__ __inline__ void copy_vector( + c10::BFloat16 *dst, const c10::BFloat16 *src) { + *dst = *src; } -template +template <> +__device__ __inline__ void copy_vector( + c10::BFloat16 *dst, const c10::BFloat16 *src) { + *((float2 *)dst) = *((float2 *)src); +} + +template <> +__device__ __inline__ void copy_vector(c10::Half *dst, + const c10::Half *src) { + *dst = *src; +} + +template <> +__device__ __inline__ void copy_vector(c10::Half *dst, + const c10::Half *src) { + *((float2 *)dst) = *((float2 *)src); +} + +template <> +__device__ __inline__ void copy_vector(uint8_t *dst, + const uint8_t *src) { + *dst = *src; +} + +template <> +__device__ __inline__ void copy_vector(uint8_t *dst, + const uint8_t *src) { + *((half2 *)dst) = *((half2 *)src); +} + +int log2_ceil(int value) { + int log2_value = 0; + while ((1 << log2_value) < value) ++log2_value; + return log2_value; +} + +template struct Add { - __device__ __forceinline__ T operator()(T a, T b) const { - return a + b; - } + __device__ __forceinline__ T operator()(T a, T b) const { return a + b; } }; -template +template struct Max { __device__ __forceinline__ T operator()(T a, T b) const { return a < b ? b : a; @@ -55,438 +71,468 @@ struct Max { }; template -__device__ __forceinline__ T WARP_SHFL_XOR_NATIVE(T value, int laneMask, int width = warpSize, unsigned int mask = 0xffffffff) -{ +__device__ __forceinline__ T +WARP_SHFL_XOR_NATIVE(T value, int laneMask, int width = warpSize, + unsigned int mask = 0xffffffff) { #if CUDA_VERSION >= 9000 - return __shfl_xor_sync(mask, value, laneMask, width); + return __shfl_xor_sync(mask, value, laneMask, width); #else - return __shfl_xor(value, laneMask, width); + return __shfl_xor(value, laneMask, width); #endif } -template class ReduceOp> -__device__ __forceinline__ void warp_reduce(acc_t* sum) { - ReduceOp r; - #pragma unroll - for (int offset = WARP_SIZE / 2; offset > 0; offset /= 2) { - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - acc_t b = WARP_SHFL_XOR_NATIVE(sum[i], offset, WARP_SIZE); - sum[i] = r(sum[i], b); - } +template class ReduceOp> +__device__ __forceinline__ void warp_reduce(acc_t *sum) { + ReduceOp r; +#pragma unroll + for (int offset = WARP_SIZE / 2; offset > 0; offset /= 2) { +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + acc_t b = WARP_SHFL_XOR_NATIVE(sum[i], offset, WARP_SIZE); + sum[i] = r(sum[i], b); } + } } /* - * Extended softmax (from native aten pytorch) with following additional features - * 1) input scaling - * 2) Explicit masking - */ -template + * Extended softmax (from native aten pytorch) with following additional + * features 1) input scaling 2) Explicit masking + */ +template __global__ void scaled_masked_softmax_warp_forward( - output_t *dst, - const input_t *src, - const uint8_t *mask, - const acc_t scale, - int micro_batch_size, - int element_count, - int pad_batches) -{ - // WARP_SIZE and WARP_BATCH must match the return values batches_per_warp and - // warp_size of method warp_softmax_forward_kernel. - constexpr int next_power_of_two = 1 << log2_elements; - constexpr int WARP_SIZE = (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; - constexpr int WARP_ITERATIONS = next_power_of_two / WARP_SIZE; - constexpr int WARP_BATCH = (next_power_of_two <= 128) ? 2 : 1; - constexpr int ELEMENTS_PER_LDG_STG = (WARP_ITERATIONS < 4) ? 1 : 4; + output_t *dst, const input_t *src, const uint8_t *mask, const acc_t scale, + int micro_batch_size, int element_count, int pad_batches) { + // WARP_SIZE and WARP_BATCH must match the return values batches_per_warp and + // warp_size of method warp_softmax_forward_kernel. + constexpr int next_power_of_two = 1 << log2_elements; + constexpr int WARP_SIZE = + (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + constexpr int WARP_ITERATIONS = next_power_of_two / WARP_SIZE; + constexpr int WARP_BATCH = (next_power_of_two <= 128) ? 2 : 1; + constexpr int ELEMENTS_PER_LDG_STG = (WARP_ITERATIONS < 4) ? 1 : 4; - // blockDim/threadIdx = (WARP_SIZE, WARPS_PER_BLOCK, ) - // gridDim/blockIdx = (seq_len, attn_heads, batches) - int first_batch = (blockDim.y * (blockIdx.x + gridDim.x * (blockIdx.y + gridDim.y * blockIdx.z))+ threadIdx.y) * WARP_BATCH; - int pad_first_batch = 0; - if (pad_batches != 1) { // bert style - pad_first_batch = (blockDim.y * (blockIdx.x + gridDim.x * blockIdx.z) + threadIdx.y) * WARP_BATCH; - } else { // gpt2 style - pad_first_batch = (blockDim.y * blockIdx.x + threadIdx.y) * WARP_BATCH; - } + // blockDim/threadIdx = (WARP_SIZE, WARPS_PER_BLOCK, ) + // gridDim/blockIdx = (seq_len, attn_heads, batches) + int first_batch = + (blockDim.y * + (blockIdx.x + gridDim.x * (blockIdx.y + gridDim.y * blockIdx.z)) + + threadIdx.y) * + WARP_BATCH; + int pad_first_batch = 0; + if (pad_batches != 1) { // bert style + pad_first_batch = + (blockDim.y * (blockIdx.x + gridDim.x * blockIdx.z) + threadIdx.y) * + WARP_BATCH; + } else { // gpt2 style + pad_first_batch = (blockDim.y * blockIdx.x + threadIdx.y) * WARP_BATCH; + } - // micro_batch_size might not be a multiple of WARP_BATCH. Check how - // many batches have to computed within this WARP. - int local_batches = micro_batch_size - first_batch; - if (local_batches > WARP_BATCH) - local_batches = WARP_BATCH; + // micro_batch_size might not be a multiple of WARP_BATCH. Check how + // many batches have to computed within this WARP. + int local_batches = micro_batch_size - first_batch; + if (local_batches > WARP_BATCH) local_batches = WARP_BATCH; - // there might be multiple batches per warp. compute the index within the batch - int local_idx = threadIdx.x; + // there might be multiple batches per warp. compute the index within the + // batch + int local_idx = threadIdx.x; - src += first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; - dst += first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; - mask += pad_first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; + src += first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; + dst += first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; + mask += pad_first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; - // load data from global memory - acc_t elements[WARP_BATCH][WARP_ITERATIONS]; - input_t temp_data[ELEMENTS_PER_LDG_STG]; - uint8_t temp_mask[ELEMENTS_PER_LDG_STG]; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - int batch_element_count = (i >= local_batches) ? 0 : element_count; + // load data from global memory + acc_t elements[WARP_BATCH][WARP_ITERATIONS]; + input_t temp_data[ELEMENTS_PER_LDG_STG]; + uint8_t temp_mask[ELEMENTS_PER_LDG_STG]; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + int batch_element_count = (i >= local_batches) ? 0 : element_count; - #pragma unroll - for (int it = 0; it < WARP_ITERATIONS; it+=ELEMENTS_PER_LDG_STG) { - int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; +#pragma unroll + for (int it = 0; it < WARP_ITERATIONS; it += ELEMENTS_PER_LDG_STG) { + int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; - if (element_index < batch_element_count) { - int itr_idx = i*element_count+it*WARP_SIZE; - copy_vector(temp_data, src + itr_idx); - copy_vector(temp_mask, mask + itr_idx); + if (element_index < batch_element_count) { + int itr_idx = i * element_count + it * WARP_SIZE; + copy_vector(temp_data, src + itr_idx); + copy_vector(temp_mask, mask + itr_idx); - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - if (temp_mask[element] != 1) { - elements[i][it + element] = (acc_t)temp_data[element] * scale; - } else { - elements[i][it + element] = -10000.0; - } - } - } else { - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - elements[i][it + element] = -std::numeric_limits::infinity(); - } - } +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + if (temp_mask[element] != 1) { + elements[i][it + element] = (acc_t)temp_data[element] * scale; + } else { + elements[i][it + element] = -10000.0; + } } - } - - // compute max_value - acc_t max_value[WARP_BATCH]; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - max_value[i] = elements[i][0]; - #pragma unroll - for (int it = 1; it < WARP_ITERATIONS; ++it) { - max_value[i] = (max_value[i] > elements[i][it]) ? max_value[i] : elements[i][it]; + } else { +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + elements[i][it + element] = -std::numeric_limits::infinity(); } + } } - warp_reduce(max_value); + } - acc_t sum[WARP_BATCH] { 0.0f }; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - #pragma unroll - for (int it = 0; it < WARP_ITERATIONS; ++it) { - elements[i][it] = std::exp((elements[i][it] - max_value[i])); - sum[i] += elements[i][it]; - } + // compute max_value + acc_t max_value[WARP_BATCH]; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + max_value[i] = elements[i][0]; +#pragma unroll + for (int it = 1; it < WARP_ITERATIONS; ++it) { + max_value[i] = + (max_value[i] > elements[i][it]) ? max_value[i] : elements[i][it]; } - warp_reduce(sum); + } + warp_reduce(max_value); - // store result - output_t out[ELEMENTS_PER_LDG_STG]; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - if (i >= local_batches) - break; - #pragma unroll - for (int it = 0; it < WARP_ITERATIONS; it+=ELEMENTS_PER_LDG_STG) { - int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; - if (element_index < element_count) { - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - out[element] = elements[i][it + element] / sum[i]; - } - copy_vector(dst + i * element_count + it * WARP_SIZE, out); - } else { - break; - } - } + acc_t sum[WARP_BATCH]{0.0f}; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { +#pragma unroll + for (int it = 0; it < WARP_ITERATIONS; ++it) { + elements[i][it] = std::exp((elements[i][it] - max_value[i])); + sum[i] += elements[i][it]; } + } + warp_reduce(sum); + + // store result + output_t out[ELEMENTS_PER_LDG_STG]; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + if (i >= local_batches) break; +#pragma unroll + for (int it = 0; it < WARP_ITERATIONS; it += ELEMENTS_PER_LDG_STG) { + int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; + if (element_index < element_count) { +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + out[element] = elements[i][it + element] / sum[i]; + } + copy_vector( + dst + i * element_count + it * WARP_SIZE, out); + } else { + break; + } + } + } } -template +template __global__ void scaled_masked_softmax_warp_backward( - output_t *gradInput, - input_t *grad, - const input_t *output, - acc_t scale, - int micro_batch_size, - int element_count) -{ - // WARP_SIZE and WARP_BATCH must match the return values batches_per_warp and - // warp_size of method warp_softmax_backward_kernel. - constexpr int next_power_of_two = 1 << log2_elements; - constexpr int WARP_SIZE = (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; - constexpr int WARP_ITERATIONS = next_power_of_two / WARP_SIZE; - constexpr int WARP_BATCH = (next_power_of_two <= 128) ? 2 : 1; - constexpr int ELEMENTS_PER_LDG_STG = (WARP_ITERATIONS < 4) ? 1 : 4; + output_t *gradInput, input_t *grad, const input_t *output, acc_t scale, + int micro_batch_size, int element_count) { + // WARP_SIZE and WARP_BATCH must match the return values batches_per_warp and + // warp_size of method warp_softmax_backward_kernel. + constexpr int next_power_of_two = 1 << log2_elements; + constexpr int WARP_SIZE = + (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + constexpr int WARP_ITERATIONS = next_power_of_two / WARP_SIZE; + constexpr int WARP_BATCH = (next_power_of_two <= 128) ? 2 : 1; + constexpr int ELEMENTS_PER_LDG_STG = (WARP_ITERATIONS < 4) ? 1 : 4; - // blockDim/threadIdx = (WARP_SIZE, WARPS_PER_BLOCK, ) - // gridDim/blockIdx = (seq_len, attn_heads, batches) - int first_batch = (blockDim.y * blockIdx.x + threadIdx.y) * WARP_BATCH; - - // micro_batch_size might not be a multiple of WARP_BATCH. Check how - // many batches have to computed within this WARP. - int local_batches = micro_batch_size - first_batch; - if (local_batches > WARP_BATCH) - local_batches = WARP_BATCH; + // blockDim/threadIdx = (WARP_SIZE, WARPS_PER_BLOCK, ) + // gridDim/blockIdx = (seq_len, attn_heads, batches) + int first_batch = (blockDim.y * blockIdx.x + threadIdx.y) * WARP_BATCH; - // there might be multiple batches per warp. compute the index within the batch - int local_idx = threadIdx.x; + // micro_batch_size might not be a multiple of WARP_BATCH. Check how + // many batches have to computed within this WARP. + int local_batches = micro_batch_size - first_batch; + if (local_batches > WARP_BATCH) local_batches = WARP_BATCH; - // the first element to process by the current thread - int thread_offset = first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; - grad += thread_offset; - output += thread_offset; - gradInput += thread_offset; + // there might be multiple batches per warp. compute the index within the + // batch + int local_idx = threadIdx.x; - // load data from global memory - acc_t grad_reg[WARP_BATCH][WARP_ITERATIONS] { 0.0f }; - acc_t output_reg[WARP_BATCH][WARP_ITERATIONS] { 0.0f }; - input_t temp_grad[ELEMENTS_PER_LDG_STG]; - input_t temp_output[ELEMENTS_PER_LDG_STG]; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - int batch_element_count = (i >= local_batches) ? 0 : element_count; + // the first element to process by the current thread + int thread_offset = + first_batch * element_count + ELEMENTS_PER_LDG_STG * local_idx; + grad += thread_offset; + output += thread_offset; + gradInput += thread_offset; - #pragma unroll - for (int it = 0; it < WARP_ITERATIONS; it+=ELEMENTS_PER_LDG_STG) { - int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; - if (element_index < batch_element_count) { - copy_vector(temp_grad, grad + i * element_count + it * WARP_SIZE); - copy_vector(temp_output, output + i * element_count + it * WARP_SIZE); + // load data from global memory + acc_t grad_reg[WARP_BATCH][WARP_ITERATIONS]{0.0f}; + acc_t output_reg[WARP_BATCH][WARP_ITERATIONS]{0.0f}; + input_t temp_grad[ELEMENTS_PER_LDG_STG]; + input_t temp_output[ELEMENTS_PER_LDG_STG]; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + int batch_element_count = (i >= local_batches) ? 0 : element_count; - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - output_reg[i][it + element] = (acc_t)temp_output[element]; - } - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - grad_reg[i][it + element] = (acc_t)temp_grad[element] * output_reg[i][it + element]; - } - } +#pragma unroll + for (int it = 0; it < WARP_ITERATIONS; it += ELEMENTS_PER_LDG_STG) { + int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; + if (element_index < batch_element_count) { + copy_vector( + temp_grad, grad + i * element_count + it * WARP_SIZE); + copy_vector( + temp_output, output + i * element_count + it * WARP_SIZE); + +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + output_reg[i][it + element] = (acc_t)temp_output[element]; } - } - - acc_t sum[WARP_BATCH]; - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - sum[i] = grad_reg[i][0]; - #pragma unroll - for (int it = 1; it < WARP_ITERATIONS; ++it) { - sum[i] += grad_reg[i][it]; +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + grad_reg[i][it + element] = + (acc_t)temp_grad[element] * output_reg[i][it + element]; } + } } - warp_reduce(sum); + } - // store result - #pragma unroll - for (int i = 0; i < WARP_BATCH; ++i) { - if (i >= local_batches) - break; - #pragma unroll - for (int it = 0; it < WARP_ITERATIONS; it+=ELEMENTS_PER_LDG_STG) { - int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; - if (element_index < element_count) { - // compute gradients - output_t out[ELEMENTS_PER_LDG_STG]; - #pragma unroll - for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { - out[element] = (output_t)(scale * (grad_reg[i][it + element] - output_reg[i][it + element] * sum[i])); - } - copy_vector(gradInput + i * element_count + it * WARP_SIZE, out); - } - } + acc_t sum[WARP_BATCH]; +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + sum[i] = grad_reg[i][0]; +#pragma unroll + for (int it = 1; it < WARP_ITERATIONS; ++it) { + sum[i] += grad_reg[i][it]; } + } + warp_reduce(sum); + +// store result +#pragma unroll + for (int i = 0; i < WARP_BATCH; ++i) { + if (i >= local_batches) break; +#pragma unroll + for (int it = 0; it < WARP_ITERATIONS; it += ELEMENTS_PER_LDG_STG) { + int element_index = ELEMENTS_PER_LDG_STG * local_idx + it * WARP_SIZE; + if (element_index < element_count) { + // compute gradients + output_t out[ELEMENTS_PER_LDG_STG]; +#pragma unroll + for (int element = 0; element < ELEMENTS_PER_LDG_STG; ++element) { + out[element] = + (output_t)(scale * (grad_reg[i][it + element] - + output_reg[i][it + element] * sum[i])); + } + copy_vector( + gradInput + i * element_count + it * WARP_SIZE, out); + } + } + } } -} // end of anonymous namespace +} // end of anonymous namespace -int get_batch_per_block(int query_seq_len, int key_seq_len, int batches, int attn_heads){ +int get_batch_per_block(int query_seq_len, int key_seq_len, int batches, + int attn_heads) { + int log2_elements = log2_ceil(key_seq_len); + const int next_power_of_two = 1 << log2_elements; + + int warp_size = + (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + int batches_per_warp = (next_power_of_two <= 128) ? 2 : 1; + + constexpr int threads_per_block = 128; + int warps_per_block = (threads_per_block / warp_size); + int batches_per_block = warps_per_block * batches_per_warp; + + return batches_per_block; +} + +template +void dispatch_scaled_masked_softmax_forward(output_t *dst, const input_t *src, + const uint8_t *mask, + const input_t scale, + int query_seq_len, int key_seq_len, + int batches, int attn_heads, + int pad_batches) { + TORCH_INTERNAL_ASSERT(key_seq_len >= 0 && key_seq_len <= 2048); + if (key_seq_len == 0) { + return; + } else { int log2_elements = log2_ceil(key_seq_len); const int next_power_of_two = 1 << log2_elements; + int batch_count = batches * attn_heads * query_seq_len; - int warp_size = (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + // This value must match the WARP_SIZE constexpr value computed inside + // softmax_warp_forward. + int warp_size = + (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + + // This value must match the WARP_BATCH constexpr value computed inside + // softmax_warp_forward. int batches_per_warp = (next_power_of_two <= 128) ? 2 : 1; + // use 128 threads per block to maximimize gpu utilization constexpr int threads_per_block = 128; + int warps_per_block = (threads_per_block / warp_size); int batches_per_block = warps_per_block * batches_per_warp; - - return batches_per_block; -} - -template -void dispatch_scaled_masked_softmax_forward( - output_t *dst, - const input_t *src, - const uint8_t *mask, - const input_t scale, - int query_seq_len, - int key_seq_len, - int batches, - int attn_heads, - int pad_batches) -{ - TORCH_INTERNAL_ASSERT(key_seq_len >= 0 && key_seq_len <= 2048 ); - if (key_seq_len == 0) { - return; - } else { - int log2_elements = log2_ceil(key_seq_len); - const int next_power_of_two = 1 << log2_elements; - int batch_count = batches * attn_heads * query_seq_len; - - // This value must match the WARP_SIZE constexpr value computed inside softmax_warp_forward. - int warp_size = (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; - - // This value must match the WARP_BATCH constexpr value computed inside softmax_warp_forward. - int batches_per_warp = (next_power_of_two <= 128) ? 2 : 1; - - // use 128 threads per block to maximimize gpu utilization - constexpr int threads_per_block = 128; - - int warps_per_block = (threads_per_block / warp_size); - int batches_per_block = warps_per_block * batches_per_warp; - TORCH_INTERNAL_ASSERT(query_seq_len%batches_per_block == 0); - dim3 blocks(query_seq_len/batches_per_block, attn_heads, batches); - dim3 threads(warp_size, warps_per_block, 1); - // Launch code would be more elegant if C++ supported FOR CONSTEXPR - switch (log2_elements) { - case 0: // 1 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 1: // 2 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 2: // 4 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 3: // 8 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 4: // 16 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 5: // 32 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 6: // 64 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 7: // 128 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 8: // 256 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 9: // 512 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 10: // 1024 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - case 11: // 2048 - scaled_masked_softmax_warp_forward - <<>>(dst, src, mask, scale, batch_count, key_seq_len, pad_batches); - break; - default: - break; - } + TORCH_INTERNAL_ASSERT(query_seq_len % batches_per_block == 0); + dim3 blocks(query_seq_len / batches_per_block, attn_heads, batches); + dim3 threads(warp_size, warps_per_block, 1); + // Launch code would be more elegant if C++ supported FOR CONSTEXPR + switch (log2_elements) { + case 0: // 1 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 1: // 2 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 2: // 4 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 3: // 8 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 4: // 16 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 5: // 32 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 6: // 64 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 7: // 128 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 8: // 256 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 9: // 512 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 10: // 1024 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + case 11: // 2048 + scaled_masked_softmax_warp_forward + <<>>( + dst, src, mask, scale, batch_count, key_seq_len, pad_batches); + break; + default: + break; } + } } -template -void dispatch_scaled_masked_softmax_backward( - output_t *grad_input, - input_t *grad, - const input_t *output, - const acc_t scale, - int query_seq_len, - int key_seq_len, - int batches, - int attn_heads) -{ - TORCH_INTERNAL_ASSERT( key_seq_len >= 0 && key_seq_len <= 2048 ); - if (key_seq_len == 0) { - return; - } else { - int log2_elements = log2_ceil(key_seq_len); - const int next_power_of_two = 1 << log2_elements; - int batch_count = batches * attn_heads * query_seq_len; +template +void dispatch_scaled_masked_softmax_backward(output_t *grad_input, + input_t *grad, + const input_t *output, + const acc_t scale, + int query_seq_len, int key_seq_len, + int batches, int attn_heads) { + TORCH_INTERNAL_ASSERT(key_seq_len >= 0 && key_seq_len <= 2048); + if (key_seq_len == 0) { + return; + } else { + int log2_elements = log2_ceil(key_seq_len); + const int next_power_of_two = 1 << log2_elements; + int batch_count = batches * attn_heads * query_seq_len; - // This value must match the WARP_SIZE constexpr value computed inside softmax_warp_backward. - int warp_size = (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; + // This value must match the WARP_SIZE constexpr value computed inside + // softmax_warp_backward. + int warp_size = + (next_power_of_two < C10_WARP_SIZE) ? next_power_of_two : C10_WARP_SIZE; - // This value must match the WARP_BATCH constexpr value computed inside softmax_warp_backward. - int batches_per_warp = (next_power_of_two <= 128) ? 2 : 1; + // This value must match the WARP_BATCH constexpr value computed inside + // softmax_warp_backward. + int batches_per_warp = (next_power_of_two <= 128) ? 2 : 1; - // use 128 threads per block to maximimize gpu utilization - constexpr int threads_per_block = 128; + // use 128 threads per block to maximimize gpu utilization + constexpr int threads_per_block = 128; - int warps_per_block = (threads_per_block / warp_size); - int batches_per_block = warps_per_block * batches_per_warp; - int blocks = batch_count/batches_per_block; - dim3 threads(warp_size, warps_per_block, 1); - // Launch code would be more elegant if C++ supported FOR CONSTEXPR - switch (log2_elements) { - case 0: // 1 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 1: // 2 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 2: // 4 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 3: // 8 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 4: // 16 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 5: // 32 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 6: // 64 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 7: // 128 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 8: // 256 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 9: // 512 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 10: // 1024 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - case 11: // 2048 - scaled_masked_softmax_warp_backward - <<>>(grad_input, grad, output, scale, batch_count, key_seq_len); - break; - default: - break; - } + int warps_per_block = (threads_per_block / warp_size); + int batches_per_block = warps_per_block * batches_per_warp; + int blocks = batch_count / batches_per_block; + dim3 threads(warp_size, warps_per_block, 1); + // Launch code would be more elegant if C++ supported FOR CONSTEXPR + switch (log2_elements) { + case 0: // 1 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 1: // 2 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 2: // 4 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 3: // 8 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 4: // 16 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 5: // 32 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 6: // 64 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 7: // 128 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 8: // 256 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 9: // 512 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 10: // 1024 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + case 11: // 2048 + scaled_masked_softmax_warp_backward + <<>>( + grad_input, grad, output, scale, batch_count, key_seq_len); + break; + default: + break; } + } }